feat: LLM-as-judge scorer, dataset auto-sampling, eval --ci baseline#78
Merged
Merged
Conversation
…65 #66 #69) - Add score_llm_judge() scorer: calls any OpenAI-compatible endpoint, parses JSON {score, reason}, clamps to [0,1], strips markdown fences - Dispatch llm_judge through run_scorer() alongside existing scorers - Add auto_populate() to dataset: 6 signal filters (has-errors, high-retry, cost-above, wide-blast, long-duration, low-eval-score), since_days window, dedup, optional label - Extend cmd_eval_ci() with --baseline, --save-baseline, --tolerance, --github-summary flags; _load_baseline/_save_baseline/_write_github_summary - GitHub summary writes PR-comment-ready Markdown with delta vs baseline - 25 new tests covering all three features (700 total, all passing) Co-authored-by: Ona <no-reply@ona.com>
This was referenced May 17, 2026
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Closes #65, #66, #69
Changes
LLM-as-judge scorer (
score_llm_judge)base_url+api_key+model){"score": float, "reason": str}from the response, strips markdown fences, clamps score to[0, 1]run_scorer("llm_judge", {...}, events)alongside existing scorersDataset auto-sampling (
eval dataset auto)auto_populate(store, path, filter, since_days, label)scans recent sessions and adds matching ones to a.jsonldatasethas-errors,high-retry,cost-above:<usd>,wide-blast,long-duration:<Ns>,low-eval-score:<threshold>agent-strace eval dataset auto --filter has-errors --since-days 7eval --ci baseline comparison
--save-baseline <path>: saves current scorer scores as a JSON baseline--baseline <path>: loads baseline and checks for regressions--tolerance <float>: allowed score drop before flagging regression (default 0.0)--github-summary: writes.agent-traces/eval-summary.mdwith a PR-comment-ready Markdown table showing score, baseline, delta, and pass/fail per scorerTests
25 new tests in
tests/test_eval_extensions.pycovering all three features. Full suite: 700 tests, all passing.