Turn any codebase into beautiful architecture diagrams — one command, no AI, no server, runs on any laptop. Zero dependencies, pure Python, agent-friendly (Cursor, Claude Code, Codex, OpenCode).
repochat-ai
Prototype for querying and understanding GitHub repositories using local LLMs
- Stars
- 1
- Forks
- 0
- Created
- Mar 13, 2026
- Last push
- Mar 13, 2026
- Default branch
- main
- Contributors
- 0
README
dsk-dev-ai/repochat-aiRepoChat AI
Chat with any GitHub repository using local AI.
Understand codebases instantly using natural language.
Built with: Python • LangChain • Ollama • ChromaDB
⭐ Star this repository if it helps you!
Demo

In this example, we run RepoChat on the Next.js repository and ask about routing:
$ repochat chat https://github.com/vercel/next.js
Cloning repository...
Loaded 80 files
Split into 600 chunks
Vector database created.
RepoChat is ready! Ask a question (type 'exit' to quit):
>>> How does routing work in Next.js?
Next.js uses a file-system based router. ... (AI-generated answer) ...
Features
- Interactive CLI chat with any GitHub repo
- Uses local open-source LLMs via Ollama
- Splits large codebases into searchable chunks
- Generates responses by retrieving context from code
Installation
pip install repochat-ai
(Requires Python 3.8+ and Ollama installed and running locally.)
Usage
repochat chat https://github.com/username/repo.git
Then ask natural language questions about the code.
Development
To run locally from source:
git clone https://github.com/USERNAME/repochat-ai.git
cd repochat-ai
pip install -e . # install dependencies
repochat chat https://github.com/vercel/next.js
Contributing
Contributions welcome! Please open issues or pull requests. This project follows standard Python packaging guidelines.
License
MIT License. See LICENSE for details.
Creating a Demo GIF
A demo GIF significantly increases interest. Steps:
Install Tools:
- On Windows: ScreenToGif or OBS Studio.
- On macOS: Kap or QuickTime + ffmpeg.
- On Linux:
gtk-recordmydesktoporPeek.
Record Steps:
- Show installing (if fresh).
- Run
repochat chat <example-repo>. - Type a sample question.
- Copy the answer from terminal.
Edit GIF:
- Crop to focus on terminal.
- Add slight delays between lines.
- Keep it short (<15s).
Save: Put the GIF at
docs/demo.gifin your repo.
Adding docs/demo.gif in README makes the project visually appealing, attracting more stars.
Git Commands (Push to GitHub)
Once everything is ready, you can push to GitHub:
git init
git add .
git commit -m "Initial RepoChat AI release with full code"
git branch -M main
git remote add origin https://github.com/YOUR_USERNAME/repochat-ai.git
git push -u origin main
Replace YOUR_USERNAME with your GitHub username. This creates a new repository and pushes your code. For future changes, just git add, commit, and push.
Related projects
Free, private image-to-text (OCR). Web UI, REST API & CLI in one repo — FastAPI + RapidOCR (ONNX). No API keys, nothing stored.
Open-source AI platform for genomics & bioinformatics — 18 free public genome research tools
177 merged pull requests across 19 repositories · data generated Oct 7, 2026