Repository navigation
Move eval harness code out of src; drop dev-only and product-specific tool surfaces - #1111
Merged
Merged
Conversation
…uct prose Move the eyewear fidelity gates and judge to eval/oracle. Remove the review_paint_peek_latest tool and the trace_from_image part and BOM detection. Neutralise product-specific skill prose and the CADGenBench comment in healStep.
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
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
Part of trimming overfit tooling:
src/should be the generic product toolset.src/: the eyewear fidelity gates and judge (src/lib/imageSimilarity/{fidelityGates,judge}.ts+ tests) move toeval/oracle/; nothing insrc/imported them.score.tsstays (used by the vision router).review_paint_peek_latestremoved from the agent toolset: it scanned hard-coded local checkout paths. kernelCAD-server defines its own hosted implementation of this tool; the local hook and/review-paintcommand don't depend on it.trace_from_imageis a pure curve tracer:detectPartsand thedetectedParts/draftBom/unmatchedLabels/partDetectionFrameoutputs and thetool.trace-from-image.part-detection-faileddiagnostic are gone, with their demo and fixture.runClosedLoop,buildRepairPrompt) stays exported: kernelCAD-server's orchestrator imports it.Merged with develop after #1108 and #1110 (registry is now 58 tools; surviving tools keep their relative order).
Local verification (Node 22): typecheck, qc:lint (0 errors), skills drift clean after
cookbook:build, 276 targeted tests (contract, registry schema, distGenerate, vision, diagnostics, tool-name consistency). cookbook:evaluate 94/96 with the two known load timeouts, which pass alone. Full suite runs in CI.