mirror of
https://github.com/NousResearch/hermes-agent.git
synced 2026-04-29 07:21:37 +08:00
1e6285c53d
Ships a complete offline eval harness at scripts/compression_eval/. Runs a real conversation fixture through ContextCompressor.compress(), asks the compressor model to answer probe questions from the compressed state, then has a judge model score each answer 0-5 on six dimensions (accuracy, context_awareness, artifact_trail, completeness, continuity, instruction_following). Methodology adapted from Factory's Dec 2025 write-up (https://factory.ai/news/evaluating-compression); the scoreboard framing is not adopted. Motivation: we edit context_compressor.py prompts and _template_sections by hand and ship with no automated check that compression still preserves file paths, error codes, or the active task. Until now there has been no signal between 'test suite green' and 'a user hits a bad summary in production.' What's shipped - DESIGN.md — full architecture, fixture/probe format, scrubber pipeline, grading rubric, open follow-ups - README.md — usage, cost expectations, when to run it - scrub_fixtures.py — reproducible pipeline that converts real sessions from ~/.hermes/sessions/*.jsonl into public-safe JSON fixtures. Applies agent.redact.redact_sensitive_text + username path normalisation + personal handle scrubbing + email/git-author normalisation + reasoning scratchpad stripping + platform-mention scrubbing + first-user paraphrase + system-prompt placeholder + orphan-message pruning + 2KB tool-output truncation - fixtures/ — three scrubbed session snapshots covering three session shapes: feature-impl-context-priority (75 msgs / ~17k tokens) debug-session-feishu-id-model (59 msgs / ~13k tokens) config-build-competitive-scouts (61 msgs / ~23k tokens) - probes/ — three probe banks (10-11 probes each) covering all four types (recall/artifact/continuation/decision) with expected_facts anchors (PR numbers, file paths, error codes, commands) - rubric.py — six-dimension grading rubric, judge-prompt builder, JSON-with-fallback response parser - compressor_driver.py — thin wrapper around ContextCompressor for forced single-shot compression (fixtures are below the default 100k threshold so we force compress() to attribute score deltas to prompt changes, not threshold-fire variance) - grader.py — two-phase continuation + grading calls via the OpenAI SDK directly against the resolved provider endpoint - report.py — markdown report renderer (paste-ready for PR bodies), --compare-to delta mode, per-run JSON dumper - run_eval.py — fire-style CLI (--fixtures, --runs, --judge-model, --compressor-model, --label, --focus-topic, --compare-to, --verbose) - tests/scripts/test_compression_eval.py — 33 hermetic unit tests covering rubric parsing edge cases, judge-prompt building, report rendering, summariser medians, per-run JSON roundtrip, fixture and probe loading, and a PII smoke check on the checked-in fixtures Non-LLM paths are covered by the 33-test suite that runs in CI. The LLM paths (continuation + grading) require credentials and real API calls, so they're exercised by running the eval itself — not by CI. Validation - 33/33 unit tests pass in 0.33s via scripts/run_tests.sh - 50/50 adjacent tests (tests/agent/test_context_compressor.py) still pass — no regression introduced - End-to-end dry run against debug-session-feishu-id-model with openai/gpt-5.4-mini via Nous Portal: Compression: 13081 -> 3055 tokens (76.6% ratio), 59 -> 10 messages Overall score: 3.25 (artifact_trail 1.50 is the weak spot, matching Factory's published observation) Specific probe misses surfaced with concrete judge notes Noise floor (one empirical data point) Same inputs re-run: overall 3.25 -> 3.17 (delta -0.08). Individual dimensions varied up to ±0.5 between two single-run medians. Confirms the DESIGN.md < 0.3 noise guidance is the right order of magnitude for single-run comparisons. Tighter noise measurement (N=10) is tracked as an open follow-up in DESIGN.md. Why scripts/ and not tests/ Requires API credentials, costs ~$0.50-1.50 per run, minutes to execute, LLM-graded (non-deterministic). Incompatible with scripts/run_tests.sh which is hermetic, parallel, credential-free. scripts/sample_and_compress.py is the existing precedent for offline credentialed tooling. Open follow-ups (tracked in DESIGN.md, not blocking this PR) 1. Iterative-merge fixture (two chained compressions on one session) 2. Precise noise-floor measurement at N=10 3. Scripted scrubber helpers to lower the cost of fixture #4+ 4. Judge model selection policy (pin vs. per-user)
75 lines
3.3 KiB
JSON
75 lines
3.3 KiB
JSON
{
|
|
"fixture": "feature-impl-context-priority",
|
|
"description": "Probes for the .hermes.md / AGENTS.md / CLAUDE.md / .cursorrules priority feature session. Anchors are the concrete facts the next assistant would need to continue: user's priority order, files modified, helper-function structure, live-test scenarios, and PR number.",
|
|
"probes": [
|
|
{
|
|
"id": "recall-priority-order",
|
|
"type": "recall",
|
|
"question": "What is the priority order the user asked for when multiple project-context files are present? List them from highest to lowest priority.",
|
|
"expected_facts": [".hermes.md", "AGENTS.md", "CLAUDE.md", ".cursorrules", "highest to lowest"]
|
|
},
|
|
{
|
|
"id": "recall-selection-mode",
|
|
"type": "recall",
|
|
"question": "When multiple context files exist in the same directory, does the agent now load all of them or pick only one?",
|
|
"expected_facts": ["only one", "priority-based selection", "highest-priority winner"]
|
|
},
|
|
{
|
|
"id": "artifact-files-modified",
|
|
"type": "artifact",
|
|
"question": "Which files in the hermes-agent repository were modified during this session? List them.",
|
|
"expected_facts": [
|
|
"agent/prompt_builder.py",
|
|
"tests/agent/test_prompt_builder.py"
|
|
]
|
|
},
|
|
{
|
|
"id": "artifact-helper-functions",
|
|
"type": "artifact",
|
|
"question": "The session introduced separate helper functions for each context-file type. What are their names?",
|
|
"expected_facts": [
|
|
"_load_hermes_md",
|
|
"_load_agents_md",
|
|
"_load_claude_md",
|
|
"_load_cursorrules"
|
|
]
|
|
},
|
|
{
|
|
"id": "artifact-test-scenarios",
|
|
"type": "artifact",
|
|
"question": "A scratch directory was created with scenario subdirectories to live-test the priority chain. Roughly how many scenarios, and what directory was it created under?",
|
|
"expected_facts": ["10 scenarios", "/tmp/context-priority-test"]
|
|
},
|
|
{
|
|
"id": "decision-claude-md-was-unsupported",
|
|
"type": "decision",
|
|
"question": "What was the finding about CLAUDE.md support in the existing loader before this session's changes?",
|
|
"expected_facts": ["CLAUDE.md was not handled", "not supported", "new handler added"]
|
|
},
|
|
{
|
|
"id": "decision-load-all-or-one",
|
|
"type": "decision",
|
|
"question": "Was the decision to load multiple context files when present, or to load only the highest-priority one? Explain the reasoning in one sentence.",
|
|
"expected_facts": ["load only one", "highest priority", "user preference", "do not want to load multiple"]
|
|
},
|
|
{
|
|
"id": "continuation-pr-number-and-status",
|
|
"type": "continuation",
|
|
"question": "A pull request was opened for this feature. What is the PR number and what is its merge status?",
|
|
"expected_facts": ["PR #2301", "merged", "squash"]
|
|
},
|
|
{
|
|
"id": "continuation-test-suite-result",
|
|
"type": "continuation",
|
|
"question": "What was the result of the full test suite run after the implementation changes?",
|
|
"expected_facts": ["5680 passed", "0 failures", "clean"]
|
|
},
|
|
{
|
|
"id": "continuation-next-step",
|
|
"type": "continuation",
|
|
"question": "If asked to pick up this session, what is the current state of main? Anything left to do?",
|
|
"expected_facts": ["merged to main", "main is current", "nothing outstanding", "pulled"]
|
|
}
|
|
]
|
|
}
|