Compare commits

...

248 Commits

Author SHA1 Message Date
Teknium
401297d166 feat(cli): add --safe-mode troubleshooting flag
Inspired by Claude Code v2.1.169 (June 2026): run Hermes with all
customizations disabled to isolate setup problems from product bugs.

--safe-mode implies --ignore-user-config and --ignore-rules, and
additionally skips plugin discovery (hermes_cli/plugins.py) and MCP
server loading (tools/mcp_tool.py) via the internal HERMES_SAFE_MODE
env bridge.
2026-06-12 17:23:30 -07:00
Teknium
9688c1a94f chore: add Kimi K2.7 code catalog slug (#45283) 2026-06-12 16:55:40 -07:00
Teknium
7e46533d9f test: compressed-summary metadata flag set in-process, stripped on wire 2026-06-12 16:47:15 -07:00
kyssta-exe
956af7f3c3 fix(agent): add metadata flag to context compression summary messages (#38389)
Summary messages (standalone insertion and merge-into-tail) now carry a
metadata flag so frontends (CLI, Desktop, gateway, TUI) can distinguish
them from real assistant/user messages without content-prefix heuristics.

Re-applied from PR #38434 onto current main (conflicted with the
_SUMMARY_END_MARKER hoist). Key renamed from the PR's
'is_compressed_summary' to '_compressed_summary': the wire sanitizers
strip underscore-prefixed message keys, so the flag stays in-process and
can never reach strict gateways (Fireworks/Mistral/Kimi reject unknown
keys with 'Extra inputs are not permitted').
2026-06-12 16:47:15 -07:00
helix4u
1899c8f507 fix(skills): run youtube transcript helper through uv 2026-06-12 16:33:46 -07:00
Teknium
8905ee6b8a fix(agent): rewind flush cursor exactly when repair compacts before the cursor
Follow-up to the #44837 clamp: a min() clamp only fixes cursor overshoot
past the new end of the list. When repair_message_sequence drops/merges
messages at indexes below the cursor, the clamp leaves the cursor pointing
past unflushed rows and the turn-end flush silently skips them.

Extract repair_message_sequence_with_cursor(): snapshot the flushed prefix
by object identity before repair, then recompute the cursor as the count
of surviving flushed messages. Falls back to the clamp when no snapshot is
available. Keeps the safety guard in _flush_messages_to_session_db.

Adds targeted tests for overshoot, before-cursor compaction, no-repair,
bare-agent, and the flush guard.
2026-06-12 16:29:01 -07:00
kyssta-exe
5d0408d9fe fix(agent): clamp flush cursor after repair_message_sequence compaction (#44837) 2026-06-12 16:29:01 -07:00
konsisumer
aec38855b5 fix(agent): preserve recent turns during compression 2026-06-12 16:26:58 -07:00
brooklyn!
bbf020e709 feat(desktop): follow streaming output at bottom + jump-to-bottom button (#45263)
Strict sticky-bottom autoscroll for the chat thread: while the viewport is
parked at the bottom, the tail follows content growth (streaming tokens, late
measurement, Shiki re-highlight) via a useLayoutEffect keyed on the
virtualizer's own size signal, pinned in the same pre-paint pass as its
scrollToFn so the two never rubber-band. The gate is a single boolean — one
upward pixel (scroll/wheel/touch) disarms follow until the user returns to the
bottom.

Adds a floating jump-to-bottom control that appears once scrolled ~10px away
(above the dim threshold so a sub-pixel settle never flashes it), positioned
above the composer with respect to the status stack, with a subtle
scale + slide in/out animation that honours prefers-reduced-motion. The button
bridges to the virtualizer's re-arm + pin path through a small nanostore
emitter.

Supersedes #43624.
2026-06-12 23:00:11 +00:00
Teknium
135fe90166 fix(profiles): backfill .env for pre-existing profiles on hermes update (#45247)
Profiles created before #44792 have no .env. Now that the Channels/Keys
endpoints are profile-scoped (no os.environ fallback), those profiles
would show everything as unconfigured. hermes update now copies the
default install's .env into each named profile that lacks one (0600,
never overwrites, placeholder fallback when the root has no .env), so
existing users keep the credentials they were effectively running with.
2026-06-12 15:42:14 -07:00
xxxigm
68536d4375 test(compressor): regression coverage for assistant-tail anchor + compaction rollup (#29824)
21 cases pinning the new ``_ensure_last_assistant_message_in_tail``
anchor and its interaction with the existing tail-cut path:

* ``TestFindLastAssistantMessageIdx`` — helper contract: prefers a
  content-bearing assistant message, skips ``tool_calls``-only
  stubs, multimodal text-block content counts, falls back to
  "any assistant" when no content-bearing reply exists, honours
  ``head_end``, returns -1 when there's none.

* ``TestEnsureLastAssistantMessageInTail`` — direct: no-op when
  already in the tail, walks ``cut_idx`` back when the reply is
  in the compressed middle, never crosses into the head region,
  re-aligns through a preceding ``tool_call`` / ``tool_result``
  group instead of orphaning it.

* ``TestFindTailCutByTokensAnchorsAssistant`` — integration:
  reporter repro (long tool-output run after the visible reply)
  now preserves the reply; user and assistant anchors compose
  in a single tail-cut call; a soft-ceiling-overrunning oversized
  tool result no longer strands the prior reply.

* ``TestCompactionRollupReproduction`` — end-to-end through
  ``compress()`` with a stubbed ``_generate_summary``: the
  visible reply text survives either as its own standalone
  assistant message (normal path) or concatenated onto the
  merged summary tail (double-collision path the WebUI then
  re-splits). The standalone-summary case is asserted strictly
  (exactly one summary row, exactly one separate assistant
  row carrying the reply) — that's the dominant path and any
  drift there reintroduces the original bug.

* ``TestSourceGuardrail`` — static asserts on
  ``agent/context_compressor.py``: the helper exists, the
  anchor is wired into ``_find_tail_cut_by_tokens`` AFTER the
  user-message anchor (so chaining is monotonic), the
  content-bearing preference is preserved, and the issue
  number is referenced so future bisects can find this fix.
2026-06-12 15:41:57 -07:00
xxxigm
2fef3e2df2 fix(webui): split merge-into-tail compaction so reply renders as its own bubble (#29824)
The compressor has a "double-collision" fallback path: when the
chosen ``summary_role`` collides with the first tail message AND
the flipped role would collide with the last head message, it can't
emit a standalone summary turn (consecutive same-role messages
break Anthropic and friends). It instead prepends the summary +
end-of-summary marker to the first tail message's content via
``_merge_summary_into_tail``.

With the matching anchor from the previous commit, that first tail
message is now usually the user's previously-visible assistant
reply — so the persisted assistant turn ends up shaped as
``[CONTEXT COMPACTION ...] ... --- END OF CONTEXT SUMMARY --- ...
THE ACTUAL REPLY``. Without splitting it, the session viewer
renders one big "Context handoff" bubble and the reply text is
buried inside the metadata blob — which is exactly the
"can't see the last reply" experience #29824 reports, just one
layer deeper.

Added ``splitCompactionContent`` that detects the merge marker
(kept in sync with ``--- END OF CONTEXT SUMMARY — respond to the
message below, not the summary above ---`` in
``agent/context_compressor.py``) and ``MessageBubble`` now
recurses on the two halves: the prefix half renders as the muted
"Context handoff" row, the remainder half renders with the
original assistant styling. Pure (non-merged) summary messages
hit the no-remainder branch and still render as a single
"Context handoff" row, preserving the original behaviour.
2026-06-12 15:41:57 -07:00
xxxigm
691ff7c188 fix(compressor): keep last visible assistant reply out of compaction summary + label handoffs in WebUI (#29824)
Two-pronged fix for the WebUI "context compaction block in place of
last assistant response" regression.

Agent layer (the real fix). ``_find_tail_cut_by_tokens`` already had
``_ensure_last_user_message_in_tail`` to keep the most recent user
request out of the compressed middle (#10896), but no symmetric
anchor for the assistant side. When the conversation has an
oversized recent tool result or a long stretch of tool-call/result
pairs *after* the assistant's last visible reply, the token-budget
walk can stop with the previously-visible reply on the wrong side
of ``cut_idx``. The summariser then rolls it into the single
``[CONTEXT COMPACTION — REFERENCE ONLY]`` block persisted as
``role="user"`` or ``role="assistant"``, and from the operator's
perspective the WebUI session viewer
(``web/src/pages/SessionsPage.tsx``) and the TUI chat panel both
suddenly show the opaque "Context compaction" block in the slot
where they were just reading the actual answer:

    User:  "i cant see the output of the last message you sent,
            i did see it previously, however now see 'context
            compaction'"

Added ``_ensure_last_assistant_message_in_tail`` mirror of the
user-side anchor. It looks for the most recent assistant message
with non-empty text content (skipping tool-call-only assistant
"stubs" which the UI renders as small "calling tool X" indicators
rather than a readable bubble) and walks ``cut_idx`` back through
the standard ``_align_boundary_backward`` so we don't split a
tool_call/result group that immediately precedes it. The two
anchors are chained — each only walks ``cut_idx`` backward, so
the tail can only grow.

Falls back to "most recent assistant of any kind" only when no
content-bearing reply exists in the compressible region (fresh
multi-step tool sequence with no prior reply) — in that case the
agent-side fix is effectively a no-op and the existing
user-message anchor carries the load.

WebUI layer (clarity). Added ``isCompactionMessage`` detector that
recognises the ``[CONTEXT COMPACTION — REFERENCE ONLY]`` (current)
and ``[CONTEXT SUMMARY]:`` (legacy) prefixes from
``agent/context_compressor.py``, and a new ``compaction`` entry
in ``MessageBubble``'s ``ROLE_STYLES`` map. Compaction blocks
now render as muted, italicised system-style rows labelled
``Context handoff`` — clearly metadata, not the assistant's
actual reply — so an operator scrolling back through a long
session can't mistake the summary for a real answer.

Keeping the detected prefixes inline (rather than importing them)
because the WebUI bundle has no Python interop. A guardrail comment
points readers at the source-of-truth constants in
``agent/context_compressor.py``.
2026-06-12 15:41:57 -07:00
Teknium
7a318aae22 fix(profiles): exclude session history, backups, and snapshots from --clone-all (#45246)
--clone-all copied the source profile's state.db, sessions/, backups/,
state-snapshots/, and checkpoints/ into the new profile. These are
per-profile history: a 49GB copy in practice (15GB snapshots + 11GB
backup archives + 16GB state.db + 6.4GB sessions), and restoring a
copied backup inside the clone would resurrect the SOURCE profile's
state. A clone is a fresh workspace; history stays with the source.

New _CLONE_ALL_HISTORY_EXCLUDE_ROOT set, applied at root level for ANY
source profile (named profiles accumulate the same artifacts), unlike
the default-gated infrastructure excludes. Nested same-name dirs still
copy. Docs and the post-create CLI message updated to match; profile
export / hermes backup remain the full-history paths.
2026-06-12 15:41:50 -07:00
Teknium
0db5cb8e75 refactor(agent): hoist summary end marker to _SUMMARY_END_MARKER; strip it on rehydration
Follow-up to the #33346 cherry-pick:
- the marker string was duplicated at both insertion sites (standalone +
  merged-into-tail); hoist to a module constant
- _strip_summary_prefix now also strips a trailing end marker so a
  rehydrated handoff body doesn't leak the boundary directive into the
  iterative-update summarizer prompt (it is re-appended on insertion)
2026-06-12 15:05:00 -07:00
Tranquil-Flow
749b7219c4 fix(compression): always append END OF CONTEXT SUMMARY marker to standalone summaries regardless of role
When the compression summary lands as an assistant-role message (head ends
with user), the end marker was not appended. Models may regurgitate the
summary text as their own visible output when there's no clear boundary
signal (#33256).

The end marker was already appended for user-role summaries (#11475, #14521)
but the assistant-role path was missed in the original fix. This ensures ALL
standalone summary messages carry the boundary marker, preventing summary
text from leaking into user-visible chat output.
2026-06-12 15:05:00 -07:00
Teknium
a118b94a85 fix(dashboard): skill installs from the dashboard silently auto-cancel (#45150)
The dashboard's /api/skills/hub/install (and the new-profile hub_skills
path) spawned `hermes skills install <id>` with stdin=DEVNULL but
without --yes. do_install()'s 'Confirm [y/N]' prompt hit EOF, defaulted
to 'n', and printed 'Installation cancelled.' into a background log the
user never sees — every dashboard install no-opped.

Pass --yes on both spawn sites, matching the uninstall endpoint which
already passed --yes. The dashboard install button is the explicit user
consent, same as the TUI/slash-command skip_confirm rationale.

Repro: spawned the exact argv with stdin=DEVNULL against a temp
HERMES_HOME — without --yes it cancels, with --yes the skill installs.
2026-06-12 12:58:36 -07:00
Teknium
bba9b519aa fix(delegation): remove the default subagent wall-clock timeout (#45149)
Subagents doing legitimate heavy work (deep code reviews, research
fan-outs, slow reasoning models) were routinely killed at the blanket
600s child_timeout_seconds cap while making steady progress (e.g. 36
API calls completed when the axe fell). Failures should come from what
the child is actually doing — API errors, tool errors, iteration
budget — not a delegation-level stopwatch.

- DEFAULT_CHILD_TIMEOUT: 600 -> None; Future.result(timeout=None)
  blocks until the child finishes
- config default delegation.child_timeout_seconds: 600 -> 0
  (0/negative = disabled; positive opts back in, floor 30s unchanged)
- stuck-child protection unchanged: the heartbeat staleness monitor
  still stops refreshing parent activity so the gateway inactivity
  timeout fires on a truly wedged worker; the 0-API-call diagnostic
  dump still works when a cap is configured
- docs updated (EN + zh-Hans)
2026-06-12 12:58:25 -07:00
Teknium
9b01c4d193 fix(update): never spawn an interactive polkit prompt when restarting a system-scope gateway (#45145)
When hermes update restarts a hermes-gateway system service as a
non-root user, the systemctl reset-failed/start/restart calls trigger
polkit's org.freedesktop.systemd1.manage-units TTY authentication
agent. That prompt runs inside a captured subprocess with a 10-15s
timeout, so it flashes and dies before the user can answer, and the
resulting TimeoutExpired was swallowed silently by the loop's blanket
except — the restart phase just vanished with no output.

- Resolve a manage-units command prefix up front: plain systemctl as
  root, sudo -n systemctl as non-root (with a targeted reset-failed
  probe so least-privilege sudoers entries scoped to hermes-gateway*
  qualify), or None when no non-interactive privilege path exists.
- Add --no-ask-password to every manage-units call in the update
  restart path so polkit can never prompt inside a captured subprocess.
- When unprivileged: after a graceful drain, rely on systemd's own
  RestartSec auto-restart (needs no privileges) with a message about
  the wait; skip the force-restart fallback with clear manual
  instructions instead of racing a doomed polkit prompt.
- Surface TimeoutExpired in the restart loop instead of passing
  silently, and add sudo to the system-scope recovery hints.
- Docs: headless-VM note recommending user service + enable-linger,
  or sudo updates / a scoped NOPASSWD sudoers entry for system
  services.
2026-06-12 12:38:15 -07:00
Teknium
fca84fe20b test: regression guard for Nous 429 fallback re-entry; AUTHOR_MAP entry 2026-06-12 12:21:29 -07:00
Aðalsteinn Helgason
2714fc8396 fix(agent): re-enter retry loop on genuine Nous 429 so fallback guard runs
The genuine-rate-limit branch set retry_count = max_retries before
continue, intending the top-of-loop Nous guard to handle fallback or
bail cleanly. But the loop condition is retry_count < max_retries, so
the guard never ran: no fallback activation, no clean rate-limit
message — just the generic retry-exhaustion error.

Set retry_count = max(0, max_retries - 1) so the loop body runs exactly
once more and the guard sees the breaker state recorded moments earlier.

Extracted from the #44061 bugfix rollup by @AIalliAI.
2026-06-12 12:21:29 -07:00
Teknium
dc467488a7 test: assert typing-stop-before-callback as an invariant, not a call count
The shared _stop_typing_refresh cleanup makes up to two bounded
stop_typing attempts; the old assertion pinned exactly one
typing-stopped event before callback-start.
2026-06-12 12:02:41 -07:00
Teknium
c2326bc3be chore: add itsflownium to AUTHOR_MAP 2026-06-12 12:02:41 -07:00
Flownium
331cb38e21 fix: stop Discord typing after replies 2026-06-12 12:02:41 -07:00
Teknium
fa5e98facb fix(send): helpful error when --file gets a binary; document MEDIA: attachments (#45116)
A user passing an image to `hermes send --file` got a raw
UnicodeDecodeError ('utf-8 codec can't decode byte 0x89...') with no
hint that media delivery goes through the MEDIA:<path> directive.

- send_cmd: catch UnicodeDecodeError separately and print a usage error
  explaining --file is for text bodies, with copy-pasteable MEDIA: and
  [[as_document]] examples using the user's own path
- --file help text + epilog now mention MEDIA:
- docs: new 'Sending images and other media' section on the hermes send
  reference page
2026-06-12 11:48:06 -07:00
Teknium
652dd9c9f2 fix: rich messages follow-ups — reply_parameters, send latch, opt-in default
- Use reply_parameters per the sendRichMessage spec instead of the
  undocumented reply_to_message_id scalar (silently ignored -> reply
  anchor quietly dropped).
- Latch rich sends off after an endpoint-capability failure (old PTB /
  server without sendRichMessage) so every later reply doesn't pay a
  doomed extra roundtrip; per-message BadRequests do NOT latch.
- Default rich_messages to OFF (opt-in) while the day-old Bot API 10.1
  endpoint is validated live; revert the prompt-hint table guidance
  until the default flips on.
- Tests: reply_parameters shape, send-latch behavior, BadRequest
  non-latch; rich tests opt in explicitly via extra.
2026-06-12 11:47:54 -07:00
ITheEqualizer
05b9c84ca4 Add Telegram Bot API 10.1 rich message support
Introduce opportunistic support for Telegram Bot API 10.1 rich messages by sending raw agent Markdown via sendRichMessage and streaming previews via sendRichMessageDraft. Implements a rich-path fast‑path in gateway/platforms/telegram.py (RICH_MESSAGE_MAX_BYTES=32768, feature gate platforms.telegram.extra.rich_messages, bot capability checks, routing/thread handling, and conservative fallback rules: permanent/capability errors fall back to the legacy MarkdownV2 path, transient/network errors are surfaced without legacy-resend). Also add a latch for draft capability failures (_rich_draft_disabled) and preserve legacy chunking and draft behavior when needed. Update agent prompt hints (telegram encourages rich Markdown/tables), add CLI config example option, update English and Chinese docs to describe rich messages and fallbacks, and add/adjust tests for rich send and draft behavior.
2026-06-12 11:47:54 -07:00
teknium1
6b4073648e fix(tui): config.yaml wins over env model seed in per-turn sync
Hosted instances set HERMES_INFERENCE_MODEL as a provision-time seed in
the container env. _config_model_target() previously went through
_resolve_model() (env-first), so on hosted VPS the sync target stayed
pinned to the seed and dashboard model changes never reached an open
chat -- the exact scenario the sync exists to fix. The sync target now
reads config.yaml first and only falls back to the env vars when config
has no model. Startup resolution (_resolve_model) is unchanged.
2026-06-12 11:03:44 -07:00
IAvecilla
bc3f4ed70f Skip redundant model switch 2026-06-12 11:03:44 -07:00
IAvecilla
8c3c08c50b Update implementation to make it cleaner 2026-06-12 11:03:44 -07:00
IAvecilla
c61815232a Update model correctly when updating from dashboard 2026-06-12 11:03:44 -07:00
ethernet
1e25358a8f refactor(desktop): use port 0 for ephemeral port discovery instead of PortPool reservation
Replace the PortPool-based port reservation system (9120-9199 range) with OS-assigned ephemeral ports via --port 0.

Before: Desktop probed a hardcoded port range, reserved ports in-process to close TOCTOU races, and passed the chosen port to the dashboard via CLI arg.

After: Desktop spawns dashboard with --port 0, parses the actual port from a stdout announcement line (HERMES_DASHBOARD_READY port=<N>), and uses that for WebSocket connections.

Changes:
- web_server.py: add --port 0 support with SO_REUSEADDR pre-bind + announcement; add EADDRINUSE preflight for explicit ports
- main.cjs: remove PortPool, PORT_FLOOR/CEILING, pickPort(), isPortAvailable(); add waitForDashboardPort() stdout parser
- Delete port-pool.cjs and port-pool.test.cjs (106 lines removed)

Net effect: eliminates the entire TOCTOU-mitigation reservation infrastructure and arbitrary port range constraints. OS handles port allocation natively.
2026-06-12 14:02:19 -04:00
ethernet
8044bf0206 fix(ci): only save test durations when tests pass
The save-durations job used `if: always()` which meant it would
run even when the test matrix failed, potentially caching duration
data from a failed/incomplete run. Changed to check
needs.test.result == 'success' so durations are only cached when
all test slices pass cleanly.
2026-06-12 13:50:52 -04:00
ethernet
4d68984ec7 fix(tests): remove no-longer-needed forensics 2026-06-12 13:42:42 -04:00
ethernet
6ff39c31ad fix(tests): guard against real 'hermes update' subprocess spawns in conftest
Extends _live_system_guard in tests/conftest.py to block any subprocess
call that would run 'hermes update' (or 'python -m hermes_cli.main update')
against the real checkout.

These commands run git fetch origin + git pull, overwriting repo files
like pyproject.toml mid-test-run and corrupting every subsequent
subprocess that reads them. The spawned process uses setsid /
start_new_session=True so it's invisible to pytest's process tree
(PPid=1) — the corruption was essentially undetectable without
explicit inotify/SHA watchdogs.

Root cause of #43703 CI failures: tests in TestUpdateCommandPlatformGate
called _handle_update_command() with HERMES_MANAGED='' and no Popen mock,
causing the code to fall through and spawn a real 'hermes update --gateway'
that overwrote pyproject.toml with origin/main's content (which still
had '--timeout=30 --timeout-method=thread' in addopts while the PR had
already removed pytest-timeout).

The guard covers all three invocation patterns:
- 'hermes update' / 'hermes update --gateway' (direct or via setsid bash -c)
- 'python -m hermes_cli.main update --gateway'
- '.venv/bin/hermes update' (absolute path variant)

Does not false-positive on: git update-index, apt-get update,
pip install --upgrade, or any command lacking 'hermes'/'hermes_cli'.
2026-06-12 13:42:42 -04:00
ethernet
c41a6534cf fix(tests): mock subprocess.Popen in all _handle_update_command tests 2026-06-12 13:42:42 -04:00
ethernet
2f9d18711f fix(ci): remove pytest-timeout, use per-file timeout only
fix(ci): write a new cache for test durations every time
change(ci): rip out error 4 retries because we found the real bug
2026-06-12 13:42:42 -04:00
brooklyn!
46d758bb3e feat(desktop): window translucency slider in Appearance settings (#45086)
A see-through-window control (0–100, off by default) that maps to the
native window opacity via setOpacity — the desktop shows through the whole
window, the same effect as the Windows shift-scroll trick. macOS + Windows;
a no-op on Linux (no runtime window opacity).

Renderer owns the value (persisted, nanostore) and mirrors it to the main
process over IPC; main persists it to translucency.json so a cold launch
applies it at window creation before the renderer reports in.
2026-06-12 12:02:38 -05:00
SHL0MS
7d4e60e44a docs(website): redirect old automation-templates URL to automation-blueprints
The Automation Blueprints rebrand (#44470) renamed the guide page from
guides/automation-templates to guides/automation-blueprints, leaving the
old URL 404ing. The site deploys to static hosting, so server-side
redirects aren't available.

Add @docusaurus/plugin-client-redirects (pinned 3.9.2, same as the other
Docusaurus packages) and a redirect entry for the old slug. The plugin
emits a static HTML page at the old path that meta-refresh/JS-redirects
to the new page, preserving query string and hash, with a canonical link
for SEO. Localized routes are handled automatically (zh-Hans verified).
2026-06-12 09:46:27 -07:00
brooklyn!
79c3ed3cc9 fix(desktop): new chat honours the active profile instead of rubberbanding to default (#45057)
The top "New Session" button (and /new, the keyboard shortcut) cleared
$newChatProfile to null, meaning "use the live gateway context". But
createBackendSessionForSend turned a null into an omitted `profile` param on
session.create. In global-remote mode one backend serves every profile, so an
omitted profile silently binds the new chat to the launch (default) profile's
home/state.db — the session "rubberbands back to default" even though the rail
still shows the selected profile. The per-profile "+" worked because it sets
$newChatProfile explicitly.

Resolve a null $newChatProfile to the active gateway profile at the single
session-creation chokepoint so session.create always carries the live profile.
Harmless for single-profile and local-pooled users: a backend resolves its own
launch profile to None (_profile_home), so passing it changes nothing.
2026-06-12 16:38:56 +00:00
brooklyn!
d62979a6f3 feat(desktop): composer status stack, live subagent windows, editable prompts (#44630)
* feat(desktop): session-scoped status stack + kill new-window theme flash

Stack subagents, background tasks, and the queue into one collapsible
"sink" above the composer, reusing the queue's chrome so every status
reads as one piece. Extracts shared StatusSection / StatusRow /
TerminalOutput primitives and a unified $statusItemsBySession store
(subagents mirrored, background owned here, merged + grouped for render).
Renames BrailleSpinner → GlyphSpinner now that it drives more than braille.

Separately, fix the white flash on every new/cmd-clicked window: macOS
`vibrancy` paints an NSVisualEffectView that follows the OS appearance and
ignores `backgroundColor`, so a dark app on a light-mode Mac flashed white
until the renderer painted over it. Pin `nativeTheme.themeSource` to the
app theme (persisted to userData so cold launches paint right before the
renderer loads), hold windows with `show:false` until `ready-to-show`, and
pre-paint the themed background via an inline script before the bundle runs.

* feat(desktop): dock the slash popover to the composer via one shared fill var

The slash·@ popover (and ? help) now docks onto the composer's edge with the
same chrome as the queue/status stack — rounded outer corners, fused borderless
edge, no shadow — but keeps its own narrow width.

Surface + drawer paint a single --composer-fill var; the state ladder
(rest / scrolled / focused / drawer-open) lives once in styles.css on
[data-slot='composer-root']. The :has() drawer-open rule is last and forces an
opaque fill, since translucent glass sampling different backdrops (thread vs
fade gradient) can never match. This replaces the focus-within !important
override that repainted the surface behind every previous matching attempt.

Also drop the chevron column from the project file tree — the folder open/closed
icon already carries the expand state.

* feat(desktop): base inset for file tree rows (post-chevron alignment)

* feat(desktop): wire the status stack's background tasks to the real process registry

The background group was UI-only (dev-mock seeded). Now it's live e2e:

- tui_gateway: new session-scoped `process.list` (registry snapshot filtered
  by the session's session_key, plus a 4KB output tail for the inline
  terminal viewer) and `process.kill` (single process, ownership-checked —
  unlike process.stop's kill_all).
- Renderer: `reconcileBackgroundProcesses` syncs snapshots into the store
  layout-stably — rows keep their position when state flips (never re-sort),
  new processes append, unchanged rows keep object identity so memoised rows
  skip re-rendering, and a dismissed-set stops the registry's retained
  finished procs from resurrecting X-ed rows.
- Refresh triggers: session open, terminal/process tool.complete,
  status.update(kind=process) from the gateway's notification poller, and a
  5s poll armed only while a running row is visible (catches silent exits).
- Stop = real `process.kill` + optimistic dismiss; Dismiss = client-side
  with resurrection guard.
- Re-keyed the stack to the RUNTIME session id: it was keyed by the stored
  session id, where neither subagent events nor process.list would ever land.
- Deleted dev-status-mocks.ts (__hermesStatusMocks) — no more seed shit.

Reconcile invariants covered in store/composer-status.test.ts.

* feat(desktop): todos + openable subagents in the status stack, self-healing file tree

- todo lists move out of the inline chat panel into the composer status stack
  (checklist icon, dashed ring = pending, spinner = in progress, check = done),
  fed live from todo tool events and seeded from history on session open
- subagent rows carry the child's real session id end-to-end
  (delegate_tool → gateway → renderer) so clicking one opens ITS session window
- status stack publishes its measured height so the thread's bottom clearance
  grows with it; card paints the shared --composer-fill so focused/scrolled
  states match the composer exactly
- file tree self-heals: ENOENT roots retry on a 3s cadence + Try again button,
  and the main process expands ~ in IPC paths (gateway cwds arrive as ~/...)
- composer drag-drop of tree entries inserts inline refs instead of attachments

* fix(desktop): file tree falls back to the workspace dir when a session's cwd is gone

Sessions record their launch cwd; deleted worktrees leave that path dead,
so opening such a session swapped the tree from the default workspace to a
directory that ENOENTs forever — the 3s retry just spun on it. On a root
read error the tree now asks main to sanitize the cwd (prefers the
configured default project dir), displays that fallback, and quietly
re-probes the original path so it switches back if the dir reappears.

* feat(desktop): working restore-checkpoint button on past user prompts

The discard icon on hover of a past user bubble was decorative — clicking
did nothing. It's now a real control: a confirmation dialog explains that
everything after the prompt is removed, then the session rewinds to that
turn and reruns the same prompt (prompt.submit with
truncate_before_user_ordinal, the same mechanism the edit composer uses).
Failures rethrow into the dialog's inline error instead of toasting.

* fix(desktop): show the restore-checkpoint button on the latest user prompt too

Restoring the most recent prompt is just 'retry this turn' — no reason to
exclude it. Stop still takes the slot while the turn is running.

* fix(desktop): finished todo lists clear themselves out of the status stack

A list whose every item is completed/cancelled lingers ~4s so the final
checkmark is visible, then the todo group drops out of the stack. A fresh
active list arriving within the linger cancels the scheduled clear.

* chore(desktop): drop dead editableCheckpoint copy, terser restore confirm

* fix(desktop): rewind clears the abandoned timeline's todos + background

Restoring to (or editing) an earlier prompt rewinds the conversation, but
the todos and background processes spawned by the now-discarded turns kept
showing in the status stack — and the real background processes kept
running. Both rewind paths now clear the session's todo rows and kill +
drop its background processes before the fresh run repopulates them. Also
drops the click-to-edit clamp transition, which flashed a half-expanded
bubble on the way into the edit composer.

* feat(desktop): user messages are always editable; edit/restore revert mid-stream

The bubble is now always click-to-edit — even while a turn streams — instead
of going inert during a run. Sending an edit acts like restore: it rewinds to
that prompt and re-runs with the new text. Both edit and restore can fire
mid-stream now; the gateway refuses prompt.submit while a turn runs (4009
"session busy"), so they interrupt the live turn first and retry the submit
until the cooperative interrupt winds it down. Restore (re-run as-is) shows on
every prompt except the latest running one, which keeps the Stop button.

* fix(desktop): label preview-pane ⌘L selections with the filename, not "zsh"

The terminal owns a global ⌘/Ctrl+L "send selection to composer" shortcut, so
selecting text in the file preview pane and hitting it fell through to the
terminal handler — which imported the right text but labelled the composer ref
"zsh:N lines" off the shell name. When the selection isn't an xterm selection,
label it with the previewed file instead.

* fix(desktop): ⌘L on a preview line selection inserts the @line ref, like dragging

The source preview lets you select lines in the gutter and drag them into the
composer as an @line:path:start-end ref. ⌘/Ctrl+L now does the same when a line
selection is active — it drops the identical ref instead of falling through to
the terminal's global handler (which grabbed the native text selection and sent
a bogus terminal block). Capture-phase + stopPropagation so it wins; with a line
selection there's no native selection, so the terminal handler stays out of it.

* chore: gitignore apps/desktop/demo/ scratch output

The desktop demo prompt writes demo/*.txt during recorded walkthroughs; it's
throwaway, never part of the app. Ignore it so it stops cluttering git status.

* feat(desktop): subagent watch windows, hard stop, sidebar hygiene

Child-session mirror for live subagent windows, delegate sessions tagged
and excluded from the sidebar, composer focus/stop polish, and WS stall
resilience on the gateway transport.

* refactor: DRY delegate SQL + trim status-stack noise

Extract shared listable-child and delegate-delete helpers in hermes_state,
collapse cancelRun busy release, and cut comment bloat in resume/status paths.

* fix(desktop): hide orphaned subagent sessions in sidebar

Cascade-delete all ephemeral children on parent delete (not just tagged rows),
run v16 backfill to tag legacy orphans, and record new delegates as source=subagent.

* fix: restore orphan contract for untagged children + lazy session eviction

Cascade-delete only _delegate_from-tagged rows (v16 backfill covers legacy),
walk marker chains recursively with FK-safe orphaning, gate lazy watch
sessions out of the still-starting eviction exemption via an explicit flag,
pass session_id to _make_agent only when resuming, and hide source=subagent
from session search.

* fix(gateway): gate child mirror off upgraded sessions + age out stale run entries

Review findings: the mirror could interleave synthetic events with a real
native stream once a watch window upgrades (prompt.submit builds an agent),
and a lost subagent.complete left _active_child_runs pinning running=true
forever. Mirror now stops when the live session owns an agent; liveness
reads ignore entries older than an hour.

* fix(gateway): reject prompt.submit into a watch session while its child runs

A lazy watch session's running flag is False (the run lives in the parent
turn), so typing mid-run sailed past the busy guard and built a second agent
racing the in-flight child on the same stored session. Busy error until the
run completes; afterwards the submit upgrades into a normal conversation.

* refactor(gateway): DRY watch-resume payload + compose listable-child SQL

Fold the duplicated child-run busy overlay into one _reuse_live_payload
helper across both resume reuse paths, collapse the twin mirror early-returns,
and build _LISTABLE_CHILD_SQL from _BRANCH_CHILD_SQL instead of restating it.

* fix(desktop): clip horizontal overflow on sidebar scroll areas

Add overflow-x-hidden alongside overflow-y-auto on session list scrollers
and the shared SidebarContent primitive — vertical scroll unchanged.
2026-06-12 08:30:06 -05:00
y0shualee
9c50521704 fix(desktop): complete backend PATH for Homebrew Codex
macOS Desktop backend processes can still miss Apple Silicon Homebrew paths even after adding Hermes-managed Node and venv bins. That leaves `/codex-runtime on` unable to find a Homebrew-installed `codex` binary at `/opt/homebrew/bin/codex`.

Add a small testable backend env helper that builds the dashboard subprocess environment in one place. It prepends Hermes-managed Node and venv bins, appends missing POSIX sane PATH entries individually, preserves caller precedence without duplicates, and keeps Windows PATH casing/delimiters intact.

Wire both source-checkout and active-install backend descriptors through the helper, and add Node regression coverage to the desktop platform test suite.
2026-06-12 03:03:44 -07:00
Teknium
88dbf95105 fix(dashboard): profile-scope Channels endpoints and seed per-profile .env (#44792)
Two halves of the same community report (dashboard Profile Builder):

1. A fresh dashboard/CLI-created profile got no .env file unless cloned,
   so it silently inherited API keys and messaging tokens from the shell
   environment / root install. create_profile() now seeds a placeholder
   .env (0600) for non-clone profiles, matching the SOUL.md seeding.

2. The Channels endpoints (/api/messaging/platforms GET/PUT/test) were
   not profile-scoped: they read/wrote the dashboard process's own .env
   via load_env()/save_env_value() regardless of the global profile
   switcher. They now accept the standard optional profile param (body
   beats query on the PUT, matching other scoped writes) and run inside
   _profile_scope(). When scoped, the payload no longer falls back to
   os.environ or load_gateway_config()'s env-override layer — both carry
   the ROOT install's credentials and would misreport them as the
   profile's. /api/messaging/platforms added to PROFILE_SCOPED_PREFIXES
   so the sidebar switcher scopes the Channels page automatically.
2026-06-12 02:09:28 -07:00
loongfay
e20e0bd744 feat(Yuanbao): support wechat forward msg (#43508)
* feat(yuanbao): support wechat forward msg

* feat(yuanbao): support wechat forward msg

---------

Co-authored-by: loongfay <izhaolongfei@gmail.com>
2026-06-12 02:06:47 -07:00
Teknium
0fd34e8c5a fix(teams): cache document/video/audio attachments and classify as DOCUMENT (#44778)
The Teams adapter only handled image/* attachments — documents (the
application/vnd.microsoft.teams.file.download.info consent-free download
payload and any direct-URL non-image attachment) never reached media_urls
at all, so run.py's document-context injection had nothing to surface.
Completes the class-wide sweep from PR #44695 (Signal/Email/SimpleX).

- download.info attachments: fetch the pre-authed SharePoint downloadUrl
  (SSRF-guarded, same guard chain as base.py cache_*_from_url) and route
  through cache_media_bytes
- direct-URL non-image attachments: same fetch + classify path
- skip Teams' text/html message-body mirror and adaptive-card attachments
- DOCUMENT > PHOTO > VIDEO > AUDIO precedence for mixed attachments,
  matching the Email precedence rationale from #44695
2026-06-12 02:05:41 -07:00
Siddharth Balyan
7ba5df0d52 feat(billing): /credits command — balance + portal top-up handoff (#44776)
* feat(billing): /usage → portal top-up browser handoff

Add the terminal side of the billing slice (phase 2a): start a top-up by
throwing the user to the portal billing page with the top-up modal open. The
terminal does not confirm, poll, or track payment — checkout completes in the
browser and the next /usage shows the new balance.

- nous_account.py: parse organisation.slug/name from /api/oauth/account into
  NousPortalAccountInfo; add nous_portal_topup_url() building the org-pinned
  {base}/orgs/{slug}/billing?topup=open with a null-slug fallback to the legacy
  {base}/billing?topup=open (never /orgs/None/...).
- portal_cli.py: 'hermes portal topup' — fresh account fetch, identity line
  (Topping up as <email> / org <name>), browser open with printed-URL fallback,
  no-wait closing copy. No polling/confirmation (deferred to 2b).
- account_usage.py: the shared /usage credits block now links the org-pinned
  top-up URL (auto-opens the modal) + points to the command.

Depends on NAS #409 (organisation.slug/name + ?topup=open). Do not merge until
that is live on the target env; until then /api/oauth/account returns
organisation: { id } only and the URL falls back to legacy.

* feat(billing): /credits command for balance + top-up handoff

Replace the standalone `hermes portal topup` subcommand with an in-session
/credits slash command — a focused money surface (balance in, top-up out) that
works in the CLI, TUI, and every messaging platform from one registry entry.

- commands.py: register /credits (Info category). Slack is at its 50-slash cap,
  so /credits is routed via /hermes credits on Slack only (new
  _SLACK_VIA_HERMES_ONLY set) to avoid clamping a canonical command off the
  native list and breaking Telegram parity; native everywhere else.
- account_usage.py: build_credits_view() — one portal fetch → balance lines +
  identity line + org-pinned top-up URL + depleted flag, consumed by all
  surfaces. Reuses the same snapshot/URL builder as /usage so numbers match.
- cli.py: _show_credits() — balance block + identity line + 3-button panel
  (Open top-up / Copy link / Cancel) via the existing prompt_toolkit modal.
  ASK, never auto-launch; headless falls back to printing the URL.
- gateway/slash_commands.py: _handle_credits_command() — renders the block +
  tappable top-up URL + no-wait copy; works on button and plain-text platforms.
- /usage credits line now points to /credits.
- Retire `hermes portal topup` (portal_cli.py back to baseline); the engine
  (slug/name parse + nous_portal_topup_url) stays as the shared core.

No polling, no payment confirmation (billing phase 2a). Depends on NAS #409.

* fix(credits): /credits works in the TUI slash-worker (non-interactive)

In the TUI, /credits runs in the slash-worker subprocess where there is no
live prompt_toolkit app and stdin is the JSON-RPC pipe. _show_credits called
the 3-button modal unconditionally, which fell back to reading stdin →
exception → slash.exec rejected → the command produced no output (only the
pre-existing 'Credit access paused' banner showed).

- _show_credits: when self._app is None (TUI worker / piped / non-interactive),
  render the text variant — balance block + tappable top-up URL + no-wait line,
  same affordance as the messaging surfaces — and skip the modal entirely. The
  3-button panel still renders in the interactive CLI.
- Depleted banner copy: 'run /usage for balance' → 'run /credits to top up'
  now that /credits is the dedicated money surface (+ tests).
- Regression tests: _show_credits with self._app=None renders text and never
  invokes the modal; logged-out path.

* feat(tui): credits.view RPC for the /credits tappable top-up button

Add a credits.view JSON-RPC method returning the structured CreditsView
(logged_in, balance_lines, identity_line, topup_url, depleted) so the TUI can
render a clickable <Link> top-up button instead of plain text. Account-
independent (portal fetch gated on a logged-in Nous account), fail-open to
{logged_in: false} on any hiccup. Mirrors session.usage's credits-block pattern.

Frontend (TUI-local /credits command + Ink component) lands separately.

* feat(tui): /credits command with keyboard-driven top-up confirm

TUI-local /credits: fetches the structured balance via the credits.view RPC,
prints the balance + identity + top-up URL, then arms the EXISTING confirm
overlay (Enter = open top-up in browser via openExternalUrl, Esc = cancel).
Reuses ConfirmReq — no new overlay component/state/input handler. Headless
(openExternalUrl returns false) falls back to printing the URL.

- gatewayTypes.ts: CreditsViewResponse.
- commands/credits.ts: the command (mirrors /status's rpc+guarded pattern).
- registry.ts: register creditsCommands.
- test: balance+overlay armed, headless fallback, no-url, logged-out (4 cases).

Matches the CLI /credits 'Enter to open' affordance. Phase 2a: no polling.
2026-06-12 08:51:10 +00:00
Teknium
4474873d2c feat(cli): persist resolved approval/clarify prompts in scrollback (#44702)
Modal prompt panels (dangerous-command approval, clarify questions)
live in the prompt_toolkit layout and vanish on the next repaint,
leaving no trace of the question or the decision in chat history.

Emit a dim one-line summary after each prompt resolves:
  ⚠ Approval: <command> → allowed for session
  ? Clarify: <question> → <answer>

Gated on display.persist_prompts (default true). Detail and outcome
are whitespace-collapsed and capped at 120 chars.
2026-06-12 01:14:35 -07:00
Teknium
8e5b7592f8 refactor(agent): hoist MEDIA-directive regex to module level
Avoid recompiling the pattern on every _serialize_for_summary call; name it
beside _PATH_MENTION_RE with the #14665 rationale.
2026-06-12 01:14:28 -07:00
Tranquil-Flow
286ecd26d8 fix(agent): strip MEDIA directives from compressor summarizer input (#14665) 2026-06-12 01:14:28 -07:00
Teknium
8b2a3c9c51 chore: add kdunn926 to AUTHOR_MAP 2026-06-12 01:07:50 -07:00
Teknium
74180ebf0b fix(gateway): classify SimpleX non-image/non-audio files as DOCUMENT
SimpleX tagged unknown files application/octet-stream in media_types
but classification only handled audio/image, leaving msg_type TEXT —
run.py never injected the document context. Same bug class as #12845.
2026-06-12 01:07:50 -07:00
Teknium
f03f161b39 fix(gateway): classify email document attachments as DOCUMENT
Email cached document attachments and placed them in media_urls, but
msg_type only flipped on image attachments — documents stayed TEXT and
run.py's document-context injection (gated on MessageType.DOCUMENT)
silently dropped them. Same bug class as Signal #12845. DOCUMENT wins
over PHOTO for mixed attachments since image handling keys off per-path
mime types while document injection gates strictly on message_type.
2026-06-12 01:07:50 -07:00
Teknium
1e29ab38c7 fix(gateway): classify Signal video attachments + catch-all DOCUMENT fallback
Widen the salvaged #12851 fix to match the established classification
pattern (WhatsApp/Slack/BlueBubbles/Mattermost): video/* -> VIDEO, and
any remaining MIME type falls through to DOCUMENT instead of TEXT, so
exotic types still trigger run.py's document-context injection.
2026-06-12 01:07:50 -07:00
Kyle Dunn
8e821cd2f5 test(gateway): verify Signal inbound text attachment sets MessageType.DOCUMENT 2026-06-12 01:07:50 -07:00
Kyle Dunn
ffef9da9b7 test(gateway): verify Signal inbound PDF attachment sets MessageType.DOCUMENT 2026-06-12 01:07:50 -07:00
Kyle Dunn
8207ae888d fix(gateway): add Signal message type classification for documents 2026-06-12 01:07:50 -07:00
teknium1
05470aa1b6 feat(messaging): expose action='unreact' in send_message + react dispatch tests
Follow-up for salvaged PR #44486: the adapter shipped remove_reaction but
the tool only exposed 'react'. Generalize _handle_react(remove=) and add
tool-level dispatch tests for react/unreact (missing from the original PR).
2026-06-12 01:07:38 -07:00
underthestars-zhy
b4e95a2efe fix(photon): add clarifying comments for Windows-safe os.kill usage 2026-06-12 01:07:38 -07:00
underthestars-zhy
23305cfeab fix(photon): normalize DM chat keys in last-inbound reaction tracker
Inbound events key the tracker by the DM chat GUID (any;-;+1555...),
but home-channel react calls address the same space by bare E.164 —
normalize both to the phone so add_reaction's last-inbound default
resolves regardless of which form the caller uses (mirrors the
sidecar's phoneTargetFromSpaceId).

Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
2026-06-12 01:07:38 -07:00
underthestars-zhy
156f4fba92 feat(photon): add agent-facing emoji reaction support
Add `action='react'` to `send_message` tool and expose `add_reaction`/
`remove_reaction` on the Photon adapter.

- Track latest inbound message id per chat (`_last_inbound_by_chat`,
  bounded to 200 entries) so the agent can react without threading
  message ids through tool calls
- New `add_reaction`/`remove_reaction` public methods on PhotonAdapter;
  unlike the lifecycle tapbacks, these are not gated by PHOTON_REACTIONS
- `send_message` gains `action='react'` with `emoji` and optional
  `message_id` params; resolves target via existing channel-directory
  and home-channel logic; requires a live gateway adapter
2026-06-12 01:07:38 -07:00
underthestars-zhy
a23c0b378c fix(photon): use per-call httpx client in _sidecar_call
Prevents "Future attached to a different loop" errors when
_sidecar_call is invoked from a worker thread via _run_async in
send_message_tool. The persistent _http_client remains in use for
the inbound streaming loop, which always runs on the gateway's loop.
2026-06-12 01:07:38 -07:00
underthestars-zhy
9bfff6e16c chore(photon): bump spectrum-ts to 3.1.0 2026-06-12 01:07:38 -07:00
underthestars-zhy
a652131c42 fix(photon): stop gateway restarts from orphaning the sidecar on its port
A hard gateway exit (crash, SIGKILL, supervisor restart) left the
detached Node sidecar running with a token the next gateway run doesn't
know, so it could never be told to /shutdown. Every replacement spawn
then died on EADDRINUSE, failing each 30→300s reconnect attempt while
the orphan kept consuming the inbound gRPC stream.

Two layers:
- Lifetime binding: the adapter now holds the sidecar's stdin as a
  pipe, and the sidecar (PHOTON_SIDECAR_WATCH_STDIN=1) shuts down on
  stdin EOF — fired by the OS on any parent death, including SIGKILL.
- Startup reaping: before spawning, the adapter probes the port and
  terminates a stale listener, but only after verifying its command
  line is a Photon sidecar; a foreign listener raises a clear error
  instead of being signalled.

Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
2026-06-12 01:07:38 -07:00
underthestars-zhy
573c4e6511 feat(photon): upgrade to spectrum-ts 3.0.0 (pinned) with markdown + reactions
Pin spectrum-ts to exactly 3.0.0 (was ^1.18.0 plus an `npm install
spectrum-ts@latest` on every setup) so breaking SDK majors can't take
down fresh installs silently; `hermes photon setup` now runs `npm ci`.
Upgrade procedure documented in the README.

Migrate resolveSpace to the v3 namespace API: `im.space.create(phone)`
for DMs and `im.space.get(id)` for everything else — group spaces are
now rehydratable from their persisted id after a sidecar restart, which
v1 could not do.

Markdown: replies go out via the v3 `markdown()` builder (iMessage
renders natively; other Spectrum platforms degrade to plain text).
`PHOTON_MARKDOWN=false` reverts to the stripped plain-text path.

Reactions, behind PHOTON_REACTIONS (default off): lifecycle tapbacks
(👀 while processing, 👍/👎 on completion) via new sidecar /react and
/unreact endpoints with per-target reaction-handle tracking, and user
tapbacks on bot-sent messages routed to the agent as synthetic
`reaction:added:<emoji>` events.

Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
2026-06-12 01:07:38 -07:00
underthestars-zhy
0a963d8c9a feat(photon): add telemetry toggle via hermes photon telemetry 2026-06-12 01:07:38 -07:00
Teknium
c196269d8d fix(credits): suppress usage gauge when top-up funds exist + add display.credits_notices toggle (#44716)
The subscription-cap usage gauge (50/75/90% bands) ignored purchased
(top-up) credits: a sub user with top-up funds got a sticky warn banner
at 90% of their cap — permanently at >=100%, alongside grant_spent —
despite being fully able to keep inferencing. The cap is the wrong
denominator for an account that can keep spending.

- evaluate_credits_notices: purchased_micros > 0 suppresses the usage
  band (grant_spent already covers the cap-reached + top-up case with
  the remaining balance). A top-up landing mid-session clears any
  showing band; spending top-up down to 0 resumes the gauge.
- New display.credits_notices config (default true): false silences all
  credits notices. State capture and /usage are unaffected. Read once
  per agent (cached) in _emit_credits_notices, fail-open true.
- Docs: configuration.md display block.
2026-06-12 01:06:46 -07:00
ethernet
906bee9cf7 fix(nix): natively compile and correctly stage node-pty for desktop app
- Add ELECTRON_SKIP_BINARY_DOWNLOAD=1 to nix/lib.nix to prevent offline download failures.
- Manually trigger native compilation of node-pty via npm rebuild --build-from-source in buildPhase.
- Run stage-native-deps.cjs to copy the natively compiled binary into build/native-deps.
- Flatten native-deps and install-stamp.json to the root of the output derivation in installPhase, matching electron-builder's extraResources behavior so main.cjs can find it at process.resourcesPath + '/native-deps/node-pty'.
- Add doCheck=true and a strict checkPhase to fail fast if the staged native binary is missing.
2026-06-12 03:55:09 -04:00
kshitij
046f444ddc Merge pull request #44738 from kshitijk4poor/salvage/memory-sync-multimodal-content
fix(memory): flatten multimodal content before provider sync
2026-06-12 00:40:31 -07:00
kshitijk4poor
15439bee47 refactor(memory): reuse _summarize_user_message_for_log instead of forking it
The original fix added agent/memory_manager.py:flatten_message_content, but
that helper was a near-exact duplicate of
agent/codex_responses_adapter.py:_summarize_user_message_for_log — same
None/str/list dispatch, same {text,input_text,output_text}/{image_url,input_image}
part sets, the identical [N image(s)] marker, and the same str() fallback. The
only difference was the join separator (newline for memory vs space for the
log/trajectory previews the existing helper already serves), and that helper is
already imported into agent/turn_finalizer.py — the same file whose call site the
memory fix touches.

Parameterize the existing helper with sep=' ' (default preserves every current
logging/trajectory caller byte-for-byte) and call it with sep='\n' at the memory
boundary; drop the forked flatten_message_content. Repoints the unit tests to the
consolidated helper and adds a case locking the default space-join.

Single source of truth for multimodal-content flattening; no behavior change for
the fix or for existing callers.
2026-06-12 12:49:18 +05:30
Erosika
87893fe4cb fix(memory): flatten multimodal content before provider sync
Multimodal turns carry message content as a list of typed parts
({type: "text"|"image_url", ...}). _sync_external_memory_for_turn
passed that list straight into MemoryManager.sync_all, and providers
feed it to regexes — Honcho's sync_turn calls sanitize_context, where
re.sub raised 'expected string or bytes-like object, got list'. Every
turn with an attached image silently never synced.

Flatten to plain text at the boundary: text parts joined, images noted
as an [N image(s)] marker so the attachment isn't erased from recall.
Fixing here covers all providers instead of patching each plugin.

(cherry picked from commit 705bdb6ffe)
2026-06-12 12:46:28 +05:30
brooklyn!
d810f2b262 Merge pull request #44676 from NousResearch/bb/fix-schema-ref-default
fix(tools): strip default from $ref nodes in tool schemas
2026-06-12 01:21:14 -05:00
teknium1
b3f5e17bb9 fix(tui): wrap long approval commands in the Ink overlay
Sibling site of the CLI approval-panel fix: the TUI ApprovalPrompt
rendered each command line with wrap="truncate-end", so a long
single-line command lost its tail at terminal width. Wrap to the
panel width via wrapAnsi before applying the 10-line preview cap.
2026-06-11 23:05:08 -07:00
墨綠BG
81cdbbddc8 🐛 fix(cli): wrap approval preview hints 2026-06-11 23:05:08 -07:00
墨綠BG
d6df38bb6b 🐛 fix(cli): wrap long approval commands in prompt 2026-06-11 23:05:08 -07:00
Teknium
c7bee8f961 refactor(agent): drop unused tail_start param from _derive_auto_focus_topic
The parameter was reserved-but-unused (del'd immediately); YAGNI. Test
call site updated.
2026-06-11 23:03:52 -07:00
konsisumer
434c684bfa fix(agent): focus automatic compression on recent user turns 2026-06-11 23:03:52 -07:00
Teknium
db7714d5f1 Merge pull request #44331 from NousResearch/hermes/hermes-6b48295e
feat(whatsapp): WhatsApp Business Cloud API adapter (salvage #43921)
2026-06-11 22:48:06 -07:00
Kyssta
343803b23c fix(cli): use subprocess on Windows for dashboard profile re-exec (#44282) (#44446)
Co-authored-by: kyssta-exe <kyssta-exe@users.noreply.github.com>
2026-06-11 22:41:39 -07:00
Kyssta
a942bfd9cc fix(gateway): reset _last_flushed_db_idx when reusing cached agent (#44327) (#44518)
Co-authored-by: kyssta-exe <kyssta-exe@users.noreply.github.com>
2026-06-11 22:41:34 -07:00
kshitij
a35b370284 Merge pull request #44674 from kshitijk4poor/fix/slack-reactions-plugin-registry-bookkeeping
fix(plugins,slack): registry bookkeeping fixes + ack reaction events (salvage #42561)
2026-06-11 22:32:59 -07:00
Brooklyn Nicholson
b2d151abe2 fix(tools): strip default from $ref nodes in tool schemas
Fireworks-hosted Kimi rejects tool requests when nullable MCP/Pydantic
schemas collapse to {"$ref": "...", "default": null}. Strip that sibling
during global schema sanitization so gateway and CLI calls succeed again.
2026-06-12 00:30:51 -05:00
kshitijk4poor
44bd478039 fix(plugins): credit shared hook/middleware/tool names to every plugin
list_plugins() attribution diffed registry names against all already-loaded
plugins, so when a plugin registered a hook / middleware / tool name an
earlier plugin had already used, the shared name was credited to the first
plugin only and later plugins under-reported (0 hooks) in hermes plugins
list. commands_registered right beside it already attributed correctly by
plugin ownership.

Snapshot per-registry counts before register() and attribute the entries
this plugin's register() actually added (per-registration delta). Add a
regression test: two plugins registering the same hook name are each
credited with 1 hook.
2026-06-12 10:57:25 +05:30
kshitijk4poor
889a13696b fix(plugins): clear _plugin_platform_names on force-rediscover
discover_and_load(force=True) cleared every per-plugin registry except
_plugin_platform_names, which register_platform() populates. A platform
plugin disabled between force-rediscovers left a stale name behind, so the
set diverged from the real platform_registry / _plugins state and never
shrank across repeated force passes.

Add the missing clear() and a regression test that seeds every per-plugin
registry, forces a rediscover, and asserts they all empty (so a future
registry addition can't silently leak across a force pass either).
2026-06-12 10:55:44 +05:30
Veritas-7
82d570165e fix(slack): ack reaction lifecycle events
Register no-op Slack event handlers for inbound reaction_added and reaction_removed events so Slack Bolt does not log unhandled-request warnings for events Hermes does not consume.
2026-06-12 10:54:07 +05:30
kshitij
c574170050 Merge pull request #44664 from kshitijk4poor/salvage/slack-plugin-action-handlers
feat(plugins): expose register_slack_action_handler API (salvage #20589)
2026-06-11 22:14:44 -07:00
kshitijk4poor
e4c168b1f4 chore: map bcsmith528 contributor email for attribution 2026-06-12 10:39:05 +05:30
Brad Smith
08e8bedae8 fix(gateway): keep plugin action wrapper signature to (ack, body, action)
The previous implementation captured loop vars via default arguments::

    async def _wrapped(ack, body, action, _cb=_cb, _plugin_name=_plugin_name):

slack_bolt's ``kwargs_injection`` introspects each listener's signature
via ``inspect.signature`` and passes ``None`` for any parameter name it
doesn't recognise (see ``slack_bolt/kwargs_injection/async_utils.py``
``build_async_required_kwargs``). That clobbered ``_cb`` to ``None`` at
dispatch time, so the wrapped plugin handler became ``NoneType`` —
``await _cb(...)`` then raised ``'NoneType' object is not callable`` and
no plugin action handler ever fired.

Replace the default-arg trick with a small closure factory so the
wrapper's public signature is exactly ``(ack, body, action)``. Add a
regression test that introspects the wrapped function's signature.

Found via real Slack click on a Block Kit button registered through
``ctx.register_slack_action_handler`` — gateway log showed
``[Slack] Plugin 'None' action handler raised: 'NoneType' object is
not callable`` despite the registration log line confirming the
handler was wired.

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
2026-06-12 10:36:14 +05:30
Brad Smith
62e937bf2b feat(plugins): expose register_slack_action_handler API
Plugins that post Block Kit messages with interactive elements (buttons,
overflow menus, datepickers, etc.) had no documented way to receive the
resulting click events. The plugin API exposed register_tool, register_hook,
register_command, register_platform, and register_context_engine, but
nothing for slack_bolt action handlers. The only workaround was to
monkey-patch SlackAdapter.connect from inside register(), which is
fragile and breaks on every Hermes update.

This change adds:

* PluginContext.register_slack_action_handler(action_id, callback) —
  validates inputs and queues the handler on the PluginManager.
  action_id accepts whatever slack_bolt.App.action() accepts (literal
  string, compiled re.Pattern, or constraint dict).
* PluginManager.get_slack_action_handlers() — accessor used by the
  Slack adapter at connect time.
* SlackAdapter.connect — after wiring its built-in approval and
  slash-confirm buttons, iterates the plugin-registered handlers
  and registers each via self._app.action(matcher)(callback). Each
  callback is wrapped defensively so a misbehaving plugin cannot
  crash slack_bolt's dispatch loop, with a best-effort ack on
  exception so Slack stops retrying the click.
* Defensive fallback when the plugin layer is unhealthy: a
  RuntimeError from get_plugin_manager() is logged and swallowed
  rather than blocking the gateway from starting.
* Test coverage in tests/gateway/test_slack_plugin_action_handlers.py
  for input validation, multi-plugin registration, the connect-time
  wiring, defensive exception handling, and the plugin-loader-
  failure fallback path.
* Documentation in website/docs/guides/build-a-hermes-plugin.md
  describing the new API alongside the existing register_command /
  dispatch_tool documentation.

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
2026-06-12 10:36:14 +05:30
brooklyn!
24f74eb888 fix(desktop): make file-preview source + markdown selectable (#44648)
body sets user-select:none for native feel and opts text back in only via
[data-selectable-text='true']; the preview's source and rendered-markdown
panes never set it, so code couldn't be selected or copied. Tag the Shiki
code column and the markdown root. The attribute stays off the SourceView
grid root so the gutter keeps its select-none and line numbers don't bleed
into copied text.
2026-06-12 04:15:06 +00:00
brooklyn!
6e41ca956b fix(desktop): bundle JetBrains Mono for the terminal pane (#44642)
The terminal listed JetBrains Mono only as a late fallback and shipped no
webfont, so on machines without SF Mono/Menlo xterm measured the grid on the
regular system face while styled SGR spans fell back to a font with different
advances — glyphs squeezed and overlapped.

Bundle the regular/bold/italic woff2 (Apache-2.0, the faces the dashboard
already ships), put the family first in the xterm stack, pin the weights, and
warm every face before mount (fonts.ready only settles already-requested
faces; bold/italic aren't asked for until styled output paints, past atlas
init). Vite emits them as hashed assets under dist/** with base './', so the
fonts ship in the asar and every install path inherits them.
2026-06-12 04:11:51 +00:00
brooklyn!
6db65e687c Merge pull request #44627 from NousResearch/bb/desktop-tool-row-copy-affordance
fix(desktop): move tool-row copy control into expanded body
2026-06-11 22:32:52 -05:00
Brooklyn Nicholson
09bcf5a937 fix(desktop): move tool-row copy control into expanded body
The per-row copy control lived in the header's trailing slot as a 24px
button that depended on a `group-hover/tool-row` group that exists nowhere
in the tree. It therefore stayed `opacity-0` yet remained clickable — an
invisible hit-target straddling the disclosure caret and duration, making
the caret hard to click without firing a copy.

Move copy into the expanded body's top-right (matching the code-block
convention) where it can't fight the caret for the right edge, and make it
actually visible (subtle at rest, full on hover/focus). The header right
edge now belongs solely to the duration label + caret.

Tradeoff: copy is only reachable once a row is expanded; rows with no
expandable body no longer surface a copy control.
2026-06-11 22:27:39 -05:00
brooklyn!
4d67ac6172 Merge pull request #44596 from NousResearch/bb/desktop-rtl-bidi
feat(desktop): auto-detect RTL/bidi text direction in chat
2026-06-11 21:44:13 -05:00
Brooklyn Nicholson
6c00077d38 feat(desktop): auto-detect RTL/bidi text direction in chat
Arabic/Hebrew/Persian/Urdu chat text rendered left-to-right and
left-aligned, and mixed RTL/English technical messages (the common case)
read backwards. Resolve each chat block's base direction from its own
first strong character (UAX#9) with pure CSS, scoped to the chat
surfaces only:

- `unicode-bidi: plaintext` + `text-align: start` on assistant prose
  blocks (p, h1-h6, li, blockquote), the user bubble's text lines, and
  both composers (main + edit share the composer-rich-input slot). RTL
  blocks read and right-align RTL; English stays LTR; mixed
  conversations resolve per block. `text-align: start` is required
  because the user bubble hardcodes `text-left`.
- Inline `code` and KaTeX are pinned `direction: ltr; unicode-bidi:
  isolate`, so the bidi first-strong heuristic skips them: a sentence
  that *starts* with a command (`./run.sh ...`) followed by Arabic
  still resolves RTL, and the command's own neutrals keep their order.
- Fenced code surfaces (code-card, user fences) are pinned LTR so they
  never mirror or right-align inside an RTL list item or blockquote.

`direction` is never forced, so app chrome, layout, and list indent
stay LTR per the issue's request not to flip the whole UI. English-only
content is byte-for-byte unchanged.

Salvaged and unified from #44065 and #44169; verified in Chromium that
isolate removes inline code from the paragraph direction vote (the
code-first case), making the JS dir-resolution in #44065 unnecessary.

Fixes #44150

Co-authored-by: Adolanium <Adolanium@users.noreply.github.com>
Co-authored-by: Adalsteinn Helgason <AIalliAI@users.noreply.github.com>
2026-06-11 21:06:26 -05:00
brooklyn!
9e484f052a Merge pull request #44559 from NousResearch/bb/persistent-terminal-env
fix(terminal): advertise persistent env state
2026-06-11 20:07:11 -05:00
Brooklyn Nicholson
ab06ef8ed6 fix(coding): teach agents terminal env state persists
Tell coding agents to activate shell setup once per session instead of re-sourcing it before every command, and pin the existing LocalEnvironment env-snapshot behavior with regression tests.
2026-06-11 19:50:08 -05:00
brooklyn!
afe53708ee Merge pull request #44545 from NousResearch/hermes-worktree-code
fix(coding): don't expose primary worktree path in coding context
2026-06-11 19:35:18 -05:00
Teknium
5affecb443 fix(mcp): capability-gate tools/list so prompt-only MCP servers can connect (#44550)
Port from anomalyco/opencode#31271: only call tools/list when the server
advertises the 'tools' capability in InitializeResult.capabilities.

Previously, _discover_tools() unconditionally called session.list_tools()
right after initialize. Prompt-only / resource-only servers (which omit
the tools capability per the MCP spec) raise McpError(-32601 Method not
found), which aborted the connection — burning all 3 initial-connect
retries and permanently failing the server even though its prompts and
resources were perfectly usable. The 180s keepalive had the same problem:
it probed with list_tools(), so even a successfully connected prompt-only
server would be torn down on the first keepalive cycle.

Changes:
- MCPServerTask._advertises_tools(): capability check with a legacy
  fallback (no captured InitializeResult -> behave as before)
- _discover_tools(): skip tools/list for non-tool servers
- keepalive: use the universal ping request for non-tool servers
- _refresh_tools(): guard against tools/list_changed from non-tool servers

E2E verified with a real stdio prompt-only FastMCP-style server: on main
it fails all 3 connection attempts with Method-not-found; with this fix
it connects, lists prompts, answers ping keepalives, and shuts down
cleanly.
2026-06-11 17:34:49 -07:00
ethernet
96cc7ee1e3 fix(coding): don't provide worktree root in context
this makes the agent frequently edit files in the wrong worktree.
what the agent doesn't know can't hurt it.
2026-06-11 20:27:06 -04:00
brooklyn!
880107ab24 Merge pull request #44529 from NousResearch/bb/desktop-profile-fallout
fix(desktop): close out the multi-profile desktop fallout — WS auth + cross-profile session reads
2026-06-11 19:06:00 -05:00
brooklyn!
4ddb03390a fix(desktop): collect + persist API key for custom OpenAI endpoints (#43896)
The desktop "Local / custom endpoint" onboarding never collected an API
key and /api/model/set silently dropped one, so an auth-gated endpoint
(e.g. a hosted vLLM behind a key) could never enumerate models — and
Settings' "Set up custom endpoint" routed `custom` into a non-existent
OAuth flow, booting the user back to the first screen (the reported loop).

Backend (web_server.py):
- /api/providers/validate accepts an optional api_key and sends it as a
  Bearer header when probing a custom endpoint's /v1/models.
- /api/model/set accepts api_key, persists it to model.api_key (same
  switch/preserve lifecycle as base_url), and registers a named
  custom_providers entry via _save_custom_provider — matching the
  `hermes model` CLI flow so the endpoint shows up as a ready picker row.

Desktop:
- ApiKeyForm shows an optional API key field for the local/custom option;
  the key is threaded through saveOnboardingLocalEndpoint → validate +
  setModelAssignment.
- New onboarding `localEndpoint` intent + startManualLocalEndpoint(); the
  Settings "Set up custom endpoint" button now opens the local-endpoint
  form (URL + key) instead of the OAuth dead-end.
- Added localApiKeyPlaceholder i18n key (en + types + zh).

Tests: api_key lifecycle on _apply_main_model_assignment, key persistence
+ custom_providers registration on /api/model/set, Bearer-header probe;
onboarding store forwards + persists the key.
2026-06-12 00:03:55 +00:00
brooklyn!
c6007e5c1a Merge pull request #44534 from NousResearch/bb/approval-allow-permanent
fix(approval): carry allow_permanent to TUI + desktop approval prompts
2026-06-11 18:49:58 -05:00
Austin Pickett
e2145a5c9c fix(ui-tui): stabilize embedded dashboard chat gateway (#44528)
Cherry-picked from #39840 by @flyinhigh and rebased cleanly on main.

- Defer config fetch in createGatewayEventHandler until gateway.ready to
  avoid render-phase RPC that can mutate transcript state and trigger
  React error 301 in embedded dashboard PTYs.
- Use undici WebSocket fallback when globalThis.WebSocket is unavailable
  (Node attach mode and sidecar mirror sockets).
- Add regression tests for both fixes.

Co-authored-by: flyinhigh <flyinhigh@users.noreply.github.com>
2026-06-11 19:47:53 -04:00
Brooklyn Nicholson
55a18e6860 chore(approval): tighten allow_permanent comments + DRY the no-always opt set
Collapse the verbose multi-line rationale comments across the TUI/desktop/
backend approval surfaces into single-line "why" notes, and derive
APPROVAL_OPTS_NO_ALWAYS from APPROVAL_OPTS instead of re-listing it.
No behavior change.
2026-06-11 18:42:59 -05:00
Brooklyn Nicholson
b097d7b033 refactor(desktop): use native fetch in dashboard-token
Node >=18 / Electron 40 ship fetch; the hand-rolled http/https.request
plumbing buys nothing. AbortSignal.timeout replaces the socket timeout,
protocol guard and >=400 rejection semantics preserved. 13/13 unit
tests and the live web_server.py repro both green over the new
transport.
2026-06-11 18:41:16 -05:00
Brooklyn Nicholson
cc726aad68 refactor(desktop): fold served-token adoption + foreign-backend refusal into one helper
Both spawn paths (startHermes, spawnPoolBackend) duplicated the same
resolve -> log-fallback -> foreign-check -> throw dance. Collapse it into
adoptServedDashboardToken(baseUrl, spawnToken, {childAlive, label}) in
dashboard-token.cjs; childAlive is a thunk so liveness is sampled after
the fetch. Drop the redundant backendPool.delete in the pool's throw
path (the child exit/error handlers already own pool eviction).

Validated end-to-end against a real web_server.py backend, not just
units: token-injection regex vs the actual served index.html, foreign
refusal (dead child + live squatter), benign drift adoption, and the
401-vs-200 token auth split on /api/sessions.
2026-06-11 18:33:05 -05:00
Brooklyn Nicholson
81436e143e fix(approval): carry allow_permanent to TUI + desktop approval prompts
When a tirith content-security warning is present the approval backend
forces allow_permanent=False and silently downgrades an "always" choice to
session scope (the persistence loop in check_all_command_guards only honors
"always" → permanent when no tirith finding exists). But the gateway notify
payload that drives the TUI and the Electron desktop app never carried that
flag, so both surfaces always rendered "Always allow" — offering a permanent
allow the backend would quietly refuse to persist.

Plumb allow_permanent end-to-end:
- tools/approval.py: include `allow_permanent: not has_tirith` in the gateway
  approval_data the notify callback emits as `approval.request`.
- ui-tui: thread `allowPermanent` through the event handler, gateway types,
  and ApprovalReq; ApprovalPrompt drops the "always" option (and renumbers the
  quick-pick keys) when it's false.
- apps/desktop: thread `allow_permanent` through the gateway payload type, the
  per-session approval store, and the inline ApprovalBar, which now hides the
  "Always allow…" dropdown item when permanent allow is disallowed — reusing
  the existing DropdownMenu / confirm-Dialog UI.

The desktop/TUI render path for approvals already landed in #38578 (the root
cause of approvals not surfacing in the GUI); this completes the salvage of
#37856 by carrying allow_permanent across both surfaces. #37856's original
thread-local _block() approach is dropped: desktop/TUI approvals resolve via
approval.respond → resolve_gateway_approval (the per-session queue), not the
_block()/request_id correlation, so a worker-thread callback waiting on _block
would never be released by the real UI.

Tests: gateway notify payload carries allow_permanent (True without tirith,
False with a tirith warning); ui-tui approvalAction reduced option set +
event-handler allowPermanent propagation; desktop store round-trip + the
ApprovalBar showing/hiding "Always allow".

Supersedes #37856
Closes #37812

Co-authored-by: LeonSGP43 <cine.dreamer.one@gmail.com>
2026-06-11 18:23:59 -05:00
Mani Saint-Victor, MD
9ff0ba0827 fix(desktop): prevent backend port-squat boot loop and pickPort self-collision
Two fixes to the Electron desktop launch path, with the port-reservation logic extracted into a unit-tested module:

1. hermes:bootstrap:reset ("Reload and retry") only cleared connectionPromise, leaving the live backend alive; the orphan kept binding PORT_FLOOR (9120) so the next startHermes() hit EADDRINUSE / "Object has been destroyed" and the window looped. Await teardownPrimaryBackendAndWait() so the reset stops the old backend before restarting.

2. pickPort() probes-then-closes a socket before the real bind happens in a separate Python child, so two concurrent spawns (primary + pool backend) could both be handed PORT_FLOOR and one died with EADDRINUSE. The reservation bookkeeping is extracted into electron/port-pool.cjs (PortPool): pickPort() reserves the chosen port until the child exits and releases it on every exit/error/throw-before-spawn path, closing the TOCTOU window.

PortPool is dependency-injected (probe passed in) and socket-free, unit-tested in electron/port-pool.test.cjs (8 cases) and wired into the test:desktop:platforms script.

(cherry picked from commit d4133945b9)
2026-06-11 18:22:54 -05:00
Brooklyn Nicholson
e3ed7722b5 fix(desktop): refuse a foreign backend's session token after readiness
The served-token fallback adopts whatever token the dashboard HTML
injects. That is correct when our own child regenerated the token (env
pin lost across a shell-wrapped spawn), but wrong when the readiness
probe answered from a process we did not spawn: /api/status is public,
so an orphaned dashboard squatting the port passes waitForHermes while
our child dies on the bind conflict. Silently adopting that process's
token would authenticate the renderer against a foreign backend,
possibly on the wrong profile.

Discriminate on child liveness: the desktop pins
HERMES_DASHBOARD_SESSION_TOKEN on every spawn, so a live child always
serves our token. Served-token mismatch + dead child = foreign backend;
fail the boot loudly instead of connecting. Mismatch + live child keeps
the adopt-served-token salvage from #43720.
2026-06-11 18:18:22 -05:00
Evis
7a2d498b9d fix(desktop): route profile session reads
(cherry picked from commit 64aaf58f5e)
2026-06-11 18:09:24 -05:00
Jeff
e96fe06e49 fix(desktop): use served dashboard token for websocket auth
(cherry picked from commit f8209f91d3)
(cherry picked from commit 72290f0809)
2026-06-11 18:07:19 -05:00
Gille
9102d4a588 fix(dashboard): show Windows 11 in host panel (#44511) 2026-06-11 19:06:29 -04:00
Andrew Fiebert
d221e369b8 fix(desktop): recover from transient assistant-ui index-lookup crash (#44493)
`@assistant-ui/store`'s index-keyed child-scope lookup (`tapClientLookup`)
throws — rather than returning undefined — when a subscriber reads an index
the message/parts list no longer has. During high-frequency store replacement
(switching sessions mid-stream, gateway reconnect replay) a subscriber from
the previous, longer list is still in React's notification queue and reads one
slot past the new, shorter array before it can unmount. The throw
(`Index N out of bounds (length: N)`, the classic index === length off-by-one)
unwinds all the way to the root error boundary and blanks the entire window,
even though the store self-heals on the very next consistent snapshot.

Wrap each virtualized message group in a tiny boundary that swallows ONLY this
transient lookup race and auto-recovers when the message signature changes
(the existing list-mutation key). Any other error re-throws to the root
boundary, so genuine bugs still surface.

Upstream-tracked and unresolved: assistant-ui/assistant-ui#4051, #3652.

Co-authored-by: mollusk <mollusk@users.noreply.github.com>
2026-06-11 22:52:37 +00:00
brooklyn!
b1fe2107d6 fix(desktop): keep named-profile desktop backends per-profile (#44510)
Desktop spawns its dashboard backend with `--profile <name>` and
`HERMES_DESKTOP=1`. cmd_dashboard's unified-launch routing treats any
named profile as a request for the shared machine dashboard: it re-execs
as the default profile (dropping HERMES_HOME) or, when one is already
listening, prints "Machine dashboard already running ... Managing profile
'<name>'" and exits 0. Either way the desktop-spawned child exits before
the app sees a ready backend, so Desktop retries forever — the Windows
named-profile boot loop in the post-mortem.

Skip the machine-dashboard reroute when HERMES_DESKTOP=1 so desktop pool
backends stay per-profile (which is what the pool expects). Carved out of
#44478.

Co-authored-by: AJ <yspdev@gmail.com>
2026-06-11 22:47:28 +00:00
brooklyn!
73969771a5 fix(desktop): discover MCP tools for dashboard /api/ws backends (#44512)
The desktop chat surface talks to the dashboard's in-process /api/ws
gateway, which builds agents through tui_gateway.server._make_agent. That
path only snapshots the existing tool registry — MCP discovery is started
by tui_gateway/entry.py (the stdio TUI), which the dashboard process never
runs. So a profile's configured MCP servers never connect under the
desktop app and sessions show no MCP tools.

Start a shared background MCP discovery thread at dashboard startup (via
hermes_cli.mcp_startup, bounded so a slow/dead server can't block boot),
and have _make_agent briefly join that thread in addition to the existing
entry-owned TUI thread before snapshotting tools.

Carved out of #44478.

Co-authored-by: AJ <yspdev@gmail.com>
2026-06-11 22:45:45 +00:00
Austin Pickett
2ee69d0579 fix(skills): let ClawHub index build walk past the 12s browse budget (#44500)
The deploy-site skills index crawl was capped at ~3k ClawHub entries
because CATALOG_WALK_BUDGET_SECONDS applied to max_items=0 walks too.
Only enforce the wall-clock budget for bounded browse requests and pass
limit=0 from build_skills_index so CI walks the full catalog.

Co-authored-by: Cursor <cursoragent@cursor.com>
2026-06-11 18:03:11 -04:00
Austin Pickett
021ed69141 docs: finish Automation Blueprints terminology rebrand (#44470)
* docs: finish Automation Blueprints terminology rebrand

Replace leftover "Automation Templates" wording from the Cron Recipes
rebrand, rename the copy-paste cookbook guide to Automation Recipes, and
point the marketing gallery link at the blueprints catalog.

Co-authored-by: Cursor <cursoragent@cursor.com>

* docs: use Automation Blueprints instead of Recipes in guide

Rename the cookbook guide from automation-recipes to
automation-blueprints so sidebar and copy match the product term.

Co-authored-by: Cursor <cursoragent@cursor.com>

* docs: rename automation-blueprints-catalog to automation-blueprints

Drop the -catalog suffix from the reference page slug and title, and
move the copy-paste cookbook to automation-blueprint-examples so the
main Automation Blueprints doc is unambiguous.

Co-authored-by: Cursor <cursoragent@cursor.com>

* Revert "docs: rename automation-blueprints-catalog to automation-blueprints"

This reverts commit 605f1eeab5.

---------

Co-authored-by: Cursor <cursoragent@cursor.com>
2026-06-11 17:22:22 -04:00
Teknium
6c752ca3a5 refactor(agent): tighten SUMMARY_PREFIX wording and fix stale doc references
Legibility pass on the consolidated prefix: collapse the topic-overlap rule
from three overlapping sentences into one WINS sentence + one discard/no-wrap-up
sentence (same constraints, less dilution), fix the module docstring to
describe the headings that actually shipped, and correct the #10896 comment's
heading name (Historical Pending User Asks).
2026-06-11 13:57:13 -07:00
Teknium
acb2954d82 fix(agent): freeze carveout-era SUMMARY_PREFIX for renormalization
The prompt consolidation above retires the carveout-era prefix. Without a
frozen copy in _HISTORICAL_SUMMARY_PREFIXES, summaries persisted by
pre-upgrade builds would lose detection (_is_context_summary_content) and
renormalization (_strip_summary_prefix) — the exact regression class the
tuple exists to prevent. Adds contract tests covering every frozen prefix.

Refs #41607 #38364 #42812
2026-06-11 13:57:13 -07:00
kyssta-exe
8f8cad7ec5 fix(agent): strengthen compression preamble against stale task execution (#41607) 2026-06-11 13:57:13 -07:00
konsisumer
d5e2fbf244 fix(agent): frame compaction handoff sections as historical context 2026-06-11 13:57:13 -07:00
brooklyn!
484f484c25 fix(desktop): carve sidebar nav rows out of the titlebar drag region (#44453)
A WSL2 user reported the top two left-sidebar items being unclickable
while the rest of the UI works. That symptom shape matches an
-webkit-app-region:drag hit-test band eating clicks, not GPU/compositing:
the shell's titlebar drag strips (app-shell.tsx) span the top 34px and
the nav group clears them by only 6px, and drag regions win hit-testing
over DOM regardless of pointer-events. Linux WCO (Electron >=32) is the
newest implementation and has known region quirks (electron#43030).

Apply the same no-drag carve-out the codebase already uses for sticky
user bubbles (USER_BUBBLE_BASE_CLASS in thread.tsx) to the sidebar nav
buttons. Harmless on every platform: the rows were never meant to be
draggable surface.
2026-06-11 15:10:09 -05:00
teknium1
114e265737 fix(plugins): don't cache a failed discovery sweep as discovered
Root-cause hardening for the stranded-empty-registry failure behind
'No web search/extract provider configured': discover_and_load() set
_discovered=True before scanning, so a sweep that raised partway was
swallowed by callers as a warning and every later call early-returned
against an empty registry for the process lifetime. The flag now acts
only as a re-entrancy guard and is reset when the sweep raises, so the
next call retries discovery.
2026-06-11 12:56:44 -07:00
xxxigm
32a73010bb test(web): cover keyless default surviving a failed plugin sweep
Pins the invariant that _ensure_web_plugins_loaded registers the keyless
Parallel default (and the wider bundled set) even when the general plugin
discovery raises, that the direct-registration fallback honors plugins.disabled,
and that it stays a no-op on the healthy path.
2026-06-11 12:56:44 -07:00
xxxigm
93764b9303 fix(web): guarantee the keyless web default registers even if discovery doesn't
web_search/web_extract are documented to work with zero setup via the bundled
keyless Parallel free-MCP backend, but that only holds when the bundled
plugins/web/* providers are registered. The dispatch relied entirely on the
general plugin sweep to do that; when the sweep finishes without registering
them (its exception swallowed as a warning, a packaged layout where it ran
before the bundled tree was importable, or a stale empty-discovery cache), the
registry is empty and BOTH tools dead-end on "No web {search,extract} provider
configured" — despite needing no setup at all.

_ensure_web_plugins_loaded now verifies the keyless default landed after the
sweep and, if not, registers the bundled web providers directly against the
registry. Idempotent, a no-op on the healthy path (one dict lookup), and honors
an explicit plugins.disabled entry.
2026-06-11 12:56:44 -07:00
Austin Pickett
c3464ecf45 fix(discord): recover from runtime gateway task exits (#44383)
* fix(discord): recover from runtime gateway task exits

Salvaged from #39416 (AMEOBIUS) — cherry-picked only the task-exit
recovery; the original PR was 1081 commits behind with 28 unrelated
commits.

A post-ready discord.py WebSocket crash left the gateway split-brained:
producers stayed active while Discord stopped responding. After this fix
the adapter calls _set_fatal_error(retryable=True) + _notify_fatal_error()
so the existing GatewayRunner reconnect watcher replaces the dead adapter.

Also adds _wait_for_ready_or_bot_exit() so startup failures (SOCKS/proxy
errors, invalid tokens) surface fast instead of burning the full ready
timeout. Because connect() no longer waits via asyncio.wait_for on that
path, test_connect_releases_token_lock_on_timeout is updated to trigger
the timeout through the new helper (same lock-release contract).

3 tests pass (2 new runtime-failure tests + the updated timeout test);
test_discord_connect.py and test_discord_slash_commands.py green.

Co-Authored-By: ameobius <ameobius@local.host>

* fix(test): patch _wait_for_ready_or_bot_exit in timeout cancel test

connect() no longer uses asyncio.wait_for for the ready handshake, so
test_connect_timeout_cancels_bot_task was hanging for 30s in CI.

Co-authored-by: Cursor <cursoragent@cursor.com>

---------

Co-authored-by: ameobius <ameobius@local.host>
Co-authored-by: Cursor <cursoragent@cursor.com>
2026-06-11 15:39:01 -04:00
ethernet
e080365a7a fix(tui): new weird typeerror 2026-06-11 15:36:39 -04:00
ethernet
5e5308d34d fix(node): fix @types/node version
TODO lock to a specific node/npm version.
this is a fix for a diff between 10 and 11.
2026-06-11 15:36:39 -04:00
teknium1
08b1c44a53 fix(discord): extend bot-task cancellation to connect()'s generic exception branch
Follow-up to #44389: the generic 'except Exception' branch in connect()
had the same orphaned-task hazard as the timeout branch. Extract the
cancel-and-await logic into _cancel_bot_task() and call it from all
three sites (timeout branch, exception branch, disconnect()).

Also adds deaneeth to AUTHOR_MAP.
2026-06-11 12:09:18 -07:00
Dineth Hettiarachchi
020ef76cf1 fix(discord): cancel _bot_task on connect() timeout to prevent zombie client
When connect() times out waiting for the Discord ready event, the background
asyncio.Task running client.start() was not cancelled. discord.py's internal
reconnect loop can ignore client.close() while a WebSocket handshake is in
flight, so the orphaned task eventually completes and fires on_ready.

A later successful reconnect then leaves two live Discord clients in the same
process — each with its own on_message handler and MessageDeduplicator instance
— so every @mention creates two threads because the per-adapter dedup caches
cannot catch cross-client duplicates.

Fix: explicitly cancel and await _bot_task in two places:
1. The asyncio.TimeoutError handler inside connect() — catches the case where
   the adapter's own inner wait_for fires before the gateway's outer timeout.
2. The start of disconnect() — the load-bearing path, always reached via
   _dispose_unused_adapter regardless of which timeout fired first.

Root cause confirmed from production logs: a Jun 8 network outage caused three
consecutive connect() timeouts. The first attempt's bot_task completed its
handshake 4 minutes later ("Connected as") with no preceding watcher line,
then the watcher's real reconnect also connected 90 seconds after that. The two
clients ran continuously for 41+ hours, confirmed by the same user message
appearing as two separate inbound events in two different thread IDs 357ms apart.

Regression tests added to tests/gateway/test_discord_connect.py:
- test_connect_timeout_cancels_bot_task: simulates a connect() timeout with a
  NeverReadyBot and asserts _bot_task is None afterward
- test_disconnect_cancels_running_bot_task: injects a live zombie task, calls
  disconnect(), and asserts the task is cancelled and the attribute cleared
2026-06-11 12:09:18 -07:00
Teknium
13650ab7f8 fix(gateway): audio attachment note no longer steers the agent into punting
Sibling site of the PDF/DOCX note fixed in PR #44175: the audio file
attachment context note led with "Ask the user what they'd like you to
do with it", steering the model into asking instead of transcribing.
Rewritten to instruct the agent to transcribe/process the file itself
when the request involves its content, only asking when intent is
genuinely unclear. Contract assertion added to the existing audio
attachment note test.
2026-06-11 11:58:19 -07:00
xxxigm
4e9be3ee32 test(gateway): cover document context note for PDF/DOCX vs text
Pin the contract for _build_document_context_note: text documents confirm the
inlined content and record the path; binary documents (PDF/DOCX/XLSX/octet-
stream) tell the agent to extract the text itself and never instruct it to ask
the user to paste the contents.
2026-06-11 11:58:19 -07:00
xxxigm
e7ae145ac4 fix(gateway): guide the agent to read attached PDF/DOCX instead of punting
When a user attached a binary document (PDF, DOCX, XLSX, …) in chat, the
context note prepended to the turn said "Ask the user what they'd like you to
do with it." That steered the model into asking the user to paste the
contents rather than extracting the text it is fully capable of reading — so
attached PDFs/DOCX appeared "unreadable" to the agent.

Rewrite the binary-document note to tell the agent the file is a non-text
format saved at the given path and to extract its text itself (e.g. via the
terminal tool or the ocr-and-documents skill) before answering. Text
documents (whose content is already inlined by the platform adapter) keep
their existing note. The note construction is pulled into a small
`_build_document_context_note` helper so it is unit-testable.
2026-06-11 11:58:19 -07:00
Austin Pickett
ce99a81123 fix(dashboard): suppress unicode-animations postinstall during npm ci
Set CI=1 in _run_npm_install_deterministic so the package's /dev/tty
postinstall demo is skipped during hermes dashboard web UI builds.

Co-authored-by: Cursor <cursoragent@cursor.com>
2026-06-11 11:49:08 -07:00
xxxigm
743c55efa3 fix(desktop): stop file tree throwing "Cannot have two HTML5 backends" on remount (#43541)
* fix(desktop): stop file tree throwing "two HTML5 backends" on remount

The Agent Workspace file tree (react-arborist) shows a permanent "TREE ERROR"
with `[error-boundary:file-tree] Cannot have two HTML5 backends at the same
time.` react-arborist mounts its own react-dnd DndProvider + HTML5Backend per
<Tree>. react-dnd v14 keeps that manager on a global, ref-counted singleton
context and nulls it when the count reaches 0. The tree is keyed on
`${cwd}:${collapseNonce}`, so changing folder / collapsing forces a fresh
<Tree>; during the remount the singleton can be torn down and recreated while
the previous HTML5Backend still owns `window.__isReactDndHtml5Backend`, so the
new backend's setup() throws. The error boundary then sticks, because "Try
again" just remounts into the same race.

Pass arborist a stable, app-lifetime `dndManager` (new getFileTreeDndManager
singleton) so it reuses one backend for the life of the app and never
double-claims the window flag. Drag/drop is already disabled on this tree;
this only changes how the (unused) dnd backend is provisioned.

Promotes dnd-core and react-dnd-html5-backend to explicit deps (already present
transitively via react-arborist's react-dnd 14.x line, so they dedupe to one
instance).

* fix(nix): bump npmDepsHash for desktop dnd deps

Adding dnd-core / react-dnd-html5-backend changed the workspace
package-lock.json, so the single workspace-root npmDepsHash in
nix/lib.nix was stale and the nix build failed. Regenerate it
(hash from the failing nix CI job's 'got:' value).

* fix(nix): update npmDepsHash for merged lockfile

After merging main, the workspace lockfile combined main's dep
changes with the desktop dnd additions, so the npmDepsHash needed
recomputing again. Hash from the nix lockfile-check job.

* fix(nix): use fetchNpmDeps hash for desktop dnd lockfile

prefetch-npm-deps reported sha256-lVnybH9RE/... but fetchNpmDeps
wants sha256-mYgKXE/FL4hnkrEvpVv+ULM/oeyIfO2AM9Ol8OrfWm0= for the
merged workspace lockfile. Use the nix build 'got:' hash so CI passes.

---------

Co-authored-by: Brooklyn Nicholson <brooklyn.bb.nicholson@gmail.com>
2026-06-11 11:47:34 -07:00
liuhao1024
93a2f680fd fix(desktop): preserve explicit hide-all choice in model visibility dialog (#43496)
When a user toggles off the last visible model for a provider group, the
effectiveVisibleKeys() function treated the missing provider prefix as
'never customized' and re-added the default models on the next render,
causing all models to snap back to enabled.

Fix: store a sentinel key (e.g. 'provider::') when the last model for a
provider is toggled off. The sentinel distinguishes 'user hid everything'
from 'user never customized', preventing the default-fallback path from
re-adding models the user explicitly chose to hide.

Fixes #43485
2026-06-11 13:27:38 -05:00
brooklyn!
8505e9d669 fix(desktop): disable spellcheck on composer inputs (#44415)
Turn off browser spellcheck, autocorrect, and autocomplete on the main chat composer and message-edit composer so code, paths, and slash commands are not flagged or altered.
2026-06-11 18:03:23 +00:00
brooklyn!
a4f179c509 fix(agent): steer GPT/Codex family to V4A for single-file edits too (#44411)
The coding-posture brief told GPT/Codex models to use patch mode='patch'
(V4A) for structured/multi-file changes but mode='replace' "for a single
small swap". That second nudge points those models at a format their
first-party harness never taught them.

Verified against openai/codex (current main): apply_patch is the ONLY file
editor in codex-rs — zero occurrences of str_replace/old_string anywhere in
the repo; the grammar (core/src/tools/handlers/apply_patch.lark) is exactly
the V4A dialect our patch_parser implements; the shipped model prompts
(gpt_5_codex, gpt-5.2-codex, gpt-5.1-codex-max + instruction templates)
explicitly say to use apply_patch "for single file edits"; and the tool is
gated per model via ModelInfo.apply_patch_tool_type, i.e. OpenAI ships
V4A-for-everything as model metadata.

The GPT-family line now steers to mode='patch' for all edits, single-file
included. The replace-family line (Claude + open-weight) is unchanged —
Claude Code's FileEdit is old_string/new_string/replace_all exact string
replacement (confirmed from Anthropic's shipped sdk-tools.d.ts, the only
file editor in its tool union), matching our mode='replace'.
2026-06-11 17:52:52 +00:00
Teknium
cb29e8a82e refactor(cron): rebrand Cron Recipes -> Automation Blueprints
Product rename across every surface: module/file names (blueprint_catalog,
tools/blueprints, blueprint_cmd), slash command /cron-recipe -> /blueprint
(alias /bp), dashboard API /api/cron/blueprints, desktop deep-link
hermes://blueprint/<key>, docs catalog page + extract script, and the
skill frontmatter block metadata.hermes.blueprint. No behavior change.
2026-06-11 10:49:47 -07:00
Teknium
3c489fda81 fix(commands): unpin /reset from Slack priority aliases — registry hit the 50-cap
CI tests the PR merged with current main, where the new /memory canonical
command filled Slack's 50-slash cap: with btw/bg/reset all pinned ahead of
canonicals, the last canonical (/debug) got clamped and the Telegram-parity
test failed. Canonical commands must win slots over alias spellings — /new
keeps its native slot and 'reset' stays reachable via /hermes reset.

Also updates test_includes_aliases_as_first_class_slashes to assert the
pinned-alias contract (_SLACK_PRIORITY_ALIASES survive) instead of a
specific unpinned alias's survival, which was the same change-detector
pattern the docstring already warned about.
2026-06-11 10:49:47 -07:00
Teknium
e8b757845d fix(cron-recipes): pre-release hardening — honest cadences, strict slot names, surface-aware UX
Review fixes for the Cron Recipes stack before release:

- hydration-move: */90 in the cron minute field silently wraps to hourly
  (croniter-verified) — 90/120-minute options never fired at their stated
  cadence. Replaced with an hour-field step (0 9-17/2 * * 1-5) and an
  interval_hours slot whose options (1/2/3h) all fire as labeled.
- fill_recipe: reject unknown slot names. A typo'd 'tiem=07:15' used to
  silently create the job at the 08:00 default; now it 422s on the dashboard
  form and errors on the slash/deep-link paths with the valid slot list.
- deliver slot: non-strict enum (options are suggestions, scheduler
  validates downstream) so slack/whatsapp/etc. users aren't locked out;
  GET /api/cron/recipes rewrites its options from cron_delivery_targets()
  so the dashboard form only offers configured platforms; help text no
  longer claims dashboard-created jobs deliver to 'the chat you set this
  up from' (the endpoint strips origin — they go to the home channel).
- gateway: success/accept messages no longer point at /cron (cli_only);
  surface-aware hint instead. Conversational fill now sends the
  'Setting up X — I'll ask you a couple of things…' ack before the agent
  turn, matching the CLI experience.
- important-mail catalog entry: reference the urgency classifier by module
  path (python3 -m cron.scripts.classify_items) instead of baking an
  absolute host path into the job prompt — stale after relocation and
  nonexistent on remote terminal backends. cron/scripts is now a real
  package and ships in the wheel (pyproject packages.find).
- export_recipe: interval schedules round-trip again — parse_schedule
  stores 'minutes' but the renderer only read 'seconds', so every interval
  job exported as the silent '0 9 * * *' fallback.
- skills_hub install: say so when a recipe suggestion is dropped
  (latched dedup or pending cap) instead of printing nothing.

Targeted tests: 58 cron/recipe + 261 web_server pass; E2E-validated all
14 recipes fill+parse, hydration cadences via croniter, typo rejection on
slash + endpoint paths, surface-aware hints, and interval export round-trip.
2026-06-11 10:49:47 -07:00
teknium1
e976faac7a feat(cron-recipes): /cron-recipe <name> seeds a conversational fill
Reworks the chat-line UX: pick a recipe by name and the agent asks you for
what it needs, one question at a time, instead of forcing you to hand-type a
slot=val command line.

- /cron-recipe                  -> lists the catalog
- /cron-recipe <name>           -> forgiving name match (exact/prefix/substring/
                                   fuzzy; ambiguous lists candidates), then seeds
                                   the agent with a natural-language fill request
                                   built from the recipe's typed slots + schedule
                                   and prompt templates. The agent asks for each
                                   value one at a time and calls the EXISTING
                                   cronjob tool. No new tool.
- /cron-recipe <name> slot=val  -> unchanged deterministic path (fill_recipe ->
                                   create_job) for the dashboard/docs/power user.

Mechanism (no new plumbing, invariant-safe — the seed enters as a normal user
turn, never a synthetic injection):
- shared handler returns RecipeCommandResult{text, agent_seed}; match_recipe()
  and build_recipe_seed() are the new shared pieces.
- gateway: dispatch rewrites event.text to the seed and falls through to the
  agent (the same pattern /steer uses).
- CLI: handler sets a one-shot self._pending_agent_seed; the interactive loop
  consumes it right after process_command() and runs it as the next turn.

The typed-slot schema stays the single source of truth (still validates the
form/inline path via fill_recipe); the agent path just renders those slots into
the questions to ask. Docs updated to lead with the name-then-ask flow.
2026-06-11 10:49:47 -07:00
teknium1
1593ca5406 feat(cron): Cron Recipes — parameterized automation templates across every surface
A 'recipe' is a one-place definition of an automation that every surface
renders natively. The slot schema (cron/recipe_catalog.py) is the single
source of truth; four renderers consume it, and all paths end at the same
cron.jobs.create_job — no second job engine.

Form where there's a screen, conversation where there's a chat line:
- Dashboard / GUI app: a Recipes sub-tab on the Cron page renders each
  recipe's typed slots as a form (time-picker, enum dropdown, free-text);
  submit POSTs /api/cron/recipes/instantiate which fills + creates the job.
- CLI / TUI / messengers: /cron-recipe lists the catalog, shows a recipe's
  fields, or fills + creates from a pasted 'key slot=val' command. The shared
  handler (hermes_cli/cron_recipe_cmd.py) names any missing/invalid slot so
  the agent can ask a targeted follow-up.
- Docs: a generated Cron Recipes catalog page (website, .mdx + React cards)
  shows each recipe with a copy-paste command and a 'Send to App' button.
- Desktop: a hermes:// URL scheme (Electron single-instance lock +
  setAsDefaultProtocolClient + open-url/second-instance) routes
  hermes://cron-recipe/<key>?slot=val into the chat composer pre-filled.

Typed slots (time/enum/text/weekdays) with defaults: users never type raw
cron — recipes parameterize time-of-day and weekday sets and translate to
cron expressions; a free-text 'schedule' slot is the full-flexibility escape
hatch. Consent-first throughout: nothing schedules without an explicit submit
or send.

Core:
- cron/recipe_catalog.py — CronRecipe + RecipeSlot, 5 curated recipes,
  recipe_form_schema / recipe_slash_command / recipe_deeplink /
  recipe_catalog_entry renderers, fill_recipe (validate + translate to
  create_job kwargs).
- hermes_cli/cron_recipe_cmd.py — shared /cron-recipe handler (CLI + TUI +
  gateway never drift). CommandDef + dispatch in commands.py / cli.py /
  gateway/run.py.

Dashboard: GET /api/cron/recipes + POST /api/cron/recipes/instantiate
(web_server.py), CronRecipes.tsx gallery+form, Segmented sub-tab on CronPage,
api.ts methods + types.

Desktop: hermes:// scheme end to end (main.cjs deep-link router + ready-queue,
preload onDeepLink/signalDeepLinkReady, global.d.ts types, desktop-controller
composer prefill, electron-builder protocols key).

Docs: extract-cron-recipes.py generator wired into prebuild.mjs,
cron-recipes-catalog.mdx + CronRecipesCatalog React component, sidebar entry.
Generated index json gitignored like skills.json.

Tests: 23 core (catalog/slots/schedule-resolution/validation/renderers/command
handler/generator) + 5 web_server endpoint tests. E2E verified end to end:
slot fill -> create_job -> persisted job with correct schedule/deliver/origin.
2026-06-11 10:49:47 -07:00
teknium1
9a09ea69fb feat(cron): Suggested Cron Jobs — one surface for proposed automations
Hermes can propose automations and let the user accept them with one tap
via /suggestions, instead of making them assemble cron jobs by hand. Every
proposal — wherever it originates — flows through one surface.

Sources (the 'where suggestions come from'):
- catalog: curated starter automations (daily briefing, important-mail
  monitor, weekly review, workday-start reminder) via /suggestions catalog
- recipe: installing a skill that carries a metadata.hermes.recipe block
  registers a suggestion instead of auto-scheduling
- usage / integration: reserved for the background-review detector and
  account-connect triggers (sources defined; emitters land next)

Pieces:
- cron/suggestions.py — the store. add/list/accept/dismiss, dedup+latch by
  key (dismissed proposals never re-offered), pending cap so it can't become
  a nag wall. Accepting calls the existing cron.jobs.create_job — there is
  NO second job engine. Mirrors jobs.py storage (atomic writes, lock, 0600).
- cron/suggestion_catalog.py — the curated set. The important-mail monitor
  entry is where the old proactive-monitor poll->classify->surface engine
  lives now (cron/scripts/classify_items.py + the 'monitor' aux task), as ONE
  catalog automation rather than a standalone feature.
- tools/recipes.py — recipe<->job bridge; register_recipe_suggestion() makes
  a recipe source 'recipe' of this surface. recipe_to_job_spec() is the single
  translation both the direct and suggestion paths share.
- hermes_cli/suggestions_cmd.py — shared /suggestions handler (CLI + gateway
  never drift); /suggestions [accept N|dismiss N|catalog|clear].
- Wired: CommandDef + CLI dispatch (cli.py) + gateway dispatch (gateway/run.py)
  + aux 'monitor' task (config.py) + recipe-install hook (skills_hub.py).

Consent-first throughout: nothing auto-schedules; acceptance is always
explicit; dismissals latch.

Supersedes #41122 (proactive-monitor) and #41127 (recipes): both fold in here
as a catalog entry and a suggestion source respectively.

Tests: store (dedup/cap/accept/dismiss/latch), catalog seeding+idempotency,
recipe->suggestion bridge, command handler, aux config. E2E: recipe SKILL.md
-> parsed -> suggested -> accepted -> real cron job persisted to jobs.json.
2026-06-11 10:49:47 -07:00
Teknium
4d6a133a9f fix(agent): gate skill-index demotion behind the opt-in focus mode (#44387)
The coding posture's names-only demotion of non-coding skill categories
(#44342) applied under the default auto mode, silently changing the skill
index for every user in a git repo. Index changes must be opt-in: demotion
now only fires under agent.coding_context=focus, alongside the toolset
collapse. auto/on leave the skill index untouched; focus semantics are
unchanged (demoted, never hidden; deny-list keeps coding-adjacent and
custom categories at full entries).
2026-06-11 10:00:57 -07:00
Teknium
c7bfc938d5 fix(dashboard): Config page header shows the switched profile's config.yaml path (#44374)
The Config page read config_path from /api/status, which is machine-global
and always reports the profile the dashboard process was started under.
After switching profiles with the global switcher, the header kept showing
the old profile's path (e.g. /root/.hermes/profiles/worker_1/config.yaml)
even though reads/writes correctly targeted the new profile.

Fix: /api/config/raw now returns the resolved path alongside the YAML
(resolved inside _profile_scope, so it follows ?profile=). ConfigPage
prefers that scoped path and only falls back to /api/status for old
servers. ProfileKeyedRoutes already remounts the page on switch, so the
header refreshes immediately.
2026-06-11 09:46:15 -07:00
yoniebans
9121834b31 fix(desktop): scope remote workspace defaults 2026-06-11 09:41:35 -07:00
yoniebans
56a0f48ba6 fix(desktop): tighten remote filesystem wiring 2026-06-11 09:41:35 -07:00
yoniebans
8878484f85 feat(desktop): wire remote filesystem browsing 2026-06-11 09:41:35 -07:00
yoniebans
db79e90130 feat(desktop): add filesystem routing facade 2026-06-11 09:41:35 -07:00
yoniebans
51f47f9a97 feat(desktop): add read-only remote filesystem API 2026-06-11 09:41:35 -07:00
helix4u
e71d746820 fix(mcp): avoid false failed startup status 2026-06-11 09:01:52 -07:00
Teknium
5508f4bc54 fix(cli): utf-8 decode for whatsapp-bridge npm install capture (sibling of #43790) 2026-06-11 09:00:55 -07:00
helix4u
b2043cf157 fix(tui): decode startup subprocess output as utf-8 2026-06-11 09:00:55 -07:00
helix4u
dca11b6650 fix(mcp): preserve stdio argv passthrough 2026-06-11 08:59:55 -07:00
brooklyn!
ee1a744ace fix(agent): demote non-coding skill categories to names-only — never hide skills (#44342)
Real-world failure with the original index pruning: under the default auto
posture, an agent-created ops skill in a demoted category vanished from the
prompt's skill index mid-project, and the agent silently fell back to a
stale sibling skill instead. The "discovery-only" premise didn't hold —
models do not reach for skills_list to rediscover what the index stops
showing them, and agent-created skills are the model's accumulated project
memory (runbooks, pitfalls, operating rules).

Gating pruning behind the opt-in focus mode was the wrong fix too: users
opening a worktree don't know the config exists, so the index-noise win
would effectively never ship.

Instead, the coding posture now DEMOTES non-coding categories rather than
hiding them: each demoted category renders as a single names-only line
("gaming [names only]: allthemons10-ops, mc-backup") with a footer note
explaining the omitted descriptions. Every skill name stays in the prompt,
so memory-anchored recall ("load <name>") keeps working in every mode,
while the description noise is still cut. Applies in auto/on/focus alike;
the general posture demotes nothing. Deny-list semantics unchanged —
unknown/custom categories and coding-adjacent ones keep full entries.

API renamed to match the honest semantics: hidden_skill_categories →
compact_skill_categories, build_skills_system_prompt(hidden_categories=) →
compact_categories=.
2026-06-11 10:25:42 -05:00
teknium1
52c7976f40 fix(whatsapp-cloud): review follow-ups for #43921
- nous_subscription: gate the STT managed-default flip on openai-audio
  entitlement and skip when a local backend (faster-whisper or custom
  command) works; new _local_stt_backend_available() helper + tests
- whatsapp_cloud: WHATSAPP_CLOUD_{DM_POLICY,ALLOW_FROM,GROUP_POLICY,
  GROUP_ALLOW_FROM} env overrides so both adapters can run in parallel;
  normalize allowlist entries (JID/punctuation) to bare wa_id
- whatsapp_cloud: wrap per-message event build in try/except (dedup-marked
  wamids would be silently dropped on Meta's batch retry otherwise)
- whatsapp_cloud: validate media_id before URL/filename interpolation,
  delete transient .ogg after voice upload, FIFO-cap interactive-button
  state dicts and per-chat wamid cache
- whatsapp_common: '# **Title**' headers no longer double-wrap asterisks
- setup wizard: read access token / app secret via getpass on TTYs
- docs: new WHATSAPP_CLOUD_* gating env vars
2026-06-11 07:51:01 -07:00
Teknium
2ecb4e62bb Merge remote-tracking branch 'origin/main' into hermes/hermes-6b48295e 2026-06-11 07:38:25 -07:00
Teknium
9c051f57c3 fix(dashboard): Anthropic API Key entry checks ANTHROPIC_API_KEY, not Claude Code creds; hide deprecated tool-progress env vars (#44286)
Two dashboard fixes:

1. The 'Anthropic API Key' OAuth catalog entry's status fn read
   ~/.claude/.credentials.json (which has its own dedicated claude-code
   entry) and never checked ANTHROPIC_API_KEY at all. It now checks the
   Hermes PKCE file, then the registry env-var order (ANTHROPIC_API_KEY
   -> ANTHROPIC_TOKEN -> CLAUDE_CODE_OAUTH_TOKEN) via get_env_value, so
   keys from .env, the shell, or Bitwarden (injected into the process
   env by load_hermes_dotenv) are all reported, with a '(from Bitwarden)'
   source suffix when applicable.

2. Deprecated HERMES_TOOL_PROGRESS / HERMES_TOOL_PROGRESS_MODE removed
   from OPTIONAL_ENV_VARS so the keys page and setup checklists stop
   offering them. Moved to _EXTRA_ENV_KEYS so .env sanitization and
   reload_env still recognize them for existing users (gateway back-compat
   fallback unchanged).
2026-06-11 07:18:15 -07:00
Teknium
e24c935cf3 fix(bedrock): fall back to non-streaming InvokeModel when IAM denies InvokeModelWithResponseStream (#44293)
IAM policies scoped to bedrock:InvokeModel only (a common least-privilege
setup) reject converse_stream() with AccessDeniedException. The agent loop
hard-prefers streaming and the denial never matched the 'stream not
supported' auto-fallback, so InvokeModel-only users looped on AccessDenied
forever.

- agent/bedrock_adapter.py: new is_streaming_access_denied_error()
  detector (ClientError code check + wrapped-SDK message match);
  call_converse_stream() falls back to converse() on denial.
- agent/chat_completion_helpers.py: bedrock_converse streaming branch
  retries inline via converse() and sets _disable_streaming so later
  turns skip the doomed stream attempt; the chat-completions retry
  block also recognizes the denial for the AnthropicBedrock SDK path
  (message pre-check avoids importing bedrock_adapter — and its lazy
  boto3 install — for unrelated providers).

Both paths print a one-line notice telling the user which IAM action
restores streaming.
2026-06-11 07:15:30 -07:00
b1af653bf6 fix(desktop): Harden local file tree paths (#43618)
* fix(desktop): Harden local file tree paths

Normalize Electron local path handling across file tree, preview, media, and git-root flows. Reject malformed and Windows device paths, recheck sensitive files after realpath resolution, and preserve external symlink traversal with stable renderer errors.

* fix(desktop): Address file tree review feedback
2026-06-11 10:05:59 -04:00
Omar Baradei
e372803554 fix(desktop): refresh session model metadata on switch (#43977)
Co-authored-by: Omar Baradei <omar@kostudios.io>
2026-06-11 10:05:32 -04:00
Austin Pickett
d0e017bac8 fix(gateway): gate oversized Telegram voice/audio before download (#44245)
* fix(gateway): gate oversized Telegram voice/audio before download

Adds a pre-download size check to the Telegram voice and audio inbound
paths. Files that exceed _max_doc_bytes (default 20 MB) are rejected
before get_file() is called, preventing silent OOM-style stalls on large
uploads. A human-readable note is appended to the event text so the
model can explain the limit to the user.

Also extends 403 entitlement detection in recover_with_credential_pool
to cover two additional cases: 'oauth authentication is currently not
allowed for this organization' and Anthropic anthropic_messages-mode 403s,
both of which should be treated as entitlement failures rather than
transient errors.

Tests: 7 new cases in test_telegram_voice_v0_regressions.py covering
the size gate (accept, reject, note text) and the STT-failure notice path.

Salvaged from #40487 (cryptopafi) — cherry-picked the Telegram voice
policy and 403 entitlement fixes; LiveKit/Discord/uv.lock workstreams
left for separate PRs.

* test(gateway): drop orphaned voice tests not backed by this PR

The cherry-picked test file from #40487 included 3 tests for STT-failure
notice and voice-mode (_handle_voice_command 'on' -> voice_only) behavior
that this PR intentionally does NOT salvage (those belong to the LiveKit/
voice-policy workstreams left in #40487). They fail on both this branch
and clean main because the feature code isn't present.

Keep only the 2 tests backed by code actually in this PR:
- test_telegram_audio_size_gate_rejects_oversized_media_before_download
  (covers the _telegram_media_size_allowed guard this PR adds)
- test_voice_tts_is_explicit_audio_reply_opt_in (matches current main)

Removed now-unused imports (MessageEvent, MessageType, AsyncMock).
2026-06-11 10:01:51 -04:00
Teknium
a09343cc96 feat(dashboard): SKILL.md editor on Skills page + attach-skill selector in cron modals (#44231)
Headless/VPS users (dashboard-over-Tailscale, no comfortable SSH) could
list/toggle/install skills and create/edit cron jobs, but not author a
custom skill or link one to a cron job — the UI set WHEN a job runs, but
not WHICH skill it uses.

- Skills page: 'New skill' button + per-row edit pencil open a SKILL.md
  editor dialog (frontmatter + body, server-side validation via the same
  _create_skill/_edit_skill path as the agent's skill_manage tool).
- New endpoints: GET /api/skills/content, POST /api/skills,
  PUT /api/skills/content — all profile-scoped via _profile_scope(),
  which now also retargets tools.skill_manager_tool's import-time
  SKILLS_DIR binding.
- Cron page: skills multi-select in both create and edit modals (parity
  with hermes cron --skill / edit --add-skill); CronJobCreate gains a
  skills field; job cards show an attached-skills badge. update_job
  already accepted skills in updates.
- Tests: 17 new endpoint tests (content read, create/edit validation +
  profile scoping + auth gate, cron skills round-trip).
2026-06-11 06:10:27 -07:00
Teknium
f456f302df fix(gateway): refuse to write service definitions with a temp-dir HERMES_HOME (#44267)
* fix(gateway): refuse to write service definitions with a temp-dir HERMES_HOME

A test/E2E harness that exports HERMES_HOME=/tmp/... and touches any
gateway service write path (install, start self-heal, restart's
refresh_systemd_unit_if_needed) bakes the throwaway home into the
production systemd unit / launchd plist. The gateway then restarts
'healthy' but pointed at an empty temp home — no platforms enabled,
deaf to every message (live incident 2026-06-11: /tmp/hermes-e2e-41264
poisoned the unit during a PR-review E2E probe; the post-update restart
produced a 7-hour zombie gateway).

The existing safety belt only sniffed pytest-shaped markers
(/pytest-of-, /hermes_test). Add a structural guard:
_temp_home_in_service_definition() extracts HERMES_HOME from the
generated systemd unit or launchd plist and refuses the write (with
actionable guidance) when it resolves under tempfile.gettempdir(),
/tmp, /var/tmp, or the macOS /private variants. Wired into all five
write sites: systemd refresh + install, launchd refresh + install +
start self-heal.

* test: patch unit generator in install tests tripped by temp-home guard

CI runs hermetic with HERMES_HOME under a tmp dir, so the real
generate_systemd_unit() output now (correctly) trips the new temp-home
write guard in three install tests. Patch the generator with synthetic
non-temp content — same pattern the existing pytest-marker guard tests
use.
2026-06-11 06:10:08 -07:00
Teknium
8972a151a4 feat(cli,tui): show time since last final agent response on the status bar (#44265)
Adds an idle clock to the context/status bar in both the prompt_toolkit CLI
and the Ink TUI: once a turn completes, a dim '✓ <elapsed>' segment shows how
long the session has been idle since the last final agent response. Hidden
while a turn is live (the per-prompt elapsed timer covers that) and before
the first turn completes.

- cli.py: track _last_turn_finished_at when the agent thread exits, surface
  it via _format_idle_since() in the snapshot, render in both the wide
  fragments path and the plain-text fallback.
- ui-tui: stamp lastTurnEndedAt when busy flips false after a live turn,
  thread it through appStatus -> StatusRule, render via a ticking IdleSince
  segment sharing the duration breakpoint/width budget.
2026-06-11 06:06:19 -07:00
Teknium
a2d7f538d4 fix(delegate): stop subagent tool completion lines leaking into parent CLI display (#44223)
Commit 550b72dd8 changed the concurrent-path tool-result rendering gate
from 'not agent.quiet_mode' to 'tool_progress_mode != off'. Subagents are
constructed with quiet_mode=True but inherit the default
tool_progress_mode='all', so every child tool call during delegate_task
started printing raw ' Tool N completed in Xs - {json...}' lines into
the parent's display, bypassing the curated tree-view relay in
_build_child_progress_callback.

Fix: require BOTH gates — quiet_mode must be off AND tool_progress_mode
must not be 'off' — restoring subagent silence while preserving the
#33860 fix (CLI verbose + tool-progress off stays suppressed). The same
combined gate is applied to the three sibling print sites in
tool_executor.py (concurrent header/args, sequential args, sequential
completion) so the whole class is consistent.
2026-06-11 05:10:10 -07:00
Teknium
9c16ca8790 fix(dashboard): normalize model assignments + confirm-modal for backup import (#44237)
Two beta-reported dashboard bugs:

1. Models page: 'Use as -> Main model' on an analytics card sends
   entry.provider, which falls back to the model's VENDOR prefix
   (modelVendor('anthropic/claude-opus-4.6') == 'anthropic') when the
   session row has no billing_provider. That persisted
   provider: anthropic + default: anthropic/claude-opus-4.6 — a
   vendor-prefixed OpenRouter slug on the NATIVE Anthropic provider.
   New sessions then 400 against api.anthropic.com and the user reads
   it as 'changing models does nothing'. Unknown vendors (moonshotai,
   poolside, ...) were worse: a provider that can never resolve
   credentials.

   Fix: _normalize_main_model_assignment() at the single write
   chokepoint — maps non-provider vendor names back to the user's
   current aggregator (else openrouter), and runs the model through
   normalize_model_for_provider() so the persisted name matches the
   target provider's API format. Wired into both /api/model/set and
   the profile-scoped _write_profile_model.

2. System page: 'Restore from backup' spawns hermes import with
   stdin=DEVNULL, so the CLI's interactive 'Continue? [y/N]' overwrite
   prompt hits EOF and auto-aborts whenever a config already exists
   (always, when the dashboard is running). Fix: ConfirmDialog in the
   dashboard owns the consent, then the endpoint passes --force so the
   restore runs non-interactively.

Validated live: dashboard on a temp HERMES_HOME, repro'd both failure
modes pre-fix (vendor-slug write verified via config.yaml + tui
session.create; import 'Aborted.' in action-import.log), then verified
post-fix (normalized writes, modal -> --force -> restored marker file).
2026-06-11 05:07:58 -07:00
Chris
4717989c10 fix(matrix): isolate room context and restore reliable inbound dispatch (#18505)
* fix(matrix): isolate room context and inbound dispatch

* test(matrix): cover room isolation and dispatch regressions

* docs(matrix): document room isolation and session scope

* fix(matrix): stabilize CI requirement checks

* test(matrix): isolate mautrix stubs in requirements tests

* fix(matrix): port room-scoped status and resume to slash commands mixin

Move Matrix /status scope output and /resume same-room guards from the
pre-refactor gateway/run.py into gateway/slash_commands.py so PR #18505
foundation behavior survives the upstream god-file decomposition.

Uses i18n keys for Matrix resume/status messages. Preserves upstream
session.py fixes (role_authorized, DM user_id isolation).

* docs(matrix): explain inbound dispatch via handle_sync loop

Document why Hermes uses an explicit sync loop with handle_sync() rather than
client.start(), aligning with upstream #7914 diagnostics while preserving
Hermes background maintenance tasks.

* fix(i18n): add Matrix resume/status keys to all locale catalogs

The Matrix /resume and /status slash-command keys added in the foundation
PR must exist in every supported locale file. tests/agent/test_i18n.py
asserts key and placeholder parity across catalogs.

Non-English locales use English strings as interim placeholders until
community translators can localize them.

* fix(matrix): restore gateway authz for allowed_users; honor config require_mention

Revert the early MATRIX_ALLOWED_USERS gate in _on_room_message so inbound
sender authorization stays in gateway authz like main. Parse require_mention
from config.extra (platforms.matrix / top-level matrix yaml) with env fallback,
matching thread_require_mention and fixing Forge when require_mention is set
only in profile config.yaml.

* fix(matrix): harden status scope and allowlisted DMs

* fix(matrix): use session store lookup for resume scope
2026-06-11 07:41:43 -04:00
Teknium
73dd584995 fix(mcp): propagate HERMES_HOME override onto the MCP event loop (#44220)
* fix(mcp): propagate HERMES_HOME override onto the MCP event loop

Closes the known limit documented in #44007: tasks scheduled via
run_coroutine_threadsafe are created INSIDE the MCP loop thread, so they
copy that thread's context — a per-request profile scope (dashboard
?profile= endpoints, e.g. the MCP 'Test server' probe) silently vanished
for anything resolving get_hermes_home() inside the coroutine. Most
visible symptom: OAuth token-store paths (HERMES_HOME/mcp-tokens/)
resolved against the process home instead of the selected profile, so
testing an OAuth MCP cross-profile read the wrong tokens.

_run_on_mcp_loop now wraps scheduled coroutines with the caller's
context-local override (_wrap_with_home_override): set inside the task's
own context on the loop, reset on completion — task-local, so concurrent
calls carrying different scopes don't interfere, and the loop thread's
default context stays untouched. No-op (coroutine passes through
unwrapped) when no override is active, i.e. every non-dashboard caller.

web_server's probe comment updated from 'known limit' to 'covered'.

Tests: override propagation (direct + factory form), OAuth token-path
resolution on the loop, loop-context cleanliness after scoped calls,
no-op passthrough. 225 green across mcp_tool + unification suites.

* test(mcp): concurrent different-scope calls don't interfere
2026-06-11 04:37:01 -07:00
Teknium
3edd09a46f fix(whatsapp): restart stale bridge processes instead of silently reusing them (#44205)
A long-lived Baileys bridge survives gateway restarts AND hermes update:
connect() adopted any bridge already listening with status connected, and
disconnect() only kills bridges the adapter spawned itself. Users who
updated to get inbound media support kept talking to a bridge process
serving months-old bridge.js — images and voice notes still arrived as
placeholders with no cached file path (refs #19105 follow-up reports).

Three fixes in the same stale-bridge class:

- Staleness handshake: bridge.js reports a sha256 self-hash in /health
  (scriptHash); connect() compares it against bridge.js on disk and
  restarts the bridge on mismatch. Pre-handshake bridges report no hash
  and are treated as stale, so every existing stale bridge gets recycled
  exactly once on the next gateway start.
- npm dep refresh: deps reinstall when package.json changes (stamp file
  in node_modules), not only when node_modules is missing — a Baileys
  pin bump now actually lands.
- Cache-dir passthrough: the gateway passes profile-aware
  HERMES_{IMAGE,AUDIO,DOCUMENT}_CACHE_DIR to the bridge instead of the
  bridge hardcoding ~/.hermes/image_cache etc., fixing media paths under
  HERMES_HOME overrides, profiles, and the new cache/ layout.
2026-06-11 03:47:29 -07:00
Teknium
875aa8f162 feat(dashboard): unify multi-profile management — one machine dashboard, global profile switcher (#44007)
* feat(dashboard): unify multi-profile management — one machine dashboard, global profile switcher

The dashboard becomes a machine-level management surface with one
write-target selector, replacing per-profile dashboard fragmentation.

Backend:
- profile param (query or body) on /api/config (get/put/raw), /api/env
  (get/put/delete/reveal), /api/mcp/servers (list/add/remove/test/enabled),
  /api/mcp/catalog (list/install), /api/model/info, /api/model/set —
  all scoped through the existing _profile_scope() context manager
- model/set restructured: expensive-model warning (await) runs before the
  scope; the config write runs sync inside the scope in a worker thread
- MCP catalog installs + git-bootstrap entries spawn 'hermes -p <profile>'
- chat PTY: ?profile= on /api/pty points the child's HERMES_HOME at the
  profile dir (its own gateway subprocess, config/skills/memory/state.db
  all profile-bound); in-process gateway attach skipped when scoped

CLI launch unification:
- '<profile> dashboard' routes to the machine dashboard: attach (open
  browser at ?profile=) when one is listening, else re-exec pinned to the
  default profile with --open-profile preselecting the launcher
- --isolated preserves the old dedicated per-profile server behavior
- start_server(initial_profile=...) appends ?profile= to the auto-open URL

Frontend:
- ProfileProvider + sidebar ProfileSwitcher: ONE global selector, URL-
  persisted (?profile=), mirrored into fetchJSON which auto-appends the
  param to the scoped endpoint families (explicit params win)
- app-wide amber banner names the managed profile
- SkillsPage's page-local selector (from the skills-scoping PR) folded
  into the global context — single source of truth
- ChatPage threads the scope into the PTY WS URL; switching profiles
  remounts the terminal into a fresh scoped session

Omitted profile keeps legacy behavior everywhere.

* docs(dashboard): document machine-level multi-profile management

- web-dashboard.md: 'Managing multiple profiles' section (switcher, URL
  deep-links, unified launch, --isolated, scoped Chat, what stays
  per-profile) + --isolated in the options table
- profiles.md: 'From the dashboard' subsection + set-as-active vs
  switcher clarification
- cli-commands.md: --isolated flag + profile-alias launch example

* fix(dashboard): address profile-unification review findings

Review findings (dev review on PR #44007):

1. HIGH — stale page state on profile switch: pages load data on mount
   and didn't consume the profile scope, so a page opened under profile A
   kept showing A's state while writes silently targeted the newly
   selected B. Fixed structurally: ProfileKeyedRoutes wraps the routed
   page tree and keys it by the selected profile, remounting every page
   (fresh state + refetch) on switch. ChatPage keeps its own remount
   (channel keyed on scopedProfile).

2. HIGH — /api/model/auxiliary read was unscoped while /api/model/set
   wrote scoped (Models page could show default's aux pins while editing
   worker's). Endpoint now takes profile + _profile_scope, added to
   PROFILE_SCOPED_PREFIXES, HTTPException re-raise so ghost profiles 404
   instead of 500. Regression test asserts read/write symmetry with
   differing worker/default aux config.

3. MEDIUM — tools post-setup spawned unscoped from the profile-aware
   drawer. Now spawns 'hermes -p <profile> tools post-setup <key>'
   (same mechanism as hub installs); drawer threads its profile prop.
   Most hooks install machine-level artifacts where the scope is inert,
   but hooks reading config/env now see the drawer's HERMES_HOME.

4. LOW — ty warnings: env Optional asserts before subscript/membership,
   fastapi import replaced with web_server.HTTPException re-use.

298 tests green across the four affected suites; tsc -b + vite build
green; aux scoping E2E-verified with real imports.

* fix(dashboard): address second profile-unification review (gille)

1. BLOCKER — profile scope dropped on sidebar navigation: ProfileProvider
   derived the selection from the current URL, and nav links are bare
   paths, so clicking Config from /skills?profile=worker silently reset
   the write target. State is now the source of truth; an effect
   re-asserts ?profile= onto the new location after every navigation
   (URL stays a synchronized projection for deep links/refresh), and an
   incoming URL param (e.g. 'Manage skills & tools' links) still wins.

2. BLOCKER — /api/model/options unscoped while model/set wrote scoped:
   the picker context (current model/provider, custom providers,
   per-profile .env auth state) now loads inside _profile_scope; added
   to PROFILE_SCOPED_PREFIXES. Test: a worker-only current-model pin
   appears in the scoped payload and not the unscoped one.

3. BLOCKER — MCP test-server probe escaped the scope after the config
   read: the probe now re-enters _profile_scope inside the worker thread
   so env-placeholder expansion resolves against the selected profile's
   .env. Known limit (documented): the probe's dedicated MCP event-loop
   thread doesn't inherit the contextvar (OAuth token paths). Test
   asserts get_hermes_home() inside the probe == the worker profile dir.

4. BLOCKER — broad excepts swallowed unknown-profile 404s: /api/model/info
   degraded to 200-with-empty-model-info and /api/mcp/catalog to a
   silently-empty catalog. Both re-raise HTTPException; 404 regression
   tests added for info/options/catalog.

Polish: scope banner clears the fixed mobile header (mt-14 lg:mt-0);
--open-profile hidden via argparse.SUPPRESS (internal re-exec flag);
attach-path test now asserts the opened ?profile= URL.

(Stale-page-state + /api/model/auxiliary findings from this review were
already fixed in 92bcd1568 — the review ran against e600f6951.)

35 tests in the two new suites + 274 in the adjacent ones, all green;
tsc -b + vite build green; scoping E2E-verified with real imports.

* docs(dashboard)+fix: self-review pass — Profiles page section, REST profile-param tip, body-beats-query precedence

Docs:
- web-dashboard.md: add the missing 'Profiles' subsection to Pages
  (cards, create/builder, manage-skills jump, set-as-active vs switcher
  distinction, editors); REST API section gets a profile-scoped-endpoints
  tip documenting ?profile= / body profile / 404 semantics / /api/pty
- (profiles.md + cli-commands.md were already updated in e600f6951)

Precedence fix: scoped endpoints taking BOTH a query param and a body
field now resolve body.profile first. The SPA's fetchJSON injects the
query param from the GLOBAL switcher; an explicit body.profile (e.g.
Profile Builder flows writing into a specific new profile) is the more
specific intent and must not be overridden by whatever the sidebar
happens to be set to. Matches the documented 'explicit beats global'
contract in api.ts.

Verified: 304 tests green across the four suites; tsc -b + vite build
green; docusaurus build green (only pre-existing broken-link warnings,
none from this PR's pages).
2026-06-11 03:29:33 -07:00
Teknium
85503dceca Merge pull request #44038 from NousResearch/hermes/hermes-fb4ee8ce
fix(cli): show quick commands in /help output
2026-06-11 03:04:30 -07:00
kshitij
955fa40062 Merge pull request #44085 from kshitijk4poor/review/pr-43754-ssh-update
fix(update): avoid SSH auth for passive official checks
2026-06-11 01:12:03 -07:00
liuhao1024
0d3e2cc539 fix(desktop): deduplicate sidebar rows by compression lineage in mergeSessionPage (#43487)
When auto-compression rotates the session tip (old #4 → new #5), the
incoming page carries the new tip but the previous list still holds the
old one. The old tip's id differs from the new tip's id, so the existing
id-only dedup in mergeSessionPage() preserves both as separate sidebar
rows.

Add lineage-level dedup: build a set of incoming lineage keys
(`_lineage_root_id ?? id`) and filter survivors whose lineage key
matches any incoming row. This mirrors the existing sessionPinId()
logic used for pin stability.

Fixes #43483
2026-06-11 01:02:27 -07:00
kshitij
c94e93a648 Merge pull request #44084 from kshitijk4poor/salvage/windows-winget-stale-reg
fix(install/windows): repair stale winget registration + refresh/merge PATH after every package manager
2026-06-11 00:25:15 -07:00
kshitij
39f40ece70 Merge pull request #44074 from kshitijk4poor/fix/archive-compressed-session-lineages-salvage
fix(sessions): archive compressed conversation lineages
2026-06-11 00:24:00 -07:00
kshitijk4poor
0edeee14c6 test(desktop): cover official-SSH remote detection for passive updates
Extract the remote-detection helpers (canonicalGitHubRemote, isSshRemote,
isOfficialSshRemote) from main.cjs into a testable update-remote.cjs sibling
module and add a node:test suite, wired into test:desktop:platforms.

main.cjs requires('electron') at load, so its inline helpers weren't unit
testable. The Python side of #43754 shipped a regression test; this gives the
desktop side the same coverage for the security-critical detection that keeps
passive update checks off the SSH origin (avoiding FIDO2/passkey touch
prompts). Tests assert SSH/HTTPS forms canonicalize equal, official SSH is
detected case-insensitively, and forks / other hosts / the HTTPS remote are
NOT misclassified.
2026-06-11 12:53:19 +05:30
kshitij
b4fbf7b93c Merge pull request #44082 from kshitijk4poor/fix/backup-staging-and-nested-skill-dirs
fix(backup): stage SQLite snapshots beside output zip (all paths) and stop excluding nested hermes-agent skill dirs
2026-06-11 00:20:52 -07:00
kshitijk4poor
9662b76d59 fix(install/windows): merge PATH in Update-ProcessPathForPackages instead of overwriting
Follow-up to the winget stale-registration fix. Update-ProcessPathForPackages
rebuilt $env:Path wholesale from the persisted User+Machine hives (plus winget's
Links dir), discarding any process-only PATH entries added earlier in the
installer run. Since the helper now runs after every package manager, that
wholesale replace is more likely to clobber a process-local entry than the
original winget-branch-only version was.

Merge instead: seed from the current process PATH, then append hive and
winget-Links entries not already present, with a case-insensitive,
order-preserving dedupe. Behaviour on a clean box is unchanged (the hive entries
are simply appended); the difference is that pre-existing process-only entries
now survive the refresh.
2026-06-11 12:49:58 +05:30
xxxigm
899acfe42f fix(install/windows): repair stale winget registration; refresh PATH after every package manager
When ripgrep/ffmpeg is missing, `winget install <id>` on a package winget
already has registered is treated as an upgrade: it finds no newer version and
exits 0x8A15002B (-1978335189, APPINSTALLER_CLI_ERROR_UPDATE_NOT_APPLICABLE)
without ensuring the binary is actually present. The installer only logged that
code and judged success by `Get-Command rg`, so a stale registration (files
removed outside winget, or a missing alias shim) became a permanent dead-end —
winget kept reporting "already installed" and the user could never reinstall.

Detect that exit code and retry once with `--force` to repair the registration
so the shim reappears.

Also refresh the process PATH after the choco and scoop fallbacks (not just
winget) via a shared helper, so a successful fallback install — or any install
on a box without winget — is no longer misreported as "not installed".
2026-06-11 12:47:59 +05:30
kshitijk4poor
ed2b9e43c8 fix(backup): stage SQLite snapshots beside output zip in pre-update path too
The pre-update / pre-migration backup path (_write_full_zip_backup) had the
same /tmp staging bug as run_backup: a small tmpfs at the default tempfile
location silently drops large *.db files from the archive. Route its SQLite
staging temp files to the output zip's directory as well, and add regression
tests (mutation-verified) for both staging paths.

Co-authored-by: liuhao1024 <sunsky.lau@gmail.com>
2026-06-11 12:45:40 +05:30
helix4u
cedd9b6d47 fix(update): avoid SSH auth for passive official checks 2026-06-11 12:45:07 +05:30
liuhao1024
dd40600e0a fix(backup): stage SQLite snapshots alongside output zip and stop excluding nested hermes-agent skill dirs
Two bugs in the backup routine:

1. SQLite safe-copy used tempfile.NamedTemporaryFile() which defaults to
   the system temp directory (/tmp).  When /tmp is a small tmpfs and the
   database is large, the copy silently fails and the resulting zip is
   missing state.db, kanban.db, and response_store.db.

   Fix: pass dir=out_path.parent so the temp file is staged alongside the
   output zip on the same filesystem.

2. _EXCLUDED_DIRS contained "hermes-agent" which matched at ANY path
   depth, accidentally excluding the Hermes Agent skill directory at
   skills/autonomous-ai-agents/hermes-agent/.

   Fix: special-case "hermes-agent" to only match when it is the first
   path component (the root-level code checkout).  All other excluded dir
   names continue to match at any depth.

Regression tests added for both fixes.
2026-06-11 12:43:39 +05:30
kshitijk4poor
5e81113d09 chore: map dschnurbusch contributor email for attribution 2026-06-11 12:34:12 +05:30
Dan Schnurbusch
04b3f19538 fix(sessions): archive compressed conversation lineages 2026-06-11 12:31:10 +05:30
Teknium
b8e2c16579 Merge origin/main into salvage branch (resolve AUTHOR_MAP conflict) 2026-06-10 23:25:54 -07:00
kshitij
4829f8d2c5 Merge pull request #44047 from kshitijk4poor/salvage/desktop-stop-stale-session
fix(desktop): recover stale session before stop
2026-06-10 23:23:38 -07:00
teknium1
cb2c13055e fix(gateway): scrub _HERMES_GATEWAY from POSIX detached restart watcher too
Follow-up to the salvaged #41264 (Windows watcher): the setsid/bash detached
restart watcher on Linux/macOS inherits _HERMES_GATEWAY=1 the same way, so
the CLI's self-restart loop guard silently refuses 'hermes gateway restart'
and the gateway never comes back. Scrub the marker from the watcher env on
the POSIX branch as well, and extend the setsid test to assert it.
2026-06-10 23:22:43 -07:00
鼬君夏纪
264ac72b67 fix(gateway,windows): preserve restart watcher env 2026-06-10 23:22:43 -07:00
helix4u
f38f7a3870 fix(desktop): recover stale session before stop
Desktop already recovers from a stale runtime session id when
`prompt.submit` returns `session not found` after a gateway restart or
sleep/wake. The stop path did not have the same recovery: `cancelRun`
called `session.interrupt` once with the stale runtime id, then surfaced
`Stop failed / session not found`.

This makes stop/cancel mirror the prompt recovery path. If
`session.interrupt` reports `session not found` and the selected stored
session id is available, Desktop resumes that durable session, updates
the active runtime ref with the recovered id, and retries
`session.interrupt` once against the recovered runtime id.

Salvaged from #43941 — rebased onto current main, dropping the unrelated
`package-lock.json` (@types/node 24.13.1->24.13.2) and `nix/lib.nix`
hash churn. That bump is a local npm 11 re-resolution artifact, not a CI
requirement: repo CI runs node 22 (npm 10) and main is green at
@types/node 24.13.1, so the lockfile and nix hash do not need to change.

Co-authored-by: helix4u <4317663+helix4u@users.noreply.github.com>
2026-06-11 11:45:08 +05:30
Teknium
2450fd7066 chore: add mvanhorn to AUTHOR_MAP 2026-06-10 22:56:17 -07:00
Matt Van Horn
0b5b7ddfd2 fix(cli): show quick commands in /help output
User-defined quick_commands from config.yaml now appear in the /help
output under a "Quick Commands" section, between skill commands and tips.

Fixes https://github.com/NousResearch/hermes-agent/issues/4090

Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
2026-06-10 22:55:52 -07:00
Shannon Sands
fa7f24e898 Enable webhooks from dashboard page 2026-06-10 22:55:06 -07:00
Teknium
13f1efdd15 fix(gateway): collapse repeated terminal headers in consecutive tool progress blocks (#43968)
When the agent runs several terminal commands back-to-back, each
progress line repeated the '💻 terminal' header above its fenced code
block, cluttering the progress bubble. Now only the first terminal call
in a streak emits the header; subsequent consecutive terminal calls
render adjacent code blocks. Any other tool (or non-block preview)
resets the streak so the next terminal call gets a fresh header.
2026-06-10 22:30:27 -07:00
brooklyn!
4d22b82933 Merge pull request #43959 from NousResearch/hermes/salvage-composer-drafts
fix(desktop): per-thread composer drafts on decoupled lifecycle (salvage #43660, supersedes #43939)
2026-06-11 00:12:23 -05:00
Brooklyn Nicholson
419c8a98a9 Merge remote-tracking branch 'origin/main' into hermes/salvage-composer-drafts 2026-06-11 00:07:07 -05:00
brooklyn!
975edd4140 fix(cli): omit --workspace when subpackage has its own package-lock.json (#42973) (#43986)
* fix(cli): omit --workspace when subpackage has its own package-lock.json

When ui-tui/ (or web/) contains its own package-lock.json, _workspace_root()
returns the subpackage directory itself.  Passing --workspace ui-tui in that
case fails because npm cannot find a workspace named 'ui-tui' inside ui-tui/.

Fix: skip the --workspace flag when npm_cwd equals the target directory,
running a plain 'npm install' from the standalone project root instead.

Applies the same fix to both _make_tui_argv (TUI) and _build_web_ui (web).

Fixes #42973

* test(cli): fix web workspace-scope fixture + cover own-lockfile fallback (#42973)

The web half of the #42977 fix broke test_npm_install_uses_workspace_web_scope,
which built its fixture with no lockfile anywhere. Without a root lockfile,
_workspace_root(web_dir) already returns web_dir, so the new
"() if npm_cwd == web_dir" branch correctly drops --workspace and the
assertion failed. Model a real workspace checkout instead: the single
package-lock.json lives at the root, so --workspace web scopes the install.

Also add the symmetric web regression test (web/ carrying its own lockfile =>
--workspace must be dropped and the install runs plainly from web_dir via
npm ci), matching the TUI coverage already in test_tui_npm_install.py.

---------

Co-authored-by: liuhao1024 <sunsky.lau@gmail.com>
2026-06-11 05:01:25 +00:00
Brooklyn Nicholson
d7d281fa37 feat(desktop): strict per-thread drafts on decoupled composer
Keyed draft stash (Map + localStorage mirror) behind the live composer:
switching threads stashes the departing draft and restores the entering
one; empty threads show an empty box. Session lifecycle never clears
composer state — the scope swap is the only coupling.

Co-authored-by: mollusk <roger@roger.local>
Co-authored-by: Teknium <127238744+teknium1@users.noreply.github.com>
2026-06-11 00:01:06 -05:00
Brooklyn Nicholson
292192f7d7 refactor(desktop): tidy composer draft persistence
- DRY the duplicated submit-restore blocks into dispatchSubmit()
- inline localStorage access (drop browserStorage indirection);
  clearPersistedComposerDraft delegates to write('')
- drop stale per-scope-stash comment in use-session-actions
2026-06-10 23:47:32 -05:00
Brooklyn Nicholson
c710868fbc refactor(desktop): decouple composer from session lifecycle entirely
The composer is a single global surface that sits ABOVE the thread: its
contents follow the user across session switches and are never touched
by session lifecycle. Switching threads doesn't change the render.

Replaces the per-scope draft choreography (scoped storage keys, attachment
stash map, skip-sentinel, restore-on-scope-change effect) with:
- one global localStorage key so an unsent draft survives app reloads
- a one-shot restore on mount
- nothing else — session switches simply don't touch the composer

Verified E2E via CDP with real sidebar clicks + real keystrokes:
typed draft survives A->B->A switching and a full page reload.
2026-06-10 23:39:35 -05:00
brooklyn!
3e74f75e41 feat(agent): coding-context posture across CLI/TUI/desktop/ACP (#43316)
* feat(agent): coding-context posture with per-model edit-format tuning

Hermes detects when it's running in a coding context — an interactive
surface (CLI, TUI, ACP, desktop) sitting in a code workspace (git repo or
recognised project root) — and shifts into a coding posture. Outside that
(chat platforms, non-workspaces) nothing changes.

The posture is modelled as a frozen RuntimeMode selected from a small
ContextProfile registry (coding/general). A profile is data: the toolset to
collapse to, the operating brief to inject, and seams for model routing and
memory. Every domain reads the same resolved object instead of re-probing
git/config on its own:

- System prompt — RuntimeMode.system_blocks(): an operating brief (gather
  context before editing, edit through tools not chat, verify with terminal,
  cap retry loops) plus a live git/workspace snapshot, built once and baked
  into the stable prompt tier so per-conversation caching is preserved.
- Per-model edit-format tuning — the brief nudges each model family toward
  the patch mode it handles best: OpenAI/Codex toward mode='patch' (V4A
  multi-file diffs), Anthropic toward mode='replace' (string replacement).
  The model id rides on RuntimeMode; unknown families keep neutral wording.
- Skill index — non-coding skill categories are pruned from the prompt's
  skill index (discovery-only; skills_list/skill_view still reach the full
  catalog, with a disclosure note).
- Toolset — only under the opt-in 'focus' mode does the posture collapse to
  the coding toolset + enabled MCP servers; the default posture is
  prompt-only and never overrides configured toolsets.

Activation via agent.coding_context: auto (default), focus, on, off.
Subagents inherit the posture for free via toolset inheritance + the shared
prompt builder. Detection is not memoized so a long-lived gateway/TUI
process can't pin a stale posture across working directories.

* feat(agent): cover new-file authoring in the coding edit-format nudge

The per-model edit-format guidance only addressed editing existing code
(patch mode='patch' vs 'replace'), but authoring a brand-new file —
write_file, not patch — is a large fraction of real coding work and the
nudge was silent on it. Surfaced when building a single-file artifact where
the dominant operation was write_file and the steering offered no guidance.

Both family lines now lead with "author new files with write_file; for
edits to existing code prefer ...". Tests assert write_file appears in each
family's brief; unknown families still get neutral wording.

* docs(agent): correct memoization docstring + clarify TUI config-load asymmetry

* feat(agent): sharpen the coding posture — verify-loop facts, wider edit steering, $HOME guard

Tuning pass on the coding posture from dogfooding it as a harness:

- Workspace snapshot now hands the model its verify loop up front:
  detected manifests + package manager (lockfile sniff), the exact
  verify commands (package.json scripts, Makefile targets,
  scripts/run_tests.sh, pytest config), and which context files
  (AGENTS.md / CLAUDE.md / .cursorrules) exist at the root. Marker-only
  (non-git) projects get the snapshot too instead of nothing. The
  "verify before claiming done" brief line was the highest-value piece
  in evals — this turns it from advice into an executable loop instead
  of making the model rediscover the test command every session. Still
  stat-cheap, size-guarded reads, built once at prompt time.

- Edit-format steering covers the families Hermes actually serves:
  Gemini and open-weight coding models (DeepSeek, Qwen, Kimi, GLM,
  Grok, Hermes, Llama, Mistral, Devstral, MiniMax) steer to
  mode='replace' — their RL scaffolds use str_replace-style editors.
  Previously only GPT/Codex and Claude families got steering; the
  models Hermes users disproportionately run all fell to neutral.

- Operating brief gains four behaviors elite harnesses encode: batch
  independent reads/searches in one turn; fix root causes and the bug
  class (sibling call paths), not the reported site; no drive-by
  refactors/renames/reformatting; never read, print, or commit secrets.
  Plus a patch-failure escalation ladder: after the same region fails
  twice, rewrite the enclosing function/file with write_file instead of
  a third patch attempt.

- $HOME dotfiles guard: a git repo rooted exactly at the home directory
  (or a marker sitting in it, e.g. a global ~/AGENTS.md) is user config,
  not a code workspace — without the guard, every session anywhere under
  a dotfiles-managed home silently flipped to the coding posture. Real
  projects under such a home still detect via their own markers/repos;
  'on' mode bypasses the guard.
2026-06-10 23:06:44 -05:00
Brooklyn Nicholson
fdc0d19566 fix(desktop): make draft persistence actually fire — new-chat sentinel, reload flush, session-switch clears
Manual testing of the salvaged draft persistence showed none of it worked
end-to-end. Three distinct bugs, all invisible to the store-level unit
tests:

1. New-chat drafts were never written. The skip-one-persist sentinel was
   reset to null after consuming, but null IS a real scope (the unsaved
   new-session draft) — so in a new chat every persist run matched the
   "consumed" sentinel and bailed. This silently killed the headline
   #38498 fix. Use undefined as the no-skip sentinel, which can never
   collide with a scope.

2. Cmd+R inside the debounce window dropped the trailing text. React does
   not run effect cleanups on a page reload, so the flush-on-unmount
   never fired; with the 400ms debounce that meant type-then-reload lost
   the draft every time. Flush pending writes on pagehide.

3. Session switch/new/resume/branch paths in use-session-actions cleared
   the composer stores synchronously with the session-id updates. React
   batches those, so by the time ChatBar's scope-change cleanup ran to
   stash the departing session's attachments, the store was already
   empty — the stash recorded [] and the chips were lost anyway. The
   composer's per-scope restore now owns composer contents wholesale on
   scope change, so drop the upstream clears (clearComposerDraft only
   touched the vestigial $composerDraft atom nothing reads).

Co-authored-by: mollusk <roger@roger.local>
Co-authored-by: Teknium <127238744+teknium1@users.noreply.github.com>
2026-06-10 22:58:50 -05:00
Teknium
7d8d000b19 revert(cron): remove per-job profile support (PR #28124) (#43956)
Fully removes the cron per-job 'profile' arg added in #28124: the
cronjob tool schema field, CLI --profile flags on cron create/edit,
job-record storage/validation, the scheduler's _job_profile_context
wrapper, and the script-runner env override. Sequential-partition
logic reverts to workdir-only.

The context-local HERMES_HOME override in hermes_constants and the
subprocess bridging in tools/environments/local.py are kept — they
now have other consumers (dashboard multi-profile, TUI gateway).
2026-06-10 20:46:17 -07:00
teknium1
68ffedb6a9 chore(release): map Spaceman-Spiffy for #35586 salvage 2026-06-10 20:45:16 -07:00
teknium1
efcbbde48c refactor: keep anthropic_content_blocks in-memory only (no state.db column)
Drop the hermes_state.py column + persistence plumbing from the salvaged
interleaved-thinking fix. The ordered-block channel covers the failure
window in-memory (turn replayed within the live conversation loop). A
session reloaded from disk after a crash falls back to reconstruction;
if that replay 400s, the thinking-signature recovery (#43667) strips
reasoning_details and retries — one degraded call in a rare resume path
instead of a schema column. Replaces the DB-roundtrip test with a
fallback-shape test.
2026-06-10 20:45:16 -07:00
RaumfahrerSpiffy
7a1eed8268 fix(anthropic): redact replayed tool inputs and broaden thinking-replay 400 recovery
Two additive hardening changes on the interleaved-thinking replay path
introduced by this PR's anthropic_content_blocks channel. Both are scoped
to that channel's blast radius; neither changes correct behavior.

1. Replay-time tool-input re-sourcing (credential safety).
   The ordered-block channel captures each tool_use `input` from the RAW
   API response in normalize_response, which is NOT credential-redacted.
   The parallel tool_calls[].function.arguments IS redacted at storage
   time (build_assistant_message, #19798). The verbatim-replay fast path
   in _convert_assistant_message replayed the raw block input, so a secret
   a model inlined into a tool call (e.g. an Authorization header value
   passed inside a terminal command) would ride back onto the wire even
   though it is redacted everywhere else in history. Re-source tool_use
   input from the redacted tool_calls map by
   sanitized id; interleave order (the reason this channel exists) is
   unaffected. Adapted from #36071, which re-sources tool inputs the same
   way on its replay path.

2. Broaden the thinking-replay 400 classifier (defense-in-depth).
   error_classifier only matched "signature" + "thinking", so the
   frozen-block variant — "thinking ... blocks in the latest assistant
   message cannot be modified. These blocks must remain as they were in
   the original response." — carried no "signature" token and fell through
   to a non-retryable abort. The anthropic_content_blocks channel prevents
   the reorder that triggers this 400 at the source, but if any future
   mutator reintroduces it, the turn now self-heals via the existing
   strip-reasoning-and-retry recovery instead of crash-looping. A negative
   case ensures an unrelated "cannot be modified" 400 (no "thinking") is
   not swept in. Mirrors the classifier broadening in #36087 and #36071.

Tests
- tests/agent/test_anthropic_thinking_block_order.py: a replay test
  asserting an inlined secret is redacted on the wire while interleave
  order is preserved.
- tests/agent/test_error_classifier.py: three cases — frozen-block 400
  native and via OpenRouter route to thinking_signature/retryable; an
  unrelated "cannot be modified" 400 does not.
Both grafts verified RED (tests fail with the change reverted) then GREEN.
Full adapter, transport, classifier and output-field-leak suites pass.

Co-authored-by: AlexanderBFoley <92330381+AlexanderBFoley@users.noreply.github.com>
2026-06-10 20:45:16 -07:00
RaumfahrerSpiffy
529bb1c3d5 fix(anthropic): strip output-only SDK fields from replayed content blocks
HTTP 400 "messages.N.content.M.text.parsed_output: Extra inputs are not
permitted" on the native Anthropic transport. Anthropic SDK 0.87.0 response
blocks carry output-only attributes the Messages *input* schema forbids: text
blocks get `parsed_output` and `citations=None`, tool_use blocks get `caller`.
normalize_response captured blocks verbatim via _to_plain_data and replayed
them as request input on the next turn, so the forbidden fields leaked back ->
400. Like the earlier thinking-block bug, one poisoned turn wedges every
subsequent request in the session (even the diagnostic turn), recoverable only
by switching models or deleting the session.

This is a defect in the anthropic_content_blocks channel added for the
interleaved-thinking fix: it preserved block ORDER correctly but copied every
SDK attribute, including output-only ones.

Fix — whitelist input-permitted fields per block type at all three leak points:
- agent/transports/anthropic.py normalize_response: sanitize at CAPTURE so the
  poison never persists to state.db (defence-in-depth).
- agent/anthropic_adapter.py _sanitize_replay_block (new): whitelist used on the
  ordered-blocks replay path; also recovers already-poisoned stored sessions.
- agent/anthropic_adapter.py _convert_content_part_to_anthropic: a stored
  `text` part is rebuilt from whitelisted fields instead of dict(part) verbatim
  (this was the exact content.N.text.parsed_output failure locus).

Whitelist not blacklist, so future SDK output-only fields can't reintroduce it.
Block order and thinking-block signatures are preserved (the reason the channel
exists). Adds tests/agent/test_anthropic_output_field_leak.py; full adapter
suite green (163 tests). Existing poisoned state.db rows scrubbed out-of-band.
2026-06-10 20:45:16 -07:00
RaumfahrerSpiffy
aaccaada28 fix(anthropic): preserve interleaved thinking/tool_use block order on replay
Interleaved-thinking turns (adaptive thinking, Claude 4.6+/Opus 4.8) emit
content blocks like:

    thinking_1(signed) tool_use_1 thinking_2(signed) tool_use_2

Anthropic signs each thinking block against the turn content preceding it
at its position. normalize_response split the turn into two parallel lists
(reasoning_details + tool_calls), discarding cross-type order, and
_convert_assistant_message rebuilt it as [all thinking][text][all tool_use].
That moved thinking_2 ahead of tool_use_1, invalidating its signature, so
Anthropic rejected the latest assistant message with HTTP 400:

    messages.N.content.M: `thinking` or `redacted_thinking` blocks in the
    latest assistant message cannot be modified.

Observed repeatedly in agent.conversation_loop against api.anthropic.com /
claude-opus-4-8, recurring across sessions on multi-thinking-block turns.

Fix: carry a verbatim, order-preserving copy of the turn's content blocks
(anthropic_content_blocks) end-to-end - capture in normalize_response,
persist/restore through state.db, and replay unchanged for the latest
assistant message. Gated to turns that actually interleave signed thinking
with tool_use, so normal turns are unaffected.

Adds 3 regression tests including a SQLite round-trip covering the
crash-recovery reload path.
2026-06-10 20:45:16 -07:00
Brooklyn Nicholson
65ddc7c4a1 fix(desktop): retain composer attachments per session scope + guard programmatic drafts
The salvaged draft persistence scoped text per session but reset the
composer's attachments to [] on every scope change, so a staged image or
file was silently dropped when you switched sessions and never restored on
return — inconsistent with the "drafts survive session switches" promise
and a real paper-cut given remote staging cost.

Retain attachments per scope in an in-memory map (keyed by the same scope
as the text draft) since blobs / object URLs / live upload state can't be
serialized to localStorage. Entering a scope restores its stashed chips;
leaving stashes the current ones; an accepted submit clears the scope.
This survives session switches (the case users hit) without pretending to
survive a full reload, which attachments fundamentally can't.

Also guard the debounced text write so browsing sent-message history or
editing a queued prompt (both swap the composer to recalled text via
loadIntoComposer) no longer clobbers the genuine in-progress draft in
storage.

Co-authored-by: mollusk <roger@roger.local>
Co-authored-by: Teknium <127238744+teknium1@users.noreply.github.com>
2026-06-10 22:41:34 -05:00
Teknium
ad9012097b fix(dashboard): dedupe useNavigate import/declaration in ProfilesPage
tsc -b (run by the Docker image build, unlike local vite-only checks)
rejected the duplicate identifier.
2026-06-10 20:34:53 -07:00
Teknium
914befa9aa feat(dashboard): profile-scoped skills & toolsets management
'Set as active' on the Profiles page only flips the sticky active_profile
file (future CLI/gateway runs) — it never retargets the running dashboard
process. The skills/toolsets endpoints called bare load_config()/
save_config(), so after 'activating' a profile in the web UI, deactivating
a skill silently wrote into the dashboard's own profile and the activated
profile was untouched.

Backend:
- _profile_scope() context manager on the skills/toolsets endpoints:
  context-local HERMES_HOME override for call-time config resolution +
  cron-style locked swap of tools.skills_tool's import-time SKILLS_DIR
- profile param on /api/skills, /api/skills/toggle, /api/tools/toolsets*
  (list/toggle/config/provider/env), hub sources/search installed-state
- hub install/uninstall/update spawn 'hermes -p <profile> skills ...' so
  the child rebinds skills_hub.SKILLS_DIR at import (the override cannot
  reach import-time globals); profile validated -> 404/400 before spawn

Frontend:
- Skills page: profile selector (deep-linkable /skills?profile=<name>),
  amber banner naming the managed profile, threaded through skill toggles,
  toolset drawer, and hub browser
- Profiles page: 'Manage skills & tools' action per card; 'Set as active'
  toast now says it applies to new CLI/gateway runs only

Omitted profile keeps legacy behavior (dashboard's own profile).
2026-06-10 20:34:53 -07:00
Teknium
3d14f01fd6 fix(desktop): debounce per-keystroke draft persistence writes
The salvaged draft-persistence effect wrote to localStorage on every
keystroke — the composer's per-keystroke path was deliberately slimmed
down previously, so debounce the write (400ms) and flush pending text on
scope change/unmount so a fast session switch can't drop trailing
keystrokes. Also add AUTHOR_MAP entry for the salvaged commit.
2026-06-10 22:34:30 -05:00
Roger
18d61bd06e fix(desktop): persist composer drafts across reloads
Save in-progress composer text to browser localStorage per chat session and restore it when the desktop composer remounts. Keep the draft when submit is rejected or throws, and clear it only after the prompt is accepted.
2026-06-10 22:34:13 -05:00
Teknium
acd7932c0f docs: cross-link write-approval gate from skills, configuration, and slash-command docs (#43801)
The memory/skill write-approval gate (#38199, #43354, #43452) was only
documented inside features/memory.md. Surface it everywhere users will
actually look:

- features/skills.md: new 'Gating agent skill writes' section under
  skill_manage, with the staging semantics, review commands, and the
  distinction from skills.guard_agent_created
- configuration.md: memory.write_approval added to the Memory
  Configuration block; new 'Write approval for skill writes' subsection
  next to the guard_agent_created scanner
- reference/slash-commands.md: /memory and /skills review subcommands in
  both the CLI and messaging tables; Notes updated since /skills
  pending/approve/reject/diff/approval now works on the gateway
- features/memory.md: cross-link to the new skills section
2026-06-10 19:54:44 -07:00
Teknium
0a5762c78d fix(web): genericize free-MCP client identity per telemetry policy
Replace the hermes-identifying clientInfo/User-Agent/session-id prefix on
the keyless Parallel Search MCP path with a neutral 'mcp-web-client'
identity. Project policy forbids third-party usage attribution without an
explicit user opt-in (see telemetry PR policy); MCP requires a clientInfo,
so a generic one satisfies the spec without attributing traffic.

Also adds the contributor AUTHOR_MAP entry and refreshes uv.lock against
current main (parallel-web 0.6.0).
2026-06-10 19:54:38 -07:00
Matt Harris
e0e2571711 feat(web): Parallel-backed web search & extract — free Search MCP when keyless, v1 REST when keyed
Make Parallel the web search/extract backend with a zero-setup free tier:

- Keyless (no PARALLEL_API_KEY): web_search/web_extract work out of the box via
  Parallel's free hosted Search MCP (search.parallel.ai/mcp), and parallel
  becomes the default backend when no other web credentials are configured
  (ahead of ddgs, which is search-only). A small hand-rolled Streamable-HTTP
  JSON-RPC client speaks the MCP's web_search/web_fetch tools; the existing
  web_search/web_extract tools are the only tools registered.
- Keyed (PARALLEL_API_KEY set): uses the Parallel v1 REST endpoints
  (client.search / client.extract with advanced_settings.full_content) — no beta.
  Bumps parallel-web 0.4.2 -> 0.6.0.
- Attribution: on the free path only, results carry provider/attribution and the
  CLI tool line reads "Parallel search" / "Parallel fetch"; the paid path is
  unbranded.
- Selection/registration: web tools register unconditionally (free MCP backstop)
  while check_web_api_key remains a real usability probe; explicit per-capability
  backends are honored (so misconfig surfaces) rather than masked by the fallback.

Tested: live web_search/web_extract against search.parallel.ai in keyless and
keyed modes; unit suites for the MCP client, backend selection, and display
labeling; full agent run shows the "Parallel search" label on the free path.
2026-06-10 19:54:38 -07:00
brooklyn!
fe54960142 desktop: un-truncate the active slash/@ row so long descriptions stay readable (#43926)
Follow-up to #42351. Slash command rows render the command label and
description with `truncate`, so skill commands and longer blurbs were
clipped with no way to read the full text. Rather than add a floating
tooltip (which overlaps the popover and only helps the mouse), the active
row — the one reached by keyboard arrows or hover, since onMouseEnter
already sets activeIndex — now drops truncation and wraps inline
(whitespace-normal break-words). Idle rows stay single-line/truncated so
the list reads compact.
2026-06-11 02:35:38 +00:00
brooklyn!
3ffbdfbcc0 desktop: registry-driven slash commands + first-class /resume & /handoff (#42351)
* desktop: surface /tools, /save, /personality and fix /help skill count

Move /tools and /save out of TERMINAL_ONLY_COMMANDS and /personality out of
ADVANCED_COMMANDS so they appear in the desktop slash palette and execute via
the existing slash.exec → command.dispatch fallback. The backend gateway already
accepts these through slash.exec (none are in _PENDING_INPUT_COMMANDS or the
skill list), so no backend change is required.

Recompute skill_count in filterDesktopCommandsCatalog from the filtered pairs.
Previously the /help footer echoed the unfiltered backend total — e.g. "60
skill commands available" while only ~29 actually appeared in the rendered
list, because the desktop hides terminal-only, picker-owned, and advanced
commands.

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>

* desktop: keep slash popover live while typing args

The trigger regex `(?:^|[\s])([@/])([^\s@/]*)$` stopped matching the moment
the user typed a space after a slash command, so the popover never showed arg
completions for `/personality`, `/tools`, etc. — even though the backend's
`complete.slash` already returns them with a `replace_from` indicator.

Split the trigger detection so `/` allows args (`/cmd arg1 arg2`) while `@`
keeps the strict no-space behavior. Restrict the slash command name to
`[a-zA-Z][\w-]*` so file paths like `src/foo/bar` don't accidentally trigger
the popover.

Rewrite arg-completion items in useSlashCompletions to insert the full
`/personality alice` token instead of stranding `/alice`: when `replace_from`
is past the command base, prepend the existing prefix to each item's text so
the chip serializer produces a coherent replacement.

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>

* cli: complete toolset names after /tools enable|disable

SlashCommandCompleter previously only auto-derived the first subcommand level
from args_hint, so `/tools enable <tab>` yielded nothing — the user had to
remember every toolset key (web, file, spotify, …) and every MCP server prefix.

Add `_tools_completions` that handles both stages: subcommand (list|disable|enable)
and tool name. Filter by current enable state so `/tools enable <tab>` only
offers disabled toolsets and `/tools disable <tab>` only offers enabled ones —
no point suggesting a no-op. MCP server prefixes (server:) come from the
saved mcp_servers config; per-tool completion under a server would require
runtime MCP introspection and is left as follow-up.

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>

* desktop: registry-driven slash commands with first-class pickers

Collapse the if/else slash dispatch into one DESKTOP_COMMAND_SPECS table
that drives popover suggestions, per-type composer pills, and execution.

- /resume, /sessions, /switch: inline session completions (like /skin) plus
  a "Browse all sessions…" entry that opens a dedicated session picker overlay
- /handoff: inline platform completion + handoff.request/handoff.state
  gateway bridge so desktop reaches CLI parity
- colored per-type pills (command/skill/theme) in the composer
- strip ANSI and fix width/alignment of slash output in the chat panel

* desktop: fold repeated slash session/output boilerplate into one helper

runExec, /title, /help and the unavailable case each re-derived the same
ensure-session → bail-with-notify → build-renderSlashOutput dance.
withSlashOutput() returns {sessionId, render} or null, so each handler is
a two-line resolve instead of an eight-line preamble.

* desktop: keep backend meta on slash arg completions

Arg suggestions (/personality <name>, /tools enable <toolset>, /handoff
<platform>) were having their meta overwritten with the parent command's
registry description: desktopSlashDescription("/personality none") canonicalizes
back to /personality and returns its blurb. Skip the lookup for arg rows so the
backend's own display_meta ("clear personality overlay", etc.) survives.

* cli: list real personalities in /personality completion

_personality_completions resolved load_config().agent.personalities — but that
schema has no agent.personalities key, so completion always returned just
`none` even though the runtime (load_cli_config().agent.personalities) ships a
dozen built-ins (helpful, kawaii, pirate, …). Read from the same source the
command actually applies, so `/personality ` surfaces the real options.

* desktop: expand bare arg-commands to their options on pick

Picking a command like /personality from the slash popover committed it
immediately instead of advancing to its argument list. Mark arg-taking
commands (/skin, /resume, /handoff, /personality, /tools) in the registry
and, when one is picked bare, insert "/cmd " as plain text and re-open the
popover on its inline options — mirroring typing "/cmd " by hand. Arg picks
(serialized text already contains a space) still commit a single pill.

Also realign trigger-popover loading test with the redesigned popover (the
/help empty-state hint shows when resolved, not while the spinner is up);
the merge from main reintroduced the pre-redesign expectation.

* tui_gateway: fold session-db close into a context manager

Both handoff RPCs repeated the same `db, close_db = _session_db_handle()`
+ `finally: if close_db: db.close()` dance. Turn the helper into a
`_session_db` contextmanager that owns the close, so callers just
`with _session_db(session) as db:`.

* desktop: unblock handoff retries and exact resume ids

Clear timed-out desktop handoffs through the gateway so retries are not stuck behind a pending row, and let typed /resume session ids bypass the loaded sidebar cache.

---------

Co-authored-by: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
2026-06-11 01:49:24 +00:00
emozilla
bfcc9f92b4 Merge commit '6110aed9b' into feat/whatsapp-cloud-api 2026-06-10 21:39:22 -04:00
xxxigm
615ad97928 fix(streaming): stop socket read timeout from preempting stale-stream detector (#43570)
* fix(streaming): stop socket read timeout from preempting stale-stream detector

The stale-stream detector is deliberately scaled to 180-300s so reasoning
models (e.g. Opus) can pause mid-stream during extended thinking. But the
httpx socket read timeout stayed at a flat 120s for cloud providers and fired
first, tearing down healthy reasoning streams before the detector (which owns
retry + diagnostics) could act. Symptom: every Copilot/Opus turn dies with
ReadTimeout at a consistent ~125s and never completes.

Floor the cloud socket read timeout at the stale-stream timeout so it can no
longer fire before the detector. Local providers and explicit
HERMES_STREAM_READ_TIMEOUT / request_timeout_seconds overrides are unchanged.

* test(streaming): pin read-timeout >= stale-stream invariant for cloud reasoning streams

Cover the contract that the httpx socket read timeout is never shorter than
the stale-stream detector for cloud providers on the default: small contexts
floor to 180s, >=50K to 240s, >=100K to 300s; explicit overrides win; local
providers and the unresolved-value fallback are unaffected.
2026-06-10 20:21:38 -05:00
Austin Pickett
9dd9ef0ec9 fix(web): profiles page modal (#43858)
* fix(web): profiles page modal

* chore: drop unrelated package-lock.json changes

Co-authored-by: Cursor <cursoragent@cursor.com>

---------

Co-authored-by: Cursor <cursoragent@cursor.com>
2026-06-10 20:43:22 -04:00
Teknium
4490c7cf8d fix: in-memory transcript blocks empty-session prune
CI caught tests/cli/test_cli_new_session.py asserting that /new keeps
the old session row when conversation history exists in memory. The
live transcript is authoritative: a session whose messages haven't
flushed to the DB yet (or whose flush failed) must not be pruned.
Guard _discard_session_if_empty on self.conversation_history and pin
the behavior with a test.
2026-06-10 17:37:34 -07:00
Teknium
e96ca1a0d3 feat(sessions): drop empty sessions on CLI exit and session rotation
Port from google-gemini/gemini-cli#27770: starting the CLI and
immediately quitting (or rotating with /new, /clear) left an empty
untitled session row behind. These ghost rows pile up in /resume,
`hermes sessions list`, and the in-chat recent-sessions browser.

- SessionDB.delete_session_if_empty(): transactional check-and-delete
  that only removes rows with no messages, no title, and no child
  sessions (delegate subagent parents are preserved). Also removes
  on-disk transcript files via the existing _remove_session_files.
- HermesCLI._discard_session_if_empty(): thin wrapper, wired into the
  cli_close shutdown path and the new_session() rotation path.
  Skipped when /exit --delete already handles removal.

Unlike the one-shot prune_empty_ghost_sessions migration (TUI-only,
24h-old rows), this prevents new ghost rows from accumulating at the
moment they would be created.
2026-06-10 17:22:27 -07:00
teknium1
d1383a6b14 fix(skills): widen HERMES_HOME-aware .env resolution to all sibling skills
Follow-up to the GitHub-skills fix: the same hardcoded ~/.hermes/.env
pattern existed across other bundled and optional skills. Under the
official Docker setup (HERMES_HOME=/opt/data, subprocess HOME=/opt/data/home)
those paths point at a nonexistent file.

- kanban-video-orchestrator setup.sh.tmpl + docs: resolve via
  ${HERMES_HOME:-$HOME/.hermes}/.env in check_key()
- telephony.py / canvas_api.py / hyperliquid_client.py: error and
  save messages now report the real resolved env path instead of a
  hardcoded literal (path resolution itself was already correct)
- godmode SKILL.md: load_dotenv snippet resolves via HERMES_HOME
- watch_github.py + ~20 SKILL.md prose mentions: document the env file
  as ${HERMES_HOME:-~/.hermes}/.env so Docker users edit the right file
2026-06-10 15:10:11 -07:00
xxxigm
0a593f132c fix(skills/github): resolve .env via HERMES_HOME, not hardcoded ~/.hermes
The GitHub skills' auth-detection fell back to reading GITHUB_TOKEN from a
hardcoded ~/.hermes/.env. In the official Docker layout HERMES_HOME=/opt/data
while tool subprocesses run with HOME=/opt/data/home, so `~/.hermes/.env`
expands to /opt/data/home/.hermes/.env — a path that does not exist — while the
real secrets file is /opt/data/.env. Result: the agent reports GITHUB_TOKEN as
"not set" even though it is present and the dashboard Keys page shows it.

Resolve the file as ${HERMES_HOME:-$HOME/.hermes}/.env (HERMES_HOME is bridged
into tool subprocess env, falling back to ~/.hermes when unset) across all six
auth-detection sites: github-auth (SKILL.md + scripts/gh-env.sh), github-issues,
github-repo-management, github-pr-workflow, github-code-review.
2026-06-10 15:10:11 -07:00
Teknium
3b4c715e1c fix(telegram): stripped-text fallbacks, re-finalize skip, and tail-only delete guard
Follow-ups on top of the two salvaged GodsBoy commits, all live-validated
against the real Telegram Bot API:

- _edit_overflow_split finalize fallbacks degrade to _strip_mdv2() clean
  text instead of putting raw **markdown** markers on screen (salvaged
  from PR #43463 minus its format-first sizing — live probes show
  Telegram's 4096 limit counts PARSED text, so MarkdownV2 escape
  inflation cannot cause MESSAGE_TOO_LONG and sizing against formatted
  wire length only causes premature splits and fragment messages).
- Skip the redundant requires-finalize edit after a got_done edit that
  split-and-delivered (salvaged from PR #43463): re-finalizing re-splits
  the full text into the adopted continuation and duplicates chunks.
- _send_fallback_final only deletes the stale partial message when the
  fallback re-sent the COMPLETE final text. When the prefix dedup sent
  only the missing tail, the partial IS the head of the answer; deleting
  it left users with only the second half of long responses (live-
  reproduced: flood-control during a long stream -> head deleted,
  ratio 0.54 of content visible). This is the third bug behind the
  'Telegram cut messages' reports and was present on main and both PRs.
2026-06-10 15:09:35 -07:00
GodsBoy
da818510ec fix(gateway): finalize best-effort delivery when stream consumer is cancelled 2026-06-10 15:09:35 -07:00
GodsBoy
590b3c0d7e fix(gateway): recover partial Telegram overflow streams 2026-06-10 15:09:35 -07:00
xxxigm
88fcf0c8c0 docs(memory): clarify that memory does not auto-compact when full
The "Persistent Memory" callout said "when memory is full, the agent
consolidates or replaces entries to make room," which reads as if the
store self-compacts automatically. It does not: the `memory` tool
returns an overflow error and the agent does the consolidation in-turn
(the design from #41755). Also note that `replace` is bound by the same
limit — swapping in a longer entry can still overflow — which is the
exact case that confused a user (replace rejected near the cap even
though the math was correct).
2026-06-10 14:39:50 -07:00
xxxigm
f7a6d6a6a1 test(cron): cover provider "custom" → providers.custom resolution
Add execution-time coverage that bare `provider="custom"` resolves a literal
providers.custom endpoint (and still falls through when none exists), plus
creation-time coverage that `_resolve_model_override` keeps a resolvable
"custom" and only pins the main provider when it is unresolvable.
2026-06-10 14:39:03 -07:00
xxxigm
acd4f34e65 fix(cron): resolve per-job provider "custom" to providers.custom instead of codex
A cron job stored with `provider: "custom"` and a matching `providers.custom`
entry in config failed at execution with `auth_unavailable: providers=codex`.
Two layers conspired:

- `_get_named_custom_provider` returned None for bare "custom" *before*
  scanning config, so a literal `providers.custom` entry was never matched and
  resolution fell through to the global default (codex). Now it scans config
  for an entry literally named "custom"; with none it still returns None,
  preserving the legacy model.base_url trust path.
- `_resolve_model_override` blindly stripped bare "custom" at job creation and
  pinned `model.provider` (e.g. codex). It now keeps "custom" when a configured
  custom endpoint resolves, pinning the main provider only when it doesn't.
2026-06-10 14:39:03 -07:00
helix4u
1e7316ced2 fix(desktop): use sudo callback without interactive env 2026-06-10 14:29:56 -07:00
Tranquil-Flow
a8f404b29f fix(gateway): probe launchd domain instead of hardcoding user/<uid> (#40831)
The previous fix for #23387 changed _launchd_domain() from gui/<uid> to
user/<uid> to support Background/SSH sessions on macOS 26+. However, this
broke Aqua sessions where gui/<uid> is the only working domain and
user/<uid> cannot bootstrap or manage the service.

Now _launchd_domain() probes which domain actually contains the loaded
service:
1. Try gui/<uid> first (Aqua sessions)
2. Fall back to user/<uid> (Background/SSH sessions)
3. Use launchctl managername as heuristic when neither has the service
4. Cache the result for the process lifetime

Regression tests cover all four paths plus caching behavior.
2026-06-10 12:39:48 -07:00
teknium1
2d75833abe chore(release): map ianculling for #36087 salvage 2026-06-10 12:39:44 -07:00
0xyg3n
9f95f72b98 fix(agent): strip api_messages in thinking-signature recovery so the retry actually omits thinking blocks
The thinking-signature recovery in agent/conversation_loop.py popped
reasoning_details from messages, then continued to retry. That had two
defects.

First, the strip never reached the wire payload. api_messages is built
once at the start of the turn by shallow-copying every entry in messages
(line 919 area). Each api_messages entry has its own reference to the
same reasoning_details list. When build_api_kwargs runs on every retry
iteration of the inner while-loop, it consumes api_messages, not
messages. Popping reasoning_details from messages left api_messages
untouched, so the retry's request still carried the same thinking
blocks Anthropic had just rejected. The classifier latched
thinking_sig_retry_attempted = True after the first attempt, and the
loop terminated with max_retries_exhausted on the same 400.

Second, the pop mutated the canonical message list. messages is the
same list _persist_session writes to state.db and the session
transcript, so a single recovery permanently wiped every signed
thinking block from the stored conversation. Subsequent turns reloaded
the stripped state, hit the same 400 ('invalid signature' or 'cannot
be modified', see #24107), and the agent stopped responding entirely.
Cascading compaction-ended sessions then chained off the corrupted
parent and the affected chat could not produce a response on any
future turn.

Move the strip onto api_messages, which is the API-call-time list
rebuilt into kwargs on every retry. messages is no longer touched, so
disk I/O stays clean and the recovery actually reaches the wire.

Observed against the native Anthropic Messages API on claude-opus-4-7
and claude-opus-4-8 with the interleaved-thinking-2025-05-14 beta on
hermes-agent 0.12.0 and 0.14.0. PR #24107 narrows the trigger; this
change makes the recovery do what it always claimed to do, and
prevents the destructive aftermath.

Tests cover the api_messages strip in isolation: pop on a shallow copy
does not affect the source, the canonical messages list survives the
strip, idempotency on a duplicate firing path, and a no-op when no
reasoning_details exist on the messages.

Related: #24107, #26959, #17861.
2026-06-10 12:39:44 -07:00
Ian Culling
86e10dd874 fix(agent): route 'thinking blocks cannot be modified' 400 to recovery
Anthropic returns a 400 when the thinking/redacted_thinking blocks in the
latest assistant message are mutated upstream: 'thinking or redacted_thinking
blocks in the latest assistant message cannot be modified. These blocks must
remain as they were in the original response.'

The classifier's thinking_signature branch only matched on the substring
'signature', so this variant fell through to a non-retryable client error
and hard-aborted the turn -- even though the existing strip-reasoning_details
-and-retry recovery would have healed it.

Broaden the 400 match to also catch 'cannot be modified' / 'must remain as
they were' (still gated on 'thinking'), routing it to the same recovery.
Adds a negative-case test so unrelated 'cannot be modified' 400s are not
swept in.

Defense-in-depth, orthogonal to the root-cause work in #35975 / #17861
(which prevent the block mutation in the first place). Only changes a
terminal-failure into a one-shot recovery.

Signed-off-by: Ian Culling <ian@culling.ca>
2026-06-10 12:39:44 -07:00
rob-maron
6110aed9be Suppress "Credit access paused" notice on free models (#43669)
* don't show credits message on free model

* PR comments
2026-06-10 23:55:06 +05:30
brooklyn!
6de3963e37 fix(desktop): keep model runtime state per session (#43702)
* fix(desktop): keep model runtime state per session

(cherry picked from commit f72ee87d99ee38cb7b5badeb9a8af869bb92073a)

* fix(desktop): keep footer model state scoped to active session

(cherry picked from commit d91942ebd4671ff857b5c8526dbf133f04782ecb)

* fix(desktop): restore stored runtime when resuming sessions

(cherry picked from commit 32b3793418257617b8da57e26151f079c2620d00)

* fix(desktop): persist live runtime changes for resume

(cherry picked from commit c58467779436dcef44a80ad55b52664752dc0837)

* fix(desktop): persist resumed endpoint runtime

* chore(attribution): map pinguarmy's commit email in AUTHOR_MAP

The salvaged commits on this branch preserve @pinguarmy's authorship
(郝鹏宇 / peterhao@Peters-MacBook-Air.local). Add the mapping so the
check-attribution CI gate resolves the email to the GitHub username.

---------

Co-authored-by: 郝鹏宇 <peterhao@Peters-MacBook-Air.local>
2026-06-10 18:16:50 +00:00
Teknium
07ac185904 fix(ci): exit-4 forensics for vanishing test files in run_tests_parallel.py (#43646)
* fix(ci): append filesystem forensics when a per-file pytest run exhausts exit-4 retries

A PR-added test file (tests/test_iron_proxy.py, PR #30179) repeatedly
failed exactly one CI shard with 'ERROR: file or directory not found'
across 4 runs (including a fresh merge SHA on fresh runners), while the
identical slice passes locally against the same merge commit and a
tree-integrity watcher confirms no sibling test mutates the repo. Three
unrelated branches showed the same one-shard signature the same day.

We currently cannot attribute these because the log only carries
pytest's exit-4 line. This adds a forensics block to the captured
output when exit-4 survives the retry loop:

- does the file exist NOW (post-retries)
- parent dir entry count + similarly-named entries
- git status --porcelain dirty-entry count + first 10 entries

Zero behavior change: rc stays 4, retries unchanged, forensics wrapped
in a broad try/except so they can never mask the failure.

Two new tests cover the exhausted-retries and genuinely-missing paths.

* chore: drop the two forensics tests — ship the runner change only
2026-06-10 10:04:17 -07:00
Shannon Sands
3acf73161f Move folder creation into dialog 2026-06-10 09:53:12 -07:00
Shannon Sands
dd60c49bb8 Add dashboard file drop upload panel 2026-06-10 09:53:12 -07:00
Shannon Sands
6fe4821926 Add dashboard file browser paths 2026-06-10 09:53:12 -07:00
Teknium
d986bb0c6d feat(dashboard): full-featured profile builder (model + skills + MCPs) (#39084)
* feat(profiles): extend create endpoint for full profile-builder (model + MCPs + skills)

Backend foundation for the dashboard profile builder. Extends POST /api/profiles
to accept, in one call, everything a profile needs beyond name/clone:

- mcp_servers[]  -> written into the new profile's config.yaml
- keep_skills[]  -> replace-semantics: disable every seeded skill not kept
- hub_skills[]   -> async install via 'hermes -p <name> skills install <id>'

All applied best-effort AFTER the profile dir exists, so a hiccup in any one
never 500s the create. Model/MCP/keep-skills writes are profile-scoped via the
HERMES_HOME context override (same mechanism as the existing _write_profile_model).
Hub installs go through a subprocess scoped with -p because skills_hub.SKILLS_DIR
is import-time-bound and the runtime override can't redirect it.

Adds two helpers (_write_profile_mcp_servers, _disable_unselected_skills) and a
TestClient test asserting all four paths land in the NEW profile's config and
the hub spawn is scoped to it. Design doc at docs/design/profile-builder.md.

* feat(dashboard): full-featured profile builder page

Adds a dedicated /profiles/new builder that composes everything a profile
needs into one stepped create flow, reusing the existing Models/Skills/MCP
data paths instead of duplicating them:

- Identity   name + description
- Model      provider+model picker (api.getModelOptions)
- Skills     keep-which-built-in/optional (replace semantics, default = full
             bundle) + skills-hub search/add (api.getSkills, searchSkillsHub)
- MCPs       add HTTP/stdio servers inline
- Review     blueprint -> single POST /api/profiles create

Nothing writes until Create; the one call commits model+MCPs+skill selection
and spawns hub-skill installs (reported in the success toast). ProfilesPage
header gets a 'Build' button (full builder) alongside 'Create' (quick modal).
Route is page-only (not in the sidebar nav). Verified with vite build (2258
modules, green).
2026-06-10 09:18:32 -07:00
ethernet
4cecb1a13a change(tooling): npm audit fix in website/ 2026-06-10 11:59:34 -04:00
ethernet
90f4b3040d change(tooling): remove react-compiler eslint, update concurrently
concurrently 9 had a critical vuln dependency,
react-compiler eslint plugin is built into react-hooks eslint plugin as
of https://react.dev/blog/2025/10/07/react-compiler-1
2026-06-10 11:59:34 -04:00
ethernet
3bfbb3f2a0 change(tooling): typecheck in CI, update ts to 6
fix(ui-tui): fix ts 6 real type errors

change(tooling): use new node everywhere
2026-06-10 11:59:34 -04:00
emozilla
984e6cb5b8 feat(whatsapp): add WhatsApp Business Cloud API adapter
Add an official, production-grade WhatsApp integration via Meta's
Business Cloud API as a complement to the existing Baileys bridge.
No bridge subprocess, no QR codes, no account-ban risk — at the cost
of a Meta Business account and a public HTTPS webhook URL.

Setup is fully wizard-driven: 'hermes whatsapp-cloud' walks through
every credential with paste-time validation (catches the #1 trap of
pasting a phone number into the Phone Number ID field), generates a
verify token, and ends with copy-paste instructions for the
cloudflared / Meta-dashboard / Business Manager pieces that can't be
automated. The wizard also points users at Meta's Business Manager
for setting the bot's display name and profile picture.

Feature set:

- Inbound: text, images (with native-vision routing), voice notes
  (STT), documents (small text inlined, larger cached), reply context.
- Outbound: text with WhatsApp-flavored markdown conversion, images,
  videos, documents, opus voice notes via ffmpeg with MP3 fallback.
- Native interactive buttons for clarify, dangerous-command approval,
  and slash-command confirmation flows — matches the Telegram /
  Discord UX, graceful degrades to plain text.
- Read receipts (blue double-checkmarks) and typing indicator,
  using Meta's combined endpoint so they fire in a single API call.
- Webhook security: X-Hub-Signature-256 HMAC verification (raw body,
  constant-time), wamid deduplication, group-shaped-message refusal
  (groups deferred to v2 — Baileys still covers them).
- Full integration with the gateway's session, cron, display-tier,
  prompt-hint, and auth-allowlist systems. Cloud and Baileys can run
  side-by-side against different phone numbers.

Also wires STT (speech-to-text) through Nous's managed audio gateway
for Nous subscribers — previously the default stt.provider=local
required a separate faster-whisper install. New subscribers now get
voice-note transcription out of the box.

Docs: 418-line user guide at website/docs/user-guide/messaging/
whatsapp-cloud.md, sidebar entry, environment-variables reference,
ADDING_A_PLATFORM.md updated with the optional interactive-UX
contract for future adapter authors.

Tests: 100 dedicated tests for the adapter, 32 for the setup wizard,
20 for the Nous subscription STT wiring, plus regression coverage
across display_config, prompt_builder, and the cron scheduler.

Known limitations (deferred until clear demand signal):
- Group chats — use the Baileys bridge if you need them.
- Message templates for 24-hour-window outside-conversation sends —
  reactive chat is unaffected; cron / delegate_task with gaps > 24h
  will fail with a clear error. The agent's system prompt warns the
  model about this so it knows to mention it when scheduling delayed
  messages.
2026-05-23 01:07:01 -04:00
546 changed files with 49900 additions and 5449 deletions

Binary file not shown.

After

Width:  |  Height:  |  Size: 428 KiB

View File

@@ -44,7 +44,7 @@ jobs:
- uses: actions/setup-node@49933ea5288caeca8642d1e84afbd3f7d6820020 # v4
with:
node-version: 20
node-version: 22
cache: npm
cache-dependency-path: website/package-lock.json

View File

@@ -90,7 +90,7 @@ jobs:
# (see `_SKIP_PARTS` in scripts/run_tests_parallel.py) because each
# shard would otherwise reach the session-scoped ``built_image``
# fixture in ``tests/docker/conftest.py`` and start a 3-7min
# ``docker build`` under a 180s pytest-timeout cap — guaranteed to
# ``docker build`` — guaranteed to
# die in fixture setup.
#
# Piggybacking here avoids a second image build: the smoke test
@@ -114,7 +114,7 @@ jobs:
run: |
uv venv .venv --python 3.11
source .venv/bin/activate
# ``dev`` extra pulls in pytest, pytest-asyncio, pytest-timeout
# ``dev`` extra pulls in pytest, pytest-asyncio —
# everything tests/docker/ needs. We deliberately avoid ``all``
# here because the docker tests only drive the container via
# subprocess and don't import hermes_agent's optional deps.

View File

@@ -18,7 +18,7 @@ jobs:
- uses: actions/setup-node@49933ea5288caeca8642d1e84afbd3f7d6820020 # v4
with:
node-version: 20
node-version: 22
cache: npm
cache-dependency-path: website/package-lock.json

View File

@@ -4,13 +4,13 @@ on:
push:
branches: [main]
paths-ignore:
- '**/*.md'
- 'docs/**'
- "**/*.md"
- "docs/**"
pull_request:
branches: [main]
paths-ignore:
- '**/*.md'
- 'docs/**'
- "**/*.md"
- "docs/**"
permissions:
contents: read
@@ -30,13 +30,17 @@ jobs:
slice: [1, 2, 3, 4, 5, 6]
steps:
- name: Checkout code
uses: actions/checkout@de0fac2e4500dabe0009e67214ff5f5447ce83dd # v6.0.2
uses: actions/checkout@de0fac2e4500dabe0009e67214ff5f5447ce83dd # v6.0.2
- name: Restore duration cache
uses: actions/cache/restore@27d5ce7f107fe9357f9df03efb73ab90386fccae # v5.0.5
uses: actions/cache/restore@27d5ce7f107fe9357f9df03efb73ab90386fccae # v5.0.5
with:
path: test_durations.json
# Single stable key. main always overwrites, PRs always find it.
# main always writes a new suffix, but jobs pick the latest one with the same prefix
# quote from https://docs.github.com/en/actions/reference/workflows-and-actions/dependency-caching#cache-hits-and-misses
# If you provide restore-keys, the cache action sequentially searches for any caches that match the list of restore-keys.
# If there are no exact matches, the action searches for partial matches of the restore keys.
# When the action finds a partial match, the most recent cache is restored to the path directory.
key: test-durations
- name: Install ripgrep (prebuilt binary)
@@ -54,7 +58,7 @@ jobs:
rg --version
- name: Install uv
uses: astral-sh/setup-uv@d4b2f3b6ecc6e67c4457f6d3e41ec42d3d0fcb86 # v5
uses: astral-sh/setup-uv@d4b2f3b6ecc6e67c4457f6d3e41ec42d3d0fcb86 # v5
with:
# Persist uv's download/wheel cache (~/.cache/uv) across runs.
# Keyed on the dependency manifests, so the cache is reused until
@@ -115,7 +119,7 @@ jobs:
NOUS_API_KEY: ""
- name: Upload per-slice durations
uses: actions/upload-artifact@043fb46d1a93c77aae656e7c1c64a875d1fc6a0a # v7.0.1
uses: actions/upload-artifact@043fb46d1a93c77aae656e7c1c64a875d1fc6a0a # v7.0.1
with:
name: test-durations-slice-${{ matrix.slice }}
path: test_durations.json
@@ -125,11 +129,11 @@ jobs:
# (including PRs) get balanced slicing.
save-durations:
needs: test
if: always() && github.ref == 'refs/heads/main'
if: needs.test.result == 'success' && github.ref == 'refs/heads/main'
runs-on: ubuntu-latest
steps:
- name: Download all slice durations
uses: actions/download-artifact@3e5f45b2cfb9172054b4087a40e8e0b5a5461e7c # v8.0.1
uses: actions/download-artifact@3e5f45b2cfb9172054b4087a40e8e0b5a5461e7c # v8.0.1
with:
pattern: test-durations-slice-*
path: durations
@@ -149,17 +153,17 @@ jobs:
"
- name: Save merged duration cache
uses: actions/cache/save@27d5ce7f107fe9357f9df03efb73ab90386fccae # v5.0.5
uses: actions/cache/save@27d5ce7f107fe9357f9df03efb73ab90386fccae # v5.0.5
with:
path: test_durations.json
key: test-durations
key: test-durations-${{ github.run_id }}
e2e:
runs-on: ubuntu-latest
timeout-minutes: 15
steps:
- name: Checkout code
uses: actions/checkout@de0fac2e4500dabe0009e67214ff5f5447ce83dd # v6.0.2
uses: actions/checkout@de0fac2e4500dabe0009e67214ff5f5447ce83dd # v6.0.2
- name: Install ripgrep (prebuilt binary)
run: |
@@ -176,7 +180,7 @@ jobs:
rg --version
- name: Install uv
uses: astral-sh/setup-uv@d4b2f3b6ecc6e67c4457f6d3e41ec42d3d0fcb86 # v5
uses: astral-sh/setup-uv@d4b2f3b6ecc6e67c4457f6d3e41ec42d3d0fcb86 # v5
with:
# Persist uv's download/wheel cache (~/.cache/uv) across runs.
# Keyed on the dependency manifests, so the cache is reused until

25
.github/workflows/typecheck.yml vendored Normal file
View File

@@ -0,0 +1,25 @@
# .github/workflows/typecheck.yml
name: Typecheck
on:
push:
branches: [main]
pull_request:
branches: [main]
jobs:
typecheck:
runs-on: ubuntu-latest
strategy:
matrix:
package:
[ui-tui, web, apps/bootstrap-installer, apps/desktop, apps/shared]
fail-fast: false # report all failures, not just the first one
steps:
- uses: actions/checkout@de0fac2e4500dabe0009e67214ff5f5447ce83dd # v6.0.2
- uses: actions/setup-node@49933ea5288caeca8642d1e84afbd3f7d6820020 # v4
with:
node-version: 22
cache: npm
- run: npm ci
- run: npm run --prefix ${{ matrix.package }} typecheck

7
.gitignore vendored
View File

@@ -89,6 +89,9 @@ website/static/api/skills-index.json
# every build).
website/static/api/skills.json
website/static/api/skills-meta.json
# automation-blueprints-index.json is a build artifact emitted by
# website/scripts/extract-automation-blueprints.py during prebuild.
website/static/api/automation-blueprints-index.json
models-dev-upstream/
# Local editor / agent tooling (machine-specific; keep in global config, not the repo)
@@ -129,3 +132,7 @@ scripts/out/
# stores the published notes. They are not a build artifact and must never be
# committed to the repo root. See the hermes-release skill.
RELEASE_v*.md
# Desktop demo-run scratch output (hermes writes demo/*.txt during recorded
# walkthroughs). Throwaway artifacts, never part of the app.
apps/desktop/demo/

View File

@@ -459,7 +459,7 @@ npm install # first time
npm run dev # watch mode (rebuilds hermes-ink + tsx --watch)
npm start # production
npm run build # full build (hermes-ink + tsc)
npm run type-check # typecheck only (tsc --noEmit)
npm run typecheck # typecheck only (tsc --noEmit)
npm run lint # eslint
npm run fmt # prettier
npm test # vitest

View File

@@ -145,7 +145,7 @@ def build_nous_credits_snapshot(account_info) -> Optional[AccountUsageSnapshot]:
account info to show (fail-open: caller just shows nothing).
"""
try:
from hermes_cli.nous_account import nous_portal_billing_url
from hermes_cli.nous_account import nous_portal_topup_url
if account_info is None or not getattr(account_info, "logged_in", False):
return None
@@ -213,7 +213,8 @@ def build_nous_credits_snapshot(account_info) -> Optional[AccountUsageSnapshot]:
if not windows and not details:
return None
details.append(f"Manage / top up: {nous_portal_billing_url(account_info)}")
details.append(f"Top up: {nous_portal_topup_url(account_info)}")
details.append("(or run /credits)")
plan = getattr(sub, "plan", None) if sub is not None else None
return AccountUsageSnapshot(
@@ -337,6 +338,93 @@ def _snapshot_from_credits_state(state) -> Optional[AccountUsageSnapshot]:
return None
@dataclass(frozen=True)
class CreditsView:
"""Surface-agnostic data for the ``/credits`` command.
One portal fetch, one parse — consumed identically by the CLI panel, the
gateway button, and any other money surface. Fail-open: when not logged in
or the portal is unreachable, ``logged_in`` is False / ``topup_url`` is None
and callers degrade gracefully.
"""
logged_in: bool
balance_lines: tuple[str, ...] = ()
identity_line: Optional[str] = None
topup_url: Optional[str] = None
depleted: bool = False
def build_credits_view(*, markdown: bool = False, timeout: float = 10.0) -> CreditsView:
"""Build the /credits view: balance block + identity line + top-up URL.
Reuses the same account fetch + snapshot + URL builder as the /usage credits
block, so the numbers always match. The balance block is the rendered
snapshot MINUS its trailing top-up/command-hint lines (the /credits surface
supplies its own affordance). Fail-open → ``CreditsView(logged_in=False)``.
"""
not_logged_in = CreditsView(logged_in=False)
try:
from hermes_cli.auth import get_provider_auth_state
tok = (get_provider_auth_state("nous") or {}).get("access_token")
if not (isinstance(tok, str) and tok.strip()):
return not_logged_in
except Exception:
return not_logged_in
try:
import concurrent.futures
from hermes_cli.nous_account import (
get_nous_portal_account_info,
nous_portal_topup_url,
)
with concurrent.futures.ThreadPoolExecutor(max_workers=1) as pool:
account = pool.submit(get_nous_portal_account_info, force_fresh=True).result(
timeout=timeout
)
except Exception:
logger.debug("credits ▸ /credits portal fetch failed (fail-open)", exc_info=True)
return not_logged_in
if account is None or not getattr(account, "logged_in", False):
return not_logged_in
snapshot = build_nous_credits_snapshot(account)
# Balance lines = the snapshot block minus the two trailing affordance lines
# ("Top up: <url>" + "(or run /credits)") that build_nous_credits_snapshot
# appends for the /usage surface. /credits renders its own button/panel.
balance_lines: list[str] = []
if snapshot is not None:
rendered = render_account_usage_lines(snapshot, markdown=markdown)
balance_lines = [
line
for line in rendered
if not line.lstrip().startswith("Top up:")
and not line.lstrip().startswith("(or run")
]
# Identity line — shown before any open (roadmap §4.4).
email = getattr(account, "email", None)
org_name = getattr(account, "org_name", None)
who: list[str] = []
if email:
who.append(str(email))
if org_name:
who.append(f"org {org_name}")
identity_line = ("Topping up as " + " / ".join(who)) if who else None
return CreditsView(
logged_in=True,
balance_lines=tuple(balance_lines),
identity_line=identity_line,
topup_url=nous_portal_topup_url(account),
depleted=getattr(account, "paid_service_access", None) is False,
)
def _resolve_codex_usage_url(base_url: str) -> str:
normalized = (base_url or "").strip().rstrip("/")
if not normalized:

View File

@@ -445,6 +445,45 @@ def repair_message_sequence(agent, messages: List[Dict]) -> int:
return repairs
def repair_message_sequence_with_cursor(agent, messages: List[Dict]) -> int:
"""Run :func:`repair_message_sequence` and keep the SessionDB flush
cursor consistent with the compacted list (#44837).
``repair_message_sequence`` merges/drops messages in place, shrinking
the list. ``_last_flushed_db_idx`` (the DB-write cursor) indexes into
that list, so after compaction it can point past the new end — the
turn-end flush would then skip the assistant/tool chain entirely — or
past unflushed messages shifted to lower indexes.
Repair preserves object identity for surviving messages, so counting
the survivors from the previously-flushed prefix gives the exact new
cursor even when messages are dropped/merged at indexes *before* the
cursor — a plain ``min()`` clamp would silently skip that many
unflushed rows. Falls back to the clamp when no prefix snapshot is
available.
Returns the number of repairs made (same as ``repair_message_sequence``).
"""
pre_repair_flushed_ids = None
flush_cursor = getattr(agent, "_last_flushed_db_idx", None)
if isinstance(flush_cursor, int) and flush_cursor > 0:
pre_repair_flushed_ids = {id(m) for m in messages[:flush_cursor]}
repairs = repair_message_sequence(agent, messages)
if repairs > 0 and hasattr(agent, "_last_flushed_db_idx"):
if pre_repair_flushed_ids is not None:
agent._last_flushed_db_idx = sum(
1 for m in messages if id(m) in pre_repair_flushed_ids
)
else:
agent._last_flushed_db_idx = min(
agent._last_flushed_db_idx, len(messages)
)
return repairs
def strip_think_blocks(agent, content: str) -> str:
"""Remove reasoning/thinking blocks from content, returning only visible text.
@@ -679,15 +718,28 @@ def recover_with_credential_pool(
# long-running TUI sessions stuck on stale tokens until the user
# exited and reopened.
is_entitlement = agent._is_entitlement_failure(error_context, status_code)
_auth_haystack = " ".join(
str(error_context.get(k) or "").lower()
for k in ("message", "reason", "code", "error")
if isinstance(error_context, dict)
)
if (
not is_entitlement
and status_code == 403
and "oauth authentication is currently not allowed for this organization" in _auth_haystack
):
is_entitlement = True
if (
not is_entitlement
and status_code == 403
and (agent.provider or "") == "anthropic"
and getattr(agent, "api_mode", "") == "anthropic_messages"
):
is_entitlement = True
if not is_entitlement and status_code == 403 and (agent.provider or "") == "xai-oauth":
_disambiguator_haystack = " ".join(
str(error_context.get(k) or "").lower()
for k in ("message", "reason", "code", "error")
if isinstance(error_context, dict)
)
_is_xai_auth_failure = (
"[wke=unauthenticated:" in _disambiguator_haystack
or "oauth2 access token could not be validated" in _disambiguator_haystack
"[wke=unauthenticated:" in _auth_haystack
or "oauth2 access token could not be validated" in _auth_haystack
)
if not _is_xai_auth_failure:
is_entitlement = True

View File

@@ -1571,6 +1571,15 @@ def _convert_content_part_to_anthropic(part: Any) -> Optional[Dict[str, Any]]:
if ptype == "input_text":
block: Dict[str, Any] = {"type": "text", "text": part.get("text", "")}
elif ptype == "text":
# A stored Anthropic text block. Rebuild from whitelisted fields only —
# SDK response text blocks carry output-only siblings (parsed_output,
# citations=None) that the Messages INPUT schema rejects with HTTP 400
# "Extra inputs are not permitted". Do NOT dict(part) it verbatim.
block = {"type": "text", "text": part.get("text", "")}
cits = part.get("citations")
if isinstance(cits, list) and cits:
block["citations"] = cits
elif ptype in {"image_url", "input_image"}:
image_value = part.get("image_url", {})
url = image_value.get("url", "") if isinstance(image_value, dict) else str(image_value or "")
@@ -1685,6 +1694,58 @@ def _content_parts_to_anthropic_blocks(parts: Any) -> List[Dict[str, Any]]:
return out
def _sanitize_replay_block(b: Dict[str, Any]) -> Optional[Dict[str, Any]]:
"""Strip output-only fields from a stored Anthropic content block so it is
valid as REQUEST input on replay.
The SDK response objects carry output-only attributes that the Messages
*input* schema forbids ("Extra inputs are not permitted"): text blocks get
``parsed_output``/``citations`` (when null), tool_use blocks get ``caller``,
etc. ``normalize_response`` captured blocks verbatim via ``_to_plain_data``,
so these leak back as input on the next turn → HTTP 400.
Whitelist per type (NOT a blacklist) so future SDK output-only fields can't
reintroduce the bug. Returns a clean block, or None to drop it.
"""
if not isinstance(b, dict):
return None
btype = b.get("type")
if btype == "text":
out: Dict[str, Any] = {"type": "text", "text": b.get("text", "")}
# citations is input-valid ONLY when it's a non-empty list; the SDK
# emits citations=None on responses, which the input schema rejects.
cits = b.get("citations")
if isinstance(cits, list) and cits:
out["citations"] = cits
if isinstance(b.get("cache_control"), dict):
out["cache_control"] = b["cache_control"]
return out
if btype == "thinking":
out = {"type": "thinking", "thinking": b.get("thinking", "")}
if b.get("signature"):
out["signature"] = b["signature"]
return out
if btype == "redacted_thinking":
# Only valid with its data payload; drop if missing.
return {"type": "redacted_thinking", "data": b["data"]} if b.get("data") else None
if btype == "tool_use":
out = {
"type": "tool_use",
"id": _sanitize_tool_id(b.get("id", "")),
"name": b.get("name", ""),
"input": b.get("input", {}),
}
if isinstance(b.get("cache_control"), dict):
out["cache_control"] = b["cache_control"]
return out
if btype == "image":
src = b.get("source")
return {"type": "image", "source": src} if isinstance(src, dict) else None
# Unknown/unsupported block type on the input path — drop rather than risk
# another "Extra inputs are not permitted".
return None
def _convert_assistant_message(m: Dict[str, Any]) -> Dict[str, Any]:
"""Convert an assistant message to Anthropic content blocks.
@@ -1692,6 +1753,55 @@ def _convert_assistant_message(m: Dict[str, Any]) -> Dict[str, Any]:
reasoning_content injection for Kimi/DeepSeek endpoints.
"""
content = m.get("content", "")
# Anthropic interleaved-thinking fast path: when this turn carries a
# verbatim, order-preserving block list (set by normalize_response only
# for turns that interleave SIGNED thinking with tool_use), replay it.
# Each block is run through _sanitize_replay_block to strip output-only
# SDK fields (parsed_output, caller, citations=None, …) that the Messages
# INPUT schema forbids — replaying them verbatim caused HTTP 400 "Extra
# inputs are not permitted" (text.parsed_output). Block ORDER is preserved
# (the reason this channel exists); only forbidden sibling fields are
# dropped, leaving thinking signatures and tool_use id/name/input intact.
ordered_blocks = m.get("anthropic_content_blocks")
if isinstance(ordered_blocks, list) and ordered_blocks:
# Re-source each tool_use input from the stored tool_calls map rather
# than the captured block. The ordered-blocks list captures tool_use
# input from the RAW API response (normalize_response), which is NOT
# credential-redacted; tool_calls[].function.arguments IS redacted at
# storage time (build_assistant_message, #19798). Replaying the raw
# block input would resurrect a secret the model inlined into a tool
# call (e.g. terminal(command="curl -H 'Authorization: Bearer sk-...'")
# onto the wire, even though the same value is redacted everywhere else
# in history. Keying by sanitized tool id preserves interleave order
# (the reason this channel exists) while swapping in the redacted
# input. Adapted from #36071 (replay-time tool-input re-sourcing).
redacted_input_by_id: Dict[str, Any] = {}
for tc in m.get("tool_calls", []) or []:
if not isinstance(tc, dict):
continue
fn = tc.get("function", {}) or {}
raw_args = fn.get("arguments", "{}")
try:
parsed_args = json.loads(raw_args) if isinstance(raw_args, str) else raw_args
except (json.JSONDecodeError, ValueError):
parsed_args = {}
redacted_input_by_id[_sanitize_tool_id(tc.get("id", ""))] = parsed_args
replayed: List[Dict[str, Any]] = []
for b in ordered_blocks:
clean = _sanitize_replay_block(b)
if clean is None:
continue
if clean.get("type") == "tool_use":
# Override raw (un-redacted) input with the redacted copy when
# we have one for this id; fall back to the sanitized block
# input only if the tool_call is missing (shape mismatch).
redacted = redacted_input_by_id.get(clean.get("id", ""))
if redacted is not None:
clean["input"] = redacted
replayed.append(clean)
if replayed:
return {"role": "assistant", "content": replayed}
blocks = _extract_preserved_thinking_blocks(m)
if content:
if isinstance(content, list):

View File

@@ -208,6 +208,41 @@ def is_stale_connection_error(exc: BaseException) -> bool:
return False
def is_streaming_access_denied_error(exc: BaseException) -> bool:
"""Return True when AWS denied the ``bedrock:InvokeModelWithResponseStream`` action.
IAM policies scoped to ``bedrock:InvokeModel`` only (a common least-privilege
setup) reject ``converse_stream()`` with an ``AccessDeniedException`` whose
message names the streaming action, e.g.::
User: arn:aws:iam::123456789012:user/x is not authorized to perform:
bedrock:InvokeModelWithResponseStream on resource: ...
This is permanent for the session — retrying the stream can never succeed —
so callers should flip to the non-streaming ``converse()`` path (which maps
to ``bedrock:InvokeModel``) instead of burning retries.
Detection is deliberately message-based: boto3 surfaces this as a
``ClientError`` with ``Error.Code == "AccessDeniedException"``, and the
AnthropicBedrock SDK wraps the same AWS response in its own exception
types, but both preserve the action name in the message.
"""
msg = str(exc).lower()
if "invokemodelwithresponsestream" not in msg:
return False
# ClientError with an explicit access-denied code is the canonical form.
try:
from botocore.exceptions import ClientError
except ImportError: # pragma: no cover — botocore always present with boto3
ClientError = None # type: ignore[assignment]
if ClientError is not None and isinstance(exc, ClientError):
code = (getattr(exc, "response", None) or {}).get("Error", {}).get("Code", "")
return code in ("AccessDeniedException", "UnauthorizedException")
# Wrapped forms (e.g. AnthropicBedrock SDK PermissionDeniedError) — match
# on the authorization-failure phrasing AWS uses.
return "not authorized" in msg or "accessdenied" in msg
# ---------------------------------------------------------------------------
# AWS credential detection
# ---------------------------------------------------------------------------
@@ -1003,6 +1038,16 @@ def call_converse_stream(
try:
response = client.converse_stream(**kwargs)
except Exception as exc:
if is_streaming_access_denied_error(exc):
# IAM allows bedrock:InvokeModel but not
# InvokeModelWithResponseStream — permanent for this session.
# Fall back to the non-streaming converse() path.
logger.info(
"bedrock: converse_stream denied by IAM on (region=%s, model=%s) — "
"falling back to non-streaming converse().",
region, model,
)
return normalize_converse_response(client.converse(**kwargs))
if is_stale_connection_error(exc):
logger.warning(
"bedrock: stale-connection error on converse_stream(region=%s, "

View File

@@ -952,6 +952,18 @@ def build_assistant_message(agent, assistant_message, finish_reason: str) -> dic
if preserved:
msg["reasoning_details"] = preserved
# Anthropic interleaved-thinking replay: when a turn interleaves signed
# thinking blocks with tool_use, the parallel reasoning_details +
# tool_calls fields lose the cross-type ordering, and reconstruction
# front-loads thinking — reordering signed blocks and triggering HTTP 400
# ("thinking ... blocks in the latest assistant message cannot be
# modified"). Carry the verbatim ordered block list so the adapter can
# replay the latest assistant message unchanged. See
# agent/transports/anthropic.py and agent/anthropic_adapter.py.
ordered_blocks = getattr(assistant_message, "anthropic_content_blocks", None)
if ordered_blocks:
msg["anthropic_content_blocks"] = ordered_blocks
# Codex Responses API: preserve encrypted reasoning items for
# multi-turn continuity. These get replayed as input on the next turn.
codex_items = getattr(assistant_message, "codex_reasoning_items", None)
@@ -1603,6 +1615,8 @@ def interruptible_streaming_api_call(agent, api_kwargs: dict, *, on_first_delta=
_get_bedrock_runtime_client,
invalidate_runtime_client,
is_stale_connection_error,
is_streaming_access_denied_error,
normalize_converse_response,
stream_converse_with_callbacks,
)
region = api_kwargs.pop("__bedrock_region__", "us-east-1")
@@ -1611,6 +1625,29 @@ def interruptible_streaming_api_call(agent, api_kwargs: dict, *, on_first_delta=
try:
raw_response = client.converse_stream(**api_kwargs)
except Exception as _bedrock_exc:
# IAM policies scoped to bedrock:InvokeModel only (no
# InvokeModelWithResponseStream) reject converse_stream()
# with AccessDeniedException. That denial is permanent for
# the session — fall back to the non-streaming converse()
# inline (it maps to bedrock:InvokeModel) and disable
# streaming for subsequent calls so we don't re-fail every
# turn.
if is_streaming_access_denied_error(_bedrock_exc):
agent._disable_streaming = True
agent._safe_print(
"\n⚠ AWS IAM denied bedrock:InvokeModelWithResponseStream — "
"falling back to non-streaming InvokeModel.\n"
" Grant that action to restore streaming output.\n"
)
logger.info(
"bedrock: converse_stream denied by IAM (%s) — "
"using non-streaming converse() for this session.",
type(_bedrock_exc).__name__,
)
result["response"] = normalize_converse_response(
client.converse(**api_kwargs)
)
return
# Evict the cached client on stale-connection failures
# so the outer retry loop builds a fresh client/pool.
if is_stale_connection_error(_bedrock_exc):
@@ -1698,6 +1735,14 @@ def interruptible_streaming_api_call(agent, api_kwargs: dict, *, on_first_delta=
# poll loop uses this to detect stale connections that keep receiving
# SSE keep-alive pings but no actual data.
last_chunk_time = {"t": time.time()}
# Stale-stream patience, shared between the httpx socket read timeout
# (built in ``_call_chat_completions`` below) and the stale-stream detector
# (computed further down, before the worker thread starts). Initialized
# here so the read-timeout builder can floor itself at the stale value and
# never fire before the detector. ``None`` until the detector value is
# resolved, so the builder degrades to its plain default if it ever runs
# first.
_stream_stale_timeout = None
def _fire_first_delta():
if not first_delta_fired["done"] and on_first_delta:
@@ -1734,6 +1779,26 @@ def interruptible_streaming_api_call(agent, api_kwargs: dict, *, on_first_delta=
"Local provider detected (%s) — stream read timeout raised to %.0fs",
agent.base_url, _stream_read_timeout,
)
elif (
_stream_read_timeout == 120.0
and _stream_stale_timeout is not None
and _stream_stale_timeout != float("inf")
and _stream_stale_timeout > _stream_read_timeout
):
# Cloud reasoning models (e.g. Opus) routinely pause mid-stream
# for minutes during extended thinking. The stale-stream
# detector is deliberately scaled up to tolerate this (180300s,
# see the stale-timeout block below), but the raw httpx socket
# read timeout defaulted to a flat 120s and fired *first* —
# tearing down a healthy reasoning stream before the stale
# detector (which owns retry + diagnostics) could act. Keep the
# socket read timeout in step with the detector so it no longer
# preempts it.
_stream_read_timeout = _stream_stale_timeout
logger.debug(
"Cloud reasoning stream — read timeout raised to %.0fs to "
"match stale-stream detector", _stream_read_timeout,
)
# Cap connect/pool at 60s even when provider timeout is higher.
# connect/pool cover TCP handshake, not model inference.
_conn_cap = min(_base_timeout, 60.0) if _provider_timeout_cfg is not None else 30.0
@@ -2384,9 +2449,34 @@ def interruptible_streaming_api_call(agent, api_kwargs: dict, *, on_first_delta=
"stream" in _err_lower
and "not supported" in _err_lower
)
if _is_stream_unsupported:
# AWS Bedrock (AnthropicBedrock SDK path): IAM policies
# with bedrock:InvokeModel but not
# InvokeModelWithResponseStream reject messages.stream()
# with a permission error naming the streaming action.
# Permanent for the session — flip to non-streaming
# (messages.create() maps to bedrock:InvokeModel).
_is_bedrock_stream_denied = False
if (
not _is_stream_unsupported
and "invokemodelwithresponsestream" in _err_lower
):
# Cheap message pre-check before importing the
# adapter — bedrock_adapter triggers a lazy boto3
# install at import time, which must not run for
# unrelated providers' stream errors.
from agent.bedrock_adapter import (
is_streaming_access_denied_error,
)
_is_bedrock_stream_denied = (
is_streaming_access_denied_error(e)
)
if _is_stream_unsupported or _is_bedrock_stream_denied:
agent._disable_streaming = True
agent._safe_print(
"\n⚠ AWS IAM denied bedrock:InvokeModelWithResponseStream. "
"Switching to non-streaming.\n"
" Grant that action to restore streaming output.\n"
if _is_bedrock_stream_denied else
"\n⚠ Streaming is not supported for this "
"model/provider. Switching to non-streaming.\n"
" To avoid this delay, set display.streaming: false "

View File

@@ -127,14 +127,21 @@ def _chat_content_to_responses_parts(content: Any, *, role: str = "user") -> Lis
return converted
def _summarize_user_message_for_log(content: Any) -> str:
"""Return a short text summary of a user message for logging/trajectory.
def _summarize_user_message_for_log(content: Any, *, sep: str = " ") -> str:
"""Flatten message content to a plain-text summary.
Multimodal messages arrive as a list of ``{type:"text"|"image_url", ...}``
parts from the API server. Logging, spinner previews, and trajectory
files all want a plain string — this helper extracts the first chunk of
text and notes any attached images. Returns an empty string for empty
lists and ``str(content)`` for unexpected scalar types.
parts from the API server. Several consumers want a plain string:
- Logging, spinner previews, and trajectory files (the default ``sep=" "``).
- External memory providers, which feed the text to regexes
(``sanitize_context``) and text APIs — a raw list crashes the sync with
``expected string or bytes-like object, got 'list'`` (use ``sep="\\n"``).
Text parts are joined with ``sep``; images become a ``[N image(s)]`` marker
so the turn isn't recorded as if the attachment never existed. Returns an
empty string for empty lists and ``str(content)`` for unexpected scalar
types.
"""
if content is None:
return ""
@@ -157,7 +164,7 @@ def _summarize_user_message_for_log(content: Any) -> str:
text_bits.append(text)
elif ptype in {"image_url", "input_image"}:
image_count += 1
summary = " ".join(text_bits).strip()
summary = sep.join(text_bits).strip()
if image_count:
note = f"[{image_count} image{'s' if image_count != 1 else ''}]"
summary = f"{note} {summary}" if summary else note

738
agent/coding_context.py Normal file
View File

@@ -0,0 +1,738 @@
"""Coding-context awareness — base Hermes, every interactive surface.
When the user runs Hermes inside a code workspace (CLI, TUI, desktop app, or an
editor over ACP), Hermes shifts into a **coding posture**. This module is the
single place that decides whether we're in that posture and what it implies,
so the rest of the codebase never re-derives "are we coding?" on its own.
Architecture — one seam, many consumers
----------------------------------------
The posture is modelled as a frozen :class:`RuntimeMode` selected from a small
:class:`ContextProfile` registry (today: ``coding`` and ``general``). A profile
is *data* — it declares the toolset to collapse to, the operating brief to
inject, and hints for other domains (model routing, memory, subagents). Every
domain reads the same resolved object instead of probing git/config itself:
* **System prompt** — ``RuntimeMode.system_blocks()`` → the operating brief +
a live git/workspace snapshot (``agent/system_prompt.py``).
* **Toolset** — ``RuntimeMode.toolset_selection()`` → the ``coding`` toolset
plus the user's enabled MCP servers (``cli.py`` / ``tui_gateway``). Only
under the opt-in ``focus`` mode: the default posture is prompt-only and
never touches the user's configured toolsets (toolsets like messaging /
smart-home / music are off-by-default anyway, and someone who explicitly
enabled image-gen or Spotify shouldn't lose it for being in a git repo).
* **Delegation** — subagents inherit the parent's toolset and run through the
same prompt builder, so the coding posture propagates to children for free.
* **Model / memory / compression** — declared on the profile
(``model_hint``, ``memory_policy``) as the extension seam; consumers read
``mode.profile`` rather than re-deciding.
Cache safety
------------
The mode is resolved **once** and is immutable. The workspace snapshot is built
once at prompt-build time and baked into the *stable* system-prompt tier — never
re-probed per turn (that would shatter the prompt cache). Branch and dirty state
drift mid-session, so the brief tells the model to re-check with ``git`` before
acting on the snapshot. A ``/coding`` flip therefore only takes effect next
session (deferred), the same contract as ``/skills install`` vs ``--now``.
Activation (config ``agent.coding_context``):
* ``auto`` (default) — posture (brief + snapshot) on an interactive coding
surface sitting in a code workspace (git repo or recognised project root).
Prompt-only; toolsets and the skill index untouched.
* ``focus`` — like ``auto``, but additionally collapses the toolset to the
``coding`` set + enabled MCP servers and demotes non-coding skill
categories to names-only in the prompt's skill index (no skill is ever
hidden). Explicit opt-in for a lean schema.
* ``on`` — force the posture anywhere (incl. non-workspaces). Prompt-only.
* ``off`` — disable entirely.
"""
from __future__ import annotations
import json
import logging
import os
import re
import subprocess
from dataclasses import dataclass
from pathlib import Path
from typing import Any, Optional
logger = logging.getLogger("hermes.coding_context")
CODING_TOOLSET = "coding"
# Surfaces where a coding posture makes sense under ``auto``. Messaging
# platforms (telegram, discord, slack, …) are intentionally absent — a chat bot
# in a group is not pair-programming.
INTERACTIVE_CODING_PLATFORMS = {"cli", "tui", "acp", "desktop", ""}
# Project-root signals that mark a directory as a code workspace even when it
# isn't (yet) a git repo. Cheap filename checks — no parsing.
_PROJECT_MARKERS = (
"pyproject.toml", "setup.py", "setup.cfg", "requirements.txt",
"package.json", "tsconfig.json", "deno.json",
"Cargo.toml", "go.mod", "pom.xml", "build.gradle", "build.gradle.kts",
"Gemfile", "composer.json", "mix.exs", "pubspec.yaml",
"CMakeLists.txt", "Makefile", "Dockerfile",
"AGENTS.md", "CLAUDE.md", ".cursorrules",
)
# Agent-instruction files surfaced separately from manifests in the snapshot.
_CONTEXT_FILES = ("AGENTS.md", "CLAUDE.md", ".cursorrules")
# Lockfile → package manager, checked in priority order.
_PY_LOCKFILES = (("uv.lock", "uv"), ("poetry.lock", "poetry"), ("Pipfile.lock", "pipenv"))
_JS_LOCKFILES = (
("pnpm-lock.yaml", "pnpm"), ("bun.lockb", "bun"), ("bun.lock", "bun"),
("yarn.lock", "yarn"), ("package-lock.json", "npm"),
)
# package.json scripts / Makefile targets worth surfacing as verify commands.
_VERIFY_TARGETS = ("test", "tests", "lint", "typecheck", "check", "build", "fmt", "format")
_MAX_VERIFY_COMMANDS = 8
_MAX_FACT_FILE_BYTES = 256 * 1024
_GIT_TIMEOUT = 2.5
# Per-model edit-format steering. Matching the edit tool format to how a model
# was trained reduces mistakes and wasted reasoning (OpenAI/Codex handle
# patch-style diffs best; Anthropic models — and most open-weight coding
# models, whose RL scaffolds use str_replace-style editors — do best with
# string-replacement). Our `patch` tool exposes both: mode="patch" (V4A
# multi-file) and mode="replace" (find-and-swap). We nudge each family toward
# its native format. Unknown families get nothing (the brief's neutral wording
# stands). Substrings match the model id; aligned with TOOL_USE_ENFORCEMENT_MODELS.
#
# GPT/Codex get V4A for ALL edits, single-file included: in codex-rs,
# apply_patch (V4A — apply_patch.lark) is the ONLY file editor, no
# str_replace-style tool exists, and the shipped model prompts say to use
# apply_patch even "for single file edits" — so a replace-mode nudge would
# steer those models toward a format their first-party harness never taught
# them.
_EDIT_FORMAT_GUIDANCE: dict[str, tuple[tuple[str, ...], str]] = {
"patch": (
("gpt", "codex"),
"- Edit format: author new files with `write_file`; for edits to "
"existing code use `patch` with `mode='patch'` (V4A diff) — including "
"single-file edits. It's the edit format you handle most reliably.",
),
"replace": (
("claude", "sonnet", "opus", "haiku",
"gemini", "gemma", "deepseek", "qwen", "kimi", "glm", "grok",
"hermes", "llama", "mistral", "devstral", "minimax"),
"- Edit format: author new files with `write_file`; for edits to "
"existing code prefer `patch` in `mode='replace'` — match a unique "
"snippet and swap it. Reach for `mode='patch'` (V4A) only when an edit "
"genuinely spans several files at once.",
),
}
def _model_family(model: Optional[str]) -> Optional[str]:
"""Classify a model id into an edit-format family key, or ``None``.
Used to steer the coding posture toward the edit tool format a model was
trained on. Family-agnostic by design: an unrecognised model gets ``None``
and the operating brief's neutral edit wording applies.
"""
if not model:
return None
lowered = model.lower()
for family, (needles, _line) in _EDIT_FORMAT_GUIDANCE.items():
if any(n in lowered for n in needles):
return family
return None
def _edit_format_line(model: Optional[str]) -> str:
"""The edit-format guidance line for this model's family (``""`` if none)."""
family = _model_family(model)
if family is None:
return ""
return _EDIT_FORMAT_GUIDANCE[family][1]
# Operating brief for the coding posture. Tool names referenced here (read_file,
# search_files, patch, write_file, terminal, todo) are in the coding toolset and
# in _HERMES_CORE_TOOLS, so they're present on every surface this fires on.
CODING_AGENT_GUIDANCE = (
"You are a coding agent pairing with the user inside their codebase. "
"Operate like a careful senior engineer.\n"
"\n"
"Gather context first:\n"
"- Read the relevant files with `read_file` and locate code with "
"`search_files` before changing anything. Trace a symbol to its definition "
"and usages rather than guessing its shape.\n"
"- Batch independent lookups: when several reads/searches don't depend on "
"each other, issue them together in one turn instead of one at a time.\n"
"- Never invent files, symbols, APIs, or imports. If you haven't seen it in "
"the repo, go look. Don't assume a library is available — check the project "
"manifest (pyproject.toml / package.json / Cargo.toml / go.mod) and how "
"neighbouring files import it.\n"
"\n"
"Make changes through the tools, not the chat:\n"
"- Edit with `patch`/`write_file`. Do NOT print code blocks to the user as "
"a substitute for editing — apply the change, then summarise it. Only show "
"code when the user explicitly asks to see it.\n"
"- Match the project's existing style and conventions; AGENTS.md / "
"CLAUDE.md / .cursorrules already in context win over your defaults. Touch "
"only what the task needs — no drive-by refactors, renames, or reformatting "
"— and add any imports/dependencies your code requires.\n"
"- If an edit fails to apply, re-read the file to get the current exact "
"contents before retrying — don't repeat a stale patch. If the same region "
"fails twice, rewrite the enclosing function or file with `write_file` "
"instead of attempting a third patch.\n"
"\n"
"Verify, and know when to stop:\n"
"- Use `terminal` for git, builds, tests, and inspection. Run the relevant "
"tests/linter/build and confirm they pass before claiming the work is done.\n"
"- Terminal state persists across calls: current directory and exported "
"environment variables carry forward. Activate a virtualenv or export setup "
"vars once, then reuse that state instead of re-sourcing it before every "
"test command.\n"
"- Fix root causes, not symptoms: when you find a bug, check sibling call "
"paths for the same flaw and fix the class, not just the reported site.\n"
"- When fixing linter/type errors on a file, stop after about three "
"attempts on the same file and ask the user rather than looping.\n"
"- Track multi-step work with `todo`. Reference code as `path:line` instead "
"of pasting whole files.\n"
"\n"
"Respect the user's repo: don't commit, push, or rewrite history unless "
"asked, and never read, print, or commit secrets — leave `.env` and "
"credential files alone unless the user explicitly asks. The Workspace "
"block below is a snapshot from session start — re-run `git status`/"
"`git branch` before relying on it. Be concise: lead with the change or "
"answer, not a preamble."
)
# ── Context profiles (declarative posture definitions) ──────────────────────
@dataclass(frozen=True)
class ContextProfile:
"""A named operating posture. Pure data — consumers read these fields.
``toolset`` — collapse to this toolset (+ enabled MCP) when no explicit
selection is pinned; ``None`` keeps the platform default.
``guidance`` — operating brief injected into the stable system prompt;
``""`` injects nothing.
``model_hint`` — routing preference key for smart model routing
(extension seam; not yet consumed by the router).
``memory_policy``— memory namespace/weighting hint (extension seam).
``compact_skill_categories`` — skill categories DEMOTED to names-only in
the system-prompt skill index under the opt-in ``focus``
mode. Never hidden: every skill name stays visible
(so memory-anchored recall keeps working) — only the
descriptions are dropped to cut index noise. Deny-list
semantics so unknown/custom categories keep full
entries.
"""
name: str
toolset: Optional[str] = None
guidance: str = ""
model_hint: Optional[str] = None
memory_policy: str = "default"
compact_skill_categories: tuple[str, ...] = ()
# Skill categories that are clearly not part of a coding workflow. Demoted to
# names-only in the prompt's skill index under the opt-in ``focus`` mode only
# (deny-list — anything not listed here, incl. custom user categories, keeps
# full entries). Coding-adjacent categories (devops, github, mcp,
# data-science, diagramming, research, security, …) are intentionally absent.
_NON_CODING_SKILL_CATEGORIES = (
"apple", "communication", "cooking", "creative", "email", "finance",
"gaming", "gifs", "health", "media", "music", "note-taking",
"productivity", "shopping", "smart-home", "social-media", "travel",
"yuanbao",
)
GENERAL_PROFILE = ContextProfile(name="general")
CODING_PROFILE = ContextProfile(
name="coding",
toolset=CODING_TOOLSET,
guidance=CODING_AGENT_GUIDANCE,
model_hint="coding",
memory_policy="project",
compact_skill_categories=_NON_CODING_SKILL_CATEGORIES,
)
_PROFILES: dict[str, ContextProfile] = {
GENERAL_PROFILE.name: GENERAL_PROFILE,
CODING_PROFILE.name: CODING_PROFILE,
}
def get_profile(name: str) -> ContextProfile:
"""Return a registered profile, falling back to ``general``."""
return _PROFILES.get(name, GENERAL_PROFILE)
# ── Helpers ─────────────────────────────────────────────────────────────────
def _coding_mode(config: Optional[dict[str, Any]]) -> str:
"""Return the normalized ``agent.coding_context`` mode (auto/focus/on/off)."""
if config is None:
try:
from hermes_cli.config import load_config
config = load_config()
except Exception:
config = {}
raw = ((config or {}).get("agent", {}) or {}).get("coding_context", "auto")
mode = str(raw).strip().lower()
if mode in {"focus", "strict", "lean"}:
return "focus"
if mode in {"on", "true", "yes", "1", "always"}:
return "on"
if mode in {"off", "false", "no", "0", "never"}:
return "off"
return "auto"
def _resolve_cwd(cwd: Optional[str | Path]) -> Path:
if cwd:
return Path(cwd).expanduser()
try:
from agent.runtime_cwd import resolve_agent_cwd
return resolve_agent_cwd()
except Exception:
return Path(os.getcwd())
def _git_root(cwd: Path) -> Optional[Path]:
current = cwd.resolve()
for parent in [current, *current.parents]:
if (parent / ".git").exists():
return parent
return None
def _home() -> Optional[Path]:
try:
return Path.home().resolve()
except (OSError, RuntimeError):
return None
def _marker_root(cwd: Path) -> Optional[Path]:
"""Nearest ancestor that looks like a project root, or ``None``.
Walks up at most a few levels so a manifest in the workspace root counts
even when the user is in a subdirectory. ``$HOME`` itself is skipped — a
Makefile or AGENTS.md sitting in the home directory is global user config,
not a project-root signal.
"""
current = cwd.resolve()
home = _home()
for depth, parent in enumerate([current, *current.parents]):
if depth > 6:
break
if parent == home:
continue
for marker in _PROJECT_MARKERS:
if (parent / marker).exists():
return parent
return None
def _detect_profile_name(mode: str, platform: str, cwd_str: str) -> str:
"""Resolve which profile applies.
``auto``/``focus``: coding when the surface is interactive AND the cwd is a
code workspace (a git repo or a recognised project root). ``on``: always
coding. ``off``: always general.
A git repo rooted at ``$HOME`` (the dotfiles pattern) is NOT a workspace
signal — without the guard, every session anywhere under a dotfiles-managed
home directory would silently flip to the coding posture.
Detection is intentionally not memoized: it's a handful of ``stat`` calls,
and callers resolve the mode once per session anyway. Caching here would
risk a stale posture if a long-lived process (gateway/TUI) serves sessions
from different working directories.
"""
if mode == "off":
return GENERAL_PROFILE.name
if mode == "on":
return CODING_PROFILE.name
if platform and platform.strip().lower() not in INTERACTIVE_CODING_PLATFORMS:
return GENERAL_PROFILE.name
cwd = Path(cwd_str)
git_root = _git_root(cwd)
if git_root is not None and git_root == _home():
git_root = None # dotfiles repo at $HOME — not a code workspace
if git_root is not None or _marker_root(cwd) is not None:
return CODING_PROFILE.name
return GENERAL_PROFILE.name
# ── RuntimeMode (the seam) ──────────────────────────────────────────────────
@dataclass(frozen=True)
class RuntimeMode:
"""The resolved operating posture for a session. Immutable by construction.
Built once via :func:`resolve_runtime_mode` and consumed by every domain
that cares about the coding/general distinction. Never mutate or re-resolve
mid-session — that would break the prompt cache.
"""
profile: ContextProfile
surface: str
cwd: Path
# The normalized ``agent.coding_context`` mode this posture was resolved
# under (auto/focus/on/off). Toolset collapse is gated on ``focus``.
config_mode: str = "auto"
# The model id this session runs (e.g. "anthropic/claude-opus-4.8"). Used
# only to steer edit-format guidance toward the model's family — see
# ``_edit_format_line``. Fixed for the session, so cache-safe.
model: Optional[str] = None
@property
def kind(self) -> str:
return self.profile.name
@property
def is_coding(self) -> bool:
return self.profile.name == CODING_PROFILE.name
def toolset_selection(self, config: Optional[dict[str, Any]] = None) -> Optional[list[str]]:
"""Toolset list for this posture, or ``None`` to keep the platform default.
Non-``None`` only under the opt-in ``focus`` mode. The default posture
is prompt-only: most strippable toolsets are off-by-default anyway, and
a user who explicitly enabled one (image-gen for frontend/game assets,
messaging for build notifications, …) keeps it while coding.
Callers apply this only when the user hasn't pinned an explicit
selection (``--toolsets``, ``HERMES_TUI_TOOLSETS``, …); they never
override a pin. Returns the profile's toolset plus enabled MCP servers.
"""
if self.config_mode != "focus":
return None
if self.profile.toolset is None:
return None
return [self.profile.toolset, *_enabled_mcp_servers(config)]
def system_blocks(self) -> list[str]:
"""Stable system-prompt blocks for this posture (brief + workspace).
The operating brief carries a model-family edit-format nudge appended
to it (one cached string, not a separate block) so the model is steered
toward the `patch` mode it handles best — see ``_edit_format_line``.
"""
if not self.is_coding:
return []
blocks: list[str] = []
if self.profile.guidance:
brief = self.profile.guidance
edit_line = _edit_format_line(self.model)
if edit_line:
brief = f"{brief}\n{edit_line}"
blocks.append(brief)
workspace = build_coding_workspace_block(self.cwd)
if workspace:
blocks.append(workspace)
return blocks
def compact_skill_categories(self) -> frozenset[str]:
"""Skill categories to demote to names-only in the prompt's skill index.
Gated on the opt-in ``focus`` mode, like the toolset collapse: the
default posture leaves the skill index untouched. Users who didn't ask
for a lean prompt keep full entries for every category — index changes
under ``auto`` proved too surprising in practice, even names-only ones
(a demoted description is information the model no longer weighs when
deciding what to load).
Demoted — never hidden — even under ``focus``. An earlier revision
fully pruned these categories from the index, which caused silent
capability loss in a real workflow: agent-created skills are the
model's accumulated project memory (server-ops runbooks, learned
pitfalls, …), and models do not reliably reach for ``skills_list`` to
rediscover what the index stopped showing them. Names-only keeps every
skill loadable on recall while still cutting the description noise.
"""
if not self.is_coding or self.config_mode != "focus":
return frozenset()
return frozenset(self.profile.compact_skill_categories)
def resolve_runtime_mode(
*,
platform: Optional[str] = None,
cwd: Optional[str | Path] = None,
config: Optional[dict[str, Any]] = None,
model: Optional[str] = None,
) -> RuntimeMode:
"""Resolve the operating posture once. Cheap — a handful of ``stat`` calls.
This is the single entry point every domain should call. The returned
object is immutable and safe to cache for the session. Detection itself is
intentionally *not* memoized (see ``_detect_profile_name``) so a long-lived
process can't pin a stale posture; callers resolve once per session and
hold the result. ``model`` is recorded only to steer edit-format guidance;
it never affects detection.
"""
resolved_cwd = _resolve_cwd(cwd)
mode = _coding_mode(config)
name = _detect_profile_name(
mode, (platform or "").strip().lower(), str(resolved_cwd)
)
return RuntimeMode(
profile=get_profile(name),
surface=platform or "",
cwd=resolved_cwd,
config_mode=mode,
model=model,
)
# ── Back-compat surface (thin wrappers over RuntimeMode) ────────────────────
def is_coding_context(
*,
platform: Optional[str] = None,
cwd: Optional[str | Path] = None,
config: Optional[dict[str, Any]] = None,
) -> bool:
"""Whether Hermes should operate in its coding posture right now."""
return resolve_runtime_mode(platform=platform, cwd=cwd, config=config).is_coding
def coding_selection(
*,
platform: Optional[str] = None,
cwd: Optional[str | Path] = None,
config: Optional[dict[str, Any]] = None,
) -> Optional[list[str]]:
"""Toolset selection for the coding posture.
``None`` unless the user opted into ``focus`` mode AND the posture is
active — the default coding posture never overrides configured toolsets.
"""
return resolve_runtime_mode(
platform=platform, cwd=cwd, config=config
).toolset_selection(config)
def coding_system_blocks(
*,
platform: Optional[str] = None,
cwd: Optional[str | Path] = None,
config: Optional[dict[str, Any]] = None,
model: Optional[str] = None,
) -> list[str]:
"""Stable system-prompt blocks for the current posture (empty when general).
``model`` steers the brief's edit-format nudge toward the model's family.
"""
return resolve_runtime_mode(
platform=platform, cwd=cwd, config=config, model=model
).system_blocks()
def coding_compact_skill_categories(
*,
platform: Optional[str] = None,
cwd: Optional[str | Path] = None,
config: Optional[dict[str, Any]] = None,
) -> frozenset[str]:
"""Skill categories the active posture demotes to names-only in the index.
Empty outside the coding posture and outside the opt-in ``focus`` mode —
the default posture never touches the skill index. Under ``focus``,
demoted — never hidden: every skill name stays in the index and remains
loadable via ``skill_view`` / ``skills_list``; only descriptions are
dropped.
"""
return resolve_runtime_mode(
platform=platform, cwd=cwd, config=config
).compact_skill_categories()
def _enabled_mcp_servers(config: Optional[dict[str, Any]]) -> list[str]:
"""Names of MCP servers the user has enabled — kept in the coding posture.
MCP servers (figma, browser, tophat, …) are explicitly configured and part
of the coding workflow, not noise to strip.
"""
try:
from hermes_cli.config import read_raw_config
from hermes_cli.tools_config import _parse_enabled_flag
servers = read_raw_config().get("mcp_servers") or {}
return [
str(name)
for name, cfg in servers.items()
if isinstance(cfg, dict)
and _parse_enabled_flag(cfg.get("enabled", True), default=True)
]
except Exception:
return []
# ── git/workspace probe ─────────────────────────────────────────────────────
def _git(cwd: Path, *args: str) -> str:
try:
out = subprocess.run(
["git", "-C", str(cwd), *args],
capture_output=True,
text=True,
timeout=_GIT_TIMEOUT,
)
except (OSError, subprocess.SubprocessError):
return ""
return out.stdout.strip() if out.returncode == 0 else ""
def _parse_status(porcelain: str) -> tuple[dict[str, str], dict[str, int]]:
"""Parse ``git status --porcelain=2 --branch`` into branch + counts."""
branch: dict[str, str] = {}
counts = {"staged": 0, "modified": 0, "untracked": 0, "conflicts": 0}
for line in porcelain.splitlines():
if line.startswith("# branch.head"):
branch["head"] = line.split(maxsplit=2)[-1]
elif line.startswith("# branch.upstream"):
branch["upstream"] = line.split(maxsplit=2)[-1]
elif line.startswith("# branch.ab"):
parts = line.split()
branch["ahead"], branch["behind"] = parts[2].lstrip("+"), parts[3].lstrip("-")
elif line.startswith(("1 ", "2 ")):
xy = line.split(maxsplit=2)[1]
if xy[0] != ".":
counts["staged"] += 1
if xy[1] != ".":
counts["modified"] += 1
elif line.startswith("u "):
counts["conflicts"] += 1
elif line.startswith("? "):
counts["untracked"] += 1
return branch, counts
def _read_small(path: Path) -> str:
"""Read a small text file, or ``""`` — never raises, never reads huge files."""
try:
if not path.is_file() or path.stat().st_size > _MAX_FACT_FILE_BYTES:
return ""
return path.read_text(encoding="utf-8", errors="replace")
except OSError:
return ""
def _project_facts(root: Path) -> list[str]:
"""Detected project facts for the workspace snapshot.
The point is to hand the model its *verify loop* up front — which manifest,
which package manager, and the exact test/lint/build commands — instead of
making it rediscover them every session. Cheap: stat calls plus reads of a
couple of small files; built once at prompt-build time (cache-safe).
"""
facts: list[str] = []
manifests = [m for m in _PROJECT_MARKERS if m not in _CONTEXT_FILES and (root / m).is_file()]
package_managers = [
pm for lock, pm in (*_PY_LOCKFILES, *_JS_LOCKFILES) if (root / lock).is_file()
]
if manifests:
line = f"- Project: {', '.join(manifests[:6])}"
if package_managers:
line += f" ({'/'.join(dict.fromkeys(package_managers))})"
facts.append(line)
verify: list[str] = []
if (root / "scripts" / "run_tests.sh").is_file():
verify.append("scripts/run_tests.sh")
if (root / "package.json").is_file():
try:
scripts = json.loads(_read_small(root / "package.json") or "{}").get("scripts") or {}
except (json.JSONDecodeError, AttributeError):
scripts = {}
js_pm = next((pm for lock, pm in _JS_LOCKFILES if (root / lock).is_file()), "npm")
verify.extend(f"{js_pm} run {name}" for name in _VERIFY_TARGETS if name in scripts)
if (root / "pytest.ini").is_file() or "[tool.pytest" in _read_small(root / "pyproject.toml"):
verify.append("pytest")
makefile = _read_small(root / "Makefile")
if makefile:
verify.extend(
f"make {name}" for name in _VERIFY_TARGETS
if re.search(rf"^{re.escape(name)}\s*:", makefile, re.MULTILINE)
)
if verify:
deduped = list(dict.fromkeys(verify))[:_MAX_VERIFY_COMMANDS]
facts.append(f"- Verify: {'; '.join(deduped)}")
context_files = [c for c in _CONTEXT_FILES if (root / c).is_file()]
if context_files:
facts.append(f"- Context files: {', '.join(context_files)}")
return facts
def build_coding_workspace_block(cwd: Optional[str | Path] = None) -> str:
"""Workspace snapshot for the system prompt (empty outside a workspace).
Git state (branch/status/commits) when the cwd is in a repo, plus detected
project facts (manifest, package manager, verify commands, context files)
— so marker-only (non-git) projects still get a snapshot.
"""
resolved = _resolve_cwd(cwd)
git_root = _git_root(resolved)
root = git_root or _marker_root(resolved)
if root is None:
return ""
lines = ["Workspace (snapshot at session start — re-check with `git` before acting on it):"]
lines.append(f"- Root: {root}")
if git_root is not None:
branch, counts = _parse_status(_git(root, "status", "--porcelain=2", "--branch"))
head = branch.get("head", "")
if head and head != "(detached)":
line = f"- Branch: {head}"
if branch.get("upstream"):
line += f" \u2192 {branch['upstream']}"
ahead, behind = branch.get("ahead", "0"), branch.get("behind", "0")
if ahead != "0" or behind != "0":
line += f" (ahead {ahead}, behind {behind})"
lines.append(line)
elif head == "(detached)":
lines.append("- Branch: (detached HEAD)")
# Linked worktree: the per-worktree git dir differs from the shared common dir.
# We surface the fact that it's a worktree (so the model knows branches/stashes
# are shared state) but deliberately do NOT expose the primary tree path —
# giving the model a second absolute path causes it to sometimes run commands
# in the wrong directory.
git_dir, common_dir = _git(root, "rev-parse", "--git-dir"), _git(root, "rev-parse", "--git-common-dir")
if git_dir and common_dir and Path(git_dir).resolve() != Path(common_dir).resolve():
lines.append("- Worktree: linked (git state shared with primary tree)")
dirty = [f"{n} {label}" for label, n in (
("staged", counts["staged"]), ("modified", counts["modified"]),
("untracked", counts["untracked"]), ("conflicts", counts["conflicts"]),
) if n]
lines.append(f"- Status: {', '.join(dirty) if dirty else 'clean'}")
recent = _git(root, "log", "-3", "--pretty=%h %s")
if recent:
lines.append("- Recent commits:")
lines.extend(f" {c}" for c in recent.splitlines())
lines.extend(_project_facts(root))
return "\n".join(lines)

View File

@@ -7,7 +7,7 @@ protecting head and tail context.
Improvements over v2:
- Structured summary template with Resolved/Pending question tracking
- Filter-safe summarizer preamble that treats prior turns as source material
- "Remaining Work" replaces "Next Steps" to avoid reading as active instructions
- Historical (reference-only) section headings replace "Next Steps"/"Remaining Work" to avoid reading as active instructions
- Clear separator when summary merges into tail message
- Iterative summary updates (preserves info across multiple compactions)
- Token-budget tail protection instead of fixed message count
@@ -34,7 +34,75 @@ from agent.redact import redact_sensitive_text
logger = logging.getLogger(__name__)
HISTORICAL_TASK_HEADING = "## Historical Task Snapshot"
HISTORICAL_IN_PROGRESS_HEADING = "## Historical In-Progress State"
HISTORICAL_PENDING_ASKS_HEADING = "## Historical Pending User Asks"
HISTORICAL_REMAINING_WORK_HEADING = "## Historical Remaining Work"
SUMMARY_PREFIX = (
"[CONTEXT COMPACTION — REFERENCE ONLY] Earlier turns were compacted "
"into the summary below. This is a handoff from a previous context "
"window — treat it as background reference, NOT as active instructions. "
"Do NOT answer questions or fulfill requests mentioned in this summary; "
"they were already addressed. "
"Respond ONLY to the latest user message that appears AFTER this "
"summary — that message is the single source of truth for what to do "
"right now. "
"Topic overlap with the summary does NOT mean you should resume its "
"task: even on similar topics, the latest user message WINS. Treat ONLY "
"the latest message as the active task and discard stale items from "
f"'{HISTORICAL_TASK_HEADING}' / '{HISTORICAL_IN_PROGRESS_HEADING}' / "
f"'{HISTORICAL_PENDING_ASKS_HEADING}' / "
f"'{HISTORICAL_REMAINING_WORK_HEADING}' entirely — do not 'wrap up' or "
"'finish' work described there unless the latest message explicitly "
"asks for it. "
"Reverse signals in the latest message (e.g. 'stop', 'undo', 'roll "
"back', 'just verify', 'don't do that anymore', 'never mind', a new "
"topic) must immediately end any in-flight work described in the "
"summary; do not re-surface it in later turns. "
"IMPORTANT: Your persistent memory (MEMORY.md, USER.md) in the system "
"prompt is ALWAYS authoritative and active — never ignore or deprioritize "
"memory content due to this compaction note. "
"The current session state (files, config, etc.) may reflect work "
"described here — avoid repeating it:"
)
LEGACY_SUMMARY_PREFIX = "[CONTEXT SUMMARY]:"
# Metadata key added to context compression summary messages so that frontends
# (CLI, Desktop, gateway, TUI) can distinguish them from real assistant/user
# messages and filter or render them appropriately without content-prefix
# heuristics. See https://github.com/NousResearch/hermes-agent/issues/38389
#
# Underscore-prefixed ON PURPOSE: the wire sanitizers
# (agent/transports/chat_completions.py convert_messages and the summary-path
# mirror in agent/chat_completion_helpers.py) strip every top-level message
# key starting with "_" before the request leaves the process. Strict
# OpenAI-compatible gateways (Fireworks, Mistral, Moonshot/Kimi, opencode-go)
# reject payloads carrying unknown keys with "Extra inputs are not permitted",
# poisoning every subsequent request in the session — a bare key like
# "is_compressed_summary" would reach the wire and trip exactly that.
COMPRESSED_SUMMARY_METADATA_KEY = "_compressed_summary"
# Appended to every standalone summary message (and to the merged-into-tail
# prefix) so the model has an unambiguous "summary ends here" boundary.
# Without it, weak models read the verbatim "## Active Task" quote as fresh
# user input (#11475, #14521) or regurgitate an assistant-role summary as
# their own output (#33256).
_SUMMARY_END_MARKER = (
"--- END OF CONTEXT SUMMARY — "
"respond to the message below, not the summary above ---"
)
# Handoff prefixes that shipped in earlier releases. A summary persisted under
# one of these can be inherited into a resumed lineage (#35344); when it is
# re-normalized on re-compaction we must strip the OLD prefix too, otherwise the
# stale directive it carried (e.g. "resume exactly from Active Task") survives
# embedded in the body and keeps hijacking replies. Keep newest-first; entries
# are matched literally. Add a frozen copy here whenever SUMMARY_PREFIX changes.
_HISTORICAL_SUMMARY_PREFIXES = (
# Carveout era (#41607/#38364/#42812): "consistent → use as background"
# licensed stale-task resumption on topic overlap.
"[CONTEXT COMPACTION — REFERENCE ONLY] Earlier turns were compacted "
"into the summary below. This is a handoff from a previous context "
"window — treat it as background reference, NOT as active instructions. "
@@ -57,17 +125,7 @@ SUMMARY_PREFIX = (
"prompt is ALWAYS authoritative and active — never ignore or deprioritize "
"memory content due to this compaction note. "
"The current session state (files, config, etc.) may reflect work "
"described here — avoid repeating it:"
)
LEGACY_SUMMARY_PREFIX = "[CONTEXT SUMMARY]:"
# Handoff prefixes that shipped in earlier releases. A summary persisted under
# one of these can be inherited into a resumed lineage (#35344); when it is
# re-normalized on re-compaction we must strip the OLD prefix too, otherwise the
# stale directive it carried (e.g. "resume exactly from Active Task") survives
# embedded in the body and keeps hijacking replies. Keep newest-first; entries
# are matched literally. Add a frozen copy here whenever SUMMARY_PREFIX changes.
_HISTORICAL_SUMMARY_PREFIXES = (
"described here — avoid repeating it:",
# Pre-#35344: contained the self-contradicting "resume exactly" directive.
"[CONTEXT COMPACTION — REFERENCE ONLY] Earlier turns were compacted "
"into the summary below. This is a handoff from a previous context "
@@ -110,10 +168,23 @@ _SUMMARY_FAILURE_COOLDOWN_SECONDS = 600
# become another unbounded transcript copy after the LLM summarizer failed.
_FALLBACK_SUMMARY_MAX_CHARS = 8_000
_FALLBACK_TURN_MAX_CHARS = 700
_AUTO_FOCUS_MAX_TURNS = 3
_AUTO_FOCUS_TURN_MAX_CHARS = 260
_AUTO_FOCUS_MAX_CHARS = 700
# Keep a short run of recent messages verbatim even when the token budget is
# already exhausted. The public ``protect_last_n`` default is intentionally
# high for small/light tails, but using all 20 as a hard floor here would bring
# back the old large-tool-output case where nothing can be compacted.
_MAX_TAIL_MESSAGE_FLOOR = 8
_PATH_MENTION_RE = re.compile(r"(?:/|~/?|[A-Za-z]:\\)[^\s`'\")\]}<>]+")
# MEDIA delivery directives must not reach the summarizer — if one leaks into
# the summary, the downstream model may re-emit it as an active directive on
# the next turn, triggering bogus attachment sends (#14665).
_MEDIA_DIRECTIVE_RE = re.compile(r"MEDIA:\S+")
def _dedupe_append(items: list[str], value: str, *, limit: int) -> None:
value = value.strip()
@@ -974,6 +1045,7 @@ class ContextCompressor(ContextEngine):
for msg in turns:
role = msg.get("role", "unknown")
content = redact_sensitive_text(msg.get("content") or "")
content = _MEDIA_DIRECTIVE_RE.sub("[media attachment]", content)
# Tool results: keep enough content for the summarizer
if role == "tool":
@@ -1155,7 +1227,7 @@ class ContextCompressor(ContextEngine):
)
reason_text = f" Summary failure reason: {reason}." if reason else ""
body = f"""## Active Task
body = f"""{HISTORICAL_TASK_HEADING}
{active_task}
## Goal
@@ -1172,7 +1244,7 @@ Recovered from a deterministic fallback because the LLM context summarizer was u
## Active State
Unknown from deterministic fallback. Inspect current repository/session state if needed.
## In Progress
{HISTORICAL_IN_PROGRESS_HEADING}
{active_task}
## Blocked
@@ -1184,13 +1256,13 @@ None recoverable from deterministic fallback.
## Resolved Questions
None recoverable from deterministic fallback.
## Pending User Asks
{HISTORICAL_PENDING_ASKS_HEADING}
{active_task}
## Relevant Files
{_bullets(relevant_files, limit=12)}
## Remaining Work
{HISTORICAL_REMAINING_WORK_HEADING}
Continue from the most recent unfulfilled user ask and protected tail messages. Verify state with tools before making claims.
## Last Dropped Turns
@@ -1312,7 +1384,7 @@ Summary generation was unavailable, so this is a best-effort deterministic fallb
_temporal_anchoring_rule = ""
# Shared structured template (used by both paths).
_template_sections = f"""## Active Task
_template_sections = f"""{HISTORICAL_TASK_HEADING}
[THE SINGLE MOST IMPORTANT FIELD. Capture the user's most recent unfulfilled
input verbatim — the exact words they used. This includes:
- Explicit task assignments ("refactor the auth module")
@@ -1359,7 +1431,7 @@ Be specific with file paths, commands, line numbers, and results.]
- Any running processes or servers
- Environment details that matter]
## In Progress
{HISTORICAL_IN_PROGRESS_HEADING}
[Work currently underway — what was being done when compaction fired]
## Blocked
@@ -1371,14 +1443,14 @@ Be specific with file paths, commands, line numbers, and results.]
## Resolved Questions
[Questions the user asked that were ALREADY answered — include the answer so it is not repeated]
## Pending User Asks
[Questions or requests from the user that have NOT yet been answered or fulfilled. If none, write "None."]
{HISTORICAL_PENDING_ASKS_HEADING}
[Questions or requests from the user that have NOT yet been answered or fulfilled. These are STALE — they were from the compacted turns. Write them here for reference only. The agent must NOT act on them unless the latest user message explicitly requests it. If none, write "None."]
## Relevant Files
[Files read, modified, or created — with brief note on each]
## Remaining Work
[What remains to be done — framed as context, not instructions]
{HISTORICAL_REMAINING_WORK_HEADING}
[What remains to be done — framed as STALE context for reference only. The agent must NOT resume this work unless the latest user message explicitly asks for it.]
## Critical Context
[Any specific values, error messages, configuration details, or data that would be lost without explicit preservation. NEVER include API keys, tokens, passwords, or credentials — write [REDACTED] instead.]
@@ -1421,7 +1493,7 @@ Use this exact structure:
prompt += f"""
FOCUS TOPIC: "{focus_topic}"
The user has requested that this compaction PRIORITISE preserving all information related to the focus topic above. For content related to "{focus_topic}", include full detail — exact values, file paths, command outputs, error messages, and decisions. For content NOT related to the focus topic, summarise more aggressively (brief one-liners or omit if truly irrelevant). The focus topic sections should receive roughly 60-70% of the summary token budget. Even for the focus topic, NEVER preserve API keys, tokens, passwords, or credentials — use [REDACTED]."""
This compaction should PRIORITISE preserving all information related to the focus topic above. For content related to "{focus_topic}", include full detail — exact values, file paths, command outputs, error messages, and decisions. For content NOT related to the focus topic, summarise more aggressively (brief one-liners or omit if truly irrelevant). The focus topic sections should receive roughly 60-70% of the summary token budget. Even for the focus topic, NEVER preserve API keys, tokens, passwords, or credentials — use [REDACTED]."""
try:
call_kwargs = {
@@ -1574,7 +1646,13 @@ The user has requested that this compaction PRIORITISE preserving all informatio
text = (summary or "").strip()
for prefix in (SUMMARY_PREFIX, LEGACY_SUMMARY_PREFIX, *_HISTORICAL_SUMMARY_PREFIXES):
if text.startswith(prefix):
return text[len(prefix):].lstrip()
text = text[len(prefix):].lstrip()
break
# Strip the trailing end marker too — a rehydrated handoff body that
# keeps it would leak the boundary directive into the iterative-update
# summarizer prompt (and the marker is re-appended on insertion anyway).
if text.endswith(_SUMMARY_END_MARKER):
text = text[: -len(_SUMMARY_END_MARKER)].rstrip()
return text
@classmethod
@@ -1590,6 +1668,52 @@ The user has requested that this compaction PRIORITISE preserving all informatio
return True
return any(text.startswith(p) for p in _HISTORICAL_SUMMARY_PREFIXES)
@staticmethod
def _has_compressed_summary_metadata(message: Any) -> bool:
"""Return True if *message* carries the compressed-summary flag.
Callers (frontends, CLI, gateway) can use this to distinguish context
compaction summaries from real assistant or user messages without
relying on content-prefix heuristics. The flag is in-process only —
the wire sanitizers strip underscore-prefixed keys before API calls.
"""
if not isinstance(message, dict):
return False
return bool(message.get(COMPRESSED_SUMMARY_METADATA_KEY))
@classmethod
def _derive_auto_focus_topic(
cls,
messages: List[Dict[str, Any]],
) -> Optional[str]:
"""Infer a compact focus hint from the most recent real user turns."""
candidates: list[str] = []
for idx in range(len(messages) - 1, -1, -1):
msg = messages[idx]
if msg.get("role") != "user":
continue
content = msg.get("content")
if cls._is_context_summary_content(content):
continue
text = redact_sensitive_text(_content_text_for_contains(content).strip())
if not text:
continue
text = " ".join(text.split())
if len(text) > _AUTO_FOCUS_TURN_MAX_CHARS:
text = text[: _AUTO_FOCUS_TURN_MAX_CHARS - 1].rstrip() + ""
candidates.append(text)
if len(candidates) >= _AUTO_FOCUS_MAX_TURNS:
break
if not candidates:
return None
candidates.reverse()
focus = "Recent user focus:\n" + "\n".join(f"- {item}" for item in candidates)
if len(focus) > _AUTO_FOCUS_MAX_CHARS:
focus = focus[: _AUTO_FOCUS_MAX_CHARS - 1].rstrip() + ""
return focus
@classmethod
def _find_latest_context_summary(
cls,
@@ -1742,6 +1866,105 @@ The user has requested that this compaction PRIORITISE preserving all informatio
return i
return -1
def _find_last_assistant_message_idx(
self, messages: List[Dict[str, Any]], head_end: int
) -> int:
"""Return the index of the last user-visible assistant reply at or
after *head_end*, or -1.
A "user-visible reply" is an assistant message with non-empty
textual content — i.e. one that the WebUI / TUI / SessionsPage
rendered as a bubble the operator could read. We deliberately
skip assistant messages that contain only ``tool_calls`` (and
no text), because those render as small "calling tool X"
indicators and aren't what the reporter means by "the output
of the last message you sent" (#29824).
Falling back to the most recent assistant message of ANY kind
only kicks in when no content-bearing assistant message exists
in the compressible region — typically a fresh session that
just started a multi-step tool sequence with no prior reply
to anchor. In that case the agent fix is a no-op and the
existing user-message anchor carries the load.
"""
last_any = -1
for i in range(len(messages) - 1, head_end - 1, -1):
msg = messages[i]
if msg.get("role") != "assistant":
continue
if last_any < 0:
last_any = i
content = msg.get("content")
if isinstance(content, str) and content.strip():
return i
if isinstance(content, list):
# Multimodal / Anthropic-style content: look for any
# text block with non-empty text.
for part in content:
if isinstance(part, dict):
text = part.get("text") or part.get("content")
if isinstance(text, str) and text.strip():
return i
return last_any
def _ensure_last_assistant_message_in_tail(
self,
messages: List[Dict[str, Any]],
cut_idx: int,
head_end: int,
) -> int:
"""Guarantee the most recent assistant message is in the protected tail.
WebUI / TUI / SessionsPage bug (#29824). Without this anchor,
``_find_tail_cut_by_tokens`` can leave the user's most recent
visible assistant response inside the compressed middle region —
especially when the conversation has a single oversized tool
result or a long stretch of tool-call/result pairs after the
last assistant reply. The summariser then rolls that reply up
into the single ``[CONTEXT COMPACTION — REFERENCE ONLY]`` block
persisted as ``role="user"`` or ``role="assistant"``. From the
operator's perspective the WebUI session viewer
(``web/src/pages/SessionsPage.tsx``) and the TUI chat panel
both suddenly show the opaque "Context compaction" block in the
slot where they were just reading the assistant's actual reply:
User: "i cant see the output of the last message you
sent, i did see it previously, however now see
'context compaction'"
Mirror of ``_ensure_last_user_message_in_tail`` but anchors on
the last assistant-role message. Re-runs the tool-group
alignment so we don't split a ``tool_call`` / ``tool_result``
group that immediately precedes the anchored message — orphaned
tool messages would otherwise be removed by
``_sanitize_tool_pairs`` and trigger the same data-loss symptom
we're trying to prevent.
"""
last_asst_idx = self._find_last_assistant_message_idx(messages, head_end)
if last_asst_idx < 0:
# No assistant message in the compressible region — nothing
# to anchor (single-turn pre-reply state, etc.).
return cut_idx
if last_asst_idx >= cut_idx:
# Already in the tail — the token-budget walk did the right
# thing on its own.
return cut_idx
# Pull cut_idx back to the assistant message, then re-align so
# we don't split a tool group that immediately precedes it
# (e.g. an ``assistant(tool_calls)`` → ``tool(result)`` →
# ``assistant(final reply)`` sequence would otherwise leave the
# ``tool`` orphan when cut lands at the final reply).
new_cut = self._align_boundary_backward(messages, last_asst_idx)
if not self.quiet_mode:
logger.debug(
"Anchoring tail cut to last assistant message at index %d "
"(was %d, aligned to %d) to keep the previously-visible "
"reply out of the compaction summary (#29824)",
last_asst_idx, cut_idx, new_cut,
)
# Safety: never go back into the head region.
return max(new_cut, head_end + 1)
def _ensure_last_user_message_in_tail(
self,
messages: List[Dict[str, Any]],
@@ -1753,7 +1976,7 @@ The user has requested that this compaction PRIORITISE preserving all informatio
Context compressor bug (#10896): ``_align_boundary_backward`` can pull
``cut_idx`` past a user message when it tries to keep tool_call/result
groups together. If the last user message ends up in the *compressed*
middle region the LLM summariser writes it into "Pending User Asks",
middle region the LLM summariser writes it into "Historical Pending User Asks",
but ``SUMMARY_PREFIX`` tells the next model to respond only to user
messages *after* the summary — so the task effectively disappears from
the active context, causing the agent to stall, repeat completed work,
@@ -1800,11 +2023,12 @@ The user has requested that this compaction PRIORITISE preserving all informatio
derived from ``summary_target_ratio * context_length``, so it
scales automatically with the model's context window.
Token budget is the primary criterion. A hard minimum of 3 messages
is always protected, but the budget is allowed to exceed by up to
1.5x to avoid cutting inside an oversized message (tool output, file
read, etc.). If even the minimum 3 messages exceed 1.5x the budget
the cut is placed right after the head so compression still runs.
Token budget is the primary criterion. A bounded message-count floor
keeps a short run of recent turns verbatim even when the budget is
exhausted, but the budget is allowed to exceed by up to 1.5x to avoid
cutting inside an oversized message (tool output, file read, etc.). If
even that floor exceeds 1.5x the budget, the cut is placed right after
the head so compression still runs.
Never cuts inside a tool_call/result group. Always ensures the most
recent user message is in the tail (see ``_ensure_last_user_message_in_tail``).
@@ -1812,8 +2036,19 @@ The user has requested that this compaction PRIORITISE preserving all informatio
if token_budget is None:
token_budget = self.tail_token_budget
n = len(messages)
# Hard minimum: always keep at least 3 messages in the tail
min_tail = min(3, n - head_end - 1) if n - head_end > 1 else 0
# Hard minimum: always keep a bounded recent-message floor in the tail.
# ``protect_last_n`` remains a minimum up to the cap; the cap avoids
# preserving a whole run of bulky tool outputs on every compaction.
available_tail = max(0, n - head_end - 1)
min_tail_floor = max(3, min(self.protect_last_n, _MAX_TAIL_MESSAGE_FLOOR))
# Leave at least two non-head messages available to summarize on short
# transcripts; otherwise compression can replace a tiny middle with a
# summary and save no messages at all.
compressible_tail_cap = max(3, available_tail - 2)
min_tail = (
min(min_tail_floor, compressible_tail_cap, available_tail)
if available_tail > 1 else 0
)
soft_ceiling = int(token_budget * 1.5)
accumulated = 0
cut_idx = n # start from beyond the end
@@ -1885,6 +2120,13 @@ The user has requested that this compaction PRIORITISE preserving all informatio
# active task is never lost to compression (fixes #10896).
cut_idx = self._ensure_last_user_message_in_tail(messages, cut_idx, head_end)
# Ensure the most recent assistant message is always in the tail
# so the previously-visible reply isn't silently rolled into the
# ``[CONTEXT COMPACTION — REFERENCE ONLY]`` block (fixes #29824).
# Each anchor only walks ``cut_idx`` backward, so chaining them is
# monotonic — the tail can only grow, never shrink.
cut_idx = self._ensure_last_assistant_message_in_tail(messages, cut_idx, head_end)
return max(cut_idx, head_end + 1)
# ------------------------------------------------------------------
@@ -2037,7 +2279,8 @@ The user has requested that this compaction PRIORITISE preserving all informatio
)
# Phase 3: Generate structured summary
summary = self._generate_summary(turns_to_summarize, focus_topic=focus_topic)
summary_focus_topic = focus_topic or self._derive_auto_focus_topic(messages)
summary = self._generate_summary(turns_to_summarize, focus_topic=summary_focus_topic)
# If summary generation failed, behavior splits on
# ``abort_on_summary_failure`` (config: compression.abort_on_summary_failure):
@@ -2117,32 +2360,33 @@ The user has requested that this compaction PRIORITISE preserving all informatio
# When the summary lands as a standalone role="user" message,
# weak models read the verbatim "## Active Task" quote of a past
# user request as fresh input (#11475, #14521). Append the explicit
# end marker — the same one used in the merge-into-tail path — so
# the model has a clear "summary above, not new input" signal.
if not _merge_summary_into_tail and summary_role == "user":
summary = (
summary
+ "\n\n--- END OF CONTEXT SUMMARY — "
"respond to the message below, not the summary above ---"
)
# user request as fresh input (#11475, #14521).
# When it lands as role="assistant", models may regurgitate the
# summary text as their own output (#33256). In both cases, append
# the explicit end marker so the model has a clear "summary ends
# here, respond to the message below" signal.
if not _merge_summary_into_tail:
summary = summary + "\n\n" + _SUMMARY_END_MARKER
if not _merge_summary_into_tail:
compressed.append({"role": summary_role, "content": summary})
compressed.append({
"role": summary_role,
"content": summary,
COMPRESSED_SUMMARY_METADATA_KEY: True,
})
for i in range(compress_end, n_messages):
msg = messages[i].copy()
if _merge_summary_into_tail and i == compress_end:
merged_prefix = (
summary
+ "\n\n--- END OF CONTEXT SUMMARY — "
"respond to the message below, not the summary above ---\n\n"
)
merged_prefix = summary + "\n\n" + _SUMMARY_END_MARKER + "\n\n"
msg["content"] = _append_text_to_content(
msg.get("content"),
merged_prefix,
prepend=True,
)
# Mark the merged message so frontends can identify it as
# containing a compression summary prefix.
msg[COMPRESSED_SUMMARY_METADATA_KEY] = True
_merge_summary_into_tail = False
compressed.append(msg)

View File

@@ -595,7 +595,11 @@ def run_conversation(
# landed after an orphan tool result). Most providers return
# empty content on malformed sequences, which would otherwise
# retrigger the empty-retry loop indefinitely.
repaired_seq = agent._repair_message_sequence(messages)
# repair_message_sequence_with_cursor also recomputes the SessionDB
# flush cursor (_last_flushed_db_idx) when repair compacts the list,
# so the turn-end flush doesn't skip the assistant/tool chain (#44837).
from agent.agent_runtime_helpers import repair_message_sequence_with_cursor
repaired_seq = repair_message_sequence_with_cursor(agent, messages)
if repaired_seq > 0:
request_logger.info(
"Repaired %s message-alternation violations before request (session=%s)",
@@ -2221,30 +2225,54 @@ def run_conversation(
print(f"{agent.log_prefix} • Legacy cleanup: hermes config set ANTHROPIC_TOKEN \"\"")
print(f"{agent.log_prefix} • Clear stale keys: hermes config set ANTHROPIC_API_KEY \"\"")
# ── Thinking block signature recovery ─────────────────
# Thinking block signature recovery.
#
# Anthropic signs thinking blocks against the full turn
# content. Any upstream mutation (context compression,
# content. Any upstream mutation (context compression,
# session truncation, message merging) invalidates the
# signature → HTTP 400. Recovery: strip reasoning_details
# from all messages so the next retry sends no thinking
# blocks at all. One-shot — don't retry infinitely.
# signature and the API replies HTTP 400 ("invalid
# signature" or "cannot be modified"). Recovery strips
# ``reasoning_details`` so the retry sends no thinking
# blocks at all. One-shot per outer loop.
#
# The strip targets ``api_messages``, which is the
# API-call-time list that ``_build_api_kwargs`` consumes
# on every retry. ``api_messages`` was populated once at
# the start of the turn from shallow copies of
# ``messages``, so mutating it does not touch the
# canonical store. The previous implementation popped
# ``reasoning_details`` from ``messages`` instead, which
# had two problems: ``api_messages`` carried its own
# reference to the field through the shallow copy, so the
# retry's wire payload still included thinking blocks and
# the recovery never reached the API; and the mutation
# persisted into ``state.db`` through any subsequent
# ``_persist_session`` call, permanently corrupting the
# conversation. Future turns would replay the stripped
# state, hit the same 400, and the agent would terminate
# with ``max_retries_exhausted``, often spawning
# cascading compaction-ended sessions chained off the
# corrupted parent.
if (
classified.reason == FailoverReason.thinking_signature
and not _retry.thinking_sig_retry_attempted
):
_retry.thinking_sig_retry_attempted = True
for _m in messages:
if isinstance(_m, dict):
_api_stripped = 0
for _m in api_messages:
if isinstance(_m, dict) and "reasoning_details" in _m:
_m.pop("reasoning_details", None)
_api_stripped += 1
agent._vprint(
f"{agent.log_prefix}⚠️ Thinking block signature invalid "
f"stripped all thinking blocks, retrying...",
f"{agent.log_prefix}⚠️ Thinking block signature invalid, "
f"stripped reasoning_details from api_messages for retry...",
force=True,
)
logger.warning(
"%sThinking block signature recovery: stripped "
"reasoning_details from %d messages",
agent.log_prefix, len(messages),
"reasoning_details from %d api_messages "
"(canonical messages unchanged)",
agent.log_prefix, _api_stripped,
)
continue
@@ -2607,10 +2635,13 @@ def run_conversation(
except Exception:
pass
if _genuine_nous_rate_limit:
# Skip straight to max_retries -- the
# top-of-loop guard will handle fallback or
# bail cleanly.
retry_count = max_retries
# Re-enter the loop exactly once so the
# top-of-loop Nous guard handles fallback or
# bails cleanly. (Setting retry_count to
# max_retries would make the while condition
# false immediately and the guard would never
# run -- no fallback, generic exhaustion error.)
retry_count = max(0, max_retries - 1)
continue
# Upstream capacity 429: fall through to normal
# retry logic. A different model (or the same

View File

@@ -194,17 +194,71 @@ class AgentNotice:
id: Optional[str] = None
# ── is_free_tier_model (local-data-only free-model check) ────────────────────
def is_free_tier_model(model: str, base_url: str = "") -> bool:
"""Return True when *model* is a Nous free-tier model, using ONLY local data.
Two signals, both zero-network:
1. The ``:free`` suffix — the canonical Nous free SKU marker (e.g.
``nvidia/nemotron-3-ultra:free``). Free by construction on the API side
(spend is forced to 0 for ``:free`` ids).
2. A peek into the in-process pricing cache in ``hermes_cli.models``
(populated when the model picker fetched ``/v1/models`` pricing for
*base_url*). PEEK ONLY — a cache miss never triggers a fetch. This is
CLI/TUI-session best-effort: gateway sessions never run the picker's
pricing fetch, so suppression there rests entirely on the ``:free``
suffix (which all Nous free SKUs carry).
Fail-open to False (the depleted notice still shows) on any error: wrongly
showing the warning is recoverable noise; wrongly hiding it on a paid model
would mask a real billing block.
"""
if not model:
return False
if model.endswith(":free"):
return True
if not base_url:
return False
try:
from hermes_cli.models import _is_model_free, _pricing_cache
# Mirror get_pricing_for_provider's key normalization: the agent's
# Nous base_url is /v1-suffixed (https://inference-api.nousresearch.com/v1)
# but the picker keys _pricing_cache on the pre-/v1 root.
key = base_url.rstrip("/")
if key.endswith("/v1"):
key = key[:-3].rstrip("/")
pricing = _pricing_cache.get(key)
if not pricing:
return False
return _is_model_free(model, pricing)
except Exception:
return False
# ── evaluate_credits_notices (pure reconciliation function) ──────────────────
def evaluate_credits_notices(
state: CreditsState,
latch: dict,
*,
model_is_free: bool = False,
) -> tuple[list[AgentNotice], list[str]]:
"""Reconcile credits notices against the latch. Mutates ``latch`` IN PLACE.
latch = {"active": set[str], "seen_below_90": bool, "usage_band": Optional[int]}.
``model_is_free``: True when the session's active model is a Nous free-tier
model (see :func:`is_free_tier_model`). Suppresses the ``credits.depleted``
notice — a depleted account on a free model can keep inferencing, so the
error banner is noise (and confuses free-tier users who never had credits).
Suppression does NOT emit the "restored" success notice; that fires only on
a genuine ``paid_access`` flip back to True.
Returns ``(to_show: list[AgentNotice], to_clear: list[str])``.
Caller emits to_clear FIRST, then to_show.
@@ -232,6 +286,16 @@ def evaluate_credits_notices(
for band in CREDITS_USAGE_BANDS: # ascending → last match wins = highest
if uf >= band[0]:
current_band = band
# Top-up suppression: when the account holds purchased (top-up) credits,
# the subscription-cap gauge is the wrong denominator — warning "90% used"
# at a user sitting on $50 of top-up is noise (and it previously stuck
# PERMANENTLY alongside grant_spent at >=100%). Suppress the usage band
# entirely; the cap-reached case is covered by the grant_spent info notice
# below, which already names the remaining top-up balance. A top-up landing
# mid-session flips current_band → None and the clear path below removes
# any showing band line.
if state.purchased_micros > 0:
current_band = None
grant_cond = (
state.denominator_kind == "subscription_cap"
and uf is not None
@@ -284,10 +348,14 @@ def evaluate_credits_notices(
active.discard("credits.grant_spent")
# ── depleted ─────────────────────────────────────────────────────────────
if depleted_cond and "credits.depleted" not in active:
# Suppressed while the active model is free: inference still works there,
# so the error banner would just alarm users (free-tier users especially,
# who never had paid credits to "lose").
show_depleted = depleted_cond and not model_is_free
if show_depleted and "credits.depleted" not in active:
to_show.append(
AgentNotice(
text="✕ Credit access paused · run /usage for balance",
text="✕ Credit access paused · run /credits to top up",
level="error",
kind=CREDITS_NOTICE_KIND,
key="credits.depleted",
@@ -295,20 +363,23 @@ def evaluate_credits_notices(
)
)
active.add("credits.depleted")
elif "credits.depleted" in active and not depleted_cond:
elif "credits.depleted" in active and not show_depleted:
to_clear.append("credits.depleted")
active.discard("credits.depleted")
# Recovery: also emit the success notice
to_show.append(
AgentNotice(
text="✓ Credit access restored",
level="success",
kind="ttl",
ttl_ms=CREDITS_RESTORED_TTL_MS,
key="credits.restored",
id="credits.restored",
if not depleted_cond:
# Genuine recovery (paid_access flipped back True): also emit the
# success notice. A clear caused by switching to a free model while
# still depleted must NOT claim access was restored.
to_show.append(
AgentNotice(
text="✓ Credit access restored",
level="success",
kind="ttl",
ttl_ms=CREDITS_RESTORED_TTL_MS,
key="credits.restored",
id="credits.restored",
)
)
)
return (to_show, to_clear)

View File

@@ -858,6 +858,20 @@ def _detect_tool_failure(tool_name: str, result: str | None) -> tuple[bool, str]
return False, ""
def _used_free_parallel(result: str | None) -> bool:
"""True when a web result came from Parallel's free Search MCP.
Only the keyless Parallel path tags its result with ``provider="parallel"``;
the paid REST path and every other provider omit it. Used to label the tool
line "Parallel search" / "Parallel fetch" exactly when the free MCP served
the call.
"""
if not isinstance(result, str) or '"provider"' not in result:
return False
data = safe_json_loads(result)
return isinstance(data, dict) and str(data.get("provider", "")).lower() == "parallel"
def get_cute_tool_message(
tool_name: str, args: dict, duration: float, result: str | None = None,
) -> str:
@@ -895,15 +909,17 @@ def get_cute_tool_message(
return f"{line}{failure_suffix}"
if tool_name == "web_search":
return _wrap(f"┊ 🔍 search {_trunc(args.get('query', ''), 42)} {dur}")
verb = "Parallel search" if _used_free_parallel(result) else "search"
return _wrap(f"┊ 🔍 {verb:<9} {_trunc(args.get('query', ''), 42)} {dur}")
if tool_name == "web_extract":
verb = "Parallel fetch" if _used_free_parallel(result) else "fetch"
urls = args.get("urls", [])
if urls:
url = urls[0] if isinstance(urls, list) else str(urls)
domain = url.replace("https://", "").replace("http://", "").split("/")[0]
extra = f" +{len(urls)-1}" if len(urls) > 1 else ""
return _wrap(f"┊ 📄 fetch {_trunc(domain, 35)}{extra} {dur}")
return _wrap(f"┊ 📄 fetch pages {dur}")
return _wrap(f"┊ 📄 {verb:<9} {_trunc(domain, 35)}{extra} {dur}")
return _wrap(f"┊ 📄 {verb:<9} pages {dur}")
if tool_name == "terminal":
return _wrap(f"┊ 💻 $ {_trunc(args.get('command', ''), 42)} {dur}")
if tool_name == "process":

View File

@@ -549,14 +549,32 @@ def classify_api_error(
should_fallback=True,
)
# Anthropic thinking block signature invalid (400).
# Anthropic thinking block recovery (400). Two distinct failure modes,
# same recovery (strip all reasoning_details and retry without thinking
# blocks — see the thinking_signature handler in conversation_loop.py):
# 1. Signature mismatch: a thinking block is signed against the full
# turn content; any upstream mutation (context compression, session
# truncation, message merging) invalidates the signature.
# Pattern: "signature" + "thinking".
# 2. Frozen-block mutation: Anthropic rejects any change to the
# thinking/redacted_thinking blocks in the *latest* assistant
# message — "`thinking` or `redacted_thinking` blocks in the latest
# assistant message cannot be modified. These blocks must remain as
# they were in the original response." This carries no "signature"
# token, so the original pattern missed it and the turn hard-aborted
# as a non-retryable client error instead of self-healing.
# Pattern: "thinking" + ("cannot be modified" | "must remain as they were").
# Don't gate on provider — OpenRouter proxies Anthropic errors, so the
# provider may be "openrouter" even though the error is Anthropic-specific.
# The message pattern ("signature" + "thinking") is unique enough.
# The combined patterns are unique enough.
if (
status_code == 400
and "signature" in error_msg
and "thinking" in error_msg
and (
"signature" in error_msg
or "cannot be modified" in error_msg
or "must remain as they were" in error_msg
)
):
return _result(
FailoverReason.thinking_signature,

View File

@@ -489,6 +489,23 @@ PLATFORM_HINTS = {
"files arrive as downloadable documents. You can also include image "
"URLs in markdown format ![alt](url) and they will be sent as photos."
),
"whatsapp_cloud": (
"You are on a text messaging communication platform, WhatsApp "
"(via Meta's official Business Cloud API). Standard markdown "
"(**bold**, ~~strike~~, # headers, [links](url)) is auto-converted "
"to WhatsApp's native syntax (*bold*, ~strike~, etc.) — feel free "
"to write in markdown. Tables are NOT supported — prefer bullet "
"lists or labeled key:value pairs. "
"You can send media files natively: include MEDIA:/absolute/path/to/file "
"in your response. Images (.jpg, .png) become photo attachments, "
"videos (.mp4) play inline, audio (.mp3, .ogg) sends as voice/audio "
"messages, other files arrive as documents. Image URLs in markdown "
"format ![alt](url) also work. "
"IMPORTANT: this platform has a 24-hour conversation window — if the "
"user hasn't messaged in 24h, free-form replies are refused by Meta "
"(error 131047). This rarely matters for live chat, but is worth "
"knowing if you're scheduling a delayed message."
),
"telegram": (
"You are on a text messaging communication platform, Telegram. "
"Standard markdown is automatically converted to Telegram format. "
@@ -1101,11 +1118,12 @@ def _skill_should_show(
def build_skills_system_prompt(
available_tools: "set[str] | None" = None,
available_toolsets: "set[str] | None" = None,
compact_categories: "frozenset[str] | None" = None,
) -> str:
"""Build a compact skill index for the system prompt.
Two-layer cache:
1. In-process LRU dict keyed by (skills_dir, tools, toolsets)
1. In-process LRU dict keyed by (skills_dir, tools, toolsets, hidden)
2. Disk snapshot (``.skills_prompt_snapshot.json``) validated by
mtime/size manifest — survives process restarts
@@ -1115,6 +1133,12 @@ def build_skills_system_prompt(
scanned alongside the local ``~/.hermes/skills/`` directory. External dirs
are read-only — they appear in the index but new skills are always created
in the local dir. Local skills take precedence when names collide.
``compact_categories`` (e.g. from the coding posture — see
agent/coding_context.py) demotes whole categories to a names-only line in
the rendered index. Nothing is ever hidden: every skill name stays
visible and loadable via ``skill_view`` / ``skills_list``; only the
descriptions are dropped, and a footer note explains the demotion.
"""
skills_dir = get_skills_dir()
external_dirs = get_all_skills_dirs()[1:] # skip local (index 0)
@@ -1139,6 +1163,7 @@ def build_skills_system_prompt(
tuple(sorted(str(ts) for ts in (available_toolsets or set()))),
_platform_hint,
tuple(sorted(disabled)),
tuple(sorted(compact_categories or ())),
)
with _SKILLS_PROMPT_CACHE_LOCK:
cached = _SKILLS_PROMPT_CACHE.get(cache_key)
@@ -1272,18 +1297,44 @@ def build_skills_system_prompt(
except Exception as e:
logger.debug("Could not read external skill description %s: %s", desc_file, e)
# Posture-driven category demotion (e.g. non-coding skills while pairing
# on code). Demoted categories stay in the index as a single names-only
# line — descriptions are dropped to cut noise, but every skill name
# remains visible so memory-anchored recall ("load <name>") keeps working.
# NEVER remove entries entirely: agent-created skills are the model's
# project memory, and models don't reach for skills_list to rediscover
# what the index stops showing them. Match on the top-level category
# segment so nested categories ("social-media/twitter") are demoted with
# their parent.
demoted = frozenset(
cat for cat in skills_by_category
if cat.split("/", 1)[0] in (compact_categories or frozenset())
)
hidden_note = ""
if demoted:
hidden_note = (
"\n(Categories marked [names only] are outside the current coding "
"context, so their descriptions are omitted — the skills work "
"normally and load with skill_view(name) as usual.)"
)
if not skills_by_category:
result = ""
else:
index_lines = []
for category in sorted(skills_by_category.keys()):
# Deduplicate and sort skills within each category
seen = set()
if category in demoted:
names = sorted({name for name, _ in skills_by_category[category]})
index_lines.append(f" {category} [names only]: {', '.join(names)}")
continue
cat_desc = category_descriptions.get(category, "")
if cat_desc:
index_lines.append(f" {category}: {cat_desc}")
else:
index_lines.append(f" {category}:")
# Deduplicate and sort skills within each category
seen = set()
for name, desc in sorted(skills_by_category[category], key=lambda x: x[0]):
if name in seen:
continue
@@ -1320,6 +1371,7 @@ def build_skills_system_prompt(
"</available_skills>\n"
"\n"
"Only proceed without loading a skill if genuinely none are relevant to the task."
+ hidden_note
)
# ── Store in LRU cache ────────────────────────────────────────────
@@ -1383,13 +1435,13 @@ def build_nous_subscription_prompt(valid_tool_names: "set[str] | None" = None) -
lines = [
"# Nous Subscription",
"Nous subscription includes managed web tools (Firecrawl), image generation (FAL), OpenAI TTS, and browser automation (Browser Use) by default. Modal execution is optional.",
"Nous subscription includes managed web tools (Firecrawl), image generation (FAL), OpenAI TTS, OpenAI Whisper STT, and browser automation (Browser Use) by default. Modal execution is optional.",
"Current capability status:",
]
lines.extend(_status_line(feature) for feature in features.items())
lines.extend(
[
"When a Nous-managed feature is active, do not ask the user for Firecrawl, FAL, OpenAI TTS, or Browser-Use API keys.",
"When a Nous-managed feature is active, do not ask the user for Firecrawl, FAL, OpenAI TTS, OpenAI Whisper, or Browser-Use API keys.",
"If the user is not subscribed and asks for a capability that Nous subscription would unlock or simplify, suggest Nous subscription as one option alongside direct setup or local alternatives.",
"Do not mention subscription unless the user asks about it or it directly solves the current missing capability.",
"Useful commands: hermes setup, hermes setup tools, hermes setup terminal, hermes status.",

View File

@@ -191,9 +191,23 @@ def build_system_prompt_parts(agent: Any, system_message: Optional[str] = None)
)
if toolset
}
# Focus mode (opt-in) demotes non-coding skill categories to
# names-only in the index (never hidden — skill_view/skills_list
# reach everything, and every name stays visible for recall). The
# default coding posture leaves the index untouched.
_compact_cats = frozenset()
try:
from agent.coding_context import coding_compact_skill_categories
_compact_cats = coding_compact_skill_categories(
platform=agent.platform, cwd=resolve_context_cwd()
)
except Exception:
_compact_cats = frozenset()
skills_prompt = _r.build_skills_system_prompt(
available_tools=agent.valid_tool_names,
available_toolsets=avail_toolsets,
compact_categories=_compact_cats or None,
)
else:
skills_prompt = ""
@@ -221,6 +235,26 @@ def build_system_prompt_parts(agent: Any, system_message: Optional[str] = None)
if _env_hints:
stable_parts.append(_env_hints)
# Coding posture (base Hermes, any interactive coding surface in a code
# workspace — see agent/coding_context.py). The operating brief + the live
# git/workspace snapshot are built once here and cached for the session;
# the snapshot is never re-probed per turn (that would break the prompt
# cache), so the brief tells the model to re-check git before relying on it.
if agent.valid_tool_names:
try:
from agent.coding_context import coding_system_blocks
stable_parts.extend(
coding_system_blocks(
platform=agent.platform,
cwd=resolve_context_cwd(),
model=agent.model,
)
)
except Exception:
# Coding-context probing must never block prompt build.
pass
# Local Python toolchain probe — names python/pip/uv/PEP-668 state when
# something is non-default so the model can pick the right install
# strategy without discovering by failure. Emits a single line; emits

View File

@@ -417,7 +417,7 @@ def execute_tool_calls_concurrent(agent, assistant_message, messages: list, effe
# ── Logging / callbacks ──────────────────────────────────────────
tool_names_str = ", ".join(name for _, name, _, _, _, _ in parsed_calls)
if not agent.quiet_mode:
if not agent.quiet_mode and getattr(agent, "tool_progress_mode", "all") != "off":
print(f" ⚡ Concurrent: {num_tools} tool calls — {tool_names_str}")
for i, (tc, name, args, middleware_trace, block_result, blocked_by_guardrail) in enumerate(parsed_calls, 1):
args_str = json.dumps(args, ensure_ascii=False)
@@ -702,7 +702,7 @@ def execute_tool_calls_concurrent(agent, assistant_message, messages: list, effe
if agent._should_emit_quiet_tool_messages():
cute_msg = _get_cute_tool_message_impl(name, args, tool_duration, result=function_result)
agent._safe_print(f" {cute_msg}")
elif getattr(agent, "tool_progress_mode", "all") != "off":
elif not agent.quiet_mode and getattr(agent, "tool_progress_mode", "all") != "off":
_preview_str = _multimodal_text_summary(function_result)
if agent.verbose_logging:
print(f" ✅ Tool {i+1} completed in {tool_duration:.2f}s")
@@ -866,7 +866,7 @@ def execute_tool_calls_sequential(agent, assistant_message, messages: list, effe
elif function_name == "skill_manage":
agent._iters_since_skill = 0
if not agent.quiet_mode:
if not agent.quiet_mode and getattr(agent, "tool_progress_mode", "all") != "off":
args_str = json.dumps(function_args, ensure_ascii=False)
if agent.verbose_logging:
print(f" 📞 Tool {i}: {function_name}({list(function_args.keys())})")
@@ -1384,7 +1384,7 @@ def execute_tool_calls_sequential(agent, assistant_message, messages: list, effe
# entire batch. The model sees it on the next API iteration.
agent._apply_pending_steer_to_tool_results(messages, 1)
if not agent.quiet_mode:
if not agent.quiet_mode and getattr(agent, "tool_progress_mode", "all") != "off":
if agent.verbose_logging:
print(f" ✅ Tool {i} completed in {tool_duration:.2f}s")
print(agent._wrap_verbose("Result: ", function_result))

View File

@@ -84,7 +84,7 @@ class AnthropicTransport(ProviderTransport):
to OpenAI finish_reason, and collects reasoning_details in provider_data.
"""
import json
from agent.anthropic_adapter import _to_plain_data
from agent.anthropic_adapter import _to_plain_data, _sanitize_replay_block
from agent.transports.types import ToolCall
strip_tool_prefix = kwargs.get("strip_tool_prefix", False)
@@ -94,14 +94,40 @@ class AnthropicTransport(ProviderTransport):
reasoning_parts = []
reasoning_details = []
tool_calls = []
# Verbatim, order-preserving copy of every content block in the turn.
# Anthropic signs each thinking block against the turn content that
# PRECEDES it at its position; when a turn interleaves thinking and
# tool_use (adaptive/interleaved thinking, Claude 4.6+), the parallel
# reasoning_details + tool_calls lists below lose that cross-type
# ordering. Replaying the latest assistant message in the wrong order
# invalidates the signatures -> HTTP 400 "thinking ... blocks in the
# latest assistant message cannot be modified". Preserve the exact
# block sequence here so the adapter can replay it unchanged. See
# tests/agent/test_anthropic_thinking_block_order.py.
ordered_blocks = []
for block in response.content:
block_dict = _to_plain_data(block)
clean_block = None
if isinstance(block_dict, dict):
# Sanitize at capture so output-only SDK fields (parsed_output,
# caller, citations=None, …) never persist to state.db and leak
# back as request input on replay → HTTP 400 "Extra inputs are
# not permitted". Defence-in-depth with the replay-side sanitize.
clean_block = _sanitize_replay_block(block_dict)
if clean_block is not None:
ordered_blocks.append(clean_block)
if block.type == "text":
text_parts.append(block.text)
elif block.type == "thinking":
reasoning_parts.append(block.thinking)
block_dict = _to_plain_data(block)
if isinstance(block_dict, dict):
elif block.type in ("thinking", "redacted_thinking"):
if block.type == "thinking":
reasoning_parts.append(block.thinking)
# Use the sanitized block (clean_block) for reasoning_details too,
# since _extract_preserved_thinking_blocks replays these on the
# non-ordered path. Falls back to raw only if sanitize dropped it.
if isinstance(clean_block, dict):
reasoning_details.append(clean_block)
elif isinstance(block_dict, dict):
reasoning_details.append(block_dict)
elif block.type == "tool_use":
name = block.name
@@ -130,6 +156,23 @@ class AnthropicTransport(ProviderTransport):
provider_data = {}
if reasoning_details:
provider_data["reasoning_details"] = reasoning_details
# Only worth carrying the ordered-blocks channel when the turn
# actually interleaves signed thinking with tool_use — that's the
# only shape the parallel lists reconstruct incorrectly. A turn that
# is purely text, or thinking-then-tools with a single leading
# thinking block, replays correctly without it.
_has_signed_thinking = any(
isinstance(b, dict)
and b.get("type") in ("thinking", "redacted_thinking")
and (b.get("signature") or b.get("data"))
for b in ordered_blocks
)
_has_tool_use = any(
isinstance(b, dict) and b.get("type") == "tool_use"
for b in ordered_blocks
)
if _has_signed_thinking and _has_tool_use:
provider_data["anthropic_content_blocks"] = ordered_blocks
return NormalizedResponse(
content="\n".join(text_parts) if text_parts else None,

View File

@@ -121,6 +121,18 @@ class NormalizedResponse:
pd = self.provider_data or {}
return pd.get("reasoning_details")
@property
def anthropic_content_blocks(self):
"""Verbatim, order-preserving Anthropic content blocks for a turn.
Present only when an Anthropic turn interleaves signed thinking with
tool_use — the one shape the parallel reasoning_details + tool_calls
lists reconstruct in the wrong order, invalidating thinking-block
signatures on replay. See agent/transports/anthropic.py.
"""
pd = self.provider_data or {}
return pd.get("anthropic_content_blocks")
@property
def codex_reasoning_items(self):
pd = self.provider_data or {}

View File

@@ -11,7 +11,8 @@
"tauri": "tauri",
"tauri:dev": "tauri dev",
"tauri:build": "tauri build",
"tauri:build:debug": "tauri build --debug"
"tauri:build:debug": "tauri build --debug",
"typecheck": "tsc -p . --noEmit"
},
"dependencies": {
"@nous-research/ui": "0.16.0",
@@ -40,7 +41,7 @@
"@types/react": "^19.2.14",
"@types/react-dom": "^19.2.3",
"@vitejs/plugin-react": "^5.2.0",
"typescript": "~5.9.3",
"typescript": "^6.0.3",
"vite": "^7.3.1"
}
}

View File

@@ -16,9 +16,8 @@
"noUnusedParameters": true,
"esModuleInterop": true,
"noFallthroughCasesInSwitch": true,
"baseUrl": ".",
"paths": {
"@/*": ["src/*"]
"@/*": ["./src/*"]
}
},
"include": ["src"],

View File

@@ -93,7 +93,7 @@ Run before opening a PR (lint may surface pre-existing warnings but must exit cl
```bash
npm run fix
npm run type-check
npm run typecheck
npm run lint
npm run test:desktop:all
```

View File

@@ -0,0 +1,101 @@
const path = require('node:path')
// Match the POSIX fallback surface used by the Python terminal environment.
// macOS apps launched from Finder/Dock often inherit only /usr/bin:/bin:/usr/sbin:/sbin,
// which misses Apple Silicon Homebrew and user-installed CLI tools such as codex.
const POSIX_SANE_PATH_ENTRIES = Object.freeze([
'/opt/homebrew/bin',
'/opt/homebrew/sbin',
'/usr/local/sbin',
'/usr/local/bin',
'/usr/sbin',
'/usr/bin',
'/sbin',
'/bin'
])
function delimiterForPlatform(platform = process.platform) {
return platform === 'win32' ? ';' : ':'
}
function pathModuleForPlatform(platform = process.platform) {
return platform === 'win32' ? path.win32 : path.posix
}
function pathEnvKey(env = process.env, platform = process.platform) {
if (platform !== 'win32') return 'PATH'
return Object.keys(env || {}).find(key => key.toUpperCase() === 'PATH') || 'PATH'
}
function currentPathValue(env = process.env, platform = process.platform) {
const key = pathEnvKey(env, platform)
return env?.[key] || ''
}
function appendUniquePathEntries(entries, { delimiter = path.delimiter } = {}) {
const seen = new Set()
const ordered = []
for (const entry of entries) {
if (!entry) continue
const parts = Array.isArray(entry) ? entry : String(entry).split(delimiter)
for (const part of parts) {
if (!part || seen.has(part)) continue
seen.add(part)
ordered.push(part)
}
}
return ordered.join(delimiter)
}
function buildDesktopBackendPath({
hermesHome,
venvRoot,
currentPath = '',
platform = process.platform,
pathModule = pathModuleForPlatform(platform)
} = {}) {
const delimiter = delimiterForPlatform(platform)
const hermesNodeBin = hermesHome ? pathModule.join(hermesHome, 'node', 'bin') : null
const venvBin = venvRoot ? pathModule.join(venvRoot, platform === 'win32' ? 'Scripts' : 'bin') : null
const saneEntries = platform === 'win32' ? [] : POSIX_SANE_PATH_ENTRIES
return appendUniquePathEntries(
[hermesNodeBin, venvBin, currentPath, saneEntries],
{ delimiter }
)
}
function buildDesktopBackendEnv({
hermesHome,
pythonPathEntries = [],
venvRoot,
currentEnv = process.env,
platform = process.platform,
pathModule = pathModuleForPlatform(platform)
} = {}) {
const delimiter = delimiterForPlatform(platform)
const currentPythonPath = currentEnv?.PYTHONPATH || ''
const key = pathEnvKey(currentEnv, platform)
return {
PYTHONPATH: appendUniquePathEntries([...pythonPathEntries, currentPythonPath], { delimiter }),
[key]: buildDesktopBackendPath({
hermesHome,
venvRoot,
currentPath: currentPathValue(currentEnv, platform),
platform,
pathModule
})
}
}
module.exports = {
POSIX_SANE_PATH_ENTRIES,
appendUniquePathEntries,
buildDesktopBackendEnv,
buildDesktopBackendPath,
delimiterForPlatform,
pathEnvKey
}

View File

@@ -0,0 +1,95 @@
const test = require('node:test')
const assert = require('node:assert/strict')
const path = require('node:path')
const {
POSIX_SANE_PATH_ENTRIES,
appendUniquePathEntries,
buildDesktopBackendEnv,
buildDesktopBackendPath,
pathEnvKey
} = require('./backend-env.cjs')
test('desktop backend PATH adds Hermes-managed bins and missing POSIX sane entries', () => {
const result = buildDesktopBackendPath({
hermesHome: '/Users/test/.hermes',
venvRoot: '/Users/test/.hermes/hermes-agent/venv',
currentPath: '/usr/bin:/bin:/usr/sbin:/sbin:/usr/local/bin',
platform: 'darwin',
pathModule: path.posix
})
const entries = result.split(':')
assert.equal(entries[0], '/Users/test/.hermes/node/bin')
assert.equal(entries[1], '/Users/test/.hermes/hermes-agent/venv/bin')
assert.ok(entries.includes('/opt/homebrew/bin'), 'Apple Silicon Homebrew bin is added')
assert.ok(entries.includes('/opt/homebrew/sbin'), 'Apple Silicon Homebrew sbin is added')
assert.ok(entries.includes('/usr/local/sbin'), 'missing standard sbin is added')
for (const expected of POSIX_SANE_PATH_ENTRIES) {
assert.ok(entries.includes(expected), `${expected} should be present`)
}
})
test('desktop backend PATH preserves first occurrence and avoids duplicates', () => {
const result = buildDesktopBackendPath({
hermesHome: '/Users/test/.hermes',
venvRoot: '/Users/test/.hermes/hermes-agent/venv',
currentPath: '/opt/homebrew/bin:/usr/bin:/opt/homebrew/bin:/bin',
platform: 'darwin',
pathModule: path.posix
})
const entries = result.split(':')
assert.equal(entries.filter(entry => entry === '/opt/homebrew/bin').length, 1)
assert.ok(
entries.indexOf('/opt/homebrew/bin') < entries.indexOf('/opt/homebrew/sbin'),
'existing Homebrew bin keeps its precedence over appended missing sane entries'
)
})
test('buildDesktopBackendEnv extends PYTHONPATH and backend PATH together', () => {
const env = buildDesktopBackendEnv({
hermesHome: '/Users/test/.hermes',
pythonPathEntries: ['/repo/hermes-agent'],
venvRoot: '/Users/test/.hermes/hermes-agent/venv',
currentEnv: {
PATH: '/usr/bin:/bin',
PYTHONPATH: '/existing/pythonpath'
},
platform: 'darwin',
pathModule: path.posix
})
assert.equal(env.PYTHONPATH, '/repo/hermes-agent:/existing/pythonpath')
assert.ok(env.PATH.startsWith('/Users/test/.hermes/node/bin:/Users/test/.hermes/hermes-agent/venv/bin:'))
assert.ok(env.PATH.includes('/opt/homebrew/bin'))
})
test('Windows PATH casing and delimiter are preserved without POSIX sane entries', () => {
const env = buildDesktopBackendEnv({
hermesHome: 'C:\\Users\\test\\AppData\\Local\\hermes',
pythonPathEntries: ['C:\\repo\\hermes-agent'],
venvRoot: 'C:\\Users\\test\\AppData\\Local\\hermes\\hermes-agent\\venv',
currentEnv: {
Path: 'C:\\Windows\\System32;C:\\Windows',
PYTHONPATH: 'C:\\existing\\pythonpath'
},
platform: 'win32',
pathModule: path.win32
})
assert.equal(pathEnvKey({ Path: 'x' }, 'win32'), 'Path')
assert.equal(env.PATH, undefined)
assert.ok(env.Path.startsWith('C:\\Users\\test\\AppData\\Local\\hermes\\node\\bin;'))
assert.ok(env.Path.includes('\\venv\\Scripts;'))
assert.ok(env.Path.includes(';C:\\Windows\\System32;C:\\Windows'))
assert.equal(env.Path.includes('/opt/homebrew/bin'), false)
})
test('appendUniquePathEntries drops empty entries and keeps first occurrence', () => {
assert.equal(
appendUniquePathEntries([':/a::/b', ['/a', '/c']], { delimiter: ':' }),
'/a:/b:/c'
)
})

View File

@@ -0,0 +1,66 @@
const _READY_RE = /^HERMES_DASHBOARD_READY port=(\d+)/m
/**
* Watch a child process's stdout for the `HERMES_DASHBOARD_READY port=<N>`
* line that web_server.py prints after uvicorn binds its socket.
*
* Returns the parsed port. Rejects if:
* - the child exits before emitting the line
* - the child emits an `error` event
* - no line arrives within the timeout
*
* A single `cleanup()` tears down every listener (data/exit/error/timeout)
* on every terminal path — resolve, reject, or timeout — so repeated
* backend spawns don't leak listener slots on the child.
*/
function waitForDashboardPort(child, timeoutMs = 45_000) {
return new Promise((resolve, reject) => {
let buf = ''
let done = false
function cleanup() {
if (done) return
done = true
clearTimeout(timer)
child.stdout.off('data', onData)
child.off('exit', onExit)
child.off('error', onError)
}
function onData(chunk) {
buf += chunk.toString()
let nl
while ((nl = buf.indexOf('\n')) !== -1) {
const line = buf.slice(0, nl)
buf = buf.slice(nl + 1)
const m = line.match(_READY_RE)
if (m) {
cleanup()
resolve(parseInt(m[1], 10))
return
}
}
}
function onExit(code, signal) {
cleanup()
reject(new Error(`Hermes backend: exited before port announcement (${signal || code})`))
}
function onError(err) {
cleanup()
reject(err)
}
const timer = setTimeout(() => {
cleanup()
reject(new Error(`Timed out waiting for Hermes backend port announcement (${timeoutMs}ms)`))
}, timeoutMs)
child.stdout.on('data', onData)
child.on('exit', onExit)
child.on('error', onError)
})
}
module.exports = { waitForDashboardPort }

View File

@@ -0,0 +1,99 @@
/**
* Helpers for local dashboard session-token discovery.
*
* The desktop main process can pass HERMES_DASHBOARD_SESSION_TOKEN when it
* spawns the local dashboard, but the dashboard is the source of truth for the
* token it actually serves to the renderer. If those drift, HTTP readiness
* probes still pass while /api/ws rejects the renderer's token.
*/
const DEFAULT_TOKEN_FETCH_TIMEOUT_MS = 3_000
async function fetchPublicText(url, options = {}) {
const { protocol } = new URL(url)
if (protocol !== 'http:' && protocol !== 'https:') {
throw new Error(`Unsupported Hermes backend URL protocol: ${protocol}`)
}
const timeoutMs = options.timeoutMs ?? DEFAULT_TOKEN_FETCH_TIMEOUT_MS
const res = await fetch(url, { signal: AbortSignal.timeout(timeoutMs) }).catch(error => {
if (error.name === 'TimeoutError') {
throw new Error(`Timed out connecting to Hermes backend after ${timeoutMs}ms`)
}
throw error
})
const text = await res.text()
if (!res.ok) throw new Error(`${res.status}: ${text || res.statusText}`)
return text
}
function extractInjectedDashboardToken(html) {
const match = /window\.__HERMES_SESSION_TOKEN__\s*=\s*("(?:\\.|[^"\\])*")/.exec(String(html || ''))
if (!match) return null
try {
return JSON.parse(match[1])
} catch {
return null
}
}
function dashboardIndexUrl(baseUrl) {
return `${String(baseUrl || '').replace(/\/+$/, '')}/`
}
async function resolveServedDashboardToken(baseUrl, fallbackToken, options = {}) {
const fetchText = options.fetchText || fetchPublicText
const html = await fetchText(dashboardIndexUrl(baseUrl), {
timeoutMs: options.timeoutMs ?? DEFAULT_TOKEN_FETCH_TIMEOUT_MS
})
const servedToken = extractInjectedDashboardToken(html)
if (servedToken && servedToken !== fallbackToken && typeof options.rememberLog === 'function') {
options.rememberLog('[boot] dashboard served a different session token; using served token for WebSocket auth')
}
return servedToken || fallbackToken
}
/**
* A served token that differs from our spawn token while our child is DEAD
* came from a process we did not spawn (orphan/port squatter that satisfied
* the public /api/status readiness probe). With a live child the mismatch is
* benign: our own backend regenerated the token because the env pin did not
* survive the spawn.
*/
function isForeignBackendToken({ servedToken, spawnToken, childAlive }) {
return Boolean(servedToken) && servedToken !== spawnToken && !childAlive
}
/**
* Resolve the token the backend actually serves, adopting benign drift and
* failing loudly on a foreign backend. `childAlive` is a thunk so liveness is
* sampled after the fetch, not before.
*/
async function adoptServedDashboardToken(baseUrl, spawnToken, { childAlive, label = 'Hermes backend', ...options }) {
const servedToken = await resolveServedDashboardToken(baseUrl, spawnToken, options).catch(error => {
options.rememberLog?.(`[boot] could not read served dashboard token (${label}): ${error.message}`)
return spawnToken
})
if (isForeignBackendToken({ servedToken, spawnToken, childAlive: childAlive() })) {
throw new Error(
`${label} exited and ${dashboardIndexUrl(baseUrl)} is served by a process we did not spawn; refusing its session token.`
)
}
return servedToken
}
module.exports = {
DEFAULT_TOKEN_FETCH_TIMEOUT_MS,
adoptServedDashboardToken,
dashboardIndexUrl,
extractInjectedDashboardToken,
fetchPublicText,
isForeignBackendToken,
resolveServedDashboardToken
}

View File

@@ -0,0 +1,142 @@
/**
* Tests for electron/dashboard-token.cjs.
*
* Run with: node --test electron/dashboard-token.test.cjs
* (Wired into npm test:desktop:platforms in package.json.)
*/
const test = require('node:test')
const assert = require('node:assert/strict')
const {
adoptServedDashboardToken,
dashboardIndexUrl,
extractInjectedDashboardToken,
fetchPublicText,
isForeignBackendToken,
resolveServedDashboardToken
} = require('./dashboard-token.cjs')
test('extractInjectedDashboardToken reads the JSON-encoded dashboard token', () => {
const html = '<script>window.__HERMES_SESSION_TOKEN__="served-token";window.__HERMES_BASE_PATH__=""</script>'
assert.equal(extractInjectedDashboardToken(html), 'served-token')
})
test('extractInjectedDashboardToken handles escaped token strings', () => {
const html = '<script>window.__HERMES_SESSION_TOKEN__="served\\\\token\\"quoted";</script>'
assert.equal(extractInjectedDashboardToken(html), 'served\\token"quoted')
})
test('extractInjectedDashboardToken returns null for missing or malformed values', () => {
assert.equal(extractInjectedDashboardToken('<html></html>'), null)
assert.equal(extractInjectedDashboardToken('<script>window.__HERMES_SESSION_TOKEN__={bad}</script>'), null)
})
test('dashboardIndexUrl preserves dashboard path prefixes', () => {
assert.equal(dashboardIndexUrl('http://127.0.0.1:9120'), 'http://127.0.0.1:9120/')
assert.equal(dashboardIndexUrl('https://host.example/hermes/'), 'https://host.example/hermes/')
})
test('resolveServedDashboardToken uses the served token and logs when it differs', async () => {
const logs = []
const token = await resolveServedDashboardToken('http://127.0.0.1:9120', 'spawn-token', {
fetchText: async url => {
assert.equal(url, 'http://127.0.0.1:9120/')
return '<script>window.__HERMES_SESSION_TOKEN__="served-token";</script>'
},
rememberLog: line => logs.push(line)
})
assert.equal(token, 'served-token')
assert.equal(logs.length, 1)
assert.match(logs[0], /served a different session token/)
})
test('resolveServedDashboardToken falls back when the served HTML has no token', async () => {
const token = await resolveServedDashboardToken('http://127.0.0.1:9120', 'spawn-token', {
fetchText: async () => '<html></html>',
rememberLog: () => {
throw new Error('should not log when no served token is present')
}
})
assert.equal(token, 'spawn-token')
})
test('resolveServedDashboardToken does not log when served token matches fallback', async () => {
const token = await resolveServedDashboardToken('http://127.0.0.1:9120', 'same-token', {
fetchText: async () => '<script>window.__HERMES_SESSION_TOKEN__="same-token";</script>',
rememberLog: () => {
throw new Error('should not log when token already matches')
}
})
assert.equal(token, 'same-token')
})
test('resolveServedDashboardToken propagates fetch errors so callers can fall back explicitly', async () => {
await assert.rejects(
() =>
resolveServedDashboardToken('http://127.0.0.1:9120', 'spawn-token', {
fetchText: async () => {
throw new Error('boom')
}
}),
/boom/
)
})
test('fetchPublicText rejects unsupported protocols', async () => {
await assert.rejects(() => fetchPublicText('file:///tmp/index.html'), /Unsupported Hermes backend URL protocol/)
})
test('isForeignBackendToken only flags a mismatched token from a dead child', () => {
const cases = [
[{ servedToken: 'other', spawnToken: 'mine', childAlive: false }, true],
// Live child + drift = our backend regenerated the token (env pin lost).
[{ servedToken: 'other', spawnToken: 'mine', childAlive: true }, false],
[{ servedToken: 'mine', spawnToken: 'mine', childAlive: false }, false],
[{ servedToken: 'mine', spawnToken: 'mine', childAlive: true }, false],
[{ servedToken: null, spawnToken: 'mine', childAlive: false }, false],
[{ servedToken: '', spawnToken: 'mine', childAlive: false }, false]
]
for (const [input, expected] of cases) {
assert.equal(isForeignBackendToken(input), expected, JSON.stringify(input))
}
})
test('adoptServedDashboardToken adopts drift from a live child', async () => {
const token = await adoptServedDashboardToken('http://127.0.0.1:9120', 'spawn-token', {
childAlive: () => true,
fetchText: async () => '<script>window.__HERMES_SESSION_TOKEN__="served-token";</script>'
})
assert.equal(token, 'served-token')
})
test('adoptServedDashboardToken refuses a foreign token when our child is dead', async () => {
await assert.rejects(
() =>
adoptServedDashboardToken('http://127.0.0.1:9120', 'spawn-token', {
childAlive: () => false,
fetchText: async () => '<script>window.__HERMES_SESSION_TOKEN__="squatter-token";</script>',
label: 'Hermes backend for profile "work"'
}),
/profile "work".*process we did not spawn/
)
})
test('adoptServedDashboardToken falls back to the spawn token when the fetch fails', async () => {
const logs = []
const token = await adoptServedDashboardToken('http://127.0.0.1:9120', 'spawn-token', {
childAlive: () => true,
fetchText: async () => {
throw new Error('boom')
},
rememberLog: line => logs.push(line)
})
assert.equal(token, 'spawn-token')
assert.equal(logs.length, 1)
assert.match(logs[0], /could not read served dashboard token \(Hermes backend\): boom/)
})

View File

@@ -0,0 +1,109 @@
'use strict'
const fs = require('node:fs')
const path = require('node:path')
const { resolveDirectoryForIpc } = require('./hardening.cjs')
const FS_READDIR_STAT_CONCURRENCY = 16
// Always-hidden noise (covers non-git projects too; gitignore catches many of
// these, but the project tree should keep the same hygiene without one).
const FS_READDIR_HIDDEN = new Set([
'.git',
'.hg',
'.svn',
'.cache',
'.next',
'.turbo',
'.venv',
'__pycache__',
'build',
'dist',
'node_modules',
'target',
'venv'
])
function direntIsDirectory(dirent) {
return typeof dirent.isDirectory === 'function' && dirent.isDirectory()
}
function direntIsFile(dirent) {
return typeof dirent.isFile === 'function' && dirent.isFile()
}
function direntIsSymbolicLink(dirent) {
return typeof dirent.isSymbolicLink === 'function' && dirent.isSymbolicLink()
}
function shouldStatDirent(dirent) {
if (direntIsDirectory(dirent)) return false
return direntIsSymbolicLink(dirent) || !direntIsFile(dirent)
}
async function entryForDirent(dirent, resolved, fsImpl) {
const fullPath = path.join(resolved, dirent.name)
let isDirectory = direntIsDirectory(dirent)
if (!isDirectory && shouldStatDirent(dirent)) {
try {
isDirectory = (await fsImpl.promises.stat(fullPath)).isDirectory()
} catch {
isDirectory = false
}
}
return { name: dirent.name, path: fullPath, isDirectory }
}
async function mapWithStatConcurrency(items, mapper) {
const results = new Array(items.length)
let nextIndex = 0
async function runWorker() {
while (nextIndex < items.length) {
const index = nextIndex
nextIndex += 1
results[index] = await mapper(items[index])
}
}
const workerCount = Math.min(FS_READDIR_STAT_CONCURRENCY, items.length)
const workers = Array.from({ length: workerCount }, () => runWorker())
await Promise.all(workers)
return results
}
async function readDirForIpc(dirPath, options = {}) {
const fsImpl = options.fs || fs
let resolved
try {
;({ resolvedPath: resolved } = await resolveDirectoryForIpc(dirPath, {
fs: fsImpl,
purpose: 'Directory read'
}))
} catch (error) {
return { entries: [], error: error?.code || 'read-error' }
}
try {
const dirents = await fsImpl.promises.readdir(resolved, { withFileTypes: true })
const visibleDirents = dirents.filter(dirent => !FS_READDIR_HIDDEN.has(dirent.name))
const entries = await mapWithStatConcurrency(visibleDirents, dirent =>
entryForDirent(dirent, resolved, fsImpl)
)
entries.sort((a, b) => Number(b.isDirectory) - Number(a.isDirectory) || a.name.localeCompare(b.name))
return { entries }
} catch (error) {
return { entries: [], error: error?.code || 'read-error' }
}
}
module.exports = {
readDirForIpc
}

View File

@@ -0,0 +1,364 @@
'use strict'
const assert = require('node:assert/strict')
const fs = require('node:fs')
const os = require('node:os')
const path = require('node:path')
const test = require('node:test')
const { pathToFileURL } = require('node:url')
const { readDirForIpc } = require('./fs-read-dir.cjs')
function mkTmpDir() {
return fs.mkdtempSync(path.join(os.tmpdir(), 'hermes-fs-read-dir-'))
}
function fakeDirent(name, flags = {}) {
return {
name,
isDirectory: () => Boolean(flags.directory),
isFile: () => Boolean(flags.file),
isSymbolicLink: () => Boolean(flags.symlink)
}
}
test('readDirForIpc hides noisy directories and files from the project tree', async () => {
const root = mkTmpDir()
try {
fs.mkdirSync(path.join(root, 'node_modules'))
fs.mkdirSync(path.join(root, 'src'))
fs.writeFileSync(path.join(root, 'target'), 'hidden file')
fs.writeFileSync(path.join(root, 'README.md'), 'visible file')
const result = await readDirForIpc(root)
assert.equal(result.error, undefined)
assert.deepEqual(
result.entries.map(entry => entry.name),
['src', 'README.md']
)
} finally {
fs.rmSync(root, { recursive: true, force: true })
}
})
test('readDirForIpc filters a hidden basename whether it is a file or directory', async () => {
const dirRoot = mkTmpDir()
const fileRoot = mkTmpDir()
try {
fs.mkdirSync(path.join(dirRoot, 'node_modules'))
fs.writeFileSync(path.join(dirRoot, 'visible.txt'), 'visible')
fs.writeFileSync(path.join(fileRoot, 'node_modules'), 'hidden file')
fs.writeFileSync(path.join(fileRoot, 'visible.txt'), 'visible')
assert.deepEqual(
(await readDirForIpc(dirRoot)).entries.map(entry => entry.name),
['visible.txt']
)
assert.deepEqual(
(await readDirForIpc(fileRoot)).entries.map(entry => entry.name),
['visible.txt']
)
} finally {
fs.rmSync(dirRoot, { recursive: true, force: true })
fs.rmSync(fileRoot, { recursive: true, force: true })
}
})
test('readDirForIpc returns directories before files and sorts by name within groups', async () => {
const root = mkTmpDir()
try {
fs.writeFileSync(path.join(root, 'z.txt'), 'z')
fs.mkdirSync(path.join(root, 'src'))
fs.writeFileSync(path.join(root, 'a.txt'), 'a')
fs.mkdirSync(path.join(root, 'lib'))
const result = await readDirForIpc(root)
assert.equal(result.error, undefined)
assert.deepEqual(
result.entries.map(entry => entry.name),
['lib', 'src', 'a.txt', 'z.txt']
)
} finally {
fs.rmSync(root, { recursive: true, force: true })
}
})
test('readDirForIpc accepts file URLs for directories', async () => {
const root = mkTmpDir()
try {
fs.mkdirSync(path.join(root, 'src'))
fs.writeFileSync(path.join(root, 'README.md'), 'visible file')
const result = await readDirForIpc(pathToFileURL(root).toString())
assert.equal(result.error, undefined)
assert.deepEqual(
result.entries.map(entry => entry.name),
['src', 'README.md']
)
} finally {
fs.rmSync(root, { recursive: true, force: true })
}
})
test('readDirForIpc returns invalid-path for blank or non-string input', async () => {
let readdirCalls = 0
const fsImpl = {
promises: {
readdir: async () => {
readdirCalls += 1
return []
}
}
}
assert.deepEqual(await readDirForIpc('', { fs: fsImpl }), { entries: [], error: 'invalid-path' })
assert.deepEqual(await readDirForIpc(' ', { fs: fsImpl }), { entries: [], error: 'invalid-path' })
assert.deepEqual(await readDirForIpc(null, { fs: fsImpl }), { entries: [], error: 'invalid-path' })
assert.equal(readdirCalls, 0)
})
test('readDirForIpc rejects Windows device paths before readdir', async () => {
let readdirCalls = 0
const fsImpl = {
promises: {
readdir: async () => {
readdirCalls += 1
return []
}
}
}
assert.deepEqual(await readDirForIpc('\\\\?\\C:\\secret', { fs: fsImpl }), {
entries: [],
error: 'device-path'
})
assert.equal(readdirCalls, 0)
})
test('readDirForIpc returns filesystem error codes instead of throwing', async () => {
const root = mkTmpDir()
try {
const result = await readDirForIpc(path.join(root, 'missing'))
assert.deepEqual(result, { entries: [], error: 'ENOENT' })
} finally {
fs.rmSync(root, { recursive: true, force: true })
}
})
test('readDirForIpc marks a symlink to a directory as a directory', async t => {
const root = mkTmpDir()
try {
fs.mkdirSync(path.join(root, 'actual-dir'))
try {
fs.symlinkSync(path.join(root, 'actual-dir'), path.join(root, 'linked-dir'), 'dir')
} catch (error) {
if (error?.code === 'EPERM' || error?.code === 'EACCES') {
t.skip(`symlink creation is not permitted on this platform (${error.code})`)
return
}
throw error
}
const result = await readDirForIpc(root)
const linked = result.entries.find(entry => entry.name === 'linked-dir')
assert.equal(result.error, undefined)
assert.equal(linked?.isDirectory, true)
} finally {
fs.rmSync(root, { recursive: true, force: true })
}
})
test('readDirForIpc marks a Windows junction to a directory as a directory', async t => {
if (process.platform !== 'win32') {
t.skip('junctions are a Windows-specific symlink type')
return
}
const root = mkTmpDir()
try {
fs.mkdirSync(path.join(root, 'actual-dir'))
try {
fs.symlinkSync(path.join(root, 'actual-dir'), path.join(root, 'junction-dir'), 'junction')
} catch (error) {
if (error?.code === 'EPERM' || error?.code === 'EACCES') {
t.skip(`junction creation is not permitted on this platform (${error.code})`)
return
}
throw error
}
const result = await readDirForIpc(root)
const junction = result.entries.find(entry => entry.name === 'junction-dir')
assert.equal(result.error, undefined)
assert.equal(junction?.isDirectory, true)
} finally {
fs.rmSync(root, { recursive: true, force: true })
}
})
test('readDirForIpc allows expanding symlink or junction directories outside the project root', async t => {
const root = mkTmpDir()
const outside = mkTmpDir()
try {
fs.writeFileSync(path.join(outside, 'outside.txt'), 'ok')
const linkPath = path.join(root, 'outside-link')
try {
fs.symlinkSync(outside, linkPath, process.platform === 'win32' ? 'junction' : 'dir')
} catch (error) {
if (error?.code === 'EPERM' || error?.code === 'EACCES') {
t.skip(`directory symlink creation is not permitted on this platform (${error.code})`)
return
}
throw error
}
const result = await readDirForIpc(linkPath)
assert.equal(result.error, undefined)
assert.deepEqual(result.entries, [
{ name: 'outside.txt', path: path.join(linkPath, 'outside.txt'), isDirectory: false }
])
} finally {
fs.rmSync(root, { recursive: true, force: true })
fs.rmSync(outside, { recursive: true, force: true })
}
})
test('readDirForIpc stats symbolic links and unknown entries without dropping the whole listing', async () => {
const input = path.join('virtual-root')
const resolved = path.resolve(input)
const statCalls = []
const fsImpl = {
promises: {
readdir: async () => [
fakeDirent('unknown-entry'),
fakeDirent('linked-dir', { symlink: true }),
fakeDirent('broken-link', { symlink: true }),
fakeDirent('plain.txt', { file: true })
],
stat: async fullPath => {
if (fullPath === resolved) {
return { isDirectory: () => true }
}
statCalls.push(fullPath)
if (fullPath.endsWith(`${path.sep}linked-dir`)) {
return { isDirectory: () => true }
}
throw Object.assign(new Error('gone'), { code: 'ENOENT' })
}
}
}
const result = await readDirForIpc(input, { fs: fsImpl })
assert.equal(result.error, undefined)
assert.deepEqual(
statCalls.sort(),
[path.join(resolved, 'broken-link'), path.join(resolved, 'linked-dir'), path.join(resolved, 'unknown-entry')].sort()
)
assert.deepEqual(result.entries, [
{ name: 'linked-dir', path: path.join(resolved, 'linked-dir'), isDirectory: true },
{ name: 'broken-link', path: path.join(resolved, 'broken-link'), isDirectory: false },
{ name: 'plain.txt', path: path.join(resolved, 'plain.txt'), isDirectory: false },
{ name: 'unknown-entry', path: path.join(resolved, 'unknown-entry'), isDirectory: false }
])
})
test('readDirForIpc bounds concurrent stats while preserving complete sorted output', async () => {
const input = path.join('virtual-root')
const resolved = path.resolve(input)
const names = Array.from({ length: 105 }, (_, index) => `entry-${String(104 - index).padStart(3, '0')}`)
const failedName = 'entry-100'
const directoryNames = new Set(names.filter((_, index) => index % 10 === 4))
const successfulDirectoryNames = new Set([...directoryNames].filter(name => name !== failedName))
const statCalls = []
let active = 0
let peak = 0
let releaseStats
let markFirstStatStarted
const statsReleased = new Promise(resolve => {
releaseStats = resolve
})
const firstStatStarted = new Promise(resolve => {
markFirstStatStarted = resolve
})
const fsImpl = {
promises: {
readdir: async () => [
fakeDirent('node_modules', { symlink: true }),
...names.map((name, index) => fakeDirent(name, { symlink: index % 2 === 0 }))
],
stat: async fullPath => {
if (fullPath === resolved) {
return { isDirectory: () => true }
}
statCalls.push(fullPath)
active += 1
peak = Math.max(peak, active)
markFirstStatStarted()
await statsReleased
active -= 1
const name = path.basename(fullPath)
if (name === failedName) {
throw Object.assign(new Error('gone'), { code: 'ENOENT' })
}
return { isDirectory: () => successfulDirectoryNames.has(name) }
}
}
}
const resultPromise = readDirForIpc(input, { fs: fsImpl })
await firstStatStarted
await new Promise(resolve => setImmediate(resolve))
releaseStats()
const result = await resultPromise
const expectedNames = [
...names.filter(name => successfulDirectoryNames.has(name)).sort(),
...names.filter(name => !successfulDirectoryNames.has(name)).sort()
]
assert.equal(result.error, undefined)
assert.equal(result.entries.length, names.length)
assert.equal(statCalls.length, names.length)
assert.equal(statCalls.some(fullPath => fullPath.endsWith(`${path.sep}node_modules`)), false)
assert.ok(peak > 1, `expected concurrent stats, observed peak ${peak}`)
assert.ok(peak <= 16, `expected at most 16 concurrent stats, observed peak ${peak}`)
assert.deepEqual(
result.entries.map(entry => entry.name),
expectedNames
)
assert.equal(result.entries.find(entry => entry.name === failedName)?.isDirectory, false)
assert.equal(
result.entries.filter(entry => entry.isDirectory).length,
successfulDirectoryNames.size
)
})

View File

@@ -0,0 +1,54 @@
'use strict'
const fs = require('node:fs')
const path = require('node:path')
const { resolveRequestedPathForIpc } = require('./hardening.cjs')
function findGitRoot(start, fsImpl = fs) {
let dir = start
for (let i = 0; i < 50; i += 1) {
try {
if (fsImpl.existsSync(path.join(dir, '.git'))) {
return dir
}
} catch {
return null
}
const parent = path.dirname(dir)
if (parent === dir) {
return null
}
dir = parent
}
return null
}
async function gitRootForIpc(startPath, options = {}) {
const fsImpl = options.fs || fs
let resolved
try {
resolved = resolveRequestedPathForIpc(startPath, { purpose: 'Git root' })
} catch {
return null
}
try {
const stat = await fsImpl.promises.stat(resolved)
const start = stat.isDirectory() ? resolved : path.dirname(resolved)
return findGitRoot(start, fsImpl)
} catch {
return findGitRoot(resolved, fsImpl)
}
}
module.exports = {
findGitRoot,
gitRootForIpc
}

View File

@@ -0,0 +1,40 @@
'use strict'
const assert = require('node:assert/strict')
const fs = require('node:fs')
const os = require('node:os')
const path = require('node:path')
const test = require('node:test')
const { pathToFileURL } = require('node:url')
const { gitRootForIpc } = require('./git-root.cjs')
function mkTmpDir() {
return fs.mkdtempSync(path.join(os.tmpdir(), 'hermes-git-root-'))
}
test('gitRootForIpc returns null for invalid and device paths', async () => {
assert.equal(await gitRootForIpc(''), null)
assert.equal(await gitRootForIpc(' '), null)
assert.equal(await gitRootForIpc(null), null)
assert.equal(await gitRootForIpc('\\\\?\\C:\\secret'), null)
assert.equal(await gitRootForIpc('file:///%E0%A4%A'), null)
})
test('gitRootForIpc resolves directories files missing descendants and file URLs', async t => {
const root = mkTmpDir()
t.after(() => fs.rmSync(root, { recursive: true, force: true }))
const gitDir = path.join(root, '.git')
const srcDir = path.join(root, 'src')
const filePath = path.join(srcDir, 'index.ts')
fs.mkdirSync(gitDir)
fs.mkdirSync(srcDir)
fs.writeFileSync(filePath, 'export {}\n', 'utf8')
assert.equal(await gitRootForIpc(root), root)
assert.equal(await gitRootForIpc(srcDir), root)
assert.equal(await gitRootForIpc(filePath), root)
assert.equal(await gitRootForIpc(pathToFileURL(filePath).toString()), root)
assert.equal(await gitRootForIpc(path.join(srcDir, 'missing.ts')), root)
})

View File

@@ -1,4 +1,5 @@
const fs = require('node:fs')
const os = require('node:os')
const path = require('node:path')
const { fileURLToPath } = require('node:url')
@@ -106,71 +107,162 @@ function sensitiveFileBlockReason(filePath) {
return null
}
function resolveRequestedFilePath(filePath, baseDir = process.cwd(), purpose = 'File read') {
const raw = String(filePath || '').trim()
function ipcPathError(code, message) {
const error = new Error(message)
error.code = code
return error
}
function rejectUnsafePathSyntax(filePath, purpose = 'File read') {
if (typeof filePath !== 'string') {
throw ipcPathError('invalid-path', `${purpose} failed: file path is required.`)
}
const raw = filePath.trim()
if (!raw) {
throw new Error(`${purpose} failed: file path is required.`)
throw ipcPathError('invalid-path', `${purpose} failed: file path is required.`)
}
if (raw.includes('\0')) {
throw new Error(`${purpose} failed: file path is invalid.`)
throw ipcPathError('invalid-path', `${purpose} failed: file path is invalid.`)
}
const normalized = raw.replace(/\\/g, '/').toLowerCase()
if (
normalized.startsWith('//?/') ||
normalized.startsWith('//./') ||
normalized.startsWith('globalroot/device/') ||
normalized.includes('/globalroot/device/')
) {
throw ipcPathError('device-path', `${purpose} blocked: Windows device paths are not allowed.`)
}
return raw
}
function resolveRequestedPathForIpc(filePath, options = {}) {
const purpose = String(options.purpose || 'File read')
let raw = rejectUnsafePathSyntax(filePath, purpose)
// Gateway-reported cwds (config `terminal.cwd`, remote sessions) routinely
// arrive as `~/...`. Node's fs has no shell — without expansion the path
// resolves under process.cwd() and every read "ENOENT"s forever.
if (raw === '~' || raw.startsWith('~/') || raw.startsWith('~\\')) {
raw = path.join(os.homedir(), raw.slice(1))
}
if (/^file:/i.test(raw)) {
let resolvedPath
try {
return fileURLToPath(raw)
const parsed = new URL(raw)
if (parsed.protocol !== 'file:') {
throw new Error('not a file URL')
}
resolvedPath = fileURLToPath(parsed)
} catch {
throw new Error(`${purpose} failed: file URL is invalid.`)
throw ipcPathError('invalid-path', `${purpose} failed: file URL is invalid.`)
}
rejectUnsafePathSyntax(resolvedPath, purpose)
return path.resolve(resolvedPath)
}
const resolvedBase = path.resolve(String(baseDir || process.cwd()))
return path.resolve(resolvedBase, raw)
const baseInput = typeof options.baseDir === 'string' && options.baseDir.trim() ? options.baseDir : process.cwd()
const safeBaseInput = rejectUnsafePathSyntax(baseInput, purpose)
const resolvedBase = path.resolve(safeBaseInput)
rejectUnsafePathSyntax(resolvedBase, purpose)
const resolvedPath = path.resolve(resolvedBase, raw)
rejectUnsafePathSyntax(resolvedPath, purpose)
return resolvedPath
}
async function statForIpc(fsImpl, resolvedPath, purpose, typeLabel) {
try {
return await fsImpl.promises.stat(resolvedPath)
} catch (error) {
const code = error && typeof error === 'object' ? error.code : ''
if (code === 'ENOENT' || code === 'ENOTDIR') {
throw ipcPathError(code || 'ENOENT', `${purpose} failed: ${typeLabel} does not exist.`)
}
throw ipcPathError(code || 'read-error', `${purpose} failed: ${error instanceof Error ? error.message : String(error)}`)
}
}
async function realpathForIpc(fsImpl, resolvedPath, purpose) {
if (typeof fsImpl.promises.realpath !== 'function') {
return resolvedPath
}
try {
const realPath = await fsImpl.promises.realpath(resolvedPath)
rejectUnsafePathSyntax(realPath, purpose)
return realPath
} catch (error) {
const code = error && typeof error === 'object' ? error.code : ''
throw ipcPathError(code || 'read-error', `${purpose} failed: ${error instanceof Error ? error.message : String(error)}`)
}
}
function rejectSensitiveFilePath(filePath, purpose) {
const blockReason = sensitiveFileBlockReason(filePath)
if (blockReason) {
throw ipcPathError('sensitive-file', `${purpose} blocked for sensitive file: ${blockReason}`)
}
}
async function resolveDirectoryForIpc(dirPath, options = {}) {
const purpose = String(options.purpose || 'Directory read')
const fsImpl = options.fs || fs
const resolvedPath = resolveRequestedPathForIpc(dirPath, { baseDir: options.baseDir, purpose })
const stat = await statForIpc(fsImpl, resolvedPath, purpose, 'directory')
if (!stat.isDirectory()) {
throw ipcPathError('ENOTDIR', `${purpose} failed: path is not a directory.`)
}
const realPath = await realpathForIpc(fsImpl, resolvedPath, purpose)
return { realPath, resolvedPath, stat }
}
async function resolveReadableFileForIpc(filePath, options = {}) {
const purpose = String(options.purpose || 'File read')
const resolvedPath = resolveRequestedFilePath(filePath, options.baseDir, purpose)
const fsImpl = options.fs || fs
const resolvedPath = resolveRequestedPathForIpc(filePath, { baseDir: options.baseDir, purpose })
if (options.blockSensitive !== false) {
const blockReason = sensitiveFileBlockReason(resolvedPath)
if (blockReason) {
throw new Error(`${purpose} blocked for sensitive file: ${blockReason}`)
}
rejectSensitiveFilePath(resolvedPath, purpose)
}
let stat
try {
stat = await fs.promises.stat(resolvedPath)
} catch (error) {
const code = error && typeof error === 'object' ? error.code : ''
if (code === 'ENOENT' || code === 'ENOTDIR') {
throw new Error(`${purpose} failed: file does not exist.`)
}
throw new Error(`${purpose} failed: ${error instanceof Error ? error.message : String(error)}`)
}
const stat = await statForIpc(fsImpl, resolvedPath, purpose, 'file')
if (stat.isDirectory()) {
throw new Error(`${purpose} failed: path points to a directory.`)
throw ipcPathError('EISDIR', `${purpose} failed: path points to a directory.`)
}
if (!stat.isFile()) {
throw new Error(`${purpose} failed: only regular files can be read.`)
throw ipcPathError('EINVAL', `${purpose} failed: only regular files can be read.`)
}
const realPath = await realpathForIpc(fsImpl, resolvedPath, purpose)
if (options.blockSensitive !== false) {
rejectSensitiveFilePath(realPath, purpose)
}
const maxBytes = Number.isFinite(options.maxBytes) && Number(options.maxBytes) > 0 ? Number(options.maxBytes) : null
if (maxBytes && stat.size > maxBytes) {
throw new Error(`${purpose} failed: file is too large (${stat.size} bytes; limit ${maxBytes} bytes).`)
throw ipcPathError('EFBIG', `${purpose} failed: file is too large (${stat.size} bytes; limit ${maxBytes} bytes).`)
}
try {
await fs.promises.access(resolvedPath, fs.constants.R_OK)
await fsImpl.promises.access(resolvedPath, fs.constants.R_OK)
} catch {
throw new Error(`${purpose} failed: file is not readable.`)
throw ipcPathError('EACCES', `${purpose} failed: file is not readable.`)
}
return { resolvedPath, stat }
return { realPath, resolvedPath, stat }
}
module.exports = {
@@ -178,7 +270,10 @@ module.exports = {
DEFAULT_FETCH_TIMEOUT_MS,
TEXT_PREVIEW_SOURCE_MAX_BYTES,
encryptDesktopSecret,
rejectUnsafePathSyntax,
resolveDirectoryForIpc,
resolveReadableFileForIpc,
resolveRequestedPathForIpc,
resolveTimeoutMs,
sensitiveFileBlockReason
}

View File

@@ -8,11 +8,20 @@ const { pathToFileURL } = require('node:url')
const {
DEFAULT_FETCH_TIMEOUT_MS,
encryptDesktopSecret,
resolveDirectoryForIpc,
resolveReadableFileForIpc,
resolveRequestedPathForIpc,
resolveTimeoutMs,
sensitiveFileBlockReason
} = require('./hardening.cjs')
async function rejectsWithCode(promise, code) {
await assert.rejects(promise, error => {
assert.equal(error?.code, code)
return true
})
}
test('resolveTimeoutMs falls back to defaults and accepts overrides', () => {
assert.equal(resolveTimeoutMs(undefined), DEFAULT_FETCH_TIMEOUT_MS)
assert.equal(resolveTimeoutMs(0), DEFAULT_FETCH_TIMEOUT_MS)
@@ -51,6 +60,65 @@ test('sensitiveFileBlockReason blocks obvious secret file patterns', () => {
assert.match(String(sensitiveFileBlockReason('/tmp/server-cert.pem')), /\.pem/)
})
test('path helpers reject blank non-string NUL and Windows device syntax', async () => {
await rejectsWithCode(resolveReadableFileForIpc('', { purpose: 'File preview' }), 'invalid-path')
await rejectsWithCode(resolveReadableFileForIpc(' ', { purpose: 'File preview' }), 'invalid-path')
await rejectsWithCode(resolveReadableFileForIpc(null, { purpose: 'File preview' }), 'invalid-path')
await rejectsWithCode(resolveReadableFileForIpc(`safe${String.fromCharCode(0)}name.txt`), 'invalid-path')
const devicePaths = [
'\\\\?\\C:\\secret.txt',
'\\\\.\\C:\\secret.txt',
'\\\\?\\UNC\\server\\share\\secret.txt',
'GLOBALROOT/Device/HarddiskVolumeShadowCopy1/secret.txt'
]
for (const devicePath of devicePaths) {
assert.throws(
() => resolveRequestedPathForIpc(devicePath, { purpose: 'File preview' }),
error => {
assert.equal(error?.code, 'device-path')
return true
}
)
await rejectsWithCode(resolveReadableFileForIpc(devicePath, { purpose: 'File preview' }), 'device-path')
}
assert.throws(
() => resolveRequestedPathForIpc('file:///%E0%A4%A', { purpose: 'File preview' }),
error => {
assert.equal(error?.code, 'invalid-path')
return true
}
)
await rejectsWithCode(resolveReadableFileForIpc('file:///%E0%A4%A', { purpose: 'File preview' }), 'invalid-path')
})
test('resolveRequestedPathForIpc resolves relative paths from the trimmed base directory', () => {
const baseDir = path.join(os.tmpdir(), 'hermes-desktop-base')
assert.equal(
resolveRequestedPathForIpc('notes.txt', {
baseDir: ` ${baseDir} `,
purpose: 'File preview'
}),
path.resolve(baseDir, 'notes.txt')
)
})
test('resolveRequestedPathForIpc expands ~ to the home directory', () => {
assert.equal(resolveRequestedPathForIpc('~', { purpose: 'Directory read' }), path.resolve(os.homedir()))
assert.equal(
resolveRequestedPathForIpc('~/www/project', { purpose: 'Directory read' }),
path.resolve(os.homedir(), 'www/project')
)
// `~user` shorthand is NOT expanded — only the caller's own home.
assert.equal(
resolveRequestedPathForIpc('~other/secret', { baseDir: os.tmpdir(), purpose: 'Directory read' }),
path.resolve(os.tmpdir(), '~other/secret')
)
})
test('resolveReadableFileForIpc validates existence type size and sensitivity', async t => {
const tempDir = fs.mkdtempSync(path.join(os.tmpdir(), 'hermes-desktop-hardening-'))
t.after(() => fs.rmSync(tempDir, { recursive: true, force: true }))
@@ -71,6 +139,13 @@ test('resolveReadableFileForIpc validates existence type size and sensitivity',
})
assert.equal(fromFileUrl.resolvedPath, textPath)
const spacedPath = path.join(tempDir, 'notes with spaces.txt')
fs.writeFileSync(spacedPath, 'space ok', 'utf8')
const fromSpacedFileUrl = await resolveReadableFileForIpc(pathToFileURL(spacedPath).toString(), {
purpose: 'File preview'
})
assert.equal(fromSpacedFileUrl.resolvedPath, spacedPath)
await assert.rejects(
resolveReadableFileForIpc('missing.txt', {
baseDir: tempDir,
@@ -114,3 +189,91 @@ test('resolveReadableFileForIpc validates existence type size and sensitivity',
})
assert.equal(envTemplate.resolvedPath, envTemplatePath)
})
test('resolveReadableFileForIpc blocks common sensitive files', async t => {
const tempDir = fs.mkdtempSync(path.join(os.tmpdir(), 'hermes-desktop-sensitive-'))
t.after(() => fs.rmSync(tempDir, { recursive: true, force: true }))
const sshDir = path.join(tempDir, '.ssh')
fs.mkdirSync(sshDir)
const blockedFiles = [
path.join(tempDir, '.env'),
path.join(tempDir, '.npmrc'),
path.join(sshDir, 'id_ed25519'),
path.join(tempDir, 'cert.pem'),
path.join(tempDir, 'cert.p12'),
path.join(tempDir, 'cert.pfx')
]
for (const filePath of blockedFiles) {
fs.writeFileSync(filePath, 'secret', 'utf8')
await rejectsWithCode(resolveReadableFileForIpc(filePath, { purpose: 'File preview' }), 'sensitive-file')
}
const allowed = path.join(tempDir, '.env.example')
fs.writeFileSync(allowed, 'EXAMPLE_TOKEN=value', 'utf8')
assert.equal((await resolveReadableFileForIpc(allowed, { purpose: 'File preview' })).resolvedPath, allowed)
})
test('resolveReadableFileForIpc blocks symlinks whose realpath is sensitive', async t => {
const tempDir = fs.mkdtempSync(path.join(os.tmpdir(), 'hermes-desktop-realpath-'))
t.after(() => fs.rmSync(tempDir, { recursive: true, force: true }))
const envPath = path.join(tempDir, '.env')
const linkPath = path.join(tempDir, 'safe-name.txt')
fs.writeFileSync(envPath, 'SECRET_TOKEN=123', 'utf8')
try {
fs.symlinkSync(envPath, linkPath, 'file')
} catch (error) {
if (error?.code === 'EPERM' || error?.code === 'EACCES') {
t.skip(`symlink creation is not permitted on this platform (${error.code})`)
return
}
throw error
}
await rejectsWithCode(resolveReadableFileForIpc(linkPath, { purpose: 'File preview' }), 'sensitive-file')
})
test('resolveDirectoryForIpc accepts directories and rejects invalid directory targets', async t => {
const tempDir = fs.mkdtempSync(path.join(os.tmpdir(), 'hermes-desktop-dir-'))
t.after(() => fs.rmSync(tempDir, { recursive: true, force: true }))
const directory = path.join(tempDir, 'project')
const filePath = path.join(tempDir, 'file.txt')
fs.mkdirSync(directory)
fs.writeFileSync(filePath, 'not a directory', 'utf8')
const resolved = await resolveDirectoryForIpc(directory)
assert.equal(resolved.resolvedPath, directory)
assert.equal(resolved.stat.isDirectory(), true)
await rejectsWithCode(resolveDirectoryForIpc(filePath), 'ENOTDIR')
await rejectsWithCode(resolveDirectoryForIpc(path.join(tempDir, 'missing')), 'ENOENT')
await rejectsWithCode(resolveDirectoryForIpc('\\\\?\\C:\\secret'), 'device-path')
})
test('resolveDirectoryForIpc accepts directory symlinks or junctions', async t => {
const tempDir = fs.mkdtempSync(path.join(os.tmpdir(), 'hermes-desktop-dir-link-'))
t.after(() => fs.rmSync(tempDir, { recursive: true, force: true }))
const directory = path.join(tempDir, 'actual-project')
const linkPath = path.join(tempDir, 'linked-project')
fs.mkdirSync(directory)
try {
fs.symlinkSync(directory, linkPath, process.platform === 'win32' ? 'junction' : 'dir')
} catch (error) {
if (error?.code === 'EPERM' || error?.code === 'EACCES') {
t.skip(`directory symlink creation is not permitted on this platform (${error.code})`)
return
}
throw error
}
const resolved = await resolveDirectoryForIpc(linkPath)
assert.equal(resolved.resolvedPath, linkPath)
assert.equal(resolved.stat.isDirectory(), true)
})

View File

@@ -22,15 +22,26 @@ const http = require('node:http')
const https = require('node:https')
const net = require('node:net')
const path = require('node:path')
const { fileURLToPath, pathToFileURL } = require('node:url')
const { pathToFileURL } = require('node:url')
const { execFileSync, spawn } = require('node:child_process')
const { detectRemoteDisplay, isWindowsBinaryPathInWsl, isWslEnvironment } = require('./bootstrap-platform.cjs')
const { runBootstrap } = require('./bootstrap-runner.cjs')
const { buildSessionWindowUrl, createSessionWindowRegistry } = require('./session-windows.cjs')
const {
buildSessionWindowUrl,
createSessionWindowRegistry,
SESSION_WINDOW_MIN_HEIGHT,
SESSION_WINDOW_MIN_WIDTH
} = require('./session-windows.cjs')
const { canImportHermesCli, verifyHermesCli } = require('./backend-probes.cjs')
const { probeGatewayWebSocket } = require('./gateway-ws-probe.cjs')
const { adoptServedDashboardToken } = require('./dashboard-token.cjs')
const { waitForDashboardPort } = require('./backend-ready.cjs')
const { serializeJsonBody, setJsonRequestHeaders } = require('./oauth-net-request.cjs')
const { fetchMarketplaceThemes, searchMarketplaceThemes } = require('./vscode-marketplace.cjs')
const { buildDesktopBackendEnv } = require('./backend-env.cjs')
const { readDirForIpc } = require('./fs-read-dir.cjs')
const { gitRootForIpc } = require('./git-root.cjs')
const { OFFICIAL_REPO_HTTPS_URL, isOfficialSshRemote } = require('./update-remote.cjs')
const {
buildPosixCleanupScript,
buildWindowsCleanupScript,
@@ -61,6 +72,7 @@ const {
TEXT_PREVIEW_SOURCE_MAX_BYTES,
encryptDesktopSecret: encryptDesktopSecretStrict,
resolveReadableFileForIpc,
resolveRequestedPathForIpc,
resolveTimeoutMs
} = require('./hardening.cjs')
@@ -86,6 +98,7 @@ try {
nodePty = require(nodePtyDir)
}
} catch {
console.log(`[terminal] failed to load node-pty from path ${nodePtyDir}`)
nodePty = null
nodePtyDir = null
}
@@ -98,8 +111,6 @@ if (USER_DATA_OVERRIDE) {
app.setPath('userData', resolvedUserData)
}
const PORT_FLOOR = 9120
const PORT_CEILING = 9199
const DEV_SERVER = process.env.HERMES_DESKTOP_DEV_SERVER
const IS_PACKAGED = app.isPackaged
const IS_MAC = process.platform === 'darwin'
@@ -333,10 +344,110 @@ const APP_ICON_PATHS = [
let rendererTitleBarTheme = null
const terminalSessions = new Map()
// Force the NATIVE window appearance (vibrancy material, titlebar, the
// pre-first-paint window background) to follow the APP theme instead of the
// OS appearance. With `vibrancy` set, macOS paints an NSVisualEffectView that
// tracks the window's effective appearance and ignores `backgroundColor` —
// so a dark-themed app on a light-mode Mac flashes a white material on every
// new window until the renderer covers it. The renderer reports its mode via
// 'hermes:native-theme' ('dark' | 'light' | 'system'); we pin
// nativeTheme.themeSource to it and persist the value so cold launches paint
// correctly before the renderer has even loaded.
const NATIVE_THEME_CONFIG_PATH = path.join(app.getPath('userData'), 'native-theme.json')
const THEME_SOURCES = new Set(['dark', 'light', 'system'])
function readPersistedThemeSource() {
try {
const parsed = JSON.parse(fs.readFileSync(NATIVE_THEME_CONFIG_PATH, 'utf8'))
if (parsed && THEME_SOURCES.has(parsed.themeSource)) {
return parsed.themeSource
}
} catch {
// Missing / malformed → follow the OS like a fresh install.
}
return 'system'
}
function writePersistedThemeSource(mode) {
try {
fs.mkdirSync(path.dirname(NATIVE_THEME_CONFIG_PATH), { recursive: true })
fs.writeFileSync(NATIVE_THEME_CONFIG_PATH, JSON.stringify({ themeSource: mode }, null, 2), 'utf8')
} catch (error) {
rememberLog(`[theme] write native theme failed: ${error.message}`)
}
}
nativeTheme.themeSource = readPersistedThemeSource()
// Window translucency (see-through window). One lever, 0100; 0 = off (the
// default). Mapped to the native window opacity so the desktop shows through
// the whole window. Persisted so a cold launch applies it at window creation,
// before the renderer reports its value. macOS + Windows only; `setOpacity` is
// a no-op on Linux. See store/translucency.
const TRANSLUCENCY_CONFIG_PATH = path.join(app.getPath('userData'), 'translucency.json')
function clampIntensity(value) {
const n = Math.round(Number(value))
return Number.isFinite(n) ? Math.min(100, Math.max(0, n)) : 0
}
function readPersistedTranslucency() {
try {
return clampIntensity(JSON.parse(fs.readFileSync(TRANSLUCENCY_CONFIG_PATH, 'utf8')).intensity)
} catch {
return 0
}
}
function writePersistedTranslucency(intensity) {
try {
fs.mkdirSync(path.dirname(TRANSLUCENCY_CONFIG_PATH), { recursive: true })
fs.writeFileSync(TRANSLUCENCY_CONFIG_PATH, JSON.stringify({ intensity }, null, 2), 'utf8')
} catch (error) {
rememberLog(`[translucency] write failed: ${error.message}`)
}
}
let translucencyIntensity = readPersistedTranslucency()
// Map the 0100 lever to a window opacity. Floor at 0.3 so the most see-through
// setting is still usable rather than nearly invisible. 0 → fully opaque.
function windowOpacity() {
return 1 - (translucencyIntensity / 100) * 0.7
}
// Re-apply translucency to a live window (runtime toggle, no recreation).
// `setOpacity` is a no-op on Linux, which is fine — it just stays opaque there.
function applyWindowTranslucency(win) {
if (!win || win.isDestroyed() || typeof win.setOpacity !== 'function') {
return
}
try {
win.setOpacity(windowOpacity())
} catch (error) {
rememberLog(`[translucency] apply failed: ${error.message}`)
}
}
function isHexColor(value) {
return typeof value === 'string' && /^#[0-9a-f]{6}$/i.test(value)
}
// Background color to paint a window with BEFORE its renderer loads, so a new
// (or reopened) window doesn't flash white/light in dark mode. Prefer the theme
// the renderer last reported; fall back to the OS preference on first launch.
function getWindowBackgroundColor() {
if (rendererTitleBarTheme && isHexColor(rendererTitleBarTheme.background)) {
return rendererTitleBarTheme.background
}
return nativeTheme.shouldUseDarkColors ? '#111111' : '#f7f7f7'
}
function getTitleBarOverlayOptions() {
if (IS_MAC) {
return { height: TITLEBAR_HEIGHT }
@@ -726,7 +837,7 @@ function openExternalUrl(rawUrl) {
if (parsed.protocol === 'file:') {
let localPath
try {
localPath = fileURLToPath(parsed.toString())
localPath = resolveRequestedPathForIpc(parsed.toString(), { purpose: 'Open external file' })
} catch {
return false
}
@@ -1149,10 +1260,14 @@ function findSystemPython() {
if (pyExe) {
for (const version of SUPPORTED_VERSIONS) {
try {
const out = execFileSync(pyExe, [`-${version}`, '-c', 'import sys; print(sys.executable)'], hiddenWindowsChildOptions({
encoding: 'utf8',
stdio: ['ignore', 'pipe', 'ignore']
}))
const out = execFileSync(
pyExe,
[`-${version}`, '-c', 'import sys; print(sys.executable)'],
hiddenWindowsChildOptions({
encoding: 'utf8',
stdio: ['ignore', 'pipe', 'ignore']
})
)
const candidate = out.trim()
if (candidate && fileExists(candidate)) return candidate
} catch {
@@ -1287,11 +1402,15 @@ function resolveUpdateRoot() {
function runGit(args, options = {}) {
return new Promise((resolve, reject) => {
const child = spawn(resolveGitBinary(), IS_WINDOWS ? ['-c', 'windows.appendAtomically=false', ...args] : args, hiddenWindowsChildOptions({
cwd: options.cwd,
env: { ...process.env, ...(options.env || {}), GIT_TERMINAL_PROMPT: '0' },
stdio: ['ignore', 'pipe', 'pipe']
}))
const child = spawn(
resolveGitBinary(),
IS_WINDOWS ? ['-c', 'windows.appendAtomically=false', ...args] : args,
hiddenWindowsChildOptions({
cwd: options.cwd,
env: { ...process.env, ...(options.env || {}), GIT_TERMINAL_PROMPT: '0' },
stdio: ['ignore', 'pipe', 'pipe']
})
)
let stdout = ''
let stderr = ''
@@ -1312,6 +1431,11 @@ function runGit(args, options = {}) {
const firstLine = text => (text || '').split('\n').find(Boolean) || ''
async function getOriginUrl(updateRoot) {
const origin = await runGit(['remote', 'get-url', 'origin'], { cwd: updateRoot })
return origin.code === 0 ? origin.stdout.trim() : ''
}
function emitUpdateProgress(payload) {
const merged = { stage: 'idle', message: '', percent: null, error: null, ...payload, at: Date.now() }
rememberLog(`[updates] ${merged.stage}: ${merged.message || merged.error || ''}`)
@@ -1331,7 +1455,9 @@ async function resolveHealedBranch(updateRoot, branch) {
return branch || 'main'
}
const probe = await runGit(['ls-remote', '--exit-code', '--heads', 'origin', branch], { cwd: updateRoot })
const originUrl = await getOriginUrl(updateRoot)
const remote = isOfficialSshRemote(originUrl) ? OFFICIAL_REPO_HTTPS_URL : 'origin'
const probe = await runGit(['ls-remote', '--exit-code', '--heads', remote, branch], { cwd: updateRoot })
if (probe.code !== 2) {
return branch
}
@@ -1359,6 +1485,40 @@ async function checkUpdates() {
}
branch = await resolveHealedBranch(updateRoot, branch)
const originUrl = await getOriginUrl(updateRoot)
if (isOfficialSshRemote(originUrl)) {
const git = args => runGit(args, { cwd: updateRoot }).then(r => r.stdout.trim())
const [currentSha, target, dirtyStr, currentBranch] = await Promise.all([
git(['rev-parse', 'HEAD']),
runGit(['ls-remote', OFFICIAL_REPO_HTTPS_URL, `refs/heads/${branch}`], { cwd: updateRoot }),
git(['status', '--porcelain']),
git(['rev-parse', '--abbrev-ref', 'HEAD'])
])
const targetSha = firstLine(target.stdout).split(/\s+/)[0] || ''
if (target.code !== 0 || !targetSha) {
return {
supported: true,
branch,
error: 'fetch-failed',
message: firstLine(target.stderr) || 'git ls-remote failed.',
hermesRoot: updateRoot,
fetchedAt: Date.now()
}
}
return {
supported: true,
branch,
currentBranch,
behind: currentSha && currentSha === targetSha ? 0 : 1,
currentSha,
targetSha,
commits: [],
dirty: dirtyStr.length > 0,
hermesRoot: updateRoot,
fetchedAt: Date.now()
}
}
const fetched = await runGit(['fetch', '--quiet', 'origin', branch], { cwd: updateRoot })
if (fetched.code !== 0) {
return {
@@ -1687,11 +1847,15 @@ function runStreamedUpdate(command, args, { cwd, env, stage } = {}) {
return new Promise(resolve => {
let child
try {
child = spawn(command, args, hiddenWindowsChildOptions({
cwd,
env: { ...process.env, ...(env || {}) },
stdio: ['ignore', 'pipe', 'pipe']
}))
child = spawn(
command,
args,
hiddenWindowsChildOptions({
cwd,
env: { ...process.env, ...(env || {}) },
stdio: ['ignore', 'pipe', 'pipe']
})
)
} catch (err) {
resolve({ code: 1, error: err.message })
return
@@ -2079,9 +2243,11 @@ function createPythonBackend(root, label, dashboardArgs, options = {}) {
label,
command: python,
args: ['-m', 'hermes_cli.main', ...dashboardArgs],
env: {
PYTHONPATH: [root, process.env.PYTHONPATH].filter(Boolean).join(path.delimiter)
},
env: buildDesktopBackendEnv({
hermesHome: HERMES_HOME,
pythonPathEntries: [root],
venvRoot: path.join(root, 'venv')
}),
root,
bootstrap: Boolean(options.bootstrap),
shell: false
@@ -2100,9 +2266,11 @@ function createActiveBackend(dashboardArgs) {
label: `Hermes at ${ACTIVE_HERMES_ROOT}`,
command: fileExists(venvPython) ? venvPython : findSystemPython(),
args: ['-m', 'hermes_cli.main', ...dashboardArgs],
env: {
PYTHONPATH: [ACTIVE_HERMES_ROOT, process.env.PYTHONPATH].filter(Boolean).join(path.delimiter)
},
env: buildDesktopBackendEnv({
hermesHome: HERMES_HOME,
pythonPathEntries: [ACTIVE_HERMES_ROOT],
venvRoot: VENV_ROOT
}),
root: ACTIVE_HERMES_ROOT,
bootstrap: true,
shell: false
@@ -2392,23 +2560,6 @@ async function ensureRuntime(backend) {
return backend
}
function isPortAvailable(port) {
return new Promise(resolve => {
const server = net.createServer()
server.once('error', () => resolve(false))
server.once('listening', () => {
server.close(() => resolve(true))
})
server.listen(port, '127.0.0.1')
})
}
async function pickPort() {
for (let port = PORT_FLOOR; port <= PORT_CEILING; port += 1) {
if (await isPortAvailable(port)) return port
}
throw new Error(`No free localhost port in ${PORT_FLOOR}-${PORT_CEILING}`)
}
function fetchJson(url, token, options = {}) {
return new Promise((resolve, reject) => {
@@ -2833,10 +2984,10 @@ async function resourceBufferFromUrl(rawUrl) {
const buffer = match[2] ? Buffer.from(encoded, 'base64') : Buffer.from(decodeURIComponent(encoded), 'utf8')
return { buffer, mimeType }
}
if (rawUrl.startsWith('file:')) {
const filePath = fileURLToPath(rawUrl)
const buffer = await fs.promises.readFile(filePath)
return { buffer, mimeType: mimeTypeForPath(filePath) }
if (/^file:/i.test(rawUrl)) {
const { resolvedPath } = await resolveReadableFileForIpc(rawUrl, { purpose: 'Image file' })
const buffer = await fs.promises.readFile(resolvedPath)
return { buffer, mimeType: mimeTypeForPath(resolvedPath) }
}
const parsed = new URL(rawUrl)
@@ -2914,11 +3065,13 @@ function expandUserPath(filePath) {
return value
}
function previewFileTarget(rawTarget, baseDir) {
async function previewFileTarget(rawTarget, baseDir) {
const raw = String(rawTarget || '').trim()
const base = baseDir ? path.resolve(expandUserPath(baseDir)) : resolveHermesCwd()
const filePath = raw.startsWith('file:') ? fileURLToPath(raw) : path.resolve(base, expandUserPath(raw))
let resolved = filePath
let resolved = resolveRequestedPathForIpc(/^file:/i.test(raw) ? raw : expandUserPath(raw), {
baseDir: base,
purpose: 'Preview target'
})
if (directoryExists(resolved)) {
resolved = path.join(resolved, 'index.html')
@@ -2929,6 +3082,8 @@ function previewFileTarget(rawTarget, baseDir) {
return null
}
;({ resolvedPath: resolved } = await resolveReadableFileForIpc(resolved, { purpose: 'Preview target' }))
const mimeType = mimeTypeForPath(resolved)
const metadata = previewFileMetadata(resolved, mimeType)
const isHtml = PREVIEW_HTML_EXTENSIONS.has(ext)
@@ -2974,7 +3129,7 @@ function previewUrlTarget(rawTarget) {
}
}
function normalizePreviewTarget(rawTarget, baseDir) {
async function normalizePreviewTarget(rawTarget, baseDir) {
const raw = String(rawTarget || '').trim()
if (!raw) {
@@ -2986,20 +3141,15 @@ function normalizePreviewTarget(rawTarget, baseDir) {
return previewUrlTarget(raw)
}
return previewFileTarget(raw, baseDir)
return await previewFileTarget(raw, baseDir)
} catch {
return null
}
}
function filePathFromPreviewUrl(rawUrl) {
const filePath = fileURLToPath(String(rawUrl || ''))
if (!fileExists(filePath)) {
throw new Error('Preview file is not readable')
}
return filePath
async function filePathFromPreviewUrl(rawUrl) {
const { resolvedPath } = await resolveReadableFileForIpc(String(rawUrl || ''), { purpose: 'Preview file' })
return resolvedPath
}
function sendPreviewFileChanged(payload) {
@@ -3009,8 +3159,8 @@ function sendPreviewFileChanged(payload) {
webContents.send('hermes:preview-file-changed', payload)
}
function watchPreviewFile(rawUrl) {
const filePath = filePathFromPreviewUrl(rawUrl)
async function watchPreviewFile(rawUrl) {
const filePath = await filePathFromPreviewUrl(rawUrl)
const watchDir = path.dirname(filePath)
const targetName = path.basename(filePath)
const id = crypto.randomBytes(12).toString('base64url')
@@ -4487,38 +4637,41 @@ async function spawnPoolBackend(profile, entry) {
}
}
const port = await pickPort()
const token = crypto.randomBytes(32).toString('base64url')
// --profile wins over the inherited HERMES_HOME env (see _apply_profile_override
// step 3 in hermes_cli/main.py), so the child re-homes to this profile.
const dashboardArgs = ['--profile', profile, 'dashboard', '--no-open', '--host', '127.0.0.1', '--port', String(port)]
// --port 0: the OS assigns an ephemeral port; the child announces it on stdout.
const dashboardArgs = ['--profile', profile, 'dashboard', '--no-open', '--host', '127.0.0.1', '--port', '0']
const backend = await ensureRuntime(resolveHermesBackend(dashboardArgs))
const hermesCwd = resolveHermesCwd()
const webDist = resolveWebDist()
rememberLog(`Starting Hermes backend for profile "${profile}" via ${backend.label}`)
const child = spawn(backend.command, backend.args, hiddenWindowsChildOptions({
cwd: hermesCwd,
env: {
...process.env,
HERMES_HOME,
...backend.env,
// Pin the gateway's tool/terminal cwd to the same directory we chose for
// the child process. Inherited TERMINAL_CWD (or a stale config bridge)
// can still point at the install dir even when spawn cwd is home.
TERMINAL_CWD: hermesCwd,
HERMES_DASHBOARD_SESSION_TOKEN: token,
// Marks this dashboard backend as desktop-spawned so it runs the cron
// scheduler tick loop (the gateway isn't running under the app).
HERMES_DESKTOP: '1',
HERMES_WEB_DIST: webDist
},
shell: backend.shell,
stdio: ['ignore', 'pipe', 'pipe']
}))
const child = spawn(
backend.command,
backend.args,
hiddenWindowsChildOptions({
cwd: hermesCwd,
env: {
...process.env,
HERMES_HOME,
...backend.env,
// Pin the gateway's tool/terminal cwd to the same directory we chose for
// the child process. Inherited TERMINAL_CWD (or a stale config bridge)
// can still point at the install dir even when spawn cwd is home.
TERMINAL_CWD: hermesCwd,
HERMES_DASHBOARD_SESSION_TOKEN: token,
// Marks this dashboard backend as desktop-spawned so it runs the cron
// scheduler tick loop (the gateway isn't running under the app).
HERMES_DESKTOP: '1',
HERMES_WEB_DIST: webDist
},
shell: backend.shell,
stdio: ['ignore', 'pipe', 'pipe']
})
)
entry.process = child
entry.port = port
entry.token = token
child.stdout.on('data', rememberLog)
@@ -4544,18 +4697,28 @@ async function spawnPoolBackend(profile, entry) {
}
})
// Discover the ephemeral port the child bound to
const port = await Promise.race([waitForDashboardPort(child), startFailed])
entry.port = port
const baseUrl = `http://127.0.0.1:${port}`
await Promise.race([waitForHermes(baseUrl, token), startFailed])
ready = true
const authToken = await adoptServedDashboardToken(baseUrl, token, {
childAlive: () => child.exitCode === null && !child.killed,
label: `Hermes backend for profile "${profile}"`,
rememberLog
})
entry.token = authToken
return {
baseUrl,
mode: 'local',
source: 'local',
authMode: 'token',
token,
token: authToken,
profile,
wsUrl: `ws://127.0.0.1:${port}/api/ws?token=${encodeURIComponent(token)}`,
wsUrl: `ws://127.0.0.1:${port}/api/ws?token=${encodeURIComponent(authToken)}`,
logs: hermesLog.slice(-80),
...getWindowState()
}
@@ -4677,10 +4840,9 @@ async function startHermes() {
}
}
await advanceBootProgress('backend.port', 'Finding an open local port', 16)
const port = await pickPort()
const token = crypto.randomBytes(32).toString('base64url')
const dashboardArgs = ['dashboard', '--no-open', '--host', '127.0.0.1', '--port', String(port)]
// --port 0: the OS assigns an ephemeral port; the child announces it on stdout.
const dashboardArgs = ['dashboard', '--no-open', '--host', '127.0.0.1', '--port', '0']
// Pin the desktop's chosen profile via the global --profile flag. This is
// deterministic (it wins over the sticky ~/.hermes/active_profile file) and
// resolves HERMES_HOME the same way `hermes -p <name>` does on the CLI. An
@@ -4698,30 +4860,34 @@ async function startHermes() {
await advanceBootProgress('backend.spawn', `Starting Hermes backend via ${backend.label}`, 84)
rememberLog(`Starting Hermes backend via ${backend.label}`)
hermesProcess = spawn(backend.command, backend.args, hiddenWindowsChildOptions({
cwd: hermesCwd,
env: {
...process.env,
// Explicitly pin HERMES_HOME for the child so Python's get_hermes_home()
// resolves to the SAME location our resolveHermesHome() picked. Without
// this pin, Python falls back to ~/.hermes on every platform — fine on
// mac/linux (where our default matches), but on Windows our default is
// %LOCALAPPDATA%\hermes, which differs from C:\Users\<u>\.hermes.
// Mismatch would split config / sessions / .env / logs across two
// directories. install.ps1 sets HERMES_HOME via setx; the desktop
// can't reliably do that, so we set it inline for every spawn.
HERMES_HOME,
...backend.env,
TERMINAL_CWD: hermesCwd,
HERMES_DASHBOARD_SESSION_TOKEN: token,
// Marks this dashboard backend as desktop-spawned so it runs the cron
// scheduler tick loop (the gateway isn't running under the app).
HERMES_DESKTOP: '1',
HERMES_WEB_DIST: webDist
},
shell: backend.shell,
stdio: ['ignore', 'pipe', 'pipe']
}))
hermesProcess = spawn(
backend.command,
backend.args,
hiddenWindowsChildOptions({
cwd: hermesCwd,
env: {
...process.env,
// Explicitly pin HERMES_HOME for the child so Python's get_hermes_home()
// resolves to the SAME location our resolveHermesHome() picked. Without
// this pin, Python falls back to ~/.hermes on every platform — fine on
// mac/linux (where our default matches), but on Windows our default is
// %LOCALAPPDATA%\hermes, which differs from C:\Users\<u>\.hermes.
// Mismatch would split config / sessions / .env / logs across two
// directories. install.ps1 sets HERMES_HOME via setx; the desktop
// can't reliably do that, so we set it inline for every spawn.
HERMES_HOME,
...backend.env,
TERMINAL_CWD: hermesCwd,
HERMES_DASHBOARD_SESSION_TOKEN: token,
// Marks this dashboard backend as desktop-spawned so it runs the cron
// scheduler tick loop (the gateway isn't running under the app).
HERMES_DESKTOP: '1',
HERMES_WEB_DIST: webDist
},
shell: backend.shell,
stdio: ['ignore', 'pipe', 'pipe']
})
)
hermesProcess.stdout.on('data', rememberLog)
hermesProcess.stderr.on('data', rememberLog)
@@ -4770,10 +4936,19 @@ async function startHermes() {
}
})
await advanceBootProgress('backend.port', 'Waiting for Hermes backend to launch', 86)
// Discover the ephemeral port the child bound to
const port = await Promise.race([waitForDashboardPort(hermesProcess), backendStartFailed])
const baseUrl = `http://127.0.0.1:${port}`
await advanceBootProgress('backend.wait', 'Waiting for Hermes backend to become ready', 90)
await Promise.race([waitForHermes(baseUrl, token), backendStartFailed])
backendReady = true
const authToken = await adoptServedDashboardToken(baseUrl, token, {
// The exit/error handlers null hermesProcess when the child dies.
childAlive: () => hermesProcess !== null && hermesProcess.exitCode === null && !hermesProcess.killed,
rememberLog
})
updateBootProgress({
phase: 'backend.ready',
message: 'Hermes backend is ready. Finalizing desktop startup',
@@ -4787,8 +4962,8 @@ async function startHermes() {
mode: 'local',
source: 'local',
authMode: 'token',
token,
wsUrl: `ws://127.0.0.1:${port}/api/ws?token=${encodeURIComponent(token)}`,
token: authToken,
wsUrl: `ws://127.0.0.1:${port}/api/ws?token=${encodeURIComponent(authToken)}`,
logs: hermesLog.slice(-80),
...getWindowState()
}
@@ -4851,21 +5026,29 @@ function focusWindow(win) {
}
// Open (or focus) a standalone window for a single chat session.
function createSessionWindow(sessionId) {
function createSessionWindow(sessionId, { watch = false } = {}) {
return sessionWindows.openOrFocus(sessionId, () => {
const icon = getAppIconPath()
const win = new BrowserWindow({
width: 480,
height: 800,
minWidth: 420,
minHeight: 620,
width: SESSION_WINDOW_MIN_WIDTH,
height: SESSION_WINDOW_MIN_HEIGHT,
minWidth: SESSION_WINDOW_MIN_WIDTH,
minHeight: SESSION_WINDOW_MIN_HEIGHT,
title: 'Hermes',
titleBarStyle: 'hidden',
titleBarOverlay: getTitleBarOverlayOptions(),
trafficLightPosition: IS_MAC ? WINDOW_BUTTON_POSITION : undefined,
vibrancy: IS_MAC ? 'sidebar' : undefined,
opacity: windowOpacity(),
icon,
backgroundColor: '#f7f7f7',
// Don't show until the renderer's first themed paint is ready. macOS
// `vibrancy` ignores `backgroundColor` and paints a translucent OS
// material (which follows the OS appearance, not the app theme), so a
// dark-themed app on a light-mode Mac flashes white until the renderer
// covers it. ready-to-show fires after the boot-time paint in
// themes/context.tsx, so the window appears already themed.
show: false,
backgroundColor: getWindowBackgroundColor(),
webPreferences: {
preload: path.join(__dirname, 'preload.cjs'),
contextIsolation: true,
@@ -4880,6 +5063,10 @@ function createSessionWindow(sessionId) {
win.setWindowButtonPosition?.(WINDOW_BUTTON_POSITION)
}
win.once('ready-to-show', () => {
if (!win.isDestroyed()) win.show()
})
win.on('will-enter-full-screen', () => sendWindowStateChanged(true))
win.on('enter-full-screen', () => sendWindowStateChanged(true))
win.on('will-leave-full-screen', () => sendWindowStateChanged(false))
@@ -4890,7 +5077,8 @@ function createSessionWindow(sessionId) {
win.loadURL(
buildSessionWindowUrl(sessionId, {
devServer: DEV_SERVER,
rendererIndexPath: DEV_SERVER ? undefined : resolveRendererIndex()
rendererIndexPath: DEV_SERVER ? undefined : resolveRendererIndex(),
watch
})
)
@@ -4916,8 +5104,13 @@ function createWindow() {
titleBarOverlay: getTitleBarOverlayOptions(),
trafficLightPosition: IS_MAC ? WINDOW_BUTTON_POSITION : undefined,
vibrancy: IS_MAC ? 'sidebar' : undefined,
opacity: windowOpacity(),
icon,
backgroundColor: '#f7f7f7',
// Hidden until the first themed paint so macOS `vibrancy` (which ignores
// `backgroundColor` and follows the OS appearance) can't flash a light
// material before the renderer paints the app theme. See createSessionWindow.
show: false,
backgroundColor: getWindowBackgroundColor(),
webPreferences: {
preload: path.join(__dirname, 'preload.cjs'),
contextIsolation: true,
@@ -4953,6 +5146,10 @@ function createWindow() {
}
}
mainWindow.once('ready-to-show', () => {
if (mainWindow && !mainWindow.isDestroyed()) mainWindow.show()
})
mainWindow.on('will-enter-full-screen', () => sendWindowStateChanged(true))
mainWindow.on('enter-full-screen', () => sendWindowStateChanged(true))
mainWindow.on('will-leave-full-screen', () => sendWindowStateChanged(false))
@@ -5064,12 +5261,12 @@ ipcMain.handle('hermes:backend:touch', async (_event, profile) => {
return { ok: true }
})
ipcMain.handle('hermes:gateway:ws-url', async (_event, profile) => freshGatewayWsUrl(profile))
ipcMain.handle('hermes:window:openSession', async (_event, sessionId) => {
ipcMain.handle('hermes:window:openSession', async (_event, sessionId, opts) => {
if (typeof sessionId !== 'string' || !sessionId.trim()) {
return { ok: false, error: 'invalid-session-id' }
}
createSessionWindow(sessionId.trim())
createSessionWindow(sessionId.trim(), { watch: opts?.watch === true })
return { ok: true }
})
@@ -5078,8 +5275,8 @@ ipcMain.handle('hermes:bootstrap:reset', async () => {
// reset connection state so the next startHermes() call restarts the
// full backend flow (including a fresh runBootstrap pass).
rememberLog('[bootstrap] reset requested by renderer; clearing latched failure')
await teardownPrimaryBackendAndWait()
bootstrapFailure = null
connectionPromise = null
bootstrapState = {
active: false,
manifest: null,
@@ -5477,6 +5674,35 @@ ipcMain.on('hermes:titlebar-theme', (_event, payload) => {
mainWindow?.setTitleBarOverlay?.(getTitleBarOverlayOptions())
})
// Pin the native appearance to the app theme (see NATIVE_THEME_CONFIG_PATH).
ipcMain.on('hermes:native-theme', (_event, mode) => {
if (!THEME_SOURCES.has(mode)) {
return
}
if (nativeTheme.themeSource !== mode) {
nativeTheme.themeSource = mode
writePersistedThemeSource(mode)
}
})
// See-through window translucency. Persist + re-apply opacity to every open
// window at runtime (no recreation, so caching/sessions are untouched).
ipcMain.on('hermes:translucency', (_event, payload) => {
const next = clampIntensity(payload && payload.intensity)
if (next === translucencyIntensity) {
return
}
translucencyIntensity = next
writePersistedTranslucency(next)
for (const win of BrowserWindow.getAllWindows()) {
applyWindowTranslucency(win)
}
})
ipcMain.handle('hermes:openExternal', (_event, url) => {
if (!openExternalUrl(url)) {
throw new Error('Invalid external URL')
@@ -5542,48 +5768,6 @@ ipcMain.handle('hermes:logs:reveal', async () => {
ipcMain.handle('hermes:logs:recent', async () => ({ path: DESKTOP_LOG_PATH, lines: hermesLog.slice(-200) }))
// Always-hidden noise (covers non-git projects too — gitignore would catch
// these anyway when present, but we want the same hygiene without one).
const FS_READDIR_HIDDEN = new Set([
'.git',
'.hg',
'.svn',
'.cache',
'.next',
'.turbo',
'.venv',
'__pycache__',
'build',
'dist',
'node_modules',
'target',
'venv'
])
function findGitRoot(start) {
let dir = start
for (let i = 0; i < 50; i += 1) {
try {
if (fs.existsSync(path.join(dir, '.git'))) {
return dir
}
} catch {
return null
}
const parent = path.dirname(dir)
if (parent === dir) {
return null
}
dir = parent
}
return null
}
function isExecutableFile(filePath) {
if (!filePath || !path.isAbsolute(filePath)) {
return false
@@ -5766,46 +5950,9 @@ function disposeTerminalSession(id) {
return true
}
ipcMain.handle('hermes:fs:readDir', async (_event, dirPath) => {
const resolved = path.resolve(String(dirPath || ''))
ipcMain.handle('hermes:fs:readDir', async (_event, dirPath) => readDirForIpc(dirPath))
if (!resolved) {
return { entries: [], error: 'invalid-path' }
}
try {
const dirents = await fs.promises.readdir(resolved, { withFileTypes: true })
const entries = dirents
.filter(d => {
if (FS_READDIR_HIDDEN.has(d.name)) {
return false
}
return true
})
.map(d => ({ name: d.name, path: path.join(resolved, d.name), isDirectory: d.isDirectory() }))
.sort((a, b) => Number(b.isDirectory) - Number(a.isDirectory) || a.name.localeCompare(b.name))
return { entries }
} catch (error) {
return { entries: [], error: error?.code || 'read-error' }
}
})
ipcMain.handle('hermes:fs:gitRoot', async (_event, startPath) => {
const input = String(startPath || '')
const resolved = input.startsWith('file:') ? fileURLToPath(input) : path.resolve(input)
try {
const stat = await fs.promises.stat(resolved)
const start = stat.isDirectory() ? resolved : path.dirname(resolved)
return findGitRoot(start)
} catch {
return findGitRoot(resolved)
}
})
ipcMain.handle('hermes:fs:gitRoot', async (_event, startPath) => gitRootForIpc(startPath))
ipcMain.handle('hermes:terminal:start', async (event, payload = {}) => {
if (!nodePty) {
@@ -5993,11 +6140,15 @@ async function getUninstallSummary() {
resolve(value)
}
try {
const child = spawn(py, ['-m', 'hermes_cli.main', 'uninstall', '--gui-summary'], hiddenWindowsChildOptions({
cwd: agentRoot,
env: { ...process.env, HERMES_HOME, NO_COLOR: '1' },
stdio: ['ignore', 'pipe', 'ignore']
}))
const child = spawn(
py,
['-m', 'hermes_cli.main', 'uninstall', '--gui-summary'],
hiddenWindowsChildOptions({
cwd: agentRoot,
env: { ...process.env, HERMES_HOME, NO_COLOR: '1' },
stdio: ['ignore', 'pipe', 'ignore']
})
)
child.stdout.on('data', chunk => {
stdout += chunk.toString()
})
@@ -6143,6 +6294,106 @@ ipcMain.handle('hermes:vscode-theme:fetch', async (_event, id) => fetchMarketpla
// Search the Marketplace for color-theme extensions (empty query = top installs).
ipcMain.handle('hermes:vscode-theme:search', async (_event, query) => searchMarketplaceThemes(String(query || ''), 20))
// ---------------------------------------------------------------------------
// hermes:// deep links (e.g. hermes://blueprint/morning-brief?time=08:00).
// A docs/dashboard "Send to App" button opens this URL; we route it into the
// running app's chat composer. Three delivery paths: macOS 'open-url',
// Win/Linux running-app 'second-instance' (argv), Win/Linux cold-start argv.
// ---------------------------------------------------------------------------
const HERMES_PROTOCOL = 'hermes'
let _pendingDeepLink = null
let _rendererReadyForDeepLink = false
function _extractDeepLink(argv) {
if (!Array.isArray(argv)) return null
return argv.find(a => typeof a === 'string' && a.startsWith(`${HERMES_PROTOCOL}://`)) || null
}
function handleDeepLink(url) {
if (!url || typeof url !== 'string') return
let parsed
try {
parsed = new URL(url)
} catch {
rememberLog(`[deeplink] ignoring malformed url: ${url}`)
return
}
// hermes://blueprint/<key>?slot=val -> host="blueprint", path="/<key>"
const kind = parsed.hostname || ''
const name = decodeURIComponent((parsed.pathname || '').replace(/^\//, ''))
const params = {}
parsed.searchParams.forEach((v, k) => {
params[k] = v
})
const payload = { kind, name, params }
if (!_rendererReadyForDeepLink || !mainWindow || mainWindow.isDestroyed()) {
_pendingDeepLink = payload
return
}
try {
if (mainWindow.isMinimized()) mainWindow.restore()
mainWindow.focus()
mainWindow.webContents.send('hermes:deep-link', payload)
rememberLog(`[deeplink] delivered ${kind}/${name}`)
} catch (err) {
rememberLog(`[deeplink] delivery failed: ${err.message}`)
}
}
// Renderer calls this (via IPC) once it has mounted its deep-link listener, so
// a link that arrived during boot/install is flushed exactly once.
ipcMain.handle('hermes:deep-link-ready', () => {
_rendererReadyForDeepLink = true
if (_pendingDeepLink) {
const queued = _pendingDeepLink
_pendingDeepLink = null
handleDeepLink(
`${HERMES_PROTOCOL}://${queued.kind}/${encodeURIComponent(queued.name)}` +
(Object.keys(queued.params).length ? '?' + new URLSearchParams(queued.params).toString() : '')
)
}
return { ok: true }
})
function registerDeepLinkProtocol() {
try {
if (process.defaultApp && process.argv.length >= 2) {
// Dev: register with the electron exec path + entry script so the OS can
// relaunch us with the URL.
app.setAsDefaultProtocolClient(HERMES_PROTOCOL, process.execPath, [path.resolve(process.argv[1])])
} else {
app.setAsDefaultProtocolClient(HERMES_PROTOCOL)
}
} catch (err) {
rememberLog(`[deeplink] protocol registration failed: ${err.message}`)
}
}
// Single-instance lock: deep links on a running app (Win/Linux) arrive as a
// second-instance argv. Without the lock a second `hermes://` launch spawns a
// whole new app instead of routing into the running one.
const _gotSingleInstanceLock = app.requestSingleInstanceLock()
if (!_gotSingleInstanceLock) {
app.quit()
} else {
app.on('second-instance', (_event, argv) => {
const url = _extractDeepLink(argv)
if (url) handleDeepLink(url)
else if (mainWindow) {
if (mainWindow.isMinimized()) mainWindow.restore()
mainWindow.focus()
}
})
}
// macOS delivers deep links via 'open-url' — register early (can fire before
// whenReady; handleDeepLink queues until the renderer is ready).
app.on('open-url', (event, url) => {
event.preventDefault()
handleDeepLink(url)
})
app.whenReady().then(() => {
if (IS_MAC) {
Menu.setApplicationMenu(buildApplicationMenu())
@@ -6151,11 +6402,16 @@ app.whenReady().then(() => {
}
installMediaPermissions()
registerMediaProtocol()
registerDeepLinkProtocol()
ensureWslWindowsFonts()
configureSpellChecker()
registerPowerResumeListeners()
createWindow()
// Win/Linux cold start: the launching hermes:// URL is in our own argv.
const _coldStartLink = _extractDeepLink(process.argv)
if (_coldStartLink) handleDeepLink(_coldStartLink)
app.on('activate', () => {
// Recreate the primary window if it's gone. Guard on mainWindow directly
// (not just total window count) so a dock click still restores the main

View File

@@ -5,7 +5,7 @@ contextBridge.exposeInMainWorld('hermesDesktop', {
revalidateConnection: () => ipcRenderer.invoke('hermes:connection:revalidate'),
touchBackend: profile => ipcRenderer.invoke('hermes:backend:touch', profile),
getGatewayWsUrl: profile => ipcRenderer.invoke('hermes:gateway:ws-url', profile),
openSessionWindow: sessionId => ipcRenderer.invoke('hermes:window:openSession', sessionId),
openSessionWindow: (sessionId, opts) => ipcRenderer.invoke('hermes:window:openSession', sessionId, opts),
getBootProgress: () => ipcRenderer.invoke('hermes:boot-progress:get'),
getConnectionConfig: profile => ipcRenderer.invoke('hermes:connection-config:get', profile),
saveConnectionConfig: payload => ipcRenderer.invoke('hermes:connection-config:save', payload),
@@ -39,6 +39,8 @@ contextBridge.exposeInMainWorld('hermesDesktop', {
watchPreviewFile: url => ipcRenderer.invoke('hermes:watchPreviewFile', url),
stopPreviewFileWatch: id => ipcRenderer.invoke('hermes:stopPreviewFileWatch', id),
setTitleBarTheme: payload => ipcRenderer.send('hermes:titlebar-theme', payload),
setNativeTheme: mode => ipcRenderer.send('hermes:native-theme', mode),
setTranslucency: payload => ipcRenderer.send('hermes:translucency', payload),
setPreviewShortcutActive: active => ipcRenderer.send('hermes:previewShortcutActive', Boolean(active)),
openExternal: url => ipcRenderer.invoke('hermes:openExternal', url),
fetchLinkTitle: url => ipcRenderer.invoke('hermes:fetchLinkTitle', url),
@@ -80,6 +82,12 @@ contextBridge.exposeInMainWorld('hermesDesktop', {
ipcRenderer.on('hermes:open-updates', listener)
return () => ipcRenderer.removeListener('hermes:open-updates', listener)
},
onDeepLink: callback => {
const listener = (_event, payload) => callback(payload)
ipcRenderer.on('hermes:deep-link', listener)
return () => ipcRenderer.removeListener('hermes:deep-link', listener)
},
signalDeepLinkReady: () => ipcRenderer.invoke('hermes:deep-link-ready'),
onWindowStateChanged: callback => {
const listener = (_event, payload) => callback(payload)
ipcRenderer.on('hermes:window-state-changed', listener)

View File

@@ -5,22 +5,30 @@
const { pathToFileURL } = require('node:url')
// Secondary windows open at the minimum usable size — a compact side panel for
// subagent watch / cmd-click session pop-out, not a second full desktop.
const SESSION_WINDOW_MIN_WIDTH = 420
const SESSION_WINDOW_MIN_HEIGHT = 620
// Build the renderer URL for a secondary window. The renderer uses a
// HashRouter, so the session route lives after the '#'. The `?win=secondary`
// flag MUST sit in the query string BEFORE the '#': anything after the '#' is
// treated as the route by HashRouter and would break routeSessionId(). The
// renderer reads the flag from window.location.search to suppress the install /
// onboarding overlays and the global session sidebar.
function buildSessionWindowUrl(sessionId, { devServer, rendererIndexPath } = {}) {
// onboarding overlays and the global session sidebar. `watch=1` marks a
// spectator window (e.g. a running subagent's session): the renderer resumes
// it lazily so the gateway never builds an agent just to stream into it.
function buildSessionWindowUrl(sessionId, { devServer, rendererIndexPath, watch } = {}) {
const query = `?win=secondary${watch ? '&watch=1' : ''}`
const route = `#/${encodeURIComponent(sessionId)}`
if (devServer) {
const base = devServer.endsWith('/') ? devServer.slice(0, -1) : devServer
return `${base}/?win=secondary${route}`
return `${base}/${query}${route}`
}
return `${pathToFileURL(rendererIndexPath).toString()}?win=secondary${route}`
return `${pathToFileURL(rendererIndexPath).toString()}${query}${route}`
}
// A small registry keyed by sessionId that guarantees one window per chat:
@@ -83,4 +91,9 @@ function createSessionWindowRegistry() {
}
}
module.exports = { buildSessionWindowUrl, createSessionWindowRegistry }
module.exports = {
buildSessionWindowUrl,
createSessionWindowRegistry,
SESSION_WINDOW_MIN_HEIGHT,
SESSION_WINDOW_MIN_WIDTH
}

View File

@@ -76,6 +76,12 @@ test('buildSessionWindowUrl builds a packaged file URL with the flag before the
assert.match(url, /^file:\/\/.*index\.html\?win=secondary#\/abc$/)
})
test('buildSessionWindowUrl adds the watch flag for spectator windows, before the hash', () => {
const url = buildSessionWindowUrl('abc', { devServer: 'http://localhost:5173', watch: true })
assert.equal(url, 'http://localhost:5173/?win=secondary&watch=1#/abc')
})
test('registry opens one window per session and focuses on re-open', () => {
const registry = createSessionWindowRegistry()
let built = 0

View File

@@ -0,0 +1,56 @@
/**
* Pure helpers for choosing a remote URL during passive update checks.
*
* A public install can end up with `origin=git@github.com:NousResearch/hermes-agent.git`.
* If the user's GitHub SSH key is FIDO2/passkey-backed, a background `git fetch
* origin` triggers an unexplained hardware-touch prompt. For passive checks
* against the official repo we substitute the public HTTPS `ls-remote` path,
* which needs no auth and cannot prompt. Active update/apply flows are left
* unchanged.
*
* Extracted from main.cjs so the security-critical remote detection is unit
* testable without booting Electron (main.cjs requires('electron') at load).
*/
const OFFICIAL_REPO_HTTPS_URL = 'https://github.com/NousResearch/hermes-agent.git'
const OFFICIAL_REPO_CANONICAL = 'github.com/nousresearch/hermes-agent'
// Normalize common GitHub remote URL forms to `host/owner/repo` (lowercased,
// no trailing slash, no .git suffix) so SSH and HTTPS forms of the same repo
// compare equal.
function canonicalGitHubRemote(url) {
if (!url) return ''
let value = String(url).trim()
if (value.startsWith('git@github.com:')) {
value = `github.com/${value.slice('git@github.com:'.length)}`
} else if (value.startsWith('ssh://git@github.com/')) {
value = `github.com/${value.slice('ssh://git@github.com/'.length)}`
} else {
try {
const parsed = new URL(value)
if (parsed.hostname && parsed.pathname) value = `${parsed.hostname}${parsed.pathname}`
} catch {
// Leave non-URL forms unchanged.
}
}
value = value.trim().replace(/\/+$/, '')
if (value.endsWith('.git')) value = value.slice(0, -4)
return value.toLowerCase()
}
function isSshRemote(url) {
const value = String(url || '').trim().toLowerCase()
return value.startsWith('git@') || value.startsWith('ssh://')
}
function isOfficialSshRemote(url) {
return isSshRemote(url) && canonicalGitHubRemote(url) === OFFICIAL_REPO_CANONICAL
}
module.exports = {
OFFICIAL_REPO_HTTPS_URL,
OFFICIAL_REPO_CANONICAL,
canonicalGitHubRemote,
isSshRemote,
isOfficialSshRemote
}

View File

@@ -0,0 +1,78 @@
/**
* Tests for electron/update-remote.cjs — the remote-detection helpers that
* keep passive update checks off the SSH origin for official installs.
*
* Run with: node --test electron/update-remote.test.cjs
* (Wired into npm test:desktop:platforms in package.json.)
*
* Why this matters: a public install can carry
* origin=git@github.com:NousResearch/hermes-agent.git. A background
* `git fetch origin` then authenticates over SSH and, with a FIDO2/passkey
* key, triggers an unexplained hardware-touch prompt. isOfficialSshRemote
* must reliably recognize the official SSH remote (in every URL form,
* case-insensitively) so the caller can swap in the anonymous HTTPS path —
* while NOT misclassifying forks, other hosts, or the HTTPS remote (which
* never prompts and should keep the normal fetch path).
*/
const test = require('node:test')
const assert = require('node:assert/strict')
const {
OFFICIAL_REPO_HTTPS_URL,
OFFICIAL_REPO_CANONICAL,
canonicalGitHubRemote,
isSshRemote,
isOfficialSshRemote
} = require('./update-remote.cjs')
test('canonicalGitHubRemote normalizes SSH and HTTPS forms to the same value', () => {
assert.equal(canonicalGitHubRemote('git@github.com:NousResearch/hermes-agent.git'), OFFICIAL_REPO_CANONICAL)
assert.equal(canonicalGitHubRemote('git@github.com:NousResearch/hermes-agent'), OFFICIAL_REPO_CANONICAL)
assert.equal(canonicalGitHubRemote('ssh://git@github.com/NousResearch/hermes-agent.git'), OFFICIAL_REPO_CANONICAL)
assert.equal(canonicalGitHubRemote('https://github.com/NousResearch/hermes-agent.git'), OFFICIAL_REPO_CANONICAL)
// Case-insensitive: an uppercased owner still canonicalizes to the same repo.
assert.equal(canonicalGitHubRemote('git@github.com:nousresearch/hermes-agent.git'), OFFICIAL_REPO_CANONICAL)
// Trailing slashes are stripped.
assert.equal(canonicalGitHubRemote('https://github.com/NousResearch/hermes-agent/'), OFFICIAL_REPO_CANONICAL)
})
test('canonicalGitHubRemote is empty for falsy input', () => {
assert.equal(canonicalGitHubRemote(''), '')
assert.equal(canonicalGitHubRemote(null), '')
assert.equal(canonicalGitHubRemote(undefined), '')
})
test('isSshRemote detects scp-like and ssh:// forms only', () => {
assert.equal(isSshRemote('git@github.com:NousResearch/hermes-agent.git'), true)
assert.equal(isSshRemote('ssh://git@github.com/NousResearch/hermes-agent.git'), true)
assert.equal(isSshRemote('https://github.com/NousResearch/hermes-agent.git'), false)
assert.equal(isSshRemote(''), false)
assert.equal(isSshRemote(null), false)
})
test('isOfficialSshRemote is true only for the official repo over SSH', () => {
assert.equal(isOfficialSshRemote('git@github.com:NousResearch/hermes-agent.git'), true)
assert.equal(isOfficialSshRemote('git@github.com:NousResearch/hermes-agent'), true)
assert.equal(isOfficialSshRemote('ssh://git@github.com/NousResearch/hermes-agent.git'), true)
// Case-insensitive owner/repo match.
assert.equal(isOfficialSshRemote('git@github.com:nousresearch/hermes-agent.git'), true)
})
test('isOfficialSshRemote does NOT match forks, other hosts, or HTTPS', () => {
// A fork over SSH belongs to the user — fetching it is their own remote,
// not the official upstream, so the SSH-avoidance swap must not apply.
assert.equal(isOfficialSshRemote('git@github.com:someuser/hermes-agent.git'), false)
// Same repo name on a different host is not the official repo.
assert.equal(isOfficialSshRemote('git@gitlab.com:NousResearch/hermes-agent.git'), false)
// HTTPS to the official repo never prompts for SSH/FIDO2, so it keeps the
// normal fetch path — must not be flagged as an official SSH remote.
assert.equal(isOfficialSshRemote('https://github.com/NousResearch/hermes-agent.git'), false)
assert.equal(isOfficialSshRemote(''), false)
assert.equal(isOfficialSshRemote(null), false)
})
test('OFFICIAL_REPO_HTTPS_URL canonicalizes to OFFICIAL_REPO_CANONICAL', () => {
// Invariant: the URL we substitute in must be the same repo we detect.
assert.equal(canonicalGitHubRemote(OFFICIAL_REPO_HTTPS_URL), OFFICIAL_REPO_CANONICAL)
})

View File

@@ -8,7 +8,7 @@ const path = require('node:path')
const ELECTRON_DIR = __dirname
function readElectronFile(name) {
return fs.readFileSync(path.join(ELECTRON_DIR, name), 'utf8')
return fs.readFileSync(path.join(ELECTRON_DIR, name), 'utf8').replace(/\r\n/g, '\n')
}
function requireHiddenChildOptions(source, needle) {

View File

@@ -3,7 +3,6 @@ import typescriptEslint from '@typescript-eslint/eslint-plugin'
import typescriptParser from '@typescript-eslint/parser'
import perfectionist from 'eslint-plugin-perfectionist'
import reactPlugin from 'eslint-plugin-react'
import reactCompiler from 'eslint-plugin-react-compiler'
import hooksPlugin from 'eslint-plugin-react-hooks'
import unusedImports from 'eslint-plugin-unused-imports'
import globals from 'globals'
@@ -47,7 +46,6 @@ export default [
'custom-rules': customRules,
perfectionist,
react: reactPlugin,
'react-compiler': reactCompiler,
'react-hooks': hooksPlugin,
'unused-imports': unusedImports
},
@@ -98,7 +96,6 @@ export default [
'perfectionist/sort-jsx-props': ['error', { order: 'asc', type: 'natural' }],
'perfectionist/sort-named-exports': ['error', { order: 'asc', type: 'natural' }],
'perfectionist/sort-named-imports': ['error', { order: 'asc', type: 'natural' }],
'react-compiler/react-compiler': 'warn',
'react-hooks/exhaustive-deps': 'warn',
'react-hooks/rules-of-hooks': 'error',
'unused-imports/no-unused-imports': 'error'

View File

@@ -9,6 +9,28 @@
<link rel="apple-touch-icon" sizes="180x180" href="/apple-touch-icon.png" />
<link rel="shortcut icon" href="/apple-touch-icon.png" />
<title>Hermes</title>
<script>
// Pre-paint the themed background before the app bundle loads. Without
// this, the first frame (which is what `ready-to-show` waits for) is the
// UA-default white page, and the real theme only lands once the whole
// module graph has executed — i.e. the "white flash" on every new
// window. applyTheme() in src/themes/context.tsx keeps these keys fresh.
try {
let bg = localStorage.getItem('hermes-boot-background')
let scheme = localStorage.getItem('hermes-boot-color-scheme')
if (!bg) {
const dark = window.matchMedia('(prefers-color-scheme: dark)').matches
bg = dark ? '#111111' : '#f7f7f7'
scheme = dark ? 'dark' : 'light'
}
document.documentElement.style.backgroundColor = bg
if (scheme === 'dark' || scheme === 'light') {
document.documentElement.style.colorScheme = scheme
}
} catch {
// localStorage unavailable — keep UA defaults.
}
</script>
</head>
<body>
<div id="root" class="scrollbar-dt"></div>

View File

@@ -18,7 +18,8 @@
"profile:main": "wait-on http://127.0.0.1:5174 && cross-env XCURSOR_SIZE=24 HERMES_DESKTOP_DEV_SERVER=http://127.0.0.1:5174 electron --inspect=9229 .",
"profile:main:cpu": "wait-on http://127.0.0.1:5174 && cross-env XCURSOR_SIZE=24 NODE_OPTIONS=--cpu-prof HERMES_DESKTOP_DEV_SERVER=http://127.0.0.1:5174 electron .",
"start": "npm run build && electron .",
"build": "node scripts/assert-root-install.cjs && node scripts/write-build-stamp.cjs && node scripts/stage-native-deps.cjs && tsc -b && vite build && node scripts/assert-dist-built.cjs",
"build": "node scripts/assert-root-install.cjs && node scripts/write-build-stamp.cjs && node scripts/stage-native-deps.cjs && tsc -b && vite build && npm run postbuild",
"postbuild": "node scripts/assert-dist-built.cjs",
"builder": "cross-env NODE_OPTIONS=--max-old-space-size=16384 electron-builder",
"pack": "npm run build && npm run builder -- --dir",
"dist": "npm run build && npm run builder",
@@ -35,8 +36,8 @@
"test:desktop:nsis": "node scripts/test-desktop.mjs nsis",
"test:desktop:existing": "node scripts/test-desktop.mjs existing",
"test:desktop:fresh": "node scripts/test-desktop.mjs fresh",
"test:desktop:platforms": "node --test electron/bootstrap-platform.test.cjs electron/hardening.test.cjs electron/backend-probes.test.cjs electron/bootstrap-runner.test.cjs electron/connection-config.test.cjs electron/gateway-ws-probe.test.cjs electron/oauth-net-request.test.cjs electron/desktop-uninstall.test.cjs electron/session-windows.test.cjs electron/workspace-cwd.test.cjs electron/windows-child-process.test.cjs",
"type-check": "tsc -b",
"test:desktop:platforms": "node --test electron/bootstrap-platform.test.cjs electron/hardening.test.cjs electron/backend-env.test.cjs electron/backend-probes.test.cjs electron/bootstrap-runner.test.cjs electron/connection-config.test.cjs electron/dashboard-token.test.cjs electron/gateway-ws-probe.test.cjs electron/oauth-net-request.test.cjs electron/desktop-uninstall.test.cjs electron/session-windows.test.cjs electron/workspace-cwd.test.cjs electron/fs-read-dir.test.cjs electron/git-root.test.cjs electron/windows-child-process.test.cjs electron/update-remote.test.cjs",
"typecheck": "tsc -p . --noEmit",
"lint": "eslint src/ electron/",
"lint:fix": "eslint src/ electron/ --fix",
"fmt": "prettier --write 'src/**/*.{ts,tsx}' 'electron/**/*.{js,cjs}' 'vite.config.ts'",
@@ -72,6 +73,7 @@
"class-variance-authority": "^0.7.1",
"clsx": "^2.1.1",
"cmdk": "^1.1.1",
"dnd-core": "^14.0.1",
"hast-util-from-html-isomorphic": "^2.0.0",
"hast-util-to-text": "^4.0.2",
"ignore": "^7.0.5",
@@ -83,6 +85,7 @@
"radix-ui": "^1.4.3",
"react": "^19.2.5",
"react-arborist": "^3.5.0",
"react-dnd-html5-backend": "^14.0.3",
"react-dom": "^19.2.5",
"react-router-dom": "^7.17.0",
"react-shiki": "^0.9.3",
@@ -103,20 +106,19 @@
"@testing-library/dom": "^10.4.0",
"@testing-library/react": "^16.3.2",
"@types/hast": "^3.0.4",
"@types/node": "^24.12.2",
"@types/node": "^24.13.2",
"@types/react": "^19.2.14",
"@types/react-dom": "^19.2.3",
"@typescript-eslint/eslint-plugin": "^8.59.1",
"@typescript-eslint/parser": "^8.59.1",
"@vitejs/plugin-react": "^6.0.1",
"concurrently": "^9.2.1",
"concurrently": "^10.0.3",
"cross-env": "^10.1.0",
"electron": "^40.9.3",
"electron-builder": "^26.8.1",
"eslint": "^9.39.4",
"eslint-plugin-perfectionist": "^5.9.0",
"eslint-plugin-react": "^7.37.5",
"eslint-plugin-react-compiler": "^19.1.0-rc.2",
"eslint-plugin-react-hooks": "^7.1.1",
"eslint-plugin-unused-imports": "^4.4.1",
"globals": "^16.5.0",
@@ -133,6 +135,14 @@
"appId": "com.nousresearch.hermes",
"productName": "Hermes",
"executableName": "Hermes",
"protocols": [
{
"name": "Hermes Protocol",
"schemes": [
"hermes"
]
}
],
"artifactName": "Hermes-${version}-${os}-${arch}.${ext}",
"icon": "assets/icon",
"directories": {

View File

@@ -3,8 +3,8 @@ import { type ReactNode, useEffect, useMemo, useState } from 'react'
import { useElapsedSeconds } from '@/components/chat/activity-timer'
import { ActivityTimerText } from '@/components/chat/activity-timer-text'
import { BrailleSpinner } from '@/components/ui/braille-spinner'
import { FadeText } from '@/components/ui/fade-text'
import { GlyphSpinner } from '@/components/ui/glyph-spinner'
import { type Translations, useI18n } from '@/i18n'
import { AlertCircle, CheckCircle2, Sparkles } from '@/lib/icons'
import { useEnterAnimation } from '@/lib/use-enter-animation'
@@ -25,7 +25,7 @@ import { OverlayView } from '../overlays/overlay-view'
function statusGlyph(status: SubagentStatus, a: Translations['agents']): ReactNode {
if (status === 'running' || status === 'queued') {
return (
<BrailleSpinner
<GlyphSpinner
ariaLabel={a.running}
className="size-3.5 shrink-0 text-[0.95rem] text-muted-foreground/80"
spinner="breathe"
@@ -290,7 +290,7 @@ function StreamLine({
<span className={cn('min-w-0 flex-1 wrap-anywhere', tone, isMono && 'font-mono text-[0.69rem]')}>
{entry.text}
{active ? (
<BrailleSpinner
<GlyphSpinner
ariaLabel={t.agents.streaming}
className="ml-1 inline-block size-2.5 align-middle text-muted-foreground/70"
spinner="breathe"
@@ -372,7 +372,9 @@ function SubagentRow({ node, depth = 0, nowMs }: { node: SubagentNode; depth?: n
{open && fileLines.length > 0 ? (
<div className="grid min-w-0 gap-0.5 pl-6">
<p className="text-[0.58rem] font-medium tracking-wider text-muted-foreground/60 uppercase">{t.agents.files}</p>
<p className="text-[0.58rem] font-medium tracking-wider text-muted-foreground/60 uppercase">
{t.agents.files}
</p>
{fileLines.slice(0, 8).map(line => (
<p className="wrap-break-word font-mono text-[0.67rem] leading-relaxed text-muted-foreground/80" key={line}>
{line}

View File

@@ -18,7 +18,7 @@ import {
} from '@/components/ui/pagination'
import { TextTab, TextTabMeta } from '@/components/ui/text-tab'
import { Tip } from '@/components/ui/tooltip'
import { getSessionMessages, listSessions } from '@/hermes'
import { getSessionMessages, listAllProfileSessions } from '@/hermes'
import { type Translations, useI18n } from '@/i18n'
import { sessionTitle } from '@/lib/chat-runtime'
import { ExternalLink, ExternalLinkIcon, hostPathLabel, urlSlugTitleLabel, useLinkTitle } from '@/lib/external-link'
@@ -388,8 +388,8 @@ export function ArtifactsView({ setStatusbarItemGroup: _setStatusbarItemGroup, .
setRefreshing(true)
try {
const sessions = (await listSessions(30, 1)).sessions
const results = await Promise.allSettled(sessions.map(session => getSessionMessages(session.id)))
const sessions = (await listAllProfileSessions(30, 1)).sessions
const results = await Promise.allSettled(sessions.map(session => getSessionMessages(session.id, session.profile)))
const nextArtifacts: ArtifactRecord[] = []
results.forEach((result, index) => {

View File

@@ -2,32 +2,21 @@ import type { Unstable_TriggerAdapter } from '@assistant-ui/core'
import { ComposerPrimitive } from '@assistant-ui/react'
import type { ReactNode } from 'react'
export const COMPLETION_DRAWER_CLASS = [
'absolute bottom-[calc(100%+0.25rem)] left-0 z-50',
'w-60 max-w-[calc(100vw-2rem)]',
'max-h-[min(23rem,calc(100vh-8rem))] overflow-y-auto overscroll-contain',
'rounded-lg border border-(--ui-stroke-secondary)',
'bg-[color-mix(in_srgb,var(--ui-bg-elevated)_96%,transparent)]',
'p-1 text-xs text-popover-foreground shadow-md',
'backdrop-blur-md'
].join(' ')
import { composerFusedDockCard } from '@/components/chat/composer-dock'
import { cn } from '@/lib/utils'
export const COMPLETION_DRAWER_BELOW_CLASS = [
'absolute left-0 top-[calc(100%+0.25rem)] z-50',
'w-60 max-w-[calc(100vw-2rem)]',
'max-h-[min(23rem,calc(100vh-8rem))] overflow-y-auto overscroll-contain',
'rounded-lg border border-(--ui-stroke-secondary)',
'bg-[color-mix(in_srgb,var(--ui-bg-elevated)_96%,transparent)]',
'p-1 text-xs text-popover-foreground shadow-md',
'backdrop-blur-md'
].join(' ')
// Same docked chrome as the queue/status stack, but its own thing: a narrow,
// left-aligned card (not full width) that fuses to the composer's edge instead
// of floating above it. `left-1` matches the stack's `mx-1` inset; the negative
// margin overlaps the seam so the composer's (now-transparent) edge border reads
// as shared. Fused (opaque) fill — the composer surface swaps to the same fill
// while a drawer is open, so the two paint as one panel.
const DRAWER_SHELL =
'absolute left-1 z-50 w-80 max-w-[calc(100%-0.5rem)] max-h-[min(22rem,calc(100vh-8rem))] overflow-y-auto overscroll-contain p-1 text-xs text-popover-foreground'
export const COMPLETION_DRAWER_ROW_CLASS = [
'relative flex cursor-default select-none items-center gap-2 rounded-md px-2 py-1',
'w-full min-w-0 text-left text-xs outline-hidden transition-colors',
'hover:bg-(--ui-bg-tertiary)',
'data-[highlighted]:bg-(--ui-bg-tertiary) data-[highlighted]:text-foreground'
].join(' ')
export const COMPLETION_DRAWER_CLASS = cn(DRAWER_SHELL, 'bottom-full -mb-[9px]', composerFusedDockCard('top'))
export const COMPLETION_DRAWER_BELOW_CLASS = cn(DRAWER_SHELL, 'top-full -mt-[9px]', composerFusedDockCard('bottom'))
export function ComposerCompletionDrawer({
adapter,

View File

@@ -11,6 +11,7 @@ import {
DropdownMenuSeparator,
DropdownMenuTrigger
} from '@/components/ui/dropdown-menu'
import { Kbd } from '@/components/ui/kbd'
import { useI18n } from '@/i18n'
import { Clipboard, FileText, FolderOpen, type IconComponent, ImageIcon, Link, MessageSquareText } from '@/lib/icons'
import { cn } from '@/lib/utils'
@@ -86,7 +87,7 @@ export function ContextMenu({
<div className="px-2 py-1 text-[0.7rem] text-muted-foreground/80">
{c.tipPre}
<kbd className="rounded bg-muted/70 px-1 py-px font-mono text-[0.65rem]">@</kbd>
<Kbd size="sm">@</Kbd>
{c.tipPost}
</div>
</DropdownMenuContent>

View File

@@ -1,5 +1,6 @@
import { Button } from '@/components/ui/button'
import { Codicon } from '@/components/ui/codicon'
import { KbdCombo } from '@/components/ui/kbd'
import { Tip } from '@/components/ui/tooltip'
import { useI18n } from '@/i18n'
import { triggerHaptic } from '@/lib/haptics'
@@ -63,7 +64,14 @@ export function ComposerControls({
}) {
const { t } = useI18n()
const c = t.composer
const steerLabel = `${c.steer} (${formatCombo('mod+enter')})`
const steerCombo = formatCombo('mod+enter')
const steerLabel = `${c.steer} (${steerCombo})`
const steerTip = (
<span className="inline-flex items-center gap-1.5">
{c.steer}
<KbdCombo combo="mod+enter" size="sm" variant="inverted" />
</span>
)
if (conversation.active) {
return <ConversationPill {...conversation} disabled={disabled} />
@@ -75,7 +83,7 @@ export function ComposerControls({
<div className="ml-auto flex shrink-0 items-center gap-(--composer-control-gap)">
<DictationButton disabled={disabled} onToggle={onDictate} state={state.voice} status={voiceStatus} />
{canSteer && (
<Tip label={steerLabel}>
<Tip label={steerTip}>
<Button
aria-label={steerLabel}
className={GHOST_ICON_BTN}

View File

@@ -10,6 +10,7 @@
* steal focus from the composer effect.
*/
import { RICH_INPUT_SLOT } from './rich-editor'
import type { InlineRefInput } from './inline-refs'
export type ComposerTarget = 'edit' | 'main'
@@ -123,3 +124,12 @@ export const focusComposerInput = (el: HTMLElement | null) => {
window.requestAnimationFrame(focus)
window.setTimeout(focus, 0)
}
/** Drop focus from the main composer input (status-stack chrome, sidebar, etc.). */
export const blurComposerInput = () => {
const el = document.querySelector(`[data-slot="${RICH_INPUT_SLOT}"]`) as HTMLElement | null
if (el && document.activeElement === el) {
el.blur()
}
}

View File

@@ -1,11 +1,23 @@
import type { ReactNode } from 'react'
import { KbdCombo } from '@/components/ui/kbd'
import { useI18n } from '@/i18n'
import { COMPLETION_DRAWER_CLASS } from './completion-drawer'
const COMMON_COMMAND_KEYS = ['/help', '/clear', '/resume', '/details', '/copy', '/quit']
const HOTKEY_KEYS = ['@', '/', '?', 'Enter', 'Cmd/Ctrl+Shift+K', 'Cmd/Ctrl+/', 'Esc', '↑ / ↓']
/** Stable ids → i18n `hotkeyDescs` keys. Combos resolve mod labels per OS. */
const COMPOSER_HOTKEY_ROWS = [
{ id: 'composer.mention', combos: ['@'] },
{ id: 'composer.slash', combos: ['/'] },
{ id: 'composer.help', combos: ['?'] },
{ id: 'composer.sendNewline', combos: ['enter', 'shift+enter'] },
{ id: 'composer.sendQueued', combos: ['mod+shift+k'] },
{ id: 'keybinds.openPanel', combos: ['mod+/'] },
{ id: 'composer.cancel', combos: ['escape'] },
{ id: 'composer.history', combos: ['up', 'down'] }
] as const
export function HelpHint() {
const { t } = useI18n()
@@ -20,8 +32,8 @@ export function HelpHint() {
</Section>
<Section title={c.hotkeys}>
{HOTKEY_KEYS.map(key => (
<Row description={c.hotkeyDescs[key] ?? ''} key={key} keyLabel={key} />
{COMPOSER_HOTKEY_ROWS.map(row => (
<HotkeyRow description={c.hotkeyDescs[row.id] ?? ''} combos={[...row.combos]} key={row.id} />
))}
</Section>
@@ -57,3 +69,16 @@ function Row({ description, keyLabel, mono = false }: { description: string; key
</div>
)
}
function HotkeyRow({ combos, description }: { combos: string[]; description: string }) {
return (
<div className="flex min-w-0 items-center gap-2 rounded-md px-2.5 py-1 text-xs">
<span className="flex shrink-0 items-center gap-1">
{combos.map(combo => (
<KbdCombo combo={combo} key={combo} size="sm" />
))}
</span>
<span className="min-w-0 truncate text-muted-foreground/80">{description}</span>
</div>
)
}

View File

@@ -5,6 +5,13 @@ export interface CompletionEntry {
text: string
display?: unknown
meta?: unknown
/** Optional section label (e.g. "Commands", "Skills"). The popover renders a
* header whenever this changes between consecutive items, so the fetcher must
* emit entries already grouped contiguously. */
group?: string
/** Optional completion-action id. When set, picking the item runs that action
* (e.g. opening an overlay) instead of inserting a chip + waiting for submit. */
action?: string
}
export interface CompletionPayload {

View File

@@ -2,12 +2,17 @@ import type { Unstable_TriggerAdapter, Unstable_TriggerItem } from '@assistant-u
import { useCallback } from 'react'
import type { HermesGateway } from '@/hermes'
import { sessionTitle } from '@/lib/chat-runtime'
import {
type CommandsCatalogLike,
desktopSkinSlashCompletions,
desktopSlashDescription,
type DesktopThemeCommandOption,
filterDesktopCommandsCatalog,
isDesktopSlashExtensionCommand,
isDesktopSlashSuggestion
} from '@/lib/desktop-slash-commands'
import { $sessions } from '@/store/session'
import type { CompletionEntry, CompletionPayload } from './use-live-completion-adapter'
import { useLiveCompletionAdapter } from './use-live-completion-adapter'
@@ -16,7 +21,10 @@ interface SlashItemMetadata extends Record<string, string> {
command: string
display: string
meta: string
group: string
rawText: string
/** Completion-action id; empty for ordinary insert-a-chip completions. */
action: string
}
function textValue(value: unknown, fallback = ''): string {
@@ -38,12 +46,21 @@ function commandText(value: string): string {
return value.startsWith('/') ? value : `/${value}`
}
/** How many recent sessions to surface inline before the "Browse all…" entry. */
const SESSION_INLINE_LIMIT = 7
/** Live `/` completions backed by the gateway's `complete.slash` RPC. */
export function useSlashCompletions(options: { gateway: HermesGateway | null }): {
export function useSlashCompletions(options: {
gateway: HermesGateway | null
/** Desktop theme list — `/skin` is owned client-side, so its arg completions
* come from here, not the backend (whose skin list is CLI/TUI-only). */
skinThemes?: DesktopThemeCommandOption[]
activeSkin?: string
}): {
adapter: Unstable_TriggerAdapter
loading: boolean
} {
const { gateway } = options
const { gateway, skinThemes, activeSkin } = options
const enabled = Boolean(gateway)
const fetcher = useCallback(
@@ -54,34 +71,136 @@ export function useSlashCompletions(options: { gateway: HermesGateway | null }):
const text = `/${query}`
// The desktop owns /skin entirely (client-side theme context). Surface its
// theme list inside this single popover instead of a bespoke one, and skip
// the backend skin completions (which describe CLI/TUI skins that don't
// apply here). Matches once we're past `/skin ` into the arg stage.
const skinArg = /^\/skin\s+(.*)$/is.exec(text)
if (skinArg && skinThemes) {
const items = desktopSkinSlashCompletions(skinThemes, activeSkin ?? '', skinArg[1] ?? '').map(entry => ({
text: entry.text,
display: entry.display,
meta: entry.meta,
group: 'Themes'
}))
return { items, query }
}
// /resume (and its aliases) completes recent sessions inline — the same
// client-side list the picker overlay shows — instead of the backend
// (whose /resume opens an interactive TUI picker we can't render here).
const sessionArg = /^\/(?:resume|sessions|switch)\s+(.*)$/is.exec(text)
if (sessionArg) {
const needle = (sessionArg[1] ?? '').trim().toLowerCase()
const matches = (
needle
? $sessions.get().filter(
session =>
sessionTitle(session).toLowerCase().includes(needle) ||
(session.preview ?? '').toLowerCase().includes(needle) ||
session.id.toLowerCase().includes(needle)
)
: $sessions.get()
).slice(0, SESSION_INLINE_LIMIT)
const items: CompletionEntry[] = matches.map(session => ({
text: `/resume ${session.id}`,
display: sessionTitle(session),
meta: (session.preview ?? '').trim(),
group: 'Sessions'
}))
// Trailing "more" affordance (Cursor-style): picking it opens the full
// session picker overlay directly. `text` stays a bare `/resume` so that
// submitting it (Enter) still opens the overlay if the action is skipped.
items.push({
text: '/resume',
display: 'Browse all sessions…',
meta: '',
group: 'Sessions',
action: 'session-picker'
})
return { items, query }
}
try {
if (!query) {
const catalog = filterDesktopCommandsCatalog(await gateway.request<CommandsCatalogLike>('commands.catalog'))
const items = (catalog.pairs ?? []).map(([command, meta]) => ({
text: command,
display: command,
meta
}))
// Prefer the categorized layout so the popover renders section headers
// (Session, Tools & Skills, ...). Fall back to the flat list when the
// backend didn't categorize.
const sections = catalog.categories?.length
? catalog.categories
: [{ name: '', pairs: catalog.pairs ?? [] }]
const items = sections.flatMap(section =>
section.pairs.map(([command, meta]) => ({
text: command,
display: command,
group: section.name || undefined,
meta
}))
)
return { items, query }
}
const result = await gateway.request<{ items?: CompletionEntry[] }>('complete.slash', { text })
const result = await gateway.request<{ items?: CompletionEntry[]; replace_from?: number }>(
'complete.slash',
{ text }
)
const items = (result.items ?? [])
.filter(item => isDesktopSlashSuggestion(item.text))
// Arg-completion items (replace_from > 1) carry just the arg stub —
// e.g. complete.slash returns `{text: "alice"}` for `/personality alic`
// with replace_from = 14. Rewrite those entries so the popover inserts
// the full `/personality alice` token instead of stranding `/alice`.
const replaceFrom = typeof result.replace_from === 'number' ? result.replace_from : 1
const isArgCompletion = replaceFrom > 1
const prefix = isArgCompletion ? text.slice(0, replaceFrom) : ''
const decorated = (result.items ?? [])
.map(item => {
if (!isArgCompletion) {
return item
}
const argText = typeof item.text === 'string' ? item.text : ''
return { ...item, text: `${prefix}${argText}` }
})
.filter(item => isArgCompletion || isDesktopSlashSuggestion(item.text))
.map(item => ({
...item,
meta: desktopSlashDescription(item.text, textValue(item.meta))
// Arg suggestions (e.g. `/handoff <platform>`) live under one
// header; otherwise split skills out from built-in commands.
group: isArgCompletion ? 'Options' : isDesktopSlashExtensionCommand(item.text) ? 'Skills' : 'Commands',
// Arg items carry their own meta (the personality/toolset/platform
// blurb). Only command rows get the registry description — looking
// one up for `/personality none` would clobber it with the parent
// command's text.
meta: isArgCompletion ? textValue(item.meta) : desktopSlashDescription(item.text, textValue(item.meta))
}))
// Keep each group contiguous so headers render once: Commands before
// Skills (stable within a group, preserving backend relevance order).
const groupOrder = ['Commands', 'Skills', 'Options']
const items = isArgCompletion
? decorated
: [...decorated].sort((a, b) => groupOrder.indexOf(a.group) - groupOrder.indexOf(b.group))
return { items, query }
} catch {
return { items: [], query }
}
},
[gateway]
[gateway, skinThemes, activeSkin]
)
const toItem = useCallback((entry: CompletionEntry, index: number): Unstable_TriggerItem => {
@@ -93,6 +212,8 @@ export function useSlashCompletions(options: { gateway: HermesGateway | null }):
command,
display,
meta,
group: textValue(entry.group),
action: textValue(entry.action),
// Provide rawText so hermesDirectiveFormatter.serialize uses the
// direct-insertion path instead of the legacy @type:id fallback.
// Without this, the item.id (which includes a "|index" suffix for

View File

@@ -13,17 +13,26 @@ import {
useState
} from 'react'
import { hermesDirectiveFormatter } from '@/components/assistant-ui/directive-text'
import { hermesDirectiveFormatter, type SlashChipKind } from '@/components/assistant-ui/directive-text'
import { composerFill, composerSurfaceGlass } from '@/components/chat/composer-dock'
import { Button } from '@/components/ui/button'
import { useMediaQuery } from '@/hooks/use-media-query'
import { useResizeObserver } from '@/hooks/use-resize-observer'
import { useI18n } from '@/i18n'
import { chatMessageText } from '@/lib/chat-messages'
import { SLASH_COMMAND_RE } from '@/lib/chat-runtime'
import { desktopSlashCommandTakesArgs } from '@/lib/desktop-slash-commands'
import { DATA_IMAGE_URL_RE } from '@/lib/embedded-images'
import { triggerHaptic } from '@/lib/haptics'
import { cn } from '@/lib/utils'
import { $composerAttachments, clearComposerAttachments, type ComposerAttachment } from '@/store/composer'
import {
$composerAttachments,
clearComposerAttachments,
clearSessionDraft,
type ComposerAttachment,
stashSessionDraft,
takeSessionDraft
} from '@/store/composer'
import {
browseBackward,
browseForward,
@@ -40,8 +49,10 @@ import {
shouldAutoDrainOnSettle,
updateQueuedPrompt
} from '@/store/composer-queue'
import { $gatewayState, $messages } from '@/store/session'
import { $statusItemsBySession } from '@/store/composer-status'
import { $gatewayState, $messages, setSessionPickerOpen } from '@/store/session'
import { $threadScrolledUp } from '@/store/thread-scroll'
import { useTheme } from '@/themes'
import { extractDroppedFiles, HERMES_PATHS_MIME, partitionDroppedFiles } from '../hooks/use-composer-actions'
@@ -71,12 +82,14 @@ import {
import { QueuePanel } from './queue-panel'
import {
composerPlainText,
normalizeComposerEditorDom,
placeCaretEnd,
refChipElement,
renderComposerContents,
RICH_INPUT_SLOT
RICH_INPUT_SLOT,
slashChipElement
} from './rich-editor'
import { SkinSlashPopover } from './skin-slash-popover'
import { ComposerStatusStack } from './status-stack'
import { detectTrigger, extractClipboardImageBlobs, textBeforeCaret, type TriggerState } from './text-utils'
import { ComposerTriggerPopover } from './trigger-popover'
import type { ChatBarProps } from './types'
@@ -95,6 +108,30 @@ const COMPOSER_FADE_BACKGROUND =
const pickPlaceholder = (pool: readonly string[]) => pool[Math.floor(Math.random() * pool.length)]
/** Completion items can carry an `action` (set in use-slash-completions) that
* runs a side effect on pick instead of inserting a chip — e.g. the session
* picker's "Browse all…" entry opens the overlay. Table-driven so new action
* items are a registry row, not a composer branch. */
const COMPLETION_ACTIONS: Record<string, () => void> = {
'session-picker': () => setSessionPickerOpen(true)
}
/** Map a picked `/` completion to its pill accent. Driven by the completion
* group set in use-slash-completions (Skills / Themes / Commands|Options). */
function slashChipKindForItem(item: Unstable_TriggerItem): SlashChipKind {
const group = (item.metadata as { group?: unknown } | undefined)?.group
if (group === 'Skills') {
return 'skill'
}
if (group === 'Themes') {
return 'theme'
}
return 'command'
}
interface QueueEditState {
attachments: ComposerAttachment[]
draft: string
@@ -104,6 +141,10 @@ interface QueueEditState {
const cloneAttachments = (attachments: ComposerAttachment[]) => attachments.map(a => ({ ...a }))
// Quiet period after the last keystroke before persisting the draft;
// unmount/pagehide flushes bypass it.
const DRAFT_PERSIST_DEBOUNCE_MS = 400
export function ChatBar({
busy,
cwd,
@@ -131,6 +172,7 @@ export function ChatBar({
const draft = useAuiState(s => s.composer.text)
const attachments = useStore($composerAttachments)
const queuedPromptsBySession = useStore($queuedPromptsBySession)
const statusItemsBySession = useStore($statusItemsBySession)
const scrolledUp = useStore($threadScrolledUp)
const sessionMessages = useStore($messages)
const activeQueueSessionKey = queueSessionKey || sessionId || null
@@ -140,11 +182,25 @@ export function ChatBar({
[activeQueueSessionKey, queuedPromptsBySession]
)
// Status items (subagents, background processes) are keyed by the RUNTIME
// session id — gateway events and process.list both speak that id. Only the
// queue uses the stored-session fallback key (prompts can queue pre-resume).
const statusSessionId = sessionId ?? null
const statusStackVisible = useMemo(
() =>
queuedPrompts.length > 0 || (statusSessionId ? (statusItemsBySession[statusSessionId]?.length ?? 0) > 0 : false),
[queuedPrompts.length, statusItemsBySession, statusSessionId]
)
const composerRef = useRef<HTMLFormElement | null>(null)
const composerSurfaceRef = useRef<HTMLDivElement | null>(null)
const editorRef = useRef<HTMLDivElement | null>(null)
const draftRef = useRef(draft)
const previousBusyRef = useRef(busy)
const pendingDraftPersistRef = useRef<{ scope: string | null; text: string } | null>(null)
const activeQueueSessionKeyRef = useRef(activeQueueSessionKey)
activeQueueSessionKeyRef.current = activeQueueSessionKey
const drainingQueueRef = useRef(false)
const urlInputRef = useRef<HTMLInputElement | null>(null)
@@ -156,14 +212,17 @@ export function ChatBar({
const [dragActive, setDragActive] = useState(false)
const [queueEdit, setQueueEdit] = useState<QueueEditState | null>(null)
const [focusRequestId, setFocusRequestId] = useState(0)
const queueEditRef = useRef(queueEdit)
queueEditRef.current = queueEdit
const dragDepthRef = useRef(0)
const composingRef = useRef(false) // true during IME composition (CJK input)
const lastSpokenIdRef = useRef<string | null>(null)
const narrow = useMediaQuery('(max-width: 30rem)')
const { availableThemes, themeName } = useTheme()
const at = useAtCompletions({ gateway: gateway ?? null, sessionId: sessionId ?? null, cwd: cwd ?? null })
const slash = useSlashCompletions({ gateway: gateway ?? null })
const slash = useSlashCompletions({ activeSkin: themeName, gateway: gateway ?? null, skinThemes: availableThemes })
const stacked = expanded || narrow || tight
const trimmedDraft = draft.trim()
@@ -171,10 +230,12 @@ export function ChatBar({
const canSubmit = busy || hasComposerPayload
const editingQueuedPrompt = queueEdit ? (queuedPrompts.find(entry => entry.id === queueEdit.entryId) ?? null) : null
const busyAction = busy && hasComposerPayload ? 'queue' : 'stop'
// Steer only makes sense mid-turn, text-only (the gateway can't carry images
// into a tool result) and never for a slash command (those execute inline).
const canSteer =
busy && !!onSteer && attachments.length === 0 && trimmedDraft.length > 0 && !SLASH_COMMAND_RE.test(trimmedDraft)
const showHelpHint = draft === '?'
const { t } = useI18n()
@@ -462,12 +523,6 @@ export function ChatBar({
})
}, [])
const selectSkinSlashCommand = (command: string) => {
draftRef.current = command
aui.composer().setText(command)
requestMainFocus()
}
const handlePaste = (event: ClipboardEvent<HTMLDivElement>) => {
const imageBlobs = extractClipboardImageBlobs(event.clipboardData)
@@ -563,9 +618,7 @@ export function ChatBar({
// (which drives `hasComposerPayload` → the send button). Shared by the input
// and compositionend paths so committed IME text reaches state through either.
const flushEditorToDraft = (editor: HTMLDivElement) => {
if (editor.childNodes.length === 1 && editor.firstChild?.nodeName === 'BR') {
editor.replaceChildren()
}
normalizeComposerEditorDom(editor)
const nextDraft = composerPlainText(editor)
@@ -620,16 +673,49 @@ export function ChatBar({
return
}
// Action items (e.g. "Browse all sessions…") run a side effect instead of
// inserting a chip: strip the typed trigger token, then fire the action.
const completionAction = (item.metadata as { action?: unknown } | undefined)?.action
const runAction = typeof completionAction === 'string' ? COMPLETION_ACTIONS[completionAction] : undefined
if (runAction) {
const current = composerPlainText(editor)
const prefix = current.slice(0, Math.max(0, current.length - trigger.tokenLength))
renderComposerContents(editor, prefix)
placeCaretEnd(editor)
draftRef.current = composerPlainText(editor)
aui.composer().setText(draftRef.current)
closeTrigger()
runAction()
requestMainFocus()
return
}
const serialized = hermesDirectiveFormatter.serialize(item)
const starter = serialized.endsWith(':')
// Picking a bare arg-taking command (e.g. `/personality`) shouldn't commit
// it — expand to its options step so the popover shows the inline list, just
// as typing `/personality ` by hand would. A serialized value with a space is
// already an arg pick (`/personality alice`), so it commits normally.
const command = (item.metadata as { command?: string } | undefined)?.command ?? ''
const expandsToArgs = trigger.kind === '/' && !serialized.includes(' ') && desktopSlashCommandTakesArgs(command)
const text = starter || serialized.endsWith(' ') ? serialized : `${serialized} `
const directive = !starter && serialized.match(/^@([^:]+):(.+)$/)
// No pill while expanding — the bare command stays plain text until an arg
// is picked, at which point a single pill is emitted for the full command.
const slashKind = !expandsToArgs && trigger.kind === '/' ? slashChipKindForItem(item) : null
const keepTriggerOpen = starter || expandsToArgs
const finish = () => {
draftRef.current = composerPlainText(editor)
aui.composer().setText(draftRef.current)
requestMainFocus()
starter ? window.setTimeout(refreshTrigger, 0) : closeTrigger()
keepTriggerOpen ? window.setTimeout(refreshTrigger, 0) : closeTrigger()
}
const sel = window.getSelection()
@@ -639,7 +725,20 @@ export function ChatBar({
if (!sel || !range || node?.nodeType !== Node.TEXT_NODE || offset < trigger.tokenLength) {
const current = composerPlainText(editor)
renderComposerContents(editor, `${current.slice(0, Math.max(0, current.length - trigger.tokenLength))}${text}`)
const prefix = current.slice(0, Math.max(0, current.length - trigger.tokenLength))
if (slashKind) {
// Two-step arg picks (e.g. `/handoff` pill already inserted, now picking
// the platform) land here because the caret sits past a contenteditable
// chip. Rebuild the prefix and re-emit a single pill for the full command.
renderComposerContents(editor, prefix)
editor.append(slashChipElement(serialized, slashKind), document.createTextNode(' '))
placeCaretEnd(editor)
return finish()
}
renderComposerContents(editor, `${prefix}${text}`)
placeCaretEnd(editor)
return finish()
@@ -650,8 +749,13 @@ export function ChatBar({
replaceRange.setEnd(node, offset)
replaceRange.deleteContents()
if (directive) {
const chip = refChipElement(directive[1], directive[2])
const chip = slashKind
? slashChipElement(serialized, slashKind)
: directive
? refChipElement(directive[1], directive[2])
: null
if (chip) {
const space = document.createTextNode(' ')
const fragment = document.createDocumentFragment()
fragment.append(chip, space)
@@ -1022,6 +1126,66 @@ export function ChatBar({
}
}
const stashAt = (scope: string | null, text = draftRef.current, attachments = $composerAttachments.get()) =>
stashSessionDraft(scope, text, attachments)
// Per-thread draft swap — the composer's only session coupling. Lifecycle
// never clears composer state; this effect alone stashes on leave, restores
// on enter. Keyed writes are idempotent, so no skip-sentinel.
useEffect(() => {
const { attachments, text } = takeSessionDraft(activeQueueSessionKey)
loadIntoComposer(text, attachments)
return () => {
const editing = queueEditRef.current
if (editing?.sessionKey === activeQueueSessionKey) {
stashAt(activeQueueSessionKey, editing.draft, editing.attachments)
} else if (!isBrowsingHistory(sessionId)) {
stashAt(activeQueueSessionKey)
}
}
}, [activeQueueSessionKey]) // eslint-disable-line react-hooks/exhaustive-deps
// Debounced stash into the active scope. Skipped while browsing history or
// editing a queued prompt — recalled text must not clobber the real draft.
useEffect(() => {
if (isBrowsingHistory(sessionId) || queueEdit) {
return
}
pendingDraftPersistRef.current = { scope: activeQueueSessionKey, text: draft }
const handle = window.setTimeout(() => {
pendingDraftPersistRef.current = null
stashAt(activeQueueSessionKey, draft)
}, DRAFT_PERSIST_DEBOUNCE_MS)
return () => window.clearTimeout(handle)
}, [activeQueueSessionKey, draft, queueEdit, sessionId])
// pagehide is load-bearing: React skips effect cleanups on reload, so Cmd+R
// inside the debounce window would drop trailing keystrokes without this.
useEffect(() => {
const flushPendingDraftPersist = () => {
const pending = pendingDraftPersistRef.current
if (!pending) {
return
}
pendingDraftPersistRef.current = null
stashAt(pending.scope, pending.text)
}
window.addEventListener('pagehide', flushPendingDraftPersist)
return () => {
window.removeEventListener('pagehide', flushPendingDraftPersist)
flushPendingDraftPersist()
}
}, [])
const beginQueuedEdit = (entry: QueuedPromptEntry) => {
if (!activeQueueSessionKey || queueEdit) {
return
@@ -1224,20 +1388,38 @@ export function ChatBar({
}
}, [busy, drainNextQueued, queuedPrompts.length])
// Clean up queue edit when its target disappears (session swap or external delete).
// Queue-edit cleanup: on session swap the scope effect already stashed the
// edit snapshot; only restore into the composer when still on the same scope.
useEffect(() => {
if (!queueEdit) {
return
}
if (queueEdit.sessionKey === activeQueueSessionKey && editingQueuedPrompt) {
return
if (queueEdit.sessionKey === activeQueueSessionKey) {
if (editingQueuedPrompt) {
return
}
loadIntoComposer(queueEdit.draft, queueEdit.attachments)
}
loadIntoComposer(queueEdit.draft, queueEdit.attachments)
setQueueEdit(null)
}, [activeQueueSessionKey, editingQueuedPrompt, queueEdit]) // eslint-disable-line react-hooks/exhaustive-deps
const dispatchSubmit = (text: string, attachments?: ComposerAttachment[]) => {
const submittedScope = activeQueueSessionKeyRef.current
const submittedAttachments = attachments ?? []
const restore = () => {
loadIntoComposer(text, submittedAttachments)
stashAt(activeQueueSessionKeyRef.current, text, submittedAttachments)
}
void Promise.resolve(attachments ? onSubmit(text, { attachments }) : onSubmit(text))
.then(accepted => void (accepted === false ? restore() : clearSessionDraft(submittedScope)))
.catch(restore)
}
const submitDraft = () => {
// Source the text from the DOM editor, not React state. The AUI composer
// state (`draft`) and the derived `hasComposerPayload` lag the DOM by a
@@ -1248,8 +1430,10 @@ export function ChatBar({
// input event; refresh it from the editor once more to also cover an
// in-flight keystroke that hasn't fired its input event yet.
const editor = editorRef.current
if (editor) {
const domText = composerPlainText(editor)
if (domText !== draftRef.current) {
draftRef.current = domText
aui.composer().setText(domText)
@@ -1270,10 +1454,9 @@ export function ChatBar({
// /send directives). Queuing them would make every slash command wait
// for the current turn to finish, which is how the TUI never behaves.
if (!attachments.length && SLASH_COMMAND_RE.test(text.trim())) {
const submitted = text
triggerHaptic('submit')
clearDraft()
void onSubmit(submitted)
dispatchSubmit(text)
} else if (payloadPresent) {
queueCurrentDraft()
} else {
@@ -1285,12 +1468,12 @@ export function ChatBar({
} else if (!payloadPresent && queuedPrompts.length > 0) {
void drainNextQueued()
} else if (payloadPresent) {
const submitted = text
const submittedAttachments = cloneAttachments(attachments)
triggerHaptic('submit')
resetBrowseState(sessionId)
clearDraft()
clearComposerAttachments()
void onSubmit(submitted, { attachments })
dispatchSubmit(text, submittedAttachments)
}
focusInput()
@@ -1457,7 +1640,7 @@ export function ChatBar({
onPaste={handlePaste}
ref={editorRef}
role="textbox"
spellCheck="true"
spellCheck={false}
suppressContentEditableWarning
/>
{/* assistant-ui requires ComposerPrimitive.Input somewhere in the tree
@@ -1476,7 +1659,15 @@ export function ChatBar({
`asChild` swaps TextareaAutosize for a Radix Slot wrapping our
plain <textarea>, which carries the binding but skips autosize. */}
<ComposerPrimitive.Input asChild submitMode="ctrlEnter" tabIndex={-1} unstable_focusOnScrollToBottom={false}>
<textarea aria-hidden className="sr-only" tabIndex={-1} />
<textarea
aria-hidden
autoCapitalize="off"
autoComplete="off"
autoCorrect="off"
className="sr-only"
spellCheck={false}
tabIndex={-1}
/>
</ComposerPrimitive.Input>
</div>
)
@@ -1488,6 +1679,7 @@ export function ChatBar({
className="group/composer absolute bottom-0 left-1/2 z-30 w-[min(var(--composer-width),calc(100%-2rem))] max-w-full -translate-x-1/2 rounded-2xl pt-2 pb-[var(--composer-shell-pad-block-end)]"
data-drag-active={dragActive ? '' : undefined}
data-slot="composer-root"
data-status-stack={statusStackVisible ? '' : undefined}
data-thread-scrolled-up={scrolledUp ? '' : undefined}
onDragEnter={handleDragEnter}
onDragLeave={handleDragLeave}
@@ -1515,27 +1707,30 @@ export function ChatBar({
onPick={replaceTriggerWithChip}
/>
)}
<SkinSlashPopover draft={draft} onSelect={selectSkinSlashCommand} />
{activeQueueSessionKey && queuedPrompts.length > 0 && (
// Out of flow so the queue never inflates the composer's measured
// height (that drives thread bottom padding → chat resizes on
// queue). Overlaps -mb-2 onto the surface's top border for a shared
// edge; capped + scrollable. Overlays the chat instead of pushing it.
<div className="absolute inset-x-0 bottom-full z-6 -mb-2 max-h-[40vh] overflow-y-auto">
<QueuePanel
busy={busy}
editingId={queueEdit?.entryId ?? null}
entries={queuedPrompts}
onDelete={id => {
if (removeQueuedPrompt(activeQueueSessionKey, id) && queueEdit?.entryId === id) {
exitQueuedEdit('cancel')
}
}}
onEdit={beginQueuedEdit}
onSendNow={id => void sendQueuedNow(id)}
/>
</div>
)}
{/* Session-scoped status stack (todos, subagents, background tasks,
queue). Out of flow so it never inflates the composer's measured
height; it overlays the chat instead of pushing it, and publishes
its own --status-stack-measured-height so the thread's clearance
accounts for it. Collapses to nothing when every status is empty. */}
<ComposerStatusStack
queue={
activeQueueSessionKey && queuedPrompts.length > 0 ? (
<QueuePanel
busy={busy}
editingId={queueEdit?.entryId ?? null}
entries={queuedPrompts}
onDelete={id => {
if (removeQueuedPrompt(activeQueueSessionKey, id) && queueEdit?.entryId === id) {
exitQueuedEdit('cancel')
}
}}
onEdit={beginQueuedEdit}
onSendNow={id => void sendQueuedNow(id)}
/>
) : null
}
sessionId={statusSessionId}
/>
<div
className="pointer-events-none absolute inset-0 rounded-[inherit]"
style={{ background: COMPOSER_FADE_BACKGROUND }}
@@ -1543,10 +1738,10 @@ export function ChatBar({
<div className="relative w-full rounded-[inherit]">
<div
className={cn(
'relative z-4 isolate rounded-[inherit] border border-[color-mix(in_srgb,var(--dt-composer-ring)_calc(18%*var(--composer-ring-strength)),var(--dt-input))] transition-[border-color] duration-200 ease-out',
'group/composer-surface relative z-4 isolate rounded-[inherit] border border-[color-mix(in_srgb,var(--dt-composer-ring)_calc(18%*var(--composer-ring-strength)),var(--dt-input))] transition-[border-color] duration-200 ease-out focus-within:border-[color-mix(in_srgb,var(--dt-composer-ring)_calc(45%*var(--composer-ring-strength)),transparent)]',
COMPOSER_DROP_FADE_CLASS,
'group-focus-within/composer:border-[color-mix(in_srgb,var(--dt-composer-ring)_calc(45%*var(--composer-ring-strength)),transparent)]',
'group-has-data-[state=open]/composer:border-t-transparent',
'group-data-[status-stack]/composer:border-t-transparent',
dragActive && COMPOSER_DROP_ACTIVE_CLASS
)}
data-slot="composer-surface"
@@ -1556,20 +1751,14 @@ export function ChatBar({
aria-hidden
className={cn(
'pointer-events-none absolute inset-0 -z-10 rounded-[inherit]',
'bg-[color-mix(in_srgb,var(--dt-card)_72%,transparent)]',
'backdrop-blur-[0.75rem] backdrop-saturate-[1.12]',
'[-webkit-backdrop-filter:blur(0.75rem)_saturate(1.12)]',
'transition-[background-color] duration-150 ease-out',
'group-data-[thread-scrolled-up]/composer:bg-[color-mix(in_srgb,var(--dt-card)_48%,transparent)]',
'group-focus-within/composer:bg-[color-mix(in_srgb,var(--dt-card)_85%,transparent)]'
composerFill,
composerSurfaceGlass
)}
/>
<div
className={cn(
'relative z-1 flex min-h-0 w-full flex-col gap-(--composer-row-gap) overflow-hidden rounded-[inherit] px-(--composer-surface-pad-x) py-(--composer-surface-pad-y) transition-opacity duration-200 ease-out',
scrolledUp
? 'opacity-30 group-hover/composer:opacity-100 group-focus-within/composer:opacity-100'
: 'opacity-100'
scrolledUp ? 'opacity-30 group-hover/composer:opacity-100 group-focus-within/composer-surface:opacity-100' : 'opacity-100'
)}
data-slot="composer-fade"
>
@@ -1644,12 +1833,8 @@ export function ChatBarFallback() {
aria-hidden
className={cn(
'pointer-events-none absolute inset-0 -z-10 rounded-[inherit]',
'bg-[color-mix(in_srgb,var(--dt-card)_72%,transparent)]',
'backdrop-blur-[0.75rem] backdrop-saturate-[1.12]',
'[-webkit-backdrop-filter:blur(0.75rem)_saturate(1.12)]',
'transition-[background-color] duration-150 ease-out',
'group-data-[thread-scrolled-up]/composer:bg-[color-mix(in_srgb,var(--dt-card)_48%,transparent)]',
'group-focus-within/composer:bg-[color-mix(in_srgb,var(--dt-card)_85%,transparent)]'
composerFill,
composerSurfaceGlass
)}
/>
</div>

View File

@@ -3,7 +3,12 @@ import { contextPath } from '@/lib/chat-runtime'
import type { DroppedFile } from '../hooks/use-composer-actions'
import { composerPlainText, escapeHtml, placeCaretEnd, refChipHtml } from './rich-editor'
import {
composerPlainText,
normalizeComposerEditorDom,
placeCaretEnd,
refChipElement
} from './rich-editor'
/** A chip to insert: a raw `@kind:value` string, or a typed value + display label. */
export type InlineRefInput = string | { kind: string; label?: string; value: string }
@@ -89,56 +94,102 @@ export function droppedFileInlineRefs(candidates: DroppedFile[], cwd: string | n
return candidates.map(candidate => droppedFileInlineRef(candidate, cwd)).filter((ref): ref is string => Boolean(ref))
}
export function insertInlineRefsIntoEditor(editor: HTMLDivElement, refs: readonly InlineRefInput[]) {
if (!refs.length) {
function parseInlineRef(ref: InlineRefInput): { kind: string; label?: string; rawValue: string } | null {
if (typeof ref !== 'string') {
return { kind: ref.kind, label: ref.label, rawValue: ref.value }
}
const match = ref.match(/^@([^:]+):(.+)$/)
if (!match) {
return null
}
const refsHtml = refs
.map(ref => {
if (typeof ref !== 'string') {
return refChipHtml(ref.kind, ref.value, ref.label)
}
return { kind: match[1] || 'file', rawValue: match[2] || '' }
}
const match = ref.match(/^@([^:]+):(.+)$/)
function plainTextInRange(editor: HTMLDivElement, range: Range, edge: 'after' | 'before') {
const slice = range.cloneRange()
slice.selectNodeContents(editor)
return match ? refChipHtml(match[1], match[2]) : escapeHtml(ref)
})
.join(' ')
if (edge === 'before') {
slice.setEnd(range.startContainer, range.startOffset)
} else {
slice.setStart(range.endContainer, range.endOffset)
}
const container = document.createElement('div')
container.appendChild(slice.cloneContents())
return composerPlainText(container)
}
function buildRefFragment(
refs: readonly { kind: string; label?: string; rawValue: string }[],
{ needsBeforeSpace, needsAfterSpace }: { needsAfterSpace: boolean; needsBeforeSpace: boolean }
) {
const fragment = document.createDocumentFragment()
if (needsBeforeSpace) {
fragment.append(document.createTextNode(' '))
}
refs.forEach((ref, index) => {
if (index > 0) {
fragment.append(document.createTextNode(' '))
}
fragment.append(refChipElement(ref.kind, ref.rawValue, ref.label))
})
if (needsAfterSpace) {
fragment.append(document.createTextNode(' '))
}
return fragment
}
export function insertInlineRefsIntoEditor(editor: HTMLDivElement, refs: readonly InlineRefInput[]) {
const parsed = refs.map(parseInlineRef).filter((ref): ref is NonNullable<typeof ref> => ref !== null)
if (!parsed.length) {
return null
}
editor.focus({ preventScroll: true })
const selection = window.getSelection()
const range =
selection?.rangeCount && editor.contains(selection.getRangeAt(0).commonAncestorContainer)
? selection.getRangeAt(0)
: null
editor.focus({ preventScroll: true })
if (range && selection) {
const beforeText = plainTextInRange(editor, range, 'before')
const afterText = plainTextInRange(editor, range, 'after')
if (range) {
const beforeRange = range.cloneRange()
beforeRange.selectNodeContents(editor)
beforeRange.setEnd(range.startContainer, range.startOffset)
const beforeContainer = document.createElement('div')
beforeContainer.appendChild(beforeRange.cloneContents())
const afterRange = range.cloneRange()
afterRange.selectNodeContents(editor)
afterRange.setStart(range.endContainer, range.endOffset)
const afterContainer = document.createElement('div')
afterContainer.appendChild(afterRange.cloneContents())
const beforeText = composerPlainText(beforeContainer)
const afterText = composerPlainText(afterContainer)
const needsBeforeSpace = beforeText.length > 0 && !/\s$/.test(beforeText)
const needsAfterSpace = afterText.length === 0 || !/^\s/.test(afterText)
document.execCommand('insertHTML', false, `${needsBeforeSpace ? ' ' : ''}${refsHtml}${needsAfterSpace ? ' ' : ''}`)
range.insertNode(
buildRefFragment(parsed, {
needsAfterSpace: afterText.length === 0 || !/^\s/.test(afterText),
needsBeforeSpace: beforeText.length > 0 && !/\s$/.test(beforeText)
})
)
range.collapse(false)
selection.removeAllRanges()
selection.addRange(range)
} else {
const current = composerPlainText(editor)
editor.append(
buildRefFragment(parsed, {
needsAfterSpace: true,
needsBeforeSpace: current.length > 0 && !/\s$/.test(current)
})
)
placeCaretEnd(editor)
document.execCommand('insertHTML', false, `${current && !/\s$/.test(current) ? ' ' : ''}${refsHtml} `)
}
normalizeComposerEditorDom(editor)
return composerPlainText(editor)
}

View File

@@ -1,10 +1,7 @@
import { useState } from 'react'
import { StatusRow } from '@/components/chat/status-row'
import { StatusSection } from '@/components/chat/status-section'
import { Button } from '@/components/ui/button'
import { DisclosureCaret } from '@/components/ui/disclosure-caret'
import { Tip } from '@/components/ui/tooltip'
import { type Translations, useI18n } from '@/i18n'
import { ArrowUp, Pencil, Trash2 } from '@/lib/icons'
import { cn } from '@/lib/utils'
import type { QueuedPromptEntry } from '@/store/composer-queue'
@@ -23,108 +20,70 @@ const entryPreview = (entry: QueuedPromptEntry, c: Translations['composer']) =>
export function QueuePanel({ busy, editingId, entries, onDelete, onEdit, onSendNow }: QueuePanelProps) {
const { t } = useI18n()
const c = t.composer
const [collapsed, setCollapsed] = useState(true)
if (entries.length === 0) {
return null
}
return (
<div className="rounded-t-2xl border border-b-0 border-border/65 bg-[color-mix(in_srgb,var(--dt-card)_70%,transparent)] pt-0.5 pb-1 mx-1">
<button
className="flex w-full items-center gap-1.5 px-2 text-left text-[0.6rem] font-medium text-muted-foreground/92 transition-colors hover:text-foreground/90"
onClick={() => setCollapsed(open => !open)}
type="button"
>
<DisclosureCaret className="shrink-0" open={!collapsed} size="1em" />
<span className="truncate">{c.queued(entries.length)}</span>
</button>
<StatusSection label={c.queued(entries.length)}>
{entries.map(entry => {
const isEditing = editingId === entry.id
const attachmentsCount = entry.attachments.length
{!collapsed && (
<div className="space-y-0.5 px-1 pb-0.5">
{entries.map(entry => {
const isEditing = editingId === entry.id
const attachmentsCount = entry.attachments.length
const sendLabel = busy ? c.sendQueuedNext : c.sendQueuedNow
return (
<div
className={cn(
'group/queue-row flex items-center gap-1.5 rounded-lg border border-transparent px-1.5 py-0.5',
'transition-colors duration-300 ease-out hover:bg-(--chrome-action-hover) hover:transition-none',
isEditing && 'border-[color-mix(in_srgb,var(--dt-composer-ring)_40%,transparent)] bg-accent/25'
)}
key={entry.id}
>
<span
aria-hidden
className="h-3.5 w-3.5 shrink-0 rounded-full border border-foreground/35 bg-transparent"
/>
<div className="min-w-0 flex-1">
<p className="truncate text-[0.73rem] leading-4 text-foreground/92">{entryPreview(entry, c)}</p>
{(attachmentsCount > 0 || isEditing) && (
<div className="mt-0.5 flex items-center gap-1.5 text-[0.64rem] text-muted-foreground/75">
{attachmentsCount > 0 && <span>{c.attachments(attachmentsCount)}</span>}
{isEditing && (
<span className="text-[color-mix(in_srgb,var(--dt-composer-ring)_78%,var(--muted-foreground))]">
{c.editingInComposer}
</span>
)}
</div>
)}
</div>
<div
className={cn(
'flex shrink-0 items-center gap-0 transition-opacity',
isEditing
? 'opacity-100'
: 'opacity-0 group-hover/queue-row:opacity-100 group-focus-within/queue-row:opacity-100'
)}
return (
<StatusRow
className={cn(
'border border-transparent',
isEditing && 'border-[color-mix(in_srgb,var(--dt-composer-ring)_40%,transparent)] bg-accent/25'
)}
key={entry.id}
leading={
<span aria-hidden className="size-3.5 shrink-0 rounded-full border border-foreground/35 bg-transparent" />
}
trailing={
<>
<Button
disabled={Boolean(editingId) && !isEditing}
onClick={() => onEdit(entry)}
size="micro"
type="button"
variant="text"
>
<Tip label={c.editQueued}>
<Button
aria-label={c.editQueued}
className="h-5 w-5 rounded-md"
disabled={Boolean(editingId) && !isEditing}
onClick={() => onEdit(entry)}
size="icon-xs"
type="button"
variant="ghost"
>
<Pencil size={11} />
</Button>
</Tip>
<Tip label={sendLabel}>
<Button
aria-label={sendLabel}
className="h-5 w-5 rounded-md"
disabled={isEditing}
onClick={() => onSendNow(entry.id)}
size="icon-xs"
type="button"
variant="ghost"
>
<ArrowUp size={11} />
</Button>
</Tip>
<Tip label={c.deleteQueued}>
<Button
aria-label={c.deleteQueued}
className="h-5 w-5 rounded-md"
onClick={() => onDelete(entry.id)}
size="icon-xs"
type="button"
variant="ghost"
>
<Trash2 size={11} />
</Button>
</Tip>
{c.queueEdit}
</Button>
<Button
disabled={isEditing}
onClick={() => onSendNow(entry.id)}
size="micro"
type="button"
variant="secondary"
>
{busy ? c.queueSendNext : c.queueSend}
</Button>
<Button onClick={() => onDelete(entry.id)} size="micro" type="button" variant="text">
{c.queueDelete}
</Button>
</>
}
trailingVisible={isEditing}
>
<div className="min-w-0 flex-1">
<p className="truncate text-[0.73rem] leading-4 text-foreground/92">{entryPreview(entry, c)}</p>
{(attachmentsCount > 0 || isEditing) && (
<div className="mt-0.5 flex items-center gap-1.5 text-[0.64rem] text-muted-foreground/75">
{attachmentsCount > 0 && <span>{c.attachments(attachmentsCount)}</span>}
{isEditing && (
<span className="text-[color-mix(in_srgb,var(--dt-composer-ring)_78%,var(--muted-foreground))]">
{c.editingInComposer}
</span>
)}
</div>
</div>
)
})}
</div>
)}
</div>
)}
</div>
</StatusRow>
)
})}
</StatusSection>
)
}

View File

@@ -1,6 +1,13 @@
import { describe, expect, it } from 'vitest'
import { composerPlainText, renderComposerContents, RICH_INPUT_SLOT } from './rich-editor'
import { insertInlineRefsIntoEditor } from './inline-refs'
import {
composerPlainText,
normalizeComposerEditorDom,
refChipElement,
renderComposerContents,
RICH_INPUT_SLOT
} from './rich-editor'
describe('renderComposerContents', () => {
it('renders refs and raw text without interpreting user text as HTML', () => {
@@ -16,3 +23,39 @@ describe('renderComposerContents', () => {
expect(composerPlainText(editor)).toBe('@file:`<img src=x onerror=alert(1)>` <b>raw</b>')
})
})
describe('normalizeComposerEditorDom', () => {
it('unwraps a single insertHTML wrapper div so plain text stays one line', () => {
const editor = document.createElement('div')
editor.dataset.slot = RICH_INPUT_SLOT
editor.innerHTML = '<div><span data-ref-text="@file:`src/foo.ts`" contenteditable="false">foo.ts</span> </div>'
normalizeComposerEditorDom(editor)
expect(composerPlainText(editor)).toBe('@file:`src/foo.ts` ')
expect(editor.querySelector(':scope > div')).toBeNull()
})
it('removes a trailing br after a ref chip', () => {
const editor = document.createElement('div')
editor.dataset.slot = RICH_INPUT_SLOT
editor.append(refChipElement('file', '`src/foo.ts`'), document.createElement('br'))
normalizeComposerEditorDom(editor)
expect(composerPlainText(editor)).toBe('@file:`src/foo.ts`')
expect(editor.querySelector('br')).toBeNull()
})
})
describe('insertInlineRefsIntoEditor', () => {
it('inserts chips without wrapper divs or spurious newlines', () => {
const editor = document.createElement('div')
editor.dataset.slot = RICH_INPUT_SLOT
insertInlineRefsIntoEditor(editor, ['@file:`src/foo.ts`'])
expect(editor.querySelector(':scope > div')).toBeNull()
expect(composerPlainText(editor)).toBe('@file:`src/foo.ts` ')
})
})

View File

@@ -10,7 +10,10 @@ import {
DIRECTIVE_CHIP_CLASS,
directiveIconElement,
directiveIconSvg,
formatRefValue
formatRefValue,
slashChipClass,
type SlashChipKind,
slashIconElement
} from '@/components/assistant-ui/directive-text'
export const RICH_INPUT_SLOT = 'composer-rich-input'
@@ -77,6 +80,24 @@ export function refChipElement(kind: string, rawValue: string, displayLabel?: st
return chip
}
/** A non-editable pill for a picked slash command (`/skin nous`, `/tropes`).
* `data-ref-text` carries the literal command so `composerPlainText` round-trips
* it back to the exact text that gets submitted. */
export function slashChipElement(command: string, kind: SlashChipKind, label?: string) {
const chip = document.createElement('span')
const text = document.createElement('span')
chip.contentEditable = 'false'
chip.dataset.refText = command
chip.dataset.slashKind = kind
chip.className = slashChipClass(kind)
text.className = 'truncate'
text.textContent = label || command
chip.append(slashIconElement(kind), text)
return chip
}
function appendTextWithBreaks(target: DocumentFragment | HTMLElement, text: string) {
const lines = text.split('\n')
@@ -163,3 +184,36 @@ export function placeCaretEnd(element: HTMLElement) {
selection?.removeAllRanges()
selection?.addRange(range)
}
/** Drop contenteditable junk that serializes as `\n` and falsely expands the composer. */
export function normalizeComposerEditorDom(editor: HTMLElement) {
if (editor.childNodes.length === 1 && editor.firstChild?.nodeName === 'BR') {
editor.replaceChildren()
return
}
if (editor.childNodes.length === 1 && editor.firstChild?.nodeType === Node.ELEMENT_NODE) {
const wrapper = editor.firstChild as HTMLElement
if (wrapper.tagName === 'DIV' && wrapper.dataset.slot !== RICH_INPUT_SLOT) {
editor.replaceChildren(...Array.from(wrapper.childNodes))
}
}
const last = editor.lastChild
if (last?.nodeName !== 'BR') {
return
}
let prev: ChildNode | null = last.previousSibling
while (prev?.nodeType === Node.TEXT_NODE && !(prev.textContent || '').trim()) {
prev = prev.previousSibling
}
if ((prev as HTMLElement | null)?.dataset.refText) {
editor.removeChild(last)
}
}

View File

@@ -1,61 +0,0 @@
import { useI18n } from '@/i18n'
import { desktopSkinSlashCompletions } from '@/lib/desktop-slash-commands'
import { triggerHaptic } from '@/lib/haptics'
import { useTheme } from '@/themes/context'
import { COMPLETION_DRAWER_CLASS, COMPLETION_DRAWER_ROW_CLASS, CompletionDrawerEmpty } from './completion-drawer'
interface SkinSlashPopoverProps {
draft: string
onSelect: (command: string) => void
}
export function SkinSlashPopover({ draft, onSelect }: SkinSlashPopoverProps) {
const { t } = useI18n()
const c = t.composer
const { availableThemes, themeName } = useTheme()
const match = draft.match(/^\/skin\s+(\S*)$/i)
if (!match) {
return null
}
const items = desktopSkinSlashCompletions(availableThemes, themeName, match[1] ?? '')
return (
<div
aria-label={c.themeSuggestions}
className={COMPLETION_DRAWER_CLASS}
data-slot="composer-skin-completion-drawer"
data-state="open"
role="listbox"
>
<div className="grid gap-0.5 pt-0.5">
{items.length === 0 ? (
<CompletionDrawerEmpty title={c.noMatchingThemes}>
{c.themeTryPre}
<span className="font-mono text-foreground/80">/skin list</span>
{c.themeTryPost}
</CompletionDrawerEmpty>
) : (
items.map(item => (
<button
className={COMPLETION_DRAWER_ROW_CLASS}
key={item.text}
onClick={() => {
triggerHaptic('selection')
onSelect(item.text)
}}
onMouseDown={event => event.preventDefault()}
role="option"
type="button"
>
<span className="shrink-0 font-mono font-medium leading-5 text-foreground">{item.display}</span>
<span className="min-w-0 truncate leading-5 text-muted-foreground/80">{item.meta}</span>
</button>
))
)}
</div>
</div>
)
}

View File

@@ -0,0 +1,194 @@
import { useStore } from '@nanostores/react'
import { type ReactNode, useEffect, useLayoutEffect, useMemo, useRef } from 'react'
import { useNavigate } from 'react-router-dom'
import { blurComposerInput } from '@/app/chat/composer/focus'
import { AGENTS_ROUTE } from '@/app/routes'
import { composerDockCard } from '@/components/chat/composer-dock'
import { StatusSection } from '@/components/chat/status-section'
import { Button } from '@/components/ui/button'
import { Codicon } from '@/components/ui/codicon'
import { type Translations, useI18n } from '@/i18n'
import { cn } from '@/lib/utils'
import {
$statusItemsBySession,
type ComposerStatusItem,
dismissBackgroundProcess,
groupStatusItems,
refreshBackgroundProcesses,
type StatusGroup,
stopBackgroundProcess
} from '@/store/composer-status'
import { $threadScrolledUp } from '@/store/thread-scroll'
import { openSessionInNewWindow } from '@/store/windows'
import { StatusItemRow } from './status-row'
// Slow safety-net poll for silent exits (processes without notify_on_complete
// emit no event when they die). Only armed while a running row is on screen.
const BACKGROUND_POLL_MS = 5_000
const groupLabel = (group: StatusGroup, s: Translations['statusStack']) => {
if (group.type === 'todo') {
return s.todos(group.items.filter(i => i.todoStatus === 'completed').length, group.items.length)
}
return group.type === 'subagent' ? s.subagents(group.items.length) : s.background(group.items.length)
}
interface ComposerStatusStackProps {
/** The queue, built by the composer (it owns the queue's callbacks). Rendered
* as the last group so it stays fused to the composer like before. */
queue: ReactNode
sessionId: null | string
}
/**
* The status "sink" above the composer: one card (the queue's chrome) holding
* every session-scoped status — subagents, background tasks, queue — grouped by
* type and separated by light dividers. Collapses to nothing when empty.
*/
export function ComposerStatusStack({ queue, sessionId }: ComposerStatusStackProps) {
const { t } = useI18n()
const navigate = useNavigate()
const itemsBySession = useStore($statusItemsBySession)
const scrolledUp = useStore($threadScrolledUp)
const groups = useMemo(
() => groupStatusItems(sessionId ? (itemsBySession[sessionId] ?? []) : []),
[itemsBySession, sessionId]
)
// Seed from the registry on session open; event-driven refreshes (terminal /
// process tool completions) live in use-message-stream.
useEffect(() => {
if (sessionId) {
void refreshBackgroundProcesses(sessionId)
}
}, [sessionId])
const hasRunningBackground = groups.some(g => g.type === 'background' && g.items.some(i => i.state === 'running'))
useEffect(() => {
if (!sessionId || !hasRunningBackground) {
return
}
const timer = setInterval(() => void refreshBackgroundProcesses(sessionId), BACKGROUND_POLL_MS)
return () => clearInterval(timer)
}, [hasRunningBackground, sessionId])
const openAgents = () => navigate(AGENTS_ROUTE)
const openSubagent = (item: ComposerStatusItem) =>
item.sessionId ? void openSessionInNewWindow(item.sessionId, { watch: true }) : openAgents()
const sections: { key: string; node: ReactNode }[] = groups.map(group => ({
key: group.type,
node: (
<StatusSection
accessory={
group.type === 'subagent' ? (
<Button
className="text-muted-foreground/75 hover:text-foreground/90"
onClick={openAgents}
size="micro"
type="button"
variant="text"
>
{t.statusStack.agents}
</Button>
) : undefined
}
defaultCollapsed={group.type !== 'todo'}
icon={
group.type === 'todo' ? (
<Codicon className="text-muted-foreground/70" name="checklist" size="0.8rem" />
) : undefined
}
label={groupLabel(group, t.statusStack)}
>
{group.items.map(item => (
<StatusItemRow
item={item}
key={item.id}
onDismiss={sessionId ? id => dismissBackgroundProcess(sessionId, id) : undefined}
onOpen={() => openSubagent(item)}
onStop={sessionId ? id => stopBackgroundProcess(sessionId, id) : undefined}
/>
))}
</StatusSection>
)
}))
if (queue) {
sections.push({ key: 'queue', node: queue })
}
const visible = sections.length > 0
const stackRef = useRef<HTMLDivElement | null>(null)
// The stack is out of flow (overlays the thread), so the composer's measured
// height never sees it. Publish our own measured height — bucketed like the
// composer's, to avoid style invalidation churn — so the thread's
// last-message clearance can add it and the stack never hides messages.
useLayoutEffect(() => {
const root = document.documentElement
const el = stackRef.current
if (!visible || !el) {
root.style.removeProperty('--status-stack-measured-height')
return
}
let last = -1
const sync = () => {
const bucket = Math.round(el.getBoundingClientRect().height / 8) * 8
if (bucket !== last) {
last = bucket
root.style.setProperty('--status-stack-measured-height', `${bucket}px`)
}
}
const observer = new ResizeObserver(sync)
observer.observe(el)
sync()
return () => {
observer.disconnect()
root.style.removeProperty('--status-stack-measured-height')
}
}, [visible])
if (!visible) {
return null
}
return (
<div
className="absolute inset-x-0 bottom-full z-6 -mb-[9px] max-h-[40vh] overflow-y-auto"
onPointerDownCapture={() => blurComposerInput()}
ref={stackRef}
>
{/* The card paints the shared --composer-fill (rest / scrolled / focused
all match the composer surface by construction); on scroll we only
ghost the CONTENT — element opacity on the card would kill the blur. */}
<div className={cn(composerDockCard('top'), 'mx-1 pt-0.5 pb-1')}>
<div
className={cn(
'transition-opacity duration-200 ease-out',
scrolledUp ? 'opacity-30 group-hover/composer:opacity-100' : 'opacity-100'
)}
>
{sections.map(section => (
<div key={section.key}>{section.node}</div>
))}
</div>
</div>
</div>
)
}

View File

@@ -0,0 +1,155 @@
import { Fragment, memo, type ReactNode, useState } from 'react'
import { StatusRow } from '@/components/chat/status-row'
import { TerminalOutput } from '@/components/chat/terminal-output'
import { Button } from '@/components/ui/button'
import { Codicon } from '@/components/ui/codicon'
import { DisclosureCaret } from '@/components/ui/disclosure-caret'
import { GlyphSpinner } from '@/components/ui/glyph-spinner'
import { Tip } from '@/components/ui/tooltip'
import { type Translations, useI18n } from '@/i18n'
import { ArrowUpRight, X } from '@/lib/icons'
import type { TodoStatus } from '@/lib/todos'
import { cn } from '@/lib/utils'
import type { ComposerStatusItem } from '@/store/composer-status'
const toolLabel = (name: string) =>
name
.split('_')
.filter(Boolean)
.map(part => part[0]!.toUpperCase() + part.slice(1))
.join(' ') || name
// Todo rows speak checkbox, not spinner-and-dot: a dashed ring while the item
// is still open (pending), codicons once it resolves, a live spinner only on
// the in-progress item.
const TODO_GLYPHS: Record<Exclude<TodoStatus, 'in_progress' | 'pending'>, { icon: string; tone: string }> = {
cancelled: { icon: 'circle-slash', tone: 'text-muted-foreground/45' },
completed: { icon: 'pass-filled', tone: 'text-emerald-500/80' }
}
// Left slot: braille spinner while running, otherwise a small status dot
// (green = done, red = failed) so the slot is always filled and rows align.
function leadingGlyph(item: ComposerStatusItem, s: Translations['statusStack']): ReactNode {
if (item.todoStatus === 'pending') {
return (
<span
aria-hidden
className="box-border size-[0.7rem] rounded-full border border-dashed border-muted-foreground/60"
/>
)
}
if (item.todoStatus && item.todoStatus !== 'in_progress') {
const glyph = TODO_GLYPHS[item.todoStatus]
return <Codicon className={glyph.tone} name={glyph.icon} size="0.8rem" />
}
if (item.state === 'running') {
return (
<GlyphSpinner
ariaLabel={s.running}
className="text-[0.9rem] leading-none text-muted-foreground/80"
spinner="braille"
/>
)
}
return (
<span
aria-hidden
className={cn('size-1.5 rounded-full', item.state === 'failed' ? 'bg-destructive/80' : 'bg-emerald-500/70')}
/>
)
}
interface StatusItemRowProps {
item: ComposerStatusItem
/** Clear a finished background task from the stack. */
onDismiss?: (id: string) => void
/** Open the subagent's own session window, livestreamed by the gateway's
* child-session mirror (Agents view fallback for older gateways). */
onOpen?: () => void
/** Cancel a running background task. */
onStop?: (id: string) => void
}
/**
* Renders one {@link ComposerStatusItem} into the shared {@link StatusRow}.
* Memoised + keyed by id so parent re-renders never remount it (the spinner
* keeps ticking instead of resetting).
*/
export const StatusItemRow = memo(function StatusItemRow({ item, onDismiss, onOpen, onStop }: StatusItemRowProps) {
const { t } = useI18n()
const s = t.statusStack
const [outputOpen, setOutputOpen] = useState(false)
const failed = item.state === 'failed'
const running = item.state === 'running'
const action =
item.type === 'background'
? running
? onStop && { label: s.stop, onClick: () => onStop(item.id) }
: onDismiss && { label: s.dismiss, onClick: () => onDismiss(item.id) }
: null
const canOpen = item.type === 'subagent' && !!onOpen
const hasOutput = item.type === 'background' && !!item.output
const onActivate = canOpen ? onOpen : hasOutput ? () => setOutputOpen(open => !open) : undefined
return (
<Fragment>
<StatusRow
leading={leadingGlyph(item, s)}
onActivate={onActivate}
trailing={
action ? (
<Tip label={action.label}>
<Button
aria-label={action.label}
className="-my-1 size-4 rounded-md text-muted-foreground/60 hover:text-foreground/90"
onClick={event => {
event.stopPropagation()
action.onClick()
}}
size="icon-xs"
type="button"
variant="ghost"
>
<X size={12} />
</Button>
</Tip>
) : canOpen ? (
<ArrowUpRight aria-hidden className="size-3.5 text-muted-foreground/55" />
) : undefined
}
>
<span
className={cn(
'min-w-0 max-w-[18rem] truncate text-[0.73rem] leading-4',
failed
? 'text-destructive/90'
: item.todoStatus && item.todoStatus !== 'in_progress'
? 'text-muted-foreground/75'
: 'text-foreground/92'
)}
>
{item.title}
</span>
{item.type === 'subagent' && item.currentTool && (
<span className="shrink-0 truncate text-[0.62rem] leading-4 text-muted-foreground/70">
{toolLabel(item.currentTool)}
</span>
)}
{failed && typeof item.exitCode === 'number' && item.exitCode !== 0 && (
<span className="shrink-0 rounded bg-destructive/15 px-1 text-[0.58rem] font-semibold text-destructive tabular-nums">
{s.exit(item.exitCode)}
</span>
)}
{hasOutput && <DisclosureCaret className="shrink-0 text-muted-foreground/45" open={outputOpen} size="0.8em" />}
</StatusRow>
{hasOutput && outputOpen && <TerminalOutput className="mx-auto mb-1 max-w-[90%]" text={item.output!} />}
</Fragment>
)
})

View File

@@ -22,6 +22,33 @@ describe('detectTrigger', () => {
it('returns null for plain text', () => {
expect(detectTrigger('hello there')).toBeNull()
})
it('keeps the slash trigger live while typing args', () => {
expect(detectTrigger('/personality ')).toEqual({
kind: '/',
query: 'personality ',
tokenLength: 13
})
expect(detectTrigger('/personality alic')).toEqual({
kind: '/',
query: 'personality alic',
tokenLength: 17
})
expect(detectTrigger('/tools enable foo')).toEqual({
kind: '/',
query: 'tools enable foo',
tokenLength: 17
})
})
it('does not treat file-style paths as slash triggers', () => {
expect(detectTrigger('src/foo/bar')).toBeNull()
expect(detectTrigger('/path/to/file')).toBeNull()
})
it('still anchors at-mention triggers strictly at the token edge', () => {
expect(detectTrigger('@file:path with space')).toBeNull()
})
})
describe('extractClipboardImageBlobs', () => {

View File

@@ -6,7 +6,13 @@ export interface TriggerState {
tokenLength: number
}
const TRIGGER_RE = /(?:^|[\s])([@/])([^\s@/]*)$/
// `@` triggers stop at the first whitespace — `@file:path` and `@diff` are
// single tokens. `/` triggers keep going so the popover stays live while the
// user types args (`/personality alic` → arg completer suggests `alice`).
// Restricting the slash command name to `[a-zA-Z][\w-]*` avoids matching file
// paths like `src/foo/bar`.
const AT_TRIGGER_RE = /(?:^|[\s])(@)([^\s@/]*)$/
const SLASH_TRIGGER_RE = /(?:^|[\s])(\/)((?:[a-zA-Z][\w-]*(?:\s+\S*)*)?)$/
/** Stable key for paste dedupe — `items` and `files` often mirror the same image as different objects. */
export function blobDedupeKey(blob: Blob): string {
@@ -97,11 +103,17 @@ export function textBeforeCaret(editor: HTMLDivElement): string | null {
}
export function detectTrigger(textBefore: string): TriggerState | null {
const match = TRIGGER_RE.exec(textBefore)
const slash = SLASH_TRIGGER_RE.exec(textBefore)
if (!match) {
return null
if (slash) {
return { kind: '/', query: slash[2], tokenLength: 1 + slash[2].length }
}
return { kind: match[1] as '@' | '/', query: match[2], tokenLength: 1 + match[2].length }
const at = AT_TRIGGER_RE.exec(textBefore)
if (at) {
return { kind: '@', query: at[2], tokenLength: 1 + at[2].length }
}
return null
}

View File

@@ -34,9 +34,17 @@ describe('ComposerTriggerPopover i18n', () => {
})
it('renders localized loading copy for slash commands', () => {
const { container } = renderPopover('/', true)
renderPopover('/', true)
// While loading the popover shows only the spinner + loading copy — the
// `/help` empty-state hint is reserved for the resolved (not-loading) state.
expect(screen.getByText('查找中…')).toBeTruthy()
})
it('renders the slash empty-state hint when not loading', () => {
const { container } = renderPopover('/')
expect(screen.getByText('没有匹配项。')).toBeTruthy()
expect(container.textContent).toContain('/help')
})
})

View File

@@ -1,15 +1,12 @@
import type { Unstable_TriggerItem } from '@assistant-ui/core'
import { Fragment } from 'react'
import { Codicon } from '@/components/ui/codicon'
import { GlyphSpinner } from '@/components/ui/glyph-spinner'
import { useI18n } from '@/i18n'
import { cn } from '@/lib/utils'
import {
COMPLETION_DRAWER_BELOW_CLASS,
COMPLETION_DRAWER_CLASS,
COMPLETION_DRAWER_ROW_CLASS,
CompletionDrawerEmpty
} from './completion-drawer'
import { COMPLETION_DRAWER_BELOW_CLASS, COMPLETION_DRAWER_CLASS, CompletionDrawerEmpty } from './completion-drawer'
const AT_ICON_BY_TYPE: Record<string, string> = {
diff: 'diff',
@@ -23,11 +20,7 @@ const AT_ICON_BY_TYPE: Record<string, string> = {
url: 'globe'
}
function completionIcon(kind: '@' | '/', item: Unstable_TriggerItem) {
if (kind === '/') {
return 'terminal'
}
function atIcon(item: Unstable_TriggerItem) {
const meta = item.metadata as { rawText?: string } | undefined
const raw = meta?.rawText || item.label
@@ -42,6 +35,18 @@ function completionIcon(kind: '@' | '/', item: Unstable_TriggerItem) {
return AT_ICON_BY_TYPE[item.type] || AT_ICON_BY_TYPE.simple
}
interface RowMeta {
display?: string
group?: string
meta?: string
}
const ROW_BASE_CLASS = [
'relative flex w-full cursor-default select-none rounded-md px-2 py-1 text-left',
'outline-hidden transition-colors hover:bg-(--ui-bg-tertiary)',
'data-[highlighted]:bg-(--ui-bg-tertiary) data-[highlighted]:text-foreground'
].join(' ')
interface ComposerTriggerPopoverProps {
activeIndex: number
items: readonly Unstable_TriggerItem[]
@@ -63,6 +68,9 @@ export function ComposerTriggerPopover({
}: ComposerTriggerPopoverProps) {
const { t } = useI18n()
const copy = t.composer
const isSlash = kind === '/'
let lastGroup: string | undefined
return (
<div
@@ -73,41 +81,94 @@ export function ComposerTriggerPopover({
role="listbox"
>
{items.length === 0 ? (
<CompletionDrawerEmpty title={loading ? copy.lookupLoading : copy.lookupNoMatches}>
{kind === '@' ? (
<>
{copy.lookupTry} <span className="font-mono text-foreground/80">@file:</span> {copy.lookupOr}{' '}
<span className="font-mono text-foreground/80">@folder:</span>.
</>
) : (
<>
{copy.lookupTry} <span className="font-mono text-foreground/80">/help</span>.
</>
)}
</CompletionDrawerEmpty>
loading ? (
<div className="flex items-center gap-2 px-2 py-1.5 text-(--ui-text-tertiary)">
<GlyphSpinner ariaLabel={copy.lookupLoading} className="text-foreground/70" spinner="braille" />
<span>{copy.lookupLoading}</span>
</div>
) : (
<CompletionDrawerEmpty title={copy.lookupNoMatches}>
{kind === '@' ? (
<>
{copy.lookupTry} <span className="font-mono text-foreground/80">@file:</span> {copy.lookupOr}{' '}
<span className="font-mono text-foreground/80">@folder:</span>.
</>
) : (
<>
{copy.lookupTry} <span className="font-mono text-foreground/80">/help</span>.
</>
)}
</CompletionDrawerEmpty>
)
) : (
items.map((item, index) => {
const meta = item.metadata as { display?: string; meta?: string } | undefined
const display = meta?.display ?? (kind === '/' ? `/${item.label}` : item.label)
const meta = item.metadata as RowMeta | undefined
const display = meta?.display ?? (isSlash ? `/${item.label}` : item.label)
const description = meta?.meta || item.description
const group = meta?.group?.trim()
const showHeader = isSlash && Boolean(group) && group !== lastGroup
const isFirstHeader = lastGroup === undefined
lastGroup = group || lastGroup
const active = index === activeIndex
return (
<button
className={cn(COMPLETION_DRAWER_ROW_CLASS, index === activeIndex && 'bg-(--ui-bg-tertiary)')}
data-highlighted={index === activeIndex ? '' : undefined}
key={item.id}
onClick={() => onPick(item)}
onMouseEnter={() => onHover(index)}
type="button"
>
<span className="grid size-3.5 shrink-0 place-items-center text-(--ui-text-tertiary)">
<Codicon name={completionIcon(kind, item)} size="0.875rem" />
</span>
<span className="min-w-0 shrink truncate font-mono font-medium leading-5 text-foreground">{display}</span>
{description && (
<span className="min-w-0 flex-1 truncate leading-5 text-(--ui-text-tertiary)">{description}</span>
<Fragment key={item.id}>
{showHeader && (
<div
className={cn(
'select-none px-2 pb-0.5 text-[0.625rem] font-semibold uppercase tracking-wider text-(--ui-text-tertiary)',
isFirstHeader ? 'pt-0.5' : 'pt-2'
)}
>
{group}
</div>
)}
</button>
<button
className={cn(ROW_BASE_CLASS, isSlash ? 'flex-col gap-0' : 'items-center gap-2')}
data-highlighted={active ? '' : undefined}
onClick={() => onPick(item)}
onMouseEnter={() => onHover(index)}
type="button"
>
{isSlash ? (
<>
{/* Active row (keyboard nav or hover) un-truncates inline so
long command names / descriptions stay readable without a
floating tooltip. */}
<span
className={cn(
'text-[0.8125rem] font-medium leading-snug text-foreground',
active ? 'whitespace-normal break-words' : 'truncate'
)}
>
{display}
</span>
{description && (
<span
className={cn(
'text-[0.6875rem] leading-snug text-(--ui-text-tertiary)',
active ? 'whitespace-normal break-words' : 'truncate'
)}
>
{description}
</span>
)}
</>
) : (
<>
<span className="grid size-4 shrink-0 place-items-center text-(--ui-text-tertiary)">
<Codicon name={atIcon(item)} size="0.875rem" />
</span>
<span className="min-w-0 shrink truncate font-mono font-medium leading-5 text-foreground">
{display}
</span>
{description && (
<span className="min-w-0 flex-1 truncate leading-5 text-(--ui-text-tertiary)">{description}</span>
)}
</>
)}
</button>
</Fragment>
)
})
)}

View File

@@ -1,6 +1,7 @@
import { useCallback } from 'react'
import { requestComposerFocus, requestComposerInsert } from '@/app/chat/composer/focus'
import { requestComposerFocus, requestComposerInsert, requestComposerInsertRefs } from '@/app/chat/composer/focus'
import { droppedFileInlineRef } from '@/app/chat/composer/inline-refs'
import { formatRefValue } from '@/components/assistant-ui/directive-text'
import { useI18n } from '@/i18n'
import { attachmentId, contextPath, pathLabel } from '@/lib/chat-runtime'
@@ -286,6 +287,26 @@ export function useComposerActions({ activeSessionId, currentCwd, requestGateway
[currentCwd]
)
const insertContextPathInlineRef = useCallback(
(path: string, isDirectory = false) => {
if (!path) {
return false
}
const ref = droppedFileInlineRef({ isDirectory, path }, currentCwd)
if (!ref) {
return false
}
requestComposerInsertRefs([ref])
requestComposerFocus('main')
return true
},
[currentCwd]
)
const attachContextFilePath = useCallback(
(filePath: string) => {
if (!filePath) {
@@ -546,6 +567,7 @@ export function useComposerActions({ activeSessionId, currentCwd, requestGateway
attachDroppedItems,
attachImageBlob,
attachImagePath,
insertContextPathInlineRef,
pasteClipboardImage,
pickContextPaths,
pickImages,

View File

@@ -43,7 +43,7 @@ import {
import type { ModelOptionsResponse } from '@/types/hermes'
import { routeSessionId } from '../routes'
import { titlebarHeaderBaseClass, titlebarHeaderShadowClass } from '../shell/titlebar'
import { titlebarHeaderBaseClass, titlebarHeaderShadowClass, titlebarHeaderTitleClass } from '../shell/titlebar'
import { ChatDropOverlay } from './chat-drop-overlay'
import { ChatSwapOverlay } from './chat-swap-overlay'
@@ -53,6 +53,7 @@ import { droppedFileInlineRefs, type SessionDragPayload, sessionInlineRef } from
import type { ChatBarState } from './composer/types'
import { type DroppedFile, partitionDroppedFiles } from './hooks/use-composer-actions'
import { useFileDropZone } from './hooks/use-file-drop-zone'
import { ScrollToBottomButton } from './scroll-to-bottom-button'
import { SessionActionsMenu } from './sidebar/session-actions-menu'
import { lastVisibleMessageIsUser, threadLoadingState } from './thread-loading'
@@ -80,6 +81,7 @@ interface ChatViewProps extends Omit<React.ComponentProps<'div'>, 'onSubmit'> {
onThreadMessagesChange: (messages: readonly ThreadMessage[]) => void
onEdit: (message: AppendMessage) => Promise<void>
onReload: (parentId: string | null) => Promise<void>
onRestoreToMessage?: (messageId: string) => Promise<void>
onTranscribeAudio?: (audio: Blob) => Promise<string>
}
@@ -124,13 +126,7 @@ function ChatHeader({
return (
<header className={cn(titlebarHeaderBaseClass, isRoutedSessionView && titlebarHeaderShadowClass)}>
<div
className="min-w-0 flex-1"
style={{
maxWidth:
'calc(100vw - var(--titlebar-content-inset,0px) - var(--titlebar-tools-right) - var(--titlebar-tools-width) - 1.5rem)'
}}
>
<div className={titlebarHeaderTitleClass}>
<SessionActionsMenu
align="start"
onDelete={selectedSessionId ? onDeleteSelectedSession : undefined}
@@ -141,7 +137,7 @@ function ChatHeader({
title={title}
>
<Button
className="pointer-events-auto flex h-6 min-w-0 max-w-full gap-1 border border-transparent bg-transparent px-2 py-0 text-(--ui-text-secondary) hover:border-(--ui-stroke-tertiary) hover:bg-(--ui-control-hover-background) hover:text-foreground data-[state=open]:border-(--ui-stroke-tertiary) data-[state=open]:bg-(--ui-control-active-background) [-webkit-app-region:no-drag]"
className="pointer-events-auto flex h-6 w-full min-w-0 max-w-full gap-1 overflow-hidden border border-transparent bg-transparent px-2 py-0 text-(--ui-text-secondary) hover:border-(--ui-stroke-tertiary) hover:bg-(--ui-control-hover-background) hover:text-foreground data-[state=open]:border-(--ui-stroke-tertiary) data-[state=open]:bg-(--ui-control-active-background) [-webkit-app-region:no-drag]"
type="button"
variant="ghost"
>
@@ -176,6 +172,7 @@ export function ChatView({
onThreadMessagesChange,
onEdit,
onReload,
onRestoreToMessage,
onTranscribeAudio
}: ChatViewProps) {
const location = useLocation()
@@ -362,6 +359,7 @@ export function ChatView({
loading={threadLoading}
onBranchInNewChat={onBranchInNewChat}
onCancel={onCancel}
onRestoreToMessage={onRestoreToMessage}
sessionId={activeSessionId}
sessionKey={threadKey}
/>
@@ -394,6 +392,7 @@ export function ChatView({
</Suspense>
)}
</AssistantRuntimeProvider>
{showChatBar && <ScrollToBottomButton />}
<ChatDropOverlay kind={dragKind} />
<ChatSwapOverlay profile={gatewaySwapTarget} />
</div>

View File

@@ -10,11 +10,16 @@ import { useEffect, useMemo, useState } from 'react'
import ShikiHighlighter from 'react-shiki'
import { Streamdown } from 'streamdown'
import { requestComposerFocus, requestComposerInsertRefs } from '@/app/chat/composer/focus'
import { droppedFileInlineRef } from '@/app/chat/composer/inline-refs'
import { HERMES_PATHS_MIME } from '@/app/chat/hooks/use-composer-actions'
import { isAddSelectionShortcut } from '@/app/right-sidebar/terminal/selection'
import { PageLoader } from '@/components/page-loader'
import { translateNow, useI18n } from '@/i18n'
import { readDesktopFileDataUrl, readDesktopFileText } from '@/lib/desktop-fs'
import { cn } from '@/lib/utils'
import type { PreviewTarget } from '@/store/preview'
import { $currentCwd } from '@/store/session'
const SHIKI_THEME = { dark: 'github-dark-default', light: 'github-light-default' } as const
const TEXT_PREVIEW_MAX_BYTES = 512 * 1024
@@ -180,15 +185,13 @@ function looksBinaryBytes(bytes: Uint8Array) {
}
async function readTextPreview(filePath: string) {
if (window.hermesDesktop.readFileText) {
try {
return await window.hermesDesktop.readFileText(filePath)
} catch (error) {
const message = error instanceof Error ? error.message : String(error)
try {
return await readDesktopFileText(filePath)
} catch (error) {
const message = error instanceof Error ? error.message : String(error)
if (!message.includes("No handler registered for 'hermes:readFileText'")) {
throw error
}
if (!message.includes("No handler registered for 'hermes:readFileText'")) {
throw error
}
}
@@ -288,7 +291,7 @@ const MARKDOWN_COMPONENTS = {
function MarkdownPreview({ text }: { text: string }) {
return (
<div className="preview-markdown mx-auto max-w-3xl px-4 py-3 text-sm text-foreground">
<div className="preview-markdown mx-auto max-w-3xl px-4 py-3 text-sm text-foreground" data-selectable-text="true">
<Streamdown components={MARKDOWN_COMPONENTS} controls={false} mode="static" parseIncompleteMarkdown={false}>
{text}
</Streamdown>
@@ -358,6 +361,38 @@ function SourceView({ filePath, language, text }: { filePath: string; language:
startLineDrag(event, filePath, inSelection(line) && selection ? selection : { end: line, start: line })
}
// ⌘/Ctrl+L with a line selection drops the same `@line:path:start-end` ref the
// gutter drag produces — so the keyboard path mirrors dragging the lines into
// the composer. Capture-phase + stopPropagation so it beats the terminal's
// global ⌘L handler (which would otherwise grab the native text selection).
useEffect(() => {
if (!selection) {
return
}
const onKeyDown = (event: KeyboardEvent) => {
if (!isAddSelectionShortcut(event)) {
return
}
const lineEnd = selection.end > selection.start ? selection.end : undefined
const ref = droppedFileInlineRef({ line: selection.start, lineEnd, path: filePath }, $currentCwd.get())
if (!ref) {
return
}
event.preventDefault()
event.stopPropagation()
requestComposerInsertRefs([ref])
requestComposerFocus('main')
}
window.addEventListener('keydown', onKeyDown, { capture: true })
return () => window.removeEventListener('keydown', onKeyDown, { capture: true })
}, [filePath, selection])
return (
<div className="grid min-w-max grid-cols-[auto_minmax(0,1fr)] font-mono text-xs leading-relaxed">
<div className="select-none py-3 text-right text-muted-foreground/55">
@@ -384,7 +419,10 @@ function SourceView({ filePath, language, text }: { filePath: string; language:
)
})}
</div>
<div className="relative [&_pre]:m-0 [&_pre]:px-3 [&_pre]:py-3 [&_pre]:bg-transparent!">
<div
className="relative [&_pre]:m-0 [&_pre]:px-3 [&_pre]:py-3 [&_pre]:bg-transparent!"
data-selectable-text="true"
>
{selection && (
<div
aria-hidden
@@ -448,7 +486,7 @@ export function LocalFilePreview({ reloadKey, target }: { reloadKey: number; tar
if (isImage) {
// Prefer bytes the caller already handed us (a pasted/dropped
// screenshot) over re-reading a path that may be transient/unreadable.
const dataUrl = target.dataUrl || (await window.hermesDesktop.readFileDataUrl(filePath))
const dataUrl = target.dataUrl || (await readDesktopFileDataUrl(filePath))
if (active) {
setState({ dataUrl, loading: false })

View File

@@ -1,11 +1,50 @@
import { act, cleanup, render } from '@testing-library/react'
import { afterEach, describe, expect, it, vi } from 'vitest'
import { afterEach, beforeEach, describe, expect, it, vi } from 'vitest'
import { $connection } from '@/store/session'
import { PreviewPane } from './preview-pane'
describe('PreviewPane console state', () => {
beforeEach(() => {
vi.stubGlobal('requestAnimationFrame', (callback: FrameRequestCallback) => window.setTimeout(() => callback(Date.now()), 0))
vi.stubGlobal('cancelAnimationFrame', (id: number) => window.clearTimeout(id))
})
afterEach(() => {
cleanup()
$connection.set(null)
vi.unstubAllGlobals()
})
it('does not watch backend-only remote filesystem previews locally', () => {
const watchPreviewFile = vi.fn(async () => ({ id: 'watch-1', path: '/remote/file.txt' }))
const onPreviewFileChanged = vi.fn(() => vi.fn())
$connection.set({ mode: 'remote' } as never)
vi.stubGlobal('window', {
...window,
hermesDesktop: {
onPreviewFileChanged,
watchPreviewFile
}
})
render(
<PreviewPane
setTitlebarToolGroup={vi.fn()}
target={{
kind: 'file',
label: 'file.txt',
path: '/remote/file.txt',
previewKind: 'text',
source: '/remote/file.txt',
url: 'file:///remote/file.txt'
}}
/>
)
expect(watchPreviewFile).not.toHaveBeenCalled()
expect(onPreviewFileChanged).not.toHaveBeenCalled()
})
it('does not rebuild the pane titlebar group for streamed console logs', () => {

View File

@@ -5,6 +5,7 @@ import { useCallback, useEffect, useRef, useState } from 'react'
import type { SetTitlebarToolGroup, TitlebarTool } from '@/app/shell/titlebar-controls'
import { Tip } from '@/components/ui/tooltip'
import { type Translations, useI18n } from '@/i18n'
import { isDesktopFsRemoteMode } from '@/lib/desktop-fs'
import { Bug } from '@/lib/icons'
import { cn } from '@/lib/utils'
import { notify, notifyError } from '@/store/notifications'
@@ -406,6 +407,7 @@ export function PreviewPane({
useEffect(() => {
if (
target.kind !== 'file' ||
isDesktopFsRemoteMode() ||
!window.hermesDesktop?.watchPreviewFile ||
!window.hermesDesktop?.onPreviewFileChanged
) {

View File

@@ -0,0 +1,58 @@
import { useStore } from '@nanostores/react'
import { useRef } from 'react'
import { Codicon } from '@/components/ui/codicon'
import { useI18n } from '@/i18n'
import { triggerHaptic } from '@/lib/haptics'
import { cn } from '@/lib/utils'
import { $threadJumpButtonVisible, requestScrollToBottom } from '@/store/thread-scroll'
/**
* Floating "jump to bottom" control. Sits centered just above the composer,
* clearing the out-of-flow status stack via the same measured-height CSS vars
* the thread's bottom clearance uses (`--composer-measured-height` +
* `--status-stack-measured-height`), so it never overlaps the queue / subagent
* / background cards. Visible only while the user has scrolled meaningfully
* away from the bottom; clicking re-arms sticky-bottom and pins the viewport.
*
* Enter/exit motion lives in styles.css under `.thread-jump-button` — a
* directional scale (contract in from 1.1, contract out to 0.9) keyed off
* `data-state`. `idle` (never-shown) stays silent so it can't flash on mount;
* `in`/`out` only swap once it has actually appeared.
*/
export function ScrollToBottomButton() {
const { t } = useI18n()
const visible = useStore($threadJumpButtonVisible)
const hasShownRef = useRef(false)
if (visible) {
hasShownRef.current = true
}
const state = visible ? 'in' : hasShownRef.current ? 'out' : 'idle'
return (
<button
aria-hidden={!visible}
aria-label={t.assistant.thread.scrollToBottom}
className={cn(
'thread-jump-button absolute left-1/2 z-20 grid size-8 place-items-center rounded-full',
'border border-border/65 bg-(--composer-fill) text-muted-foreground hover:text-foreground',
'backdrop-blur-[0.75rem] [-webkit-backdrop-filter:blur(0.75rem)]',
!visible && 'pointer-events-none'
)}
data-state={state}
onClick={() => {
triggerHaptic('selection')
requestScrollToBottom()
}}
style={{
bottom: 'calc(var(--composer-measured-height) + var(--status-stack-measured-height) + 0.625rem)'
}}
tabIndex={visible ? 0 : -1}
type="button"
>
<Codicon name="arrow-down" size="1rem" />
</button>
)
}

View File

@@ -168,7 +168,7 @@ export function SidebarCronJobsSection({
</button>
</div>
{open && (
<SidebarGroupContent className="flex max-h-72 flex-col gap-px overflow-y-auto overscroll-contain pb-1.75 compact:max-h-none compact:overflow-visible">
<SidebarGroupContent className="flex max-h-72 flex-col gap-px overflow-x-hidden overflow-y-auto overscroll-contain pb-1.75 compact:max-h-none compact:overflow-visible">
{shown.map(job => (
<CronJobSidebarRow
expanded={peekJobId === job.id}

View File

@@ -39,6 +39,7 @@ import { Tip } from '@/components/ui/tooltip'
import { searchSessions, type SessionInfo, type SessionSearchResult } from '@/hermes'
import { useI18n } from '@/i18n'
import { profileColor } from '@/lib/profile-color'
import { comboTokens } from '@/lib/keybinds/combo'
import { sessionMatchesSearch } from '@/lib/session-search'
import { normalizeSessionSource, sessionSourceLabel } from '@/lib/session-source'
import { cn } from '@/lib/utils'
@@ -108,11 +109,7 @@ const VIRTUALIZE_THRESHOLD = 25
const NON_SESSION_INITIAL_ROWS = 3
const NON_SESSION_LOAD_STEP = 10
// Render the modifier key the user actually presses on this platform. The
// global accelerator is bound to both Cmd+N (macOS) and Ctrl+N (everywhere
// else) in desktop-controller.tsx, but the hint should match muscle memory.
const NEW_SESSION_KBD: readonly string[] =
typeof navigator !== 'undefined' && navigator.platform.toLowerCase().includes('mac') ? ['⌘', 'N'] : ['Ctrl', 'N']
const NEW_SESSION_KBD = comboTokens('mod+n')
const SIDEBAR_NAV: SidebarNavItem[] = [
{
@@ -144,8 +141,11 @@ const GROUP_DND_ID_PREFIX = 'group:'
// the next — the flexbox `min-height: auto` overlap trap that caused the bug.
const COMPACT_FLAT = 'compact:max-h-none compact:overflow-visible'
// Vertical scroll only — never a horizontal bar from glow bleed, long titles, etc.
const SCROLL_Y = 'overflow-y-auto overflow-x-hidden overscroll-contain'
// A non-session group's scroll body: own scroller when tall, flattened when compact.
const GROUP_BODY = cn('overflow-y-auto overscroll-contain', COMPACT_FLAT)
const GROUP_BODY = cn(SCROLL_Y, COMPACT_FLAT)
const groupDndId = (id: string) => `${GROUP_DND_ID_PREFIX}${id}`
@@ -797,7 +797,14 @@ export function ChatSidebar({
<SidebarMenuButton
aria-disabled={!isInteractive}
className={cn(
'flex h-7 w-full justify-start gap-2 rounded-md border border-transparent px-2 text-left text-[0.8125rem] font-medium text-(--ui-text-secondary) transition-colors duration-100 ease-out hover:bg-(--ui-control-hover-background) hover:text-foreground hover:transition-none',
// no-drag: these rows sit directly under the titlebar's
// [-webkit-app-region:drag] strips (app-shell.tsx), with only
// 6px of clearance. Drag regions win hit-testing over DOM
// (pointer-events can't override), and on Linux/WSLg the
// resolved region has been observed to swallow clicks on the
// top rows. Same carve-out as USER_BUBBLE_BASE_CLASS in
// thread.tsx.
'flex h-7 w-full justify-start gap-2 rounded-md border border-transparent px-2 text-left text-[0.8125rem] font-medium text-(--ui-text-secondary) transition-colors duration-100 ease-out [-webkit-app-region:no-drag] hover:bg-(--ui-control-hover-background) hover:text-foreground hover:transition-none',
active &&
'border-(--ui-stroke-tertiary) bg-(--ui-control-active-background) text-foreground shadow-none hover:border-(--ui-stroke-tertiary)!',
!isInteractive &&
@@ -823,8 +830,9 @@ export function ChatSidebar({
<span className="min-w-0 flex-1 truncate">{s.nav[item.id] ?? item.label}</span>
{isNewSession && (
<KbdGroup
className={cn('ml-auto', newSessionKbdFlash && 'opacity-100!')}
className={cn('ml-auto opacity-55', newSessionKbdFlash && 'opacity-100!')}
keys={[...NEW_SESSION_KBD]}
size="sm"
/>
)}
</>
@@ -850,11 +858,11 @@ export function ChatSidebar({
)}
{contentVisible && showSessionSections && (
<div className="flex min-h-0 flex-1 flex-col overflow-y-auto overscroll-contain pb-1.75">
<div className={cn('flex min-h-0 flex-1 flex-col pb-1.75', SCROLL_Y)}>
{trimmedQuery && (
<SidebarSessionsSection
activeSessionId={activeSidebarSessionId}
contentClassName="flex min-h-0 flex-1 flex-col gap-px overflow-y-auto overscroll-contain pb-1.75"
contentClassName={cn('flex min-h-0 flex-1 flex-col gap-px pb-1.75', SCROLL_Y)}
emptyState={
<div className="grid min-h-24 place-items-center rounded-lg px-2 text-center text-xs text-(--ui-text-tertiary)">
{s.noMatch(trimmedQuery)}
@@ -901,7 +909,8 @@ export function ChatSidebar({
<SidebarSessionsSection
activeSessionId={activeSidebarSessionId}
contentClassName={cn(
'flex min-h-0 flex-1 flex-col overflow-y-auto overscroll-contain pb-1.75',
'flex min-h-0 flex-1 flex-col pb-1.75',
SCROLL_Y,
// Separate profile sections clearly in the ALL view; rows inside
// each group keep their own tight gap-px rhythm.
showAllProfiles ? 'gap-3' : 'gap-px',

View File

@@ -88,7 +88,7 @@ function useSessionActions({ sessionId, title, pinned = false, profile, onPin, o
label: r.export,
onSelect: () => {
triggerHaptic('selection')
void exportSession(sessionId, { title })
void exportSession(sessionId, { profile, title })
}
},
{

View File

@@ -102,7 +102,7 @@ export const VirtualSessionList: FC<VirtualSessionListProps> = ({
})
const list = (
<div className={cn('relative min-h-0 flex-1 overflow-y-auto overscroll-contain', className)} ref={scrollerRef}>
<div className={cn('relative min-h-0 flex-1 overflow-x-hidden overflow-y-auto overscroll-contain', className)} ref={scrollerRef}>
<div className="grid gap-px" style={{ paddingBottom: `${paddingBottom}px`, paddingTop: `${paddingTop}px` }}>
{rows}
</div>

View File

@@ -7,8 +7,8 @@ import { useNavigate } from 'react-router-dom'
import { HUD_HEADING, HUD_ITEM, HUD_POSITION, HUD_SURFACE, HUD_TEXT } from '@/app/floating-hud'
import { setTerminalTakeover } from '@/app/right-sidebar/store'
import { Command, CommandEmpty, CommandGroup, CommandInput, CommandItem, CommandList } from '@/components/ui/command'
import { KbdGroup } from '@/components/ui/kbd'
import { getHermesConfigRecord, listSessions } from '@/hermes'
import { KbdCombo } from '@/components/ui/kbd'
import { getHermesConfigRecord, listAllProfileSessions } from '@/hermes'
import { useI18n } from '@/i18n'
import { sessionTitle } from '@/lib/chat-runtime'
import {
@@ -38,7 +38,6 @@ import {
Wrench,
Zap
} from '@/lib/icons'
import { comboTokens } from '@/lib/keybinds/combo'
import { cn } from '@/lib/utils'
import { $commandPaletteOpen, closeCommandPalette, setCommandPaletteOpen } from '@/store/command-palette'
import { $bindings } from '@/store/keybinds'
@@ -119,7 +118,7 @@ const paletteFilter = (value: string, search: string, keywords?: string[]): numb
return needle.split(/\s+/).every(term => haystack.includes(term)) ? 1 : 0
}
type SessionRow = Awaited<ReturnType<typeof listSessions>>['sessions'][number]
type SessionRow = Awaited<ReturnType<typeof listAllProfileSessions>>['sessions'][number]
const toSessionEntry = (session: SessionRow): SessionEntry => ({
id: session.id,
@@ -218,13 +217,13 @@ export function CommandPalette() {
const sessionsQuery = useQuery({
queryKey: ['command-palette', 'sessions'],
queryFn: () => listSessions(200, 1, 'exclude'),
queryFn: () => listAllProfileSessions(200, 1, 'exclude'),
enabled: open
})
const archivedQuery = useQuery({
queryKey: ['command-palette', 'archived'],
queryFn: () => listSessions(200, 0, 'only'),
queryFn: () => listAllProfileSessions(200, 0, 'only'),
enabled: open
})
@@ -620,7 +619,6 @@ export function CommandPalette() {
{group.items.map(item => {
const Icon = item.icon
const combo = item.action ? bindings[item.action]?.[0] : undefined
const keys = combo ? comboTokens(combo) : null
return (
<CommandItem
@@ -632,10 +630,10 @@ export function CommandPalette() {
>
<Icon className="size-3.5 shrink-0 text-muted-foreground" />
<span className="truncate">{item.label}</span>
{keys && <KbdGroup className="ml-auto" keys={keys} />}
{combo && <KbdCombo className="ml-auto opacity-55" combo={combo} size="sm" />}
{item.to && (
<ChevronRight
className={cn('size-3.5 shrink-0 text-muted-foreground/70', !keys && 'ml-auto')}
className={cn('size-3.5 shrink-0 text-muted-foreground/70', !combo && 'ml-auto')}
/>
)}
</CommandItem>

View File

@@ -20,6 +20,7 @@ import {
MESSAGING_SESSION_SOURCE_IDS,
normalizeSessionSource
} from '../lib/session-source'
import { latestSessionTodos } from '../lib/todos'
import { setCronFocusJobId, setCronJobs } from '../store/cron'
import {
$panesFlipped,
@@ -75,10 +76,12 @@ import {
setSessionsLoading,
setSessionsTotal
} from '../store/session'
import { clearSessionTodos, setSessionTodos, todoListActive } from '../store/todos'
import { openUpdatesWindow, startUpdatePoller, stopUpdatePoller } from '../store/updates'
import { isSecondaryWindow } from '../store/windows'
import { ChatView } from './chat'
import { requestComposerFocus, requestComposerInsert } from './chat/composer/focus'
import { useComposerActions } from './chat/hooks/use-composer-actions'
import {
ChatPreviewRail,
@@ -98,6 +101,7 @@ import { RightSidebarPane } from './right-sidebar'
import { $terminalTakeover } from './right-sidebar/store'
import { PersistentTerminal, TerminalSlot } from './right-sidebar/terminal/persistent'
import { CRON_ROUTE, NEW_CHAT_ROUTE, routeSessionId, sessionRoute, SETTINGS_ROUTE } from './routes'
import { SessionPickerOverlay } from './session-picker-overlay'
import { SessionSwitcher } from './session-switcher'
import { useContextSuggestions } from './session/hooks/use-context-suggestions'
import { useCwdActions } from './session/hooks/use-cwd-actions'
@@ -139,7 +143,7 @@ const CRON_POLL_INTERVAL_MS = 30_000
// self-managed sidebar section (refreshMessagingSessions). Excluding both here
// keeps "Load more" paging through interactive local chats instead of
// interleaving gateway threads that bury them.
const SIDEBAR_EXCLUDED_SOURCES = ['cron', ...MESSAGING_SESSION_SOURCE_IDS]
const SIDEBAR_EXCLUDED_SOURCES = ['cron', 'subagent', 'tool', ...MESSAGING_SESSION_SOURCE_IDS]
// The messaging slice is the inverse: drop cron + every local source so only
// external-platform conversations remain, then split per platform in the UI.
const MESSAGING_EXCLUDED_SOURCES = ['cron', ...LOCAL_SESSION_SOURCE_IDS]
@@ -265,6 +269,36 @@ export function DesktopController() {
}
}, [])
// hermes:// deep links (e.g. a docs "Send to App" button for an automation blueprint).
// Build the equivalent /blueprint slash command from the payload and drop
// it into the composer — the user reviews/edits, then sends; the agent (or
// the shared command handler) creates the job. Signal readiness so a link
// that arrived during boot is flushed exactly once.
useEffect(() => {
const unsubscribe = window.hermesDesktop?.onDeepLink?.(payload => {
if (!payload || payload.kind !== 'blueprint' || !payload.name) {
return
}
const slots = Object.entries(payload.params || {})
.map(([k, v]) => {
const sval = /\s/.test(v) ? `"${v.replace(/"/g, '\\"')}"` : v
return `${k}=${sval}`
})
.join(' ')
const command = `/blueprint ${payload.name}${slots ? ' ' + slots : ''}`
requestComposerInsert(command, { mode: 'block', target: 'main' })
requestComposerFocus('main')
})
// Tell the main process the renderer is ready to receive deep links.
void window.hermesDesktop?.signalDeepLinkReady?.()
return () => unsubscribe?.()
}, [])
useEffect(() => {
const onKeyDown = (event: KeyboardEvent) => {
if (!$filePreviewTarget.get() && !$previewTarget.get()) {
@@ -520,20 +554,34 @@ export function DesktopController() {
return
}
const storedProfile = $sessions.get().find(session => session.id === storedSessionId)?.profile
const storedProfile = $sessions
.get()
.find(session => session.id === storedSessionId || session._lineage_root_id === storedSessionId)?.profile
for (let index = 0; index < Math.max(1, attempts); index += 1) {
try {
const latest = await getSessionMessages(storedSessionId, storedProfile)
const messages = toChatMessages(latest.messages)
updateSessionState(
runtimeSessionId,
state => ({
...state,
messages: preserveLocalAssistantErrors(toChatMessages(latest.messages), state.messages)
messages: preserveLocalAssistantErrors(messages, state.messages)
}),
storedSessionId
)
// Seed the status stack's todo group from history — but only while
// the plan is still in flight, so reopening an old chat doesn't pin
// its finished todo list above the composer forever.
const todos = latestSessionTodos(messages)
if (todos && todoListActive(todos)) {
setSessionTodos(runtimeSessionId, todos)
} else {
clearSessionTodos(runtimeSessionId)
}
return
} catch {
// Best-effort fallback when live stream payloads are empty.
@@ -553,6 +601,7 @@ export function DesktopController() {
queryClient,
refreshHermesConfig,
refreshSessions,
sessionStateByRuntimeIdRef,
updateSessionState
})
@@ -682,6 +731,7 @@ export function DesktopController() {
editMessage,
handleThreadMessagesChange,
reloadFromMessage,
restoreToMessage,
steerPrompt,
submitText,
transcribeVoiceAudio
@@ -694,6 +744,7 @@ export function DesktopController() {
handleSkinCommand,
refreshSessions,
requestGateway,
resumeStoredSession: resumeSession,
selectedStoredSessionIdRef,
startFreshSessionDraft,
sttEnabled,
@@ -743,6 +794,13 @@ export function DesktopController() {
}
}, [gatewayState, refreshCronJobs])
useEffect(() => {
if (gatewayState === 'open' && !activeSessionId && freshDraftReady) {
void refreshCurrentModel()
void refreshHermesConfig()
}
}, [activeSessionId, freshDraftReady, gatewayState, refreshCurrentModel, refreshHermesConfig])
useRouteResume({
activeSessionId,
activeSessionIdRef,
@@ -822,6 +880,7 @@ export function DesktopController() {
/>
)}
<ModelPickerOverlay gateway={gatewayRef.current || undefined} onSelect={selectModel} />
<SessionPickerOverlay onResume={resumeSession} />
<ModelVisibilityOverlay gateway={gatewayRef.current || undefined} onOpenProviders={openProviderSettings} />
<UpdatesOverlay />
<GatewayConnectingOverlay />
@@ -907,6 +966,7 @@ export function DesktopController() {
onPickImages={() => void composer.pickImages()}
onReload={reloadFromMessage}
onRemoveAttachment={id => void composer.removeAttachment(id)}
onRestoreToMessage={restoreToMessage}
onSteer={steerPrompt}
onSubmit={submitText}
onThreadMessagesChange={handleThreadMessagesChange}
@@ -952,8 +1012,8 @@ export function DesktopController() {
width={FILE_BROWSER_DEFAULT_WIDTH}
>
<RightSidebarPane
onActivateFile={composer.attachContextFilePath}
onActivateFolder={composer.attachContextFolderPath}
onActivateFile={path => composer.insertContextPathInlineRef(path)}
onActivateFolder={path => composer.insertContextPathInlineRef(path, true)}
onChangeCwd={changeSessionCwd}
/>
</Pane>

View File

@@ -3,6 +3,7 @@ import { useEffect, useRef } from 'react'
import type { HermesConnection } from '@/global'
import { HermesGateway } from '@/hermes'
import { translateNow } from '@/i18n'
import { desktopDefaultCwd } from '@/lib/desktop-fs'
import { isGatewayReauthRequired, resolveGatewayWsUrl } from '@/lib/gateway-ws-url'
import {
$desktopBoot,
@@ -25,12 +26,16 @@ import {
import { notify, notifyError } from '@/store/notifications'
import { $activeGatewayProfile, normalizeProfileKey, touchActiveGatewayBackend } from '@/store/profile'
import {
$activeSessionId,
$attentionSessionIds,
$connection,
$currentCwd,
$sessions,
$workingSessionIds,
ensureDefaultWorkspaceCwd,
setConnection,
setCurrentBranch,
setCurrentCwd,
setSessionsLoading
} from '@/store/session'
import type { RpcEvent } from '@/types/hermes'
@@ -353,6 +358,11 @@ export function useGatewayBoot({
progress: 97
})
await ensureDefaultWorkspaceCwd()
const remoteDefault = await desktopDefaultCwd().catch(() => null)
if (remoteDefault?.cwd && !$activeSessionId.get() && !$currentCwd.get()) {
setCurrentCwd(remoteDefault.cwd)
setCurrentBranch(remoteDefault.branch || '')
}
await callbacksRef.current.refreshHermesConfig()
if (cancelled) {

View File

@@ -0,0 +1,27 @@
import { createDragDropManager, type DragDropManager } from 'dnd-core'
import { HTML5Backend } from 'react-dnd-html5-backend'
let manager: DragDropManager | null = null
/**
* A single, app-lifetime react-dnd manager for the file tree.
*
* react-arborist mounts its own react-dnd `DndProvider` with `HTML5Backend`
* inside every `<Tree>`. react-dnd v14 stores that provider's manager on a
* global, ref-counted singleton context and nulls it when the count hits 0.
* On a keyed remount (cwd / collapse changes force a fresh `<Tree>`), the
* singleton can be torn down and recreated while the previous `HTML5Backend`
* still owns the `window.__isReactDndHtml5Backend` setup flag — so the new
* backend's `setup()` throws "Cannot have two HTML5 backends at the same
* time." and trips the file-tree error boundary (it never recovers, because
* "Try again" just remounts into the same race).
*
* Passing arborist a stable `dndManager` makes it skip the global-singleton
* path entirely and reuse one backend for the lifetime of the app, so the
* window flag is never double-claimed.
*/
export function getFileTreeDndManager(): DragDropManager {
manager ??= createDragDropManager(HTML5Backend)
return manager
}

View File

@@ -0,0 +1,100 @@
/// <reference types="node" />
import { Buffer } from 'node:buffer'
import { afterEach, beforeEach, describe, expect, it, vi } from 'vitest'
import type { HermesReadDirEntry, HermesReadDirResult } from '@/global'
import { clearProjectDirCache, readProjectDir } from './ipc'
const readDir = vi.fn<(path: string) => Promise<HermesReadDirResult>>()
const readFileDataUrl = vi.fn<(path: string) => Promise<string>>()
const gitRoot = vi.fn<(path: string) => Promise<string | null>>()
function ok(entries: HermesReadDirEntry[]): HermesReadDirResult {
return { entries }
}
function dataUrl(text: string) {
return `data:text/plain;base64,${Buffer.from(text, 'utf8').toString('base64')}`
}
function installBridge() {
;(
window as unknown as {
hermesDesktop: {
gitRoot: typeof gitRoot
readDir: typeof readDir
readFileDataUrl: typeof readFileDataUrl
}
}
).hermesDesktop = { gitRoot, readDir, readFileDataUrl }
}
describe('readProjectDir', () => {
beforeEach(() => {
clearProjectDirCache()
readDir.mockReset()
readFileDataUrl.mockReset()
gitRoot.mockReset()
installBridge()
})
afterEach(() => {
clearProjectDirCache()
delete (window as unknown as { hermesDesktop?: unknown }).hermesDesktop
})
it('returns no-bridge when the desktop bridge is unavailable', async () => {
delete (window as unknown as { hermesDesktop?: unknown }).hermesDesktop
await expect(readProjectDir('/repo')).resolves.toEqual({ entries: [], error: 'no-bridge' })
})
it('filters gitignored entries when readDir returns Windows-style paths', async () => {
gitRoot.mockResolvedValue('C:\\repo')
readDir.mockImplementation(async path => {
if (path === 'C:\\repo\\src') {
return ok([
{ name: 'debug.log', path: 'C:\\repo\\src\\debug.log', isDirectory: false },
{ name: '临时.txt', path: 'C:\\repo\\src\\临时.txt', isDirectory: false },
{ name: 'keep.ts', path: 'C:\\repo\\src\\keep.ts', isDirectory: false }
])
}
if (path === 'C:/repo') {
return ok([{ name: '.gitignore', path: 'C:/repo/.gitignore', isDirectory: false }])
}
if (path === 'C:/repo/src') {
return ok([])
}
return ok([])
})
readFileDataUrl.mockResolvedValue(dataUrl('# Unicode 路径规则\nsrc/*.log\nsrc/临时.txt\n'))
const result = await readProjectDir('C:\\repo\\src', 'C:\\repo')
expect(result.entries.map(entry => entry.name)).toEqual(['keep.ts'])
expect(gitRoot).toHaveBeenCalledWith('C:/repo')
expect(readFileDataUrl).toHaveBeenCalledWith('C:/repo/.gitignore')
})
it('does not fetch .gitignore contents when listings do not contain .gitignore', async () => {
gitRoot.mockResolvedValue('/repo')
readDir.mockImplementation(async path => {
if (path === '/repo/src') {
return ok([{ name: 'debug.log', path: '/repo/src/debug.log', isDirectory: false }])
}
return ok([])
})
const result = await readProjectDir('/repo/src', '/repo')
expect(result.entries.map(entry => entry.name)).toEqual(['debug.log'])
expect(readFileDataUrl).not.toHaveBeenCalled()
})
})

View File

@@ -1,5 +1,6 @@
import ignore from 'ignore'
import { desktopFsCacheKey, desktopGitRoot, readDesktopDir, readDesktopFileDataUrl } from '@/lib/desktop-fs'
import type { HermesReadDirEntry, HermesReadDirResult } from '@/global'
export type ProjectTreeEntry = HermesReadDirEntry
@@ -27,7 +28,7 @@ function decodeDataUrl(dataUrl: string) {
}
function clean(path: string) {
return path.replace(/\/+$/, '') || '/'
return path.replace(/\\/g, '/').replace(/\/+$/, '') || '/'
}
/** Strict POSIX-style relative path; null if `child` is not inside `root`. */
@@ -63,15 +64,11 @@ function ancestorDirs(root: string, dir: string) {
}
async function gitRootFor(start: string) {
if (!window.hermesDesktop?.gitRoot) {
return null
}
const key = clean(start)
const key = `${desktopFsCacheKey()}:${clean(start)}`
let cached = gitRootCache.get(key)
if (!cached) {
cached = window.hermesDesktop.gitRoot(key)
cached = desktopGitRoot(start)
gitRootCache.set(key, cached)
}
@@ -80,18 +77,14 @@ async function gitRootFor(start: string) {
/** Read .gitignore at `dir` if it actually exists — never probe missing files. */
async function readGitignore(dir: string): Promise<GitignoreRule | null> {
if (!window.hermesDesktop?.readDir || !window.hermesDesktop.readFileDataUrl) {
return null
}
try {
const listing = await window.hermesDesktop.readDir(dir)
const listing = await readDesktopDir(dir)
if (!listing.entries.some(e => e.name === '.gitignore' && !e.isDirectory)) {
return null
}
const text = decodeDataUrl(await window.hermesDesktop.readFileDataUrl(`${dir}/.gitignore`))
const text = decodeDataUrl(await readDesktopFileDataUrl(`${dir}/.gitignore`))
return { base: dir, ig: ignore().add(text) }
} catch {
@@ -100,11 +93,11 @@ async function readGitignore(dir: string): Promise<GitignoreRule | null> {
}
async function gitignoreFor(dir: string) {
const key = clean(dir)
const key = `${desktopFsCacheKey()}:${clean(dir)}`
let cached = gitignoreCache.get(key)
if (!cached) {
cached = readGitignore(key)
cached = readGitignore(clean(dir))
gitignoreCache.set(key, cached)
}
@@ -142,9 +135,10 @@ export async function readProjectDir(dirPath: string, rootPath = dirPath): Promi
return { entries: [], error: 'no-bridge' }
}
const result = await window.hermesDesktop.readDir(dirPath)
const result = await readDesktopDir(dirPath)
const entries = result?.entries ?? []
return { ...result, entries: await filterIgnored(result.entries, rootPath, dirPath) }
return { ...result, entries: await filterIgnored(entries, rootPath, dirPath) }
}
export function clearProjectDirCache(rootPath?: string) {
@@ -155,7 +149,7 @@ export function clearProjectDirCache(rootPath?: string) {
return
}
const key = clean(rootPath)
const key = `${desktopFsCacheKey()}:${clean(rootPath)}`
gitRootCache.delete(key)
gitignoreCache.delete(key)
}

View File

@@ -0,0 +1,177 @@
import { useEffect, useMemo, useState } from 'react'
import { Button } from '@/components/ui/button'
import { Codicon } from '@/components/ui/codicon'
import { Dialog, DialogContent, DialogDescription, DialogTitle } from '@/components/ui/dialog'
import { useI18n } from '@/i18n'
import { readDesktopDir, setDesktopFsRemotePicker } from '@/lib/desktop-fs'
import { cn } from '@/lib/utils'
function clean(path: string) {
return path.replace(/\/+$/, '') || '/'
}
function parentDir(path: string) {
const value = clean(path)
if (value === '/') {
return '/'
}
const parent = value.slice(0, value.lastIndexOf('/'))
return parent || '/'
}
function pathName(path: string) {
return path.split('/').filter(Boolean).pop() || path
}
interface PendingSelection {
defaultPath: string
resolve: (paths: string[]) => void
title: string
}
export function RemoteFolderPicker() {
const { t } = useI18n()
const r = t.rightSidebar
const [pending, setPending] = useState<PendingSelection | null>(null)
const [currentPath, setCurrentPath] = useState('/')
const [entries, setEntries] = useState<Array<{ name: string; path: string }>>([])
const [error, setError] = useState<string | null>(null)
const [loading, setLoading] = useState(false)
useEffect(() => {
setDesktopFsRemotePicker({
selectPaths: options =>
new Promise(resolve => {
const defaultPath = clean(options?.defaultPath || '/')
setCurrentPath(defaultPath)
setPending({ defaultPath, resolve, title: options?.title || r.remotePickerTitle })
})
})
return () => setDesktopFsRemotePicker(null)
}, [r.remotePickerTitle])
useEffect(() => {
if (!pending) {
return
}
let active = true
setLoading(true)
setError(null)
void readDesktopDir(currentPath)
.then(result => {
if (!active) {
return
}
if (result.error) {
setError(result.error)
setEntries([])
return
}
setEntries(result.entries.filter(entry => entry.isDirectory).map(entry => ({ name: entry.name, path: entry.path })))
})
.catch(err => {
if (active) {
setError(err instanceof Error ? err.message : String(err))
setEntries([])
}
})
.finally(() => {
if (active) {
setLoading(false)
}
})
return () => {
active = false
}
}, [currentPath, pending])
const crumbs = useMemo(() => {
const parts = clean(currentPath).split('/').filter(Boolean)
const out = [{ label: '/', path: '/' }]
let acc = ''
for (const part of parts) {
acc += `/${part}`
out.push({ label: part, path: acc })
}
return out
}, [currentPath])
const close = (paths: string[] = []) => {
pending?.resolve(paths)
setPending(null)
setEntries([])
setError(null)
}
return (
<Dialog onOpenChange={open => !open && close()} open={Boolean(pending)}>
<DialogContent className="max-w-lg gap-0 overflow-hidden p-0">
<div className="border-b border-border/70 px-4 py-3">
<DialogTitle className="text-sm">{pending?.title || r.remotePickerTitle}</DialogTitle>
<DialogDescription className="mt-1 text-xs">{r.remotePickerDescription}</DialogDescription>
</div>
<div className="flex min-h-[22rem] flex-col">
<div className="flex flex-wrap items-center gap-1 border-b border-border/50 px-3 py-2 text-xs text-muted-foreground">
{crumbs.map((crumb, index) => (
<button
className={cn('rounded px-1.5 py-0.5 hover:bg-muted hover:text-foreground', index === crumbs.length - 1 && 'text-foreground')}
key={crumb.path}
onClick={() => setCurrentPath(crumb.path)}
type="button"
>
{crumb.label}
</button>
))}
</div>
<div className="min-h-0 flex-1 overflow-y-auto p-2">
<FolderRow disabled={currentPath === '/'} name=".." onClick={() => setCurrentPath(parentDir(currentPath))} />
{loading ? (
<div className="flex items-center gap-2 px-2 py-3 text-xs text-muted-foreground">
<Codicon name="loading" size="0.8rem" spinning />
{r.loadingFiles}
</div>
) : error ? (
<div className="px-2 py-3 text-xs text-destructive">{r.unreadableBody(error)}</div>
) : entries.length === 0 ? (
<div className="px-2 py-3 text-xs text-muted-foreground">{r.emptyBody}</div>
) : (
entries.map(entry => <FolderRow key={entry.path} name={pathName(entry.path)} onClick={() => setCurrentPath(entry.path)} />)
)}
</div>
</div>
<div className="flex items-center justify-between gap-2 border-t border-border/70 px-4 py-3">
<div className="min-w-0 truncate text-xs text-muted-foreground">{currentPath}</div>
<div className="flex shrink-0 items-center gap-2">
<Button onClick={() => close()} size="sm" variant="ghost">
{t.common.cancel}
</Button>
<Button onClick={() => close([currentPath])} size="sm">
{r.remotePickerSelect}
</Button>
</div>
</div>
</DialogContent>
</Dialog>
)
}
function FolderRow({ disabled = false, name, onClick }: { disabled?: boolean; name: string; onClick: () => void }) {
return (
<button
className="flex w-full items-center gap-2 rounded-md px-2 py-1.5 text-left text-xs text-(--ui-text-secondary) hover:bg-(--ui-row-hover-background) hover:text-foreground disabled:pointer-events-none disabled:opacity-40"
disabled={disabled}
onClick={onClick}
type="button"
>
<Codicon name="folder" size="0.875rem" />
<span className="min-w-0 truncate">{name}</span>
</button>
)
}

View File

@@ -7,10 +7,13 @@ import { useResizeObserver } from '@/hooks/use-resize-observer'
import { useI18n } from '@/i18n'
import { cn } from '@/lib/utils'
import { getFileTreeDndManager } from './dnd-manager'
import type { TreeNode } from './use-project-tree'
const ROW_HEIGHT = 22
const INDENT = 10
/** Base inset for every row; react-arborist owns paddingLeft for depth indent. */
const TREE_ROW_INSET = 12
interface ProjectTreeProps {
collapseNonce: number
@@ -94,6 +97,7 @@ export function ProjectTree({
disableDrag
disableDrop
disableEdit
dndManager={getFileTreeDndManager()}
height={size.height}
indent={INDENT}
initialOpenState={openState}
@@ -145,7 +149,8 @@ function ProjectTreeRow({
}
const isFolder = node.data.isDirectory
const isPlaceholder = node.data.id.endsWith('::__loading__')
const isPlaceholder = Boolean(node.data.placeholder)
const isErrorPlaceholder = node.data.placeholder === 'error'
return (
<div
@@ -197,21 +202,21 @@ function ProjectTreeRow({
event.dataTransfer.setData('text/plain', node.data.id)
}}
ref={dragHandle}
style={style}
style={{
...style,
paddingLeft:
(typeof style.paddingLeft === 'number'
? style.paddingLeft
: Number.parseFloat(String(style.paddingLeft ?? 0)) || 0) + TREE_ROW_INSET
}}
>
{isFolder && !isPlaceholder && (
<span aria-hidden className="flex w-3 items-center justify-center">
<Codicon
className="text-(--ui-text-tertiary)"
name={node.isOpen ? 'chevron-down' : 'chevron-right'}
size="0.75rem"
/>
</span>
)}
{!isFolder && <span aria-hidden className="w-3 shrink-0" />}
{/* No chevron column — the folder icon (open/closed) already carries the
expand state, so the extra glyph was pure noise. */}
<span aria-hidden className="flex w-3.5 items-center justify-center text-(--ui-text-tertiary)">
{isPlaceholder ? (
{isPlaceholder && !isErrorPlaceholder ? (
<Codicon name="loading" size="0.75rem" spinning />
) : isErrorPlaceholder ? (
<Codicon name="warning" size="0.75rem" />
) : isFolder ? (
<Codicon name={node.isOpen ? 'folder-opened' : 'folder'} size="0.875rem" />
) : (

View File

@@ -1,19 +1,24 @@
import { act, renderHook, waitFor } from '@testing-library/react'
import { act, cleanup, renderHook, waitFor } from '@testing-library/react'
import { afterEach, beforeEach, describe, expect, it, vi } from 'vitest'
import { $connection } from '@/store/session'
import type { HermesReadDirResult } from '@/global'
import { clearProjectDirCache, readProjectDir } from './ipc'
import { resetProjectTreeState, useProjectTree } from './use-project-tree'
const readDir = vi.fn<(path: string) => Promise<HermesReadDirResult>>()
beforeEach(() => {
$connection.set(null)
resetProjectTreeState()
readDir.mockReset()
;(window as unknown as { hermesDesktop: { readDir: typeof readDir } }).hermesDesktop = { readDir }
})
afterEach(() => {
cleanup()
$connection.set(null)
resetProjectTreeState()
delete (window as unknown as { hermesDesktop?: unknown }).hermesDesktop
})
@@ -106,7 +111,37 @@ describe('useProjectTree', () => {
expect(readDir).toHaveBeenCalledTimes(1)
})
it('captures per-folder error code and leaves the folder expandable but empty', async () => {
it('reads gitignore from the real path while caching per connection', async () => {
const readFileDataUrl = vi.fn(async () => `data:text/plain;base64,${btoa('ignored.log\n')}`)
const gitRoot = vi.fn(async () => '/repo')
readDir.mockImplementation(async path => {
if (path === '/repo') return ok([{ name: '.gitignore', path: '/repo/.gitignore', isDirectory: false }])
if (path === '/repo/src') {
return ok([
{ name: 'app.ts', path: '/repo/src/app.ts', isDirectory: false },
{ name: 'ignored.log', path: '/repo/src/ignored.log', isDirectory: false }
])
}
throw new Error(`unexpected path ${path}`)
})
;(window as unknown as { hermesDesktop: unknown }).hermesDesktop = { gitRoot, readDir, readFileDataUrl }
$connection.set({ baseUrl: 'local-a', mode: 'local' } as never)
await expect(readProjectDir('/repo/src', '/repo')).resolves.toMatchObject({
entries: [{ name: 'app.ts', path: '/repo/src/app.ts', isDirectory: false }]
})
expect(readDir).toHaveBeenCalledWith('/repo')
expect(readDir).not.toHaveBeenCalledWith(expect.stringContaining('local-a'))
$connection.set({ baseUrl: 'local-b', mode: 'local' } as never)
clearProjectDirCache()
await expect(readProjectDir('/repo/src', '/repo')).resolves.toMatchObject({
entries: [{ name: 'app.ts', path: '/repo/src/app.ts', isDirectory: false }]
})
expect(readDir.mock.calls.filter(([path]) => path === '/repo')).toHaveLength(2)
})
it('captures per-folder error code and shows an error placeholder child', async () => {
readDir.mockResolvedValueOnce(ok([{ name: 'priv', path: '/p/priv', isDirectory: true }]))
readDir.mockResolvedValueOnce({ entries: [], error: 'EACCES' })
@@ -119,7 +154,14 @@ describe('useProjectTree', () => {
})
expect(result.current.data[0].error).toBe('EACCES')
expect(result.current.data[0].children).toEqual([])
expect(result.current.data[0].children).toEqual([
{
id: '/p/priv::__error__',
isDirectory: false,
name: 'Unable to read (EACCES)',
placeholder: 'error'
}
])
})
it('dedupes concurrent loadChildren calls for the same id', async () => {
@@ -179,6 +221,36 @@ describe('useProjectTree', () => {
expect(readDir).toHaveBeenLastCalledWith('/b')
})
it('falls back to the sanitized workspace dir when the session cwd is gone', async () => {
const sanitizeWorkspaceCwd = vi.fn(async () => ({ cwd: '/home/me/projects', sanitized: true }))
readDir.mockImplementation(async path => {
if (path === '/deleted/worktree') return { entries: [], error: 'ENOENT' }
if (path === '/home/me/projects') return ok([{ name: 'repo', path: '/home/me/projects/repo', isDirectory: true }])
throw new Error(`unexpected path ${path}`)
})
;(window as unknown as { hermesDesktop: unknown }).hermesDesktop = { readDir, sanitizeWorkspaceCwd }
const { result } = renderHook(() => useProjectTree('/deleted/worktree'))
await waitFor(() => expect(result.current.data.length).toBe(1))
expect(sanitizeWorkspaceCwd).toHaveBeenCalledWith('/deleted/worktree')
expect(result.current.rootError).toBeNull()
expect(result.current.effectiveCwd).toBe('/home/me/projects')
expect(result.current.data[0]?.name).toBe('repo')
})
it('keeps the root error when sanitize offers no usable fallback', async () => {
const sanitizeWorkspaceCwd = vi.fn(async () => ({ cwd: '/deleted/worktree', sanitized: false }))
readDir.mockResolvedValue({ entries: [], error: 'ENOENT' })
;(window as unknown as { hermesDesktop: unknown }).hermesDesktop = { readDir, sanitizeWorkspaceCwd }
const { result } = renderHook(() => useProjectTree('/deleted/worktree'))
await waitFor(() => expect(result.current.rootError).toBe('ENOENT'))
expect(result.current.effectiveCwd).toBe('/deleted/worktree')
})
it('returns no-bridge gracefully when window.hermesDesktop is missing', async () => {
delete (window as unknown as { hermesDesktop?: unknown }).hermesDesktop

View File

@@ -2,6 +2,8 @@ import { useStore } from '@nanostores/react'
import { atom } from 'nanostores'
import { useCallback, useEffect, useMemo } from 'react'
import { $connection } from '@/store/session'
import { clearProjectDirCache, readProjectDir } from './ipc'
export interface TreeNode {
@@ -14,11 +16,14 @@ export interface TreeNode {
children?: TreeNode[]
/** True while a readDir for this folder is in flight. */
loading?: boolean
/** Synthetic loading/error rows are not real filesystem entries. */
placeholder?: 'error' | 'loading'
/** Last error code from readDir (e.g. EACCES). Cleared on next successful load. */
error?: string
}
const PLACEHOLDER_ID = '__loading__'
const ERROR_PLACEHOLDER_ID = '__error__'
function makeNode(path: string, name: string, isDirectory: boolean): TreeNode {
return { id: path, isDirectory, name }
@@ -43,13 +48,26 @@ function patchNode(nodes: TreeNode[] | undefined | null, id: string, patch: (n:
}
function placeholderChild(parentId: string): TreeNode {
return { id: `${parentId}::${PLACEHOLDER_ID}`, isDirectory: false, name: 'Loading…' }
return { id: `${parentId}::${PLACEHOLDER_ID}`, isDirectory: false, name: 'Loading…', placeholder: 'loading' }
}
function errorChild(parentId: string, error: string | undefined): TreeNode {
return {
id: `${parentId}::${ERROR_PLACEHOLDER_ID}`,
isDirectory: false,
name: `Unable to read (${error || 'read-error'})`,
placeholder: 'error'
}
}
export interface UseProjectTreeResult {
/** Bumped by collapseAll so callers can remount the tree fully collapsed. */
collapseNonce: number
data: TreeNode[]
/** Directory actually displayed — differs from the requested cwd when the
* session's recorded cwd no longer exists and we fell back to the default
* workspace dir. */
effectiveCwd: string
openState: Record<string, boolean>
rootError: string | null
rootLoading: boolean
@@ -66,6 +84,8 @@ interface ProjectTreeState {
loaded: boolean
openState: Record<string, boolean>
requestId: number
/** Directory the displayed entries were read from ('' until first load). */
resolvedCwd: string
rootError: string | null
rootLoading: boolean
}
@@ -77,6 +97,7 @@ const initialState: ProjectTreeState = {
loaded: false,
openState: {},
requestId: 0,
resolvedCwd: '',
rootError: null,
rootLoading: false
}
@@ -84,6 +105,12 @@ const initialState: ProjectTreeState = {
const inflight = new Set<string>()
const $projectTree = atom<ProjectTreeState>(initialState)
let nextRootRequestId = 0
let lastConnectionKey = ''
// While the root is errored (ENOENT during a session's cwd race, a folder that
// reappears after a checkout, a remote that wasn't ready), keep retrying on a
// slow cadence so the tree self-heals instead of staying "UNREADABLE" forever.
const ROOT_ERROR_RETRY_MS = 3_000
function setProjectTree(updater: (current: ProjectTreeState) => ProjectTreeState) {
$projectTree.set(updater($projectTree.get()))
@@ -95,6 +122,31 @@ function clearProjectTree() {
$projectTree.set({ ...initialState, requestId: nextRootRequestId })
}
/** Sessions record their launch cwd; deleted worktrees and remote-backend
* paths arrive here as directories that don't exist on this machine. Rather
* than bricking the tree, display the sanitized workspace fallback (main
* prefers the configured default project dir). Local connections only —
* remote trees are read through the remote bridge. */
async function fallbackRootFor(cwd: string): Promise<string | null> {
if ($connection.get()?.mode === 'remote') {
return null
}
const sanitize = window.hermesDesktop?.sanitizeWorkspaceCwd
if (!sanitize) {
return null
}
try {
const { cwd: fallback, sanitized } = await sanitize(cwd)
return sanitized && fallback && fallback !== cwd ? fallback : null
} catch {
return null
}
}
async function loadRoot(cwd: string, { force = false }: { force?: boolean } = {}) {
if (!cwd) {
clearProjectTree()
@@ -123,11 +175,27 @@ async function loadRoot(cwd: string, { force = false }: { force?: boolean } = {}
loaded: false,
openState: current.cwd === cwd ? current.openState : {},
requestId,
resolvedCwd: '',
rootError: null,
rootLoading: true
})
const { entries, error } = await readProjectDir(cwd, cwd)
let resolvedCwd = cwd
let { entries, error } = await readProjectDir(cwd, cwd)
if (error) {
const fallback = await fallbackRootFor(cwd)
if (fallback) {
const retry = await readProjectDir(fallback, fallback)
if (!retry.error) {
resolvedCwd = fallback
entries = retry.entries
error = undefined
}
}
}
setProjectTree(latest => {
if (latest.cwd !== cwd || latest.requestId !== requestId) {
@@ -138,6 +206,7 @@ async function loadRoot(cwd: string, { force = false }: { force?: boolean } = {}
...latest,
data: error ? [] : entries.map(e => makeNode(e.path, e.name, e.isDirectory)),
loaded: true,
resolvedCwd,
rootError: error || null,
rootLoading: false
}
@@ -145,6 +214,7 @@ async function loadRoot(cwd: string, { force = false }: { force?: boolean } = {}
}
export function resetProjectTreeState() {
lastConnectionKey = ''
clearProjectTree()
clearProjectDirCache()
}
@@ -158,6 +228,8 @@ export function resetProjectTreeState() {
*/
export function useProjectTree(cwd: string): UseProjectTreeResult {
const state = useStore($projectTree)
const connection = useStore($connection)
const connectionKey = `${connection?.mode || 'local'}:${connection?.profile || ''}:${connection?.baseUrl || ''}`
const refreshRoot = useCallback(() => loadRoot(cwd, { force: true }), [cwd])
@@ -212,7 +284,8 @@ export function useProjectTree(cwd: string): UseProjectTreeResult {
}
})
const { entries, error } = await readProjectDir(id, cwd)
const rootPath = $projectTree.get().resolvedCwd || cwd
const { entries, error } = await readProjectDir(id, rootPath)
inflight.delete(id)
@@ -227,7 +300,7 @@ export function useProjectTree(cwd: string): UseProjectTreeResult {
...n,
loading: false,
error: error || undefined,
children: error ? [] : entries.map(e => makeNode(e.path, e.name, e.isDirectory))
children: error ? [errorChild(n.id, error)] : entries.map(e => makeNode(e.path, e.name, e.isDirectory))
}))
}
})
@@ -236,14 +309,64 @@ export function useProjectTree(cwd: string): UseProjectTreeResult {
)
useEffect(() => {
const connectionChanged = lastConnectionKey !== '' && lastConnectionKey !== connectionKey
lastConnectionKey = connectionKey
if (connectionChanged) {
clearProjectDirCache()
void loadRoot(cwd, { force: true })
return
}
void loadRoot(cwd)
}, [cwd])
}, [connectionKey, cwd])
// Self-heal: an errored root re-probes every few seconds while the tree is
// mounted. Each attempt bumps requestId, so a persistent error re-arms the
// timer; a success clears rootError and stops it.
useEffect(() => {
if (!cwd || state.cwd !== cwd || !state.rootError) {
return
}
const timer = window.setTimeout(() => void loadRoot(cwd, { force: true }), ROOT_ERROR_RETRY_MS)
return () => window.clearTimeout(timer)
}, [cwd, state.cwd, state.requestId, state.rootError])
// While showing the fallback root, quietly re-probe the session's real cwd
// (a worktree re-created, a checkout restored) and switch back when it
// reappears. The probe never touches state, so there's no flicker.
const usingFallback = state.cwd === cwd && Boolean(state.resolvedCwd) && state.resolvedCwd !== cwd
useEffect(() => {
if (!cwd || !usingFallback) {
return
}
let cancelled = false
const timer = window.setInterval(() => {
void readProjectDir(cwd, cwd).then(({ error }) => {
if (!cancelled && !error) {
void loadRoot(cwd, { force: true })
}
})
}, ROOT_ERROR_RETRY_MS)
return () => {
cancelled = true
window.clearInterval(timer)
}
}, [cwd, usingFallback])
return useMemo(
() => ({
collapseAll,
collapseNonce: state.cwd === cwd ? state.collapseNonce : 0,
data: state.cwd === cwd ? state.data : [],
effectiveCwd: state.cwd === cwd && state.resolvedCwd ? state.resolvedCwd : cwd,
loadChildren,
openState: state.cwd === cwd ? state.openState : {},
refreshRoot,
@@ -261,6 +384,7 @@ export function useProjectTree(cwd: string): UseProjectTreeResult {
state.cwd,
state.data,
state.openState,
state.resolvedCwd,
state.rootError,
state.rootLoading
]

View File

@@ -5,8 +5,8 @@ import { ErrorBoundary } from '@/components/error-boundary'
import { Button } from '@/components/ui/button'
import { Codicon } from '@/components/ui/codicon'
import { Loader } from '@/components/ui/loader'
import { Tip } from '@/components/ui/tooltip'
import { useI18n } from '@/i18n'
import { selectDesktopPaths } from '@/lib/desktop-fs'
import { normalizeOrLocalPreviewTarget } from '@/lib/local-preview'
import { cn } from '@/lib/utils'
import { $panesFlipped } from '@/store/layout'
@@ -16,6 +16,7 @@ import { $currentCwd } from '@/store/session'
import { SidebarPanelLabel } from '../shell/sidebar-label'
import { RemoteFolderPicker } from './files/remote-picker'
import { ProjectTree } from './files/tree'
import { useProjectTree } from './files/use-project-tree'
@@ -32,17 +33,11 @@ export function RightSidebarPane({ onActivateFile, onActivateFolder, onChangeCwd
const currentCwd = useStore($currentCwd).trim()
const hasCwd = currentCwd.length > 0
const cwdName = hasCwd
? (currentCwd
.split(/[\\/]+/)
.filter(Boolean)
.pop() ?? currentCwd)
: r.noFolderSelected
const {
collapseAll,
collapseNonce,
data,
effectiveCwd,
loadChildren,
openState,
refreshRoot,
@@ -51,11 +46,18 @@ export function RightSidebarPane({ onActivateFile, onActivateFolder, onChangeCwd
setNodeOpen
} = useProjectTree(currentCwd)
const cwdName = hasCwd
? (effectiveCwd
.split(/[\\/]+/)
.filter(Boolean)
.pop() ?? effectiveCwd)
: r.noFolderSelected
const canCollapse = Object.values(openState).some(Boolean)
const chooseFolder = async () => {
const selected = await window.hermesDesktop?.selectPaths({
defaultPath: hasCwd ? currentCwd : undefined,
const selected = await selectDesktopPaths({
defaultPath: hasCwd ? effectiveCwd : undefined,
directories: true,
multiple: false,
title: r.changeCwdTitle
@@ -68,7 +70,7 @@ export function RightSidebarPane({ onActivateFile, onActivateFolder, onChangeCwd
const previewFile = async (path: string) => {
try {
const preview = await normalizeOrLocalPreviewTarget(path, currentCwd || undefined)
const preview = await normalizeOrLocalPreviewTarget(path, effectiveCwd || undefined)
if (!preview) {
throw new Error(r.couldNotPreview(path))
@@ -90,10 +92,12 @@ export function RightSidebarPane({ onActivateFile, onActivateFolder, onChangeCwd
: 'border-l shadow-[inset_0.0625rem_0_0_color-mix(in_srgb,white_18%,transparent)]'
)}
>
<RemoteFolderPicker />
<FilesystemTab
canCollapse={canCollapse}
collapseNonce={collapseNonce}
cwd={currentCwd}
cwd={effectiveCwd}
cwdName={cwdName}
data={data}
error={rootError}
@@ -122,13 +126,12 @@ interface FilesystemTabProps extends FileTreeBodyProps {
onRefresh: () => void
}
// Sidebar-specific color/hover treatment only — size, radius, cursor and the
// base focus ring come from <Button size="icon-xs">. This constant exists
// purely to share the sidebar palette + the hover-reveal behavior below.
// Sidebar palette + hover-reveal: refresh tracks label hover; collapse-all
// stays visible while any folder is expanded.
const HEADER_ACTION_CLASS =
'text-sidebar-foreground/70 hover:bg-sidebar-accent! hover:text-sidebar-accent-foreground! focus-visible:ring-sidebar-ring'
const HEADER_ACTION_REVEAL_CLASS = `${HEADER_ACTION_CLASS} pointer-events-none opacity-0 transition-opacity focus-visible:opacity-100 group-focus-within/project-header:pointer-events-auto group-focus-within/project-header:opacity-100 group-hover/project-header:pointer-events-auto group-hover/project-header:opacity-100`
const HEADER_ACTION_LABEL_REVEAL = `${HEADER_ACTION_CLASS} pointer-events-none opacity-0 transition-opacity focus-visible:pointer-events-auto focus-visible:opacity-100 peer-focus-visible/project-label:pointer-events-auto peer-focus-visible/project-label:opacity-100 peer-hover/project-label:pointer-events-auto peer-hover/project-label:opacity-100`
function FilesystemTab({
canCollapse,
@@ -153,20 +156,20 @@ function FilesystemTab({
const r = t.rightSidebar
return (
<div className="group/project-header flex min-h-0 flex-1 flex-col">
<div className="flex min-h-0 flex-1 flex-col">
<RightSidebarSectionHeader>
<Tip label={hasCwd ? r.folderTip(cwd) : r.openFolder}>
<div className="peer/project-label flex min-w-0 flex-1">
<button
className="flex min-w-0 flex-1 items-center rounded-md text-left hover:text-(--ui-text-secondary)"
className="flex w-full min-w-0 items-center rounded-md text-left hover:text-(--ui-text-secondary)"
onClick={() => void onChangeFolder()}
type="button"
>
<SidebarPanelLabel>{cwdName}</SidebarPanelLabel>
</button>
</Tip>
</div>
<Button
aria-label={r.refreshTree}
className={HEADER_ACTION_CLASS}
className={HEADER_ACTION_LABEL_REVEAL}
disabled={!hasCwd || loading}
onClick={onRefresh}
size="icon-xs"
@@ -185,7 +188,7 @@ function FilesystemTab({
</Button>
<Button
aria-label={r.collapseAll}
className={HEADER_ACTION_REVEAL_CLASS}
className={cn(HEADER_ACTION_CLASS, !canCollapse && 'pointer-events-none opacity-0')}
disabled={!hasCwd || !canCollapse}
onClick={onCollapseAll}
size="icon-xs"
@@ -205,6 +208,7 @@ function FilesystemTab({
onLoadChildren={onLoadChildren}
onNodeOpenChange={onNodeOpenChange}
onPreviewFile={onPreviewFile}
onRetry={onRefresh}
openState={openState}
/>
</div>
@@ -226,6 +230,9 @@ interface FileTreeBodyProps {
onLoadChildren: (id: string) => void | Promise<void>
onNodeOpenChange: (id: string, open: boolean) => void
onPreviewFile?: (path: string) => void
/** Force-reload the root. The hook also auto-retries while errored, so this
* is the impatient-user path. */
onRetry?: () => void
openState: ReturnType<typeof useProjectTree>['openState']
}
@@ -240,6 +247,7 @@ function FileTreeBody({
onLoadChildren,
onNodeOpenChange,
onPreviewFile,
onRetry,
openState
}: FileTreeBodyProps) {
const { t } = useI18n()
@@ -250,7 +258,20 @@ function FileTreeBody({
}
if (error) {
return <EmptyState body={r.unreadableBody(error)} title={r.unreadableTitle} />
return (
<div className="flex min-h-0 flex-1 flex-col items-center justify-center gap-2 px-4 text-center">
<EmptyState body={r.unreadableBody(error)} title={r.unreadableTitle} />
{onRetry && (
<button
className="text-[0.68rem] font-medium text-muted-foreground transition hover:text-foreground"
onClick={onRetry}
type="button"
>
{r.tryAgain}
</button>
)}
</div>
)
}
if (loading && data.length === 0) {

View File

@@ -9,7 +9,7 @@ import { useI18n } from '@/i18n'
import { SidebarPanelLabel } from '../../shell/sidebar-label'
import { setTerminalTakeover } from '../store'
import { addSelectionShortcutLabel } from './selection'
import { KbdCombo } from '@/components/ui/kbd'
import { useTerminalSession } from './use-terminal-session'
interface TerminalTabProps {
@@ -69,7 +69,7 @@ export function TerminalTab({ cwd, onAddSelectionToChat }: TerminalTabProps) {
variant="secondary"
>
{t.rightSidebar.addToChat}
<span className="ml-1 text-[0.6rem] text-(--ui-text-tertiary)">{addSelectionShortcutLabel()}</span>
<KbdCombo className="ml-1 opacity-70" combo="mod+l" size="sm" />
</Button>
</div>
)}

View File

@@ -99,8 +99,6 @@ export function resolveSurfaceColor(fallback: string): string {
export const isMacPlatform = () => navigator.platform.toLowerCase().includes('mac')
export const addSelectionShortcutLabel = () => (isMacPlatform() ? '⌘L' : 'Ctrl+L')
export function isAddSelectionShortcut(event: KeyboardEvent) {
const mod = isMacPlatform() ? event.metaKey : event.ctrlKey

View File

@@ -7,6 +7,7 @@ import { useCallback, useEffect, useMemo, useRef, useState } from 'react'
import type { CSSProperties } from 'react'
import { triggerHaptic } from '@/lib/haptics'
import { $filePreviewTarget, $previewTarget } from '@/store/preview'
import { useTheme } from '@/themes/context'
import { makeTerminalReader, setActiveTerminalReader } from './buffer'
@@ -20,6 +21,17 @@ import {
type TerminalStatus = 'closed' | 'open' | 'starting'
// ⌘/Ctrl+L is a global shortcut, so a text selection in the file preview pane
// lands in this handler with no xterm selection. Label those with the previewed
// file's name instead of the shell, so the composer ref reads as a file quote
// rather than a bogus "zsh:N lines".
function previewSelectionLabel(): string {
const target = $filePreviewTarget.get() ?? $previewTarget.get()
const source = target?.path || target?.url || ''
return source.split(/[\\/]/).filter(Boolean).pop() || target?.label?.trim() || ''
}
const HERMES_PATHS_MIME = 'application/x-hermes-paths'
function readEscapeSequence(data: string, index: number) {
@@ -257,16 +269,20 @@ export function useTerminalSession({ cwd, onAddSelectionToChat }: UseTerminalSes
)
const addSelectionToChat = useCallback(() => {
const selectedText = readSelection() || selectionRef.current
const termSelection = (termRef.current?.getSelection() || selectionRef.current).trim()
const selectedText = termSelection || window.getSelection()?.toString() || ''
const trimmed = selectedText.trim()
if (!trimmed) {
return
}
const label =
selectionLabelRef.current ||
(termRef.current ? terminalSelectionLabel(termRef.current, shellNameRef.current, selectedText) : 'selection')
// Terminal selection → shell-anchored label; anything else came from the
// preview pane sharing this global shortcut → label it with the file.
const label = termSelection
? selectionLabelRef.current ||
(termRef.current ? terminalSelectionLabel(termRef.current, shellNameRef.current, selectedText) : 'selection')
: previewSelectionLabel() || 'selection'
onAddSelectionToChatRef.current(trimmed, label)
termRef.current?.clearSelection()
@@ -275,7 +291,7 @@ export function useTerminalSession({ cwd, onAddSelectionToChat }: UseTerminalSes
setSelection('')
setSelectionStyle(null)
triggerHaptic('selection')
}, [readSelection])
}, [])
// Always listen — gating on the React selection state misses selections the
// TUI redraw races. Only swallow ⌘/Ctrl+L when there's text to send, else it
@@ -315,8 +331,11 @@ export function useTerminalSession({ cwd, onAddSelectionToChat }: UseTerminalSes
allowTransparency: true,
convertEol: true,
cursorBlink: true,
fontFamily: "'SF Mono', 'Menlo', 'Cascadia Code', 'JetBrains Mono', monospace",
fontFamily: "'JetBrains Mono', 'Cascadia Code', 'SF Mono', Menlo, Consolas, monospace",
fontSize: 11,
fontWeight: '400',
fontWeightBold: '700',
letterSpacing: 0,
lineHeight: 1.12,
// Full-screen TUIs (hermes --tui, vim) grab the mouse, so a plain drag
// can't select — ⌥-drag (macOS) / Shift-drag (else) forces a native
@@ -598,13 +617,13 @@ export function useTerminalSession({ cwd, onAddSelectionToChat }: UseTerminalSes
startSession()
}
const fonts = typeof document !== 'undefined' ? document.fonts : undefined
// fonts.ready settles only already-requested faces; bold/italic aren't asked
// for until styled output paints (past atlas init), so warm them up front.
const warm = document.fonts?.load
? Promise.allSettled(['400', '700', 'italic 400'].map(v => document.fonts.load(`${v} 11px 'JetBrains Mono'`)))
: Promise.resolve()
if (fonts?.ready) {
void fonts.ready.then(mount, mount)
} else {
mount()
}
void warm.then(mount, mount)
return () => {
disposed = true

View File

@@ -0,0 +1,32 @@
import { useStore } from '@nanostores/react'
import { SessionPickerDialog } from '@/components/session-picker'
import { $gatewayState, $selectedStoredSessionId, $sessionPickerOpen, setSessionPickerOpen } from '@/store/session'
interface SessionPickerOverlayProps {
onResume: (storedSessionId: string) => void
}
/**
* Mounts the session picker that `/resume` (and `/sessions`, `/switch`) opens —
* the desktop equivalent of the TUI's sessions overlay. Resuming runs through
* the same `resumeSession` path the sidebar uses.
*/
export function SessionPickerOverlay({ onResume }: SessionPickerOverlayProps) {
const open = useStore($sessionPickerOpen)
const gatewayOpen = useStore($gatewayState) === 'open'
const activeStoredSessionId = useStore($selectedStoredSessionId)
if (!gatewayOpen) {
return null
}
return (
<SessionPickerDialog
activeStoredSessionId={activeStoredSessionId}
onOpenChange={setSessionPickerOpen}
onResume={onResume}
open={open}
/>
)
}

View File

@@ -18,7 +18,9 @@ import { coerceGatewayText, coerceThinkingText, normalizePersonalityValue } from
import { gatewayEventRequiresSessionId } from '@/lib/gateway-events'
import { triggerHaptic } from '@/lib/haptics'
import { isProviderSetupErrorMessage } from '@/lib/provider-setup-errors'
import { parseTodos } from '@/lib/todos'
import { setClarifyRequest } from '@/store/clarify'
import { refreshBackgroundProcesses } from '@/store/composer-status'
import { $gateway } from '@/store/gateway'
import { notify } from '@/store/notifications'
import { requestDesktopOnboarding } from '@/store/onboarding'
@@ -37,6 +39,7 @@ import {
setYoloActive
} from '@/store/session'
import { clearSessionSubagents, pruneDelegateFallbackSubagents, upsertSubagent } from '@/store/subagents'
import { setSessionTodos } from '@/store/todos'
import { recordToolDiff } from '@/store/tool-diffs'
import type { RpcEvent } from '@/types/hermes'
@@ -52,6 +55,7 @@ interface MessageStreamOptions {
queryClient: QueryClient
refreshHermesConfig: () => Promise<void>
refreshSessions: () => Promise<void>
sessionStateByRuntimeIdRef: MutableRefObject<Map<string, ClientSessionState>>
updateSessionState: (
sessionId: string,
updater: (state: ClientSessionState) => ClientSessionState,
@@ -64,6 +68,59 @@ interface QueuedStreamDeltas {
reasoning: string
}
type SessionRuntimeStatePatch = Partial<
Pick<
ClientSessionState,
'branch' | 'cwd' | 'fast' | 'model' | 'personality' | 'provider' | 'reasoningEffort' | 'serviceTier' | 'yolo'
>
>
function sessionInfoStatePatch(payload: GatewayEventPayload | undefined): SessionRuntimeStatePatch {
const patch: SessionRuntimeStatePatch = {}
if (typeof payload?.model === 'string') {
patch.model = payload.model || ''
}
if (typeof payload?.provider === 'string') {
patch.provider = payload.provider || ''
}
if (typeof payload?.cwd === 'string') {
patch.cwd = payload.cwd
}
if (typeof payload?.branch === 'string') {
patch.branch = payload.branch
}
if (typeof payload?.personality === 'string') {
patch.personality = normalizePersonalityValue(payload.personality)
}
if (typeof payload?.reasoning_effort === 'string') {
patch.reasoningEffort = payload.reasoning_effort
}
if (typeof payload?.service_tier === 'string') {
patch.serviceTier = payload.service_tier
}
if (typeof payload?.fast === 'boolean') {
patch.fast = payload.fast
}
if (typeof payload?.yolo === 'boolean') {
patch.yolo = payload.yolo
}
return patch
}
function hasSessionInfoStatePatch(patch: SessionRuntimeStatePatch): boolean {
return Object.keys(patch).length > 0
}
// Minimum gap between two assistant-text flushes during a stream. Was 16ms
// (rAF only), which at typical LLM token rates of ~30-80 tok/sec meant every
// token got its own React commit + Streamdown markdown re-parse, scaling
@@ -192,8 +249,14 @@ export function useMessageStream({
queryClient,
refreshHermesConfig,
refreshSessions,
sessionStateByRuntimeIdRef,
updateSessionState
}: MessageStreamOptions) {
const sessionInterrupted = useCallback(
(sessionId: string) => sessionStateByRuntimeIdRef.current.get(sessionId)?.interrupted ?? false,
[sessionStateByRuntimeIdRef]
)
// Patch the in-flight assistant message (or seed it). Centralises the
// streamId/groupId bookkeeping every event callback would otherwise repeat.
const mutateStream = useCallback(
@@ -417,6 +480,20 @@ export function useMessageStream({
// a tool part can't jump ahead of the text that preceded it.
flushQueuedDeltas(sessionId)
if (sessionInterrupted(sessionId)) {
return
}
// The composer status stack owns todo display now (no inline panel) —
// mirror every todo state the tool reports into its session store.
if (payload?.name === 'todo') {
const todos = parseTodos(payload.todos) ?? parseTodos(payload.result) ?? parseTodos(payload.args)
if (todos) {
setSessionTodos(sessionId, todos)
}
}
if (!nativeSubagentSessionsRef.current.has(sessionId)) {
for (const subagentPayload of delegateTaskPayloads(payload, phase, sourceEventType)) {
upsertSubagent(
@@ -435,7 +512,7 @@ export function useMessageStream({
{ pending: m => phase !== 'complete' || (m.pending ?? false) }
)
},
[flushQueuedDeltas, mutateStream]
[flushQueuedDeltas, mutateStream, sessionInterrupted]
)
const completeAssistantMessage = useCallback(
@@ -616,9 +693,11 @@ export function useMessageStream({
(event: RpcEvent) => {
const payload = event.payload as GatewayEventPayload | undefined
const explicitSid = event.session_id || ''
if (!explicitSid && gatewayEventRequiresSessionId(event.type)) {
return
}
const sessionId = explicitSid || activeSessionIdRef.current
const isActiveEvent = !!sessionId && sessionId === activeSessionIdRef.current
@@ -628,13 +707,13 @@ export function useMessageStream({
// Apply session-scoped fields when the event targets the active
// session, OR when it's a global broadcast and we have no session.
const apply = explicitSid ? isActiveEvent : !activeSessionIdRef.current
const statePatch = sessionInfoStatePatch(payload)
const hasStatePatch = hasSessionInfoStatePatch(statePatch)
const modelChanged = typeof payload?.model === 'string'
const providerChanged = typeof payload?.provider === 'string'
const runningChanged = typeof payload?.running === 'boolean'
if (apply) {
const runtimeInfo: { branch?: string; cwd?: string } = {}
if (modelChanged) {
setCurrentModel(payload!.model || '')
}
@@ -645,20 +724,10 @@ export function useMessageStream({
if (typeof payload?.cwd === 'string') {
setCurrentCwd(payload.cwd)
runtimeInfo.cwd = payload.cwd
}
if (typeof payload?.branch === 'string') {
setCurrentBranch(payload.branch)
runtimeInfo.branch = payload.branch
}
if (sessionId && (runtimeInfo.cwd !== undefined || runtimeInfo.branch !== undefined)) {
updateSessionState(sessionId, state => ({
...state,
branch: runtimeInfo.branch ?? state.branch,
cwd: runtimeInfo.cwd ?? state.cwd
}))
}
if (typeof payload?.personality === 'string') {
@@ -680,7 +749,18 @@ export function useMessageStream({
if (typeof payload?.yolo === 'boolean') {
setYoloActive(payload.yolo)
}
}
if (sessionId && hasStatePatch) {
updateSessionState(sessionId, state => ({
...state,
...statePatch,
branch: statePatch.branch ?? state.branch,
cwd: statePatch.cwd ?? state.cwd
}))
}
if (apply) {
if (runningChanged && sessionId) {
updateSessionState(sessionId, state => {
const busy = Boolean(payload!.running)
@@ -813,13 +893,22 @@ export function useMessageStream({
// the sidebar indicator clears as soon as it's answered, not only at
// message.complete.
updateSessionState(sessionId, state => (state.needsInput ? { ...state, needsInput: false } : state))
// terminal/process tool calls are the only things that spawn or reap
// background processes — sync the composer status stack right after.
if (
!sessionInterrupted(sessionId) &&
(payload?.name === 'terminal' || payload?.name === 'process')
) {
void refreshBackgroundProcesses(sessionId)
}
}
if (typeof payload?.inline_diff === 'string' && payload.inline_diff.trim()) {
recordToolDiff(payload.tool_id || payload.name || '', payload.inline_diff)
}
} else if (SUBAGENT_EVENT_TYPES.has(event.type)) {
if (sessionId && payload) {
if (sessionId && payload && !sessionInterrupted(sessionId)) {
if (!nativeSubagentSessionsRef.current.has(sessionId)) {
pruneDelegateFallbackSubagents(sessionId)
}
@@ -871,6 +960,8 @@ export function useMessageStream({
// raise it and wait — the sidebar flags "needs input" and the inline bar
// surfaces once the user focuses that chat.
setApprovalRequest({
// false only when a tirith warning forbids it; backend omits the field otherwise.
allowPermanent: payload?.allow_permanent !== false,
command: typeof payload?.command === 'string' ? payload.command : '',
description: typeof payload?.description === 'string' ? payload.description : 'dangerous command',
sessionId: sessionId ?? null
@@ -923,6 +1014,12 @@ export function useMessageStream({
text: result ? JSON.stringify(result) : ''
})
}
} else if (event.type === 'status.update') {
// The gateway's notification poller announces background process
// completions / watch matches here — re-sync the status stack.
if (sessionId && payload?.kind === 'process') {
void refreshBackgroundProcesses(sessionId)
}
} else if (event.type === 'error') {
const errorMessage = payload?.message || 'Hermes reported an error'
const looksLikeProviderSetup = isProviderSetupErrorMessage(errorMessage)
@@ -963,6 +1060,7 @@ export function useMessageStream({
flushQueuedDeltas,
queryClient,
refreshHermesConfig,
sessionInterrupted,
updateSessionState,
upsertToolCall
]

View File

@@ -0,0 +1,77 @@
import { renderHook } from '@testing-library/react'
import { QueryClient } from '@tanstack/react-query'
import { afterEach, beforeEach, describe, expect, it, vi } from 'vitest'
import { getGlobalModelInfo } from '@/hermes'
import {
$activeSessionId,
$currentModel,
$currentProvider,
setCurrentModel,
setCurrentProvider
} from '@/store/session'
import { useModelControls } from './use-model-controls'
vi.mock('@/hermes', () => ({
getGlobalModelInfo: vi.fn(),
setGlobalModel: vi.fn()
}))
describe('useModelControls.refreshCurrentModel', () => {
beforeEach(() => {
$activeSessionId.set(null)
setCurrentModel('')
setCurrentProvider('')
})
afterEach(() => {
vi.restoreAllMocks()
$activeSessionId.set(null)
setCurrentModel('')
setCurrentProvider('')
})
it('applies the global model when there is no active runtime session', async () => {
vi.mocked(getGlobalModelInfo).mockResolvedValue({
model: 'openai/gpt-5.5',
provider: 'openai-codex'
})
const { result } = renderHook(() =>
useModelControls({
activeSessionId: null,
queryClient: new QueryClient(),
requestGateway: vi.fn()
})
)
await result.current.refreshCurrentModel()
expect($currentModel.get()).toBe('openai/gpt-5.5')
expect($currentProvider.get()).toBe('openai-codex')
})
it('does not clobber the active session footer state with global model info', async () => {
setCurrentModel('deepseek/deepseek-v4-pro')
setCurrentProvider('deepseek')
$activeSessionId.set('runtime-1')
vi.mocked(getGlobalModelInfo).mockResolvedValue({
model: 'openai/gpt-5.5',
provider: 'openai-codex'
})
const { result } = renderHook(() =>
useModelControls({
activeSessionId: 'runtime-1',
queryClient: new QueryClient(),
requestGateway: vi.fn()
})
)
await result.current.refreshCurrentModel()
expect($currentModel.get()).toBe('deepseek/deepseek-v4-pro')
expect($currentProvider.get()).toBe('deepseek')
})
})

View File

@@ -4,7 +4,13 @@ import { useCallback } from 'react'
import { getGlobalModelInfo, setGlobalModel } from '@/hermes'
import { useI18n } from '@/i18n'
import { notifyError } from '@/store/notifications'
import { $currentModel, $currentProvider, setCurrentModel, setCurrentProvider } from '@/store/session'
import {
$activeSessionId,
$currentModel,
$currentProvider,
setCurrentModel,
setCurrentProvider
} from '@/store/session'
import type { ModelOptionsResponse } from '@/types/hermes'
interface ModelSelection {
@@ -39,6 +45,13 @@ export function useModelControls({ activeSessionId, queryClient, requestGateway
try {
const result = await getGlobalModelInfo()
// A resumed/live session owns the footer model state. Global config
// refreshes (gateway boot, profile swap, settings save) must not clobber
// the active chat's runtime model/provider in the status bar.
if ($activeSessionId.get()) {
return
}
if (typeof result.model === 'string') {
setCurrentModel(result.model)
}

View File

@@ -1,10 +1,11 @@
import { cleanup, render, waitFor } from '@testing-library/react'
import type { MutableRefObject } from 'react'
import { useEffect } from 'react'
import { useEffect, useRef } from 'react'
import { afterEach, beforeEach, describe, expect, it, vi } from 'vitest'
import { textPart } from '@/lib/chat-messages'
import { $composerAttachments, type ComposerAttachment } from '@/store/composer'
import { $connection, $sessions, setSessions } from '@/store/session'
import { $busy, $connection, $messages, $sessions, setSessions } from '@/store/session'
import type { SessionInfo } from '@/types/hermes'
import { uploadComposerAttachment, usePromptActions } from './use-prompt-actions'
@@ -42,6 +43,8 @@ function sessionInfo(overrides: Partial<SessionInfo> = {}): SessionInfo {
}
interface HarnessHandle {
cancelRun: () => Promise<void>
restoreToMessage: (messageId: string) => Promise<void>
steerPrompt: (text: string) => Promise<boolean>
submitText: (
text: string,
@@ -55,6 +58,8 @@ function Harness({
onSeedState,
refreshSessions,
requestGateway,
resumeStoredSession,
seedMessages,
storedSessionId
}: {
busyRef?: MutableRefObject<boolean>
@@ -62,6 +67,8 @@ function Harness({
onSeedState?: (state: Record<string, unknown>) => void
refreshSessions: () => Promise<void>
requestGateway: <T>(method: string, params?: Record<string, unknown>) => Promise<T>
resumeStoredSession?: (storedSessionId: string) => Promise<void> | void
seedMessages?: unknown[]
storedSessionId?: null | string
}) {
const activeSessionIdRef: MutableRefObject<string | null> = { current: RUNTIME_SESSION_ID }
@@ -69,6 +76,12 @@ function Harness({
current: storedSessionId === undefined ? RUNTIME_SESSION_ID : storedSessionId
}
const localBusyRef = busyRef ?? { current: false }
const stateRef = useRef({
messages: seedMessages ?? [],
busy: false,
awaitingResponse: false,
interrupted: true
} as never)
const actions = usePromptActions({
activeSessionId: RUNTIME_SESSION_ID,
@@ -79,17 +92,14 @@ function Harness({
handleSkinCommand: () => '',
refreshSessions,
requestGateway,
resumeStoredSession: resumeStoredSession ?? (() => undefined),
selectedStoredSessionIdRef,
startFreshSessionDraft: () => undefined,
sttEnabled: false,
updateSessionState: (_sessionId, updater) => {
// Seed with interrupted:true so we can prove a fresh submit clears it.
const next = updater({
messages: [],
busy: false,
awaitingResponse: false,
interrupted: true
} as never) as unknown as Record<string, unknown>
const next = updater(stateRef.current) as unknown as Record<string, unknown>
stateRef.current = next as never
onSeedState?.(next)
return next as never
@@ -97,8 +107,13 @@ function Harness({
})
useEffect(() => {
onReady({ steerPrompt: actions.steerPrompt, submitText: actions.submitText })
}, [actions.steerPrompt, actions.submitText, onReady])
onReady({
cancelRun: actions.cancelRun,
restoreToMessage: actions.restoreToMessage,
steerPrompt: actions.steerPrompt,
submitText: actions.submitText
})
}, [actions.cancelRun, actions.restoreToMessage, actions.steerPrompt, actions.submitText, onReady])
return null
}
@@ -190,6 +205,68 @@ describe('usePromptActions /title', () => {
})
})
describe('usePromptActions desktop slash pickers', () => {
beforeEach(() => {
setSessions(() => [sessionInfo({ id: '20260610_120000_abcdef', title: 'Loaded session' })])
})
afterEach(() => {
cleanup()
vi.useRealTimers()
vi.restoreAllMocks()
})
it('resumes an exact session id even when it is not in the loaded sidebar cache', async () => {
const resumeStoredSession = vi.fn(async () => undefined)
const requestGateway = vi.fn(async () => ({}) as never)
let handle: HarnessHandle | null = null
render(
<Harness
onReady={h => (handle = h)}
refreshSessions={async () => undefined}
requestGateway={requestGateway}
resumeStoredSession={resumeStoredSession}
/>
)
await handle!.submitText('/resume 20260610_130000_123abc')
expect(resumeStoredSession).toHaveBeenCalledWith('20260610_130000_123abc')
expect(requestGateway).not.toHaveBeenCalledWith('slash.exec', expect.anything())
})
it('marks a timed-out handoff as failed so the next attempt can retry', async () => {
vi.useFakeTimers()
const calls: { method: string; params?: Record<string, unknown> }[] = []
const requestGateway = vi.fn(async (method: string, params?: Record<string, unknown>) => {
calls.push({ method, params })
if (method === 'handoff.state') {
return { state: 'pending' } as never
}
return {} as never
})
let handle: HarnessHandle | null = null
render(<Harness onReady={h => (handle = h)} refreshSessions={async () => undefined} requestGateway={requestGateway} />)
const result = handle!.submitText('/handoff telegram')
await vi.advanceTimersByTimeAsync(61_000)
await result
expect(calls.some(call => call.method === 'handoff.request')).toBe(true)
expect(calls).toContainEqual({
method: 'handoff.fail',
params: {
error: expect.stringContaining('Timed out'),
session_id: RUNTIME_SESSION_ID
}
})
})
})
describe('usePromptActions submit / queue drain semantics', () => {
afterEach(() => {
cleanup()
@@ -323,6 +400,125 @@ describe('usePromptActions steerPrompt', () => {
})
})
describe('usePromptActions restoreToMessage', () => {
beforeEach(() => {
$busy.set(false)
$messages.set([
{ id: 'u1', role: 'user', parts: [textPart('first prompt')] },
{ id: 'a1', role: 'assistant', parts: [textPart('first answer')] },
{ id: 'u2', role: 'user', parts: [textPart('second prompt')] },
{ id: 'a2', role: 'assistant', parts: [textPart('second answer')] }
])
})
afterEach(() => {
cleanup()
$busy.set(false)
$messages.set([])
vi.restoreAllMocks()
})
it('rewinds to the target user turn and resubmits its text', async () => {
const requestGateway = vi.fn(async () => ({}) as never)
let lastState: Record<string, unknown> = {}
let handle: HarnessHandle | null = null
render(
<Harness
onReady={h => (handle = h)}
onSeedState={state => (lastState = state)}
refreshSessions={async () => undefined}
requestGateway={requestGateway}
seedMessages={$messages.get()}
/>
)
await handle!.restoreToMessage('u1')
// Ordinal 0 = "truncate before the first visible user message": the gateway
// drops that turn and everything after, then runs the same text again.
expect(requestGateway).toHaveBeenCalledWith('prompt.submit', {
session_id: RUNTIME_SESSION_ID,
text: 'first prompt',
truncate_before_user_ordinal: 0
})
expect((lastState.messages as { id: string }[]).map(m => m.id)).toEqual(['u1'])
expect(lastState.busy).toBe(true)
})
it('rethrows gateway failures and clears the busy flags for the dialog to surface', async () => {
const requestGateway = vi.fn(async () => {
throw new Error('gateway exploded')
})
let lastState: Record<string, unknown> = {}
let handle: HarnessHandle | null = null
render(
<Harness
onReady={h => (handle = h)}
onSeedState={state => (lastState = state)}
refreshSessions={async () => undefined}
requestGateway={requestGateway}
/>
)
await expect(handle!.restoreToMessage('u2')).rejects.toThrow('gateway exploded')
expect(lastState.busy).toBe(false)
})
it('interrupts the live turn and retries past "session busy" when reverting mid-stream', async () => {
$busy.set(true)
let submitAttempts = 0
const requestGateway = vi.fn(async (method: string) => {
if (method === 'prompt.submit') {
submitAttempts += 1
// The cooperative interrupt hasn't wound the turn down yet on the first
// try; the second attempt lands once the gateway reports idle.
if (submitAttempts === 1) {
throw new Error('session busy')
}
}
return {} as never
})
let handle: HarnessHandle | null = null
render(
<Harness
onReady={h => (handle = h)}
refreshSessions={async () => undefined}
requestGateway={requestGateway}
seedMessages={$messages.get()}
/>
)
await handle!.restoreToMessage('u1')
expect(requestGateway).toHaveBeenCalledWith('session.interrupt', { session_id: RUNTIME_SESSION_ID })
expect(submitAttempts).toBe(2)
expect(requestGateway).toHaveBeenCalledWith('prompt.submit', {
session_id: RUNTIME_SESSION_ID,
text: 'first prompt',
truncate_before_user_ordinal: 0
})
})
it('ignores non-user targets and unknown ids without touching the gateway', async () => {
const requestGateway = vi.fn(async () => ({}) as never)
let handle: HarnessHandle | null = null
render(<Harness onReady={h => (handle = h)} refreshSessions={async () => undefined} requestGateway={requestGateway} />)
await handle!.restoreToMessage('a1')
await handle!.restoreToMessage('missing')
expect(requestGateway).not.toHaveBeenCalled()
})
})
describe('usePromptActions file attachment sync', () => {
afterEach(() => {
cleanup()
@@ -562,6 +758,43 @@ describe('usePromptActions sleep/wake session recovery', () => {
expect(calls[2]?.params).toEqual({ session_id: RECOVERED_SESSION_ID, text: 'message after wake' })
})
it('resumes the stored session and retries once when session.interrupt reports "session not found"', async () => {
const calls: { method: string; params?: Record<string, unknown> }[] = []
let interruptAttempts = 0
const requestGateway = vi.fn(async (method: string, params?: Record<string, unknown>) => {
calls.push({ method, params })
if (method === 'session.interrupt') {
interruptAttempts += 1
if (interruptAttempts === 1) {
throw new Error('session not found')
}
return {} as never
}
if (method === 'session.resume') {
return { session_id: RECOVERED_SESSION_ID } as never
}
return {} as never
})
let handle: HarnessHandle | null = null
render(
<Harness
onReady={h => (handle = h)}
refreshSessions={async () => undefined}
requestGateway={requestGateway}
storedSessionId={STORED_SESSION_ID}
/>
)
await waitFor(() => expect(handle).not.toBeNull())
await handle!.cancelRun()
expect(calls.map(c => c.method)).toEqual(['session.interrupt', 'session.resume', 'session.interrupt'])
expect(calls[0]?.params).toEqual({ session_id: RUNTIME_SESSION_ID })
expect(calls[1]?.params).toEqual({ session_id: STORED_SESSION_ID })
expect(calls[2]?.params).toEqual({ session_id: RECOVERED_SESSION_ID })
})
it('surfaces the original error (no resume) when the failure is not "session not found"', async () => {
const calls: string[] = []
const states: Record<string, unknown>[] = []
@@ -751,4 +984,3 @@ describe('uploadComposerAttachment remote read failures', () => {
).rejects.toThrow('ENOENT: no such file')
})
})

File diff suppressed because it is too large Load Diff

Some files were not shown because too many files have changed in this diff Show More