Cache identical warm evaluator responses by input revision

Objective

Cache identical warm evaluator responses by input revision.

Notes: Cached up to 128 deterministic evaluator payloads by canonical configuration plus registry/DB revision; deep copies isolate callers. Focused tests prove reuse, mutation isolation, and DB invalidation. Internal-only FEATURE_COMPLETION exception documented. Live t302 latency samples remain external.

Approach

[reconstructed — uncertain] No dedicated findings Markdown located for t307. Summary reconstructed from research/findings/task-history.json entry and git history. Uncertain fields: detailed approach, exact file list, and metrics beyond the task note.

Files / code / data changed

• No commit SHA confidently associated with t307 in sampled log [reconstructed — uncertain].
• Task-history entry: research/findings/task-history.json#t307

Results

Task note: Cached up to 128 deterministic evaluator payloads by canonical configuration plus registry/DB revision; deep copies isolate callers. Focused tests prove reuse, mutation isolation, and DB invalidation. Internal-only FEATURE_COMPLETION exception documented. Live t302 latency samples remain external. [reconstructed — uncertain] No quantitative tables recovered for this entry; see git diff and task-history for grounds.

Conclusions

[reconstructed — uncertain] Outcome inferred from status done and task note.

Problems / follow-ups

• See source report or PR discussion for follow-ups. No unsupported follow-ups fabricated.

Links

• Task-history: research/findings/task-history.json id t307