Glevel 39019e2574 feat: analyze teacher comments before rework (defend solution) ...

Add analyze_comments() that calls LLM to evaluate each teacher comment:
- valid criticism -> fixes list (LLM will fix these)
- incorrect/misunderstood -> defenses list (LLM adds NOTE: comments in code)
Verdict: needs_fixes | already_correct | mixed

solve() now prints analysis verdict and counts before generation.
Commit message on rework includes verdict + fix/defense counts.

Co-Authored-By: Claude Sonnet 4.5 <noreply@anthropic.com>

2026-06-04 19:21:39 +03:00

26 KiB

Raw Blame History

View Raw