Add deterministic eval suite (+ README/manifest refresh)#4
Closed
drakulavich wants to merge 16 commits into
Closed
Add deterministic eval suite (+ README/manifest refresh)#4drakulavich wants to merge 16 commits into
drakulavich wants to merge 16 commits into
Conversation
Deterministic, keyless benchmark of diagram-type selection, abstention, and heuristic Mermaid validity over a hybrid synthetic+real PR corpus. Publishes to BENCHMARKS.md + README badge. Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
…k keywords Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Centered logo/title, Tests/npm/MIT/Bun badges, bold-lead feature bullets, and a footer flourish. Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Mermaid rejects identifiers starting with a digit, so a migration named 003_orders.sql produced an unparseable `003_ORDERS` entity. Strip leading digits and separators before uppercasing. Caught by the new eval suite.
Owner
Author
|
Closing in favor of a clean skill-first reorientation: dropping the GitHub Action (and the eval suite that tested it), refocusing Iago as a Claude Code / Codex skill. New PR to follow. |
4 tasks
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.
Summary
Two strands of work, bundled:
1. Eval suite (
evals/) — a deterministic, offline, keyless benchmark for Iago's diagram pipeline, plus published results.classify_diagram_type/heuristic_diagramfromaction/scripts/run.py— benchmarks shipping code, no reimplementation.mermaid.parse(), no headless browser).python evals/run.pyregenerates marker-delimited result blocks inBENCHMARKS.md+README.mdand writesbadge.json/results.json;--checkgates CI on thresholds (accuracy ≥ baseline, validity 100%) and published-artifact drift.evals/capture.pysnapshots real public PRs with provenance as the corpus grows.evalsjob in.github/workflows/test.yml.BENCHMARKS.mdis explicit about what is not measured yet (LLM-output faithfulness / LLM-Mermaid validity — spot-checked only, LLM-judge eval on the roadmap).Caught a real bug:
fix(action)— the heuristic ER generator emitted003_ORDERSfor a migration named003_orders.sql; Mermaid rejects identifiers starting with a digit. Fixed by stripping numeric migration prefixes; covered by a regression test.2. Docs/manifest refresh
plugin.json+marketplace.json; addedgreptile/squawkkeywords.docs/superpowers/.Test Plan
python -m pytest evals/tests→ 23 passedpython evals/run.py --check→ exit 0 (100% accuracy, 4/4 validity)--checkfail (exit 1), restoring clears itevalsjob green on this PR