experiment-iterative-coder

Automate iterative code refinement by guiding autonomous agents through repeatable plan, code, evaluate, and refine cycles to catch issues early. This approach applies phase-based decomposition to handle complex multi-file tasks, improving overall code quality and production readiness.

Self-evaluation works by applying linting with ruff and unit testing with pytest during each cycle to validate correctness. The structured evaluation loop assesses code against requirements, triggering further refinement iterations if quality thresholds are not met.

Structure phase planning by decomposing complex tasks into distinct phases with configurable iteration limits. You can execute up to three iterations per phase and set a global limit of ten total iterations to ensure scalable and controlled code generation.

Yes, iterative code generation integrates directly with pytest and ruff for validation. It uses ruff for linting to enforce code style and pytest for unit testing to verify correctness, applying both tools within its structured evaluation loop.

The iteration limits for autonomous code refinement cycles allow up to three iterations per phase and a maximum of ten total iterations. These configurable limits prevent infinite loops while ensuring sufficient attempts for code generation and self-evaluation.