Repomind
RepoMind is a context-aware AI code review system built on a multi-agent architecture. Unlike traditional automated reviewers that operate only on the diff, RepoMind first indexes the entire repository using AST parsing (tree-sitter), vector embeddings, and a dependency graph, then retrieves semantically relevant context before reviewing each Pull Request. The review pipeline Indexer, Contextualizer, Reviewer, Critic, and Reporter is orchestrated with LangGraph and streams results live to the frontend. An interactive 3D knowledge graph (Three.js) visualizes the repository architecture and highlights exactly which files the RAG retrieved during each review, making the AI reasoning transparent and auditable. Built with privacy first: runs fully local via Ollama, so source code never leaves the developer's machine. Switching LLM providers (Ollama, AMD Developer Cloud, Groq) requires zero code changes just editing the .env file. Stack: FastAPI, React, LangGraph, ChromaDB, tree-sitter, NetworkX, Three.js.
Category tags:.png&w=96&q=75){kind=small}
For Innovators & Creators