fix(learning): preserve research feedback and measured evidence - #190
Merged
Conversation
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
Knowledge candidate callbacks could not use earlier diagnosis or update results because development and final measurement ran separate lifecycles.
Research also stopped when storage requirements passed, even when its driver had unfinished work.
This change carries one lifecycle through diagnosis, acquisition, update, frozen final measurement, and the promotion decision.
It deletes the post-hoc lifecycle merge and keeps final data outside adaptive updates.
It rejects disabled required phases before candidate work.
The research loop also requires an optional driver completion result and permits steering with zero storage gaps.
Default evaluator version 2 omits unmeasured answer, promotion, and blocking-readiness dimensions.
It averages measured dimensions equally and labels structural-only results.
A candidate-ready result remains detached from the live knowledge base.
Release 14.0.0 records the exported ResearchDriver interface change and the evaluator semantics.
The release uses agent-eval 0.174.x for its corrected complete-method result contract.
Validation before submission used Node 24.11.1 and the published Eval 0.174.0 package.
Source and contract typechecks passed; all 249 source and test files passed lint.
The five focused test files passed all 85 cases.
Build and packed-package checks passed, including clean install, declarations, all imports, CLI version, and re-pack.
The API record contains 995 exports across six entrypoints; only ResearchDriver changes shape.
The built artifact completed the expected two-round research flow, retained steering with zero storage gaps, and rejected unsupported candidate execution on macOS.
Ubuntu 24.04 CI passed 862 tests with seven conditional skips across 90 files, including exact Linux candidate snapshots.
A separate packed-package job passed both official GEPA and SkillOpt integrations.
Both jobs are terminal success: https://github.com/tangle-network/agent-knowledge/actions/runs/34013062769