I built the AI Code Reviewer & Bug Fixing Agent repository on GitHub (https://github.com/Nazakath-Kanz/ai-code-reviewer-agent). It reviews Python code, identifies bugs using both static analysis and LLM reasoning, proposes fixes, and verifies those fixes by actually executing them in a sandbox — retrying automatically if a fix doesn't work.
The system uses a LangGraph state machine with conditional retry routing.
I chose Groq API (openai/gpt-oss-120b) as the LLM provider.
Verification runs the corrected code in a Python subprocess with isolated temporary directories and hard timeouts.
Demo frontend URL: https://auto-code-fix.lovable.app
Backend Swagger URL: https://ai-code-reviewer-agent.onrender.com/docs
Render free‑tier cold start delay: 30‑60 seconds for the first request after idle.