snapA2A

snapA2A is a multi-agent coding pipeline where several LLM agents collaborate on a programming task. A Writer generates code, a Reviewer looks for bugs and edge cases, an Optimizer produces the improved version, and a Judge scores the result for correctness, readability, and efficiency.

The core behavior is the retry loop. If the Judge scores the output below the acceptance threshold, the LangGraph pipeline sends the accumulated feedback back to the Writer for another iteration. Each run produces both final Python code and a report containing the task, feedback, scores, and iteration history.