Code Review Agent

CodeMind is a hackathon-built AI code review agent that goes beyond linting by using repository history, past reviews, bug reports, and team patterns.

The system combines retrieval-augmented memory with LLM analysis to deliver context-aware feedback instead of generic static comments.

Built with Streamlit, LiteLLM, Groq, Qwen3-32B, and a vector-backed memory layer, it aims to reflect how a specific team actually works.

The team hit an async context manager timeout when memory retrieval ran alongside LLM inference, then refactored the architecture to fix it.

The project shows how code review tools can become historically informed assistants that improve consistency and catch recurring issues better.