← Back to all projects
🧠

My product Open source

RepoMind

Ask any repo — get answers with file & line citations

An assistant that answers questions about any GitHub repository in plain English and cites the exact file and line it used.

AST chunking9 LLM providersfile:line citations

Built for: Developers joining a new codebase, reviewers and maintainers.

↓ Scroll to explore the problem, architecture and stack

😩

The problem

Getting up to speed on an unfamiliar codebase means hours of searching, and generic chatbots answer confidently without showing where the answer came from.

💡

Why it matters

Every answer points to real code, so you can verify it in seconds — and onboarding onto a new repo takes minutes instead of days.

What you can do with it

01Point it at a GitHub repository and ask questions in plain English.
02Get answers that cite the exact file and line numbers.
03Search by whole functions and classes, not random text slices.
04Combine indexed code with live GitHub API lookups.
05Run with cloud models or locally with Ollama.

How it’s built

A simplified view of the architecture — follow the dots as a request moves down the layers.

You
React UIask a question
API
FastAPI
LangGraph workflowplan → retrieve → answer
Knowledge
QdrantAST code chunks
GitHub APIlive lookups
Models
LLM router9 providers · Ollama
LangSmithtracing

A request, step by step

  1. The repository is parsed into functions and classes (AST chunks) and embedded into Qdrant.
  2. Your question enters a LangGraph workflow that decides what to look up.
  3. It retrieves matching code from Qdrant and, when needed, fresh data from the GitHub API.
  4. An LLM writes the answer and cites the exact file and line ranges it used.

Tech stack

Backend
PythonFastAPI
AI
LangGraph📚RAGOllamaLangSmith
Data
QdrantGitHub API
Frontend
React

What I built

  • A LangGraph workflow behind a FastAPI API that combines vector search with live GitHub API lookups.
  • AST-based code chunking, so search works on whole functions and classes.
  • Qdrant vector search over the indexed code.
  • LLM fallback across 9 providers, with local Ollama support.
  • A React interface and LangSmith tracing for debugging answers.