RepoChat

A full-stack application that lets you index any public GitHub repository and chat with its codebase using AI. Ask natural language questions and get answers grounded in the actual source code.

Features

• Repository Ingestion — Clone and index any public GitHub repo into a vector store
• Semantic + Keyword Search — Hybrid retrieval using ChromaDB for accurate context
• Streaming Chat — Real-time streamed responses powered by Groq (Llama 3.3 70B)
• Chat History — Persistent conversation history per repo session
• Authentication — JWT-based user accounts with login/register
• Previously Ingested Repos — Quick access to repos you've already indexed

Tech Stack

Frontend

• SvelteKit 5 with TypeScript
• Tailwind CSS + shadcn-svelte components
• marked for markdown rendering in chat

Backend

• Django 6 + Django REST Framework
• Celery + Redis for async ingestion tasks
• ChromaDB for vector storage
• LangChain + Groq (Llama 3.3 70B) for LLM streaming
• PostgreSQL for persistence
• JWT authentication via djangorestframework-simplejwt

Architecture

User → SvelteKit Frontend
         ↓
   Django REST API
    ├── Ingestion (POST /api/ingest/)
    │     └── Celery Task → Clone repo → Chunk files → Embed into ChromaDB
    ├── Chat (GET /api/chat/)
    │     └── Hybrid search ChromaDB → Stream LLM response via SSE
    ├── Auth (/api/auth/register/, /api/auth/token/, /api/auth/refresh/)
    └── Repos (GET /api/repos/)

Getting Started

Prerequisites

• Docker & Docker Compose
• A Groq API key

Running with Docker Compose

1. Clone the repository:

   git clone https://github.com/shv-ng/repochat.git
   cd repochat

2. Create a .env file in the project root (or set environment variables):

   GROQ_API_KEY=your_groq_api_key_here

3. Start all services:

   docker compose up --build

4. Open http://localhost:3000 in your browser.

Running Locally (without Docker)

Backend

cd backend
# Install dependencies (requires uv)
uv sync

# Set environment variables
export GROQ_API_KEY=your_key
export DB_NAME=postgres
export DB_USER=postgres
export DB_PASSWORD=postgres
export DB_HOST=localhost
export REDIS_URL=redis://localhost:6379/0

# Run migrations
uv run python manage.py migrate

# Start Django
uv run python manage.py runserver

# Start Celery worker (separate terminal)
uv run celery -A config worker -l info

Frontend

cd frontend
cp .env.example .env   # set PUBLIC_BASE_URL=http://localhost:8000
npm install
npm run dev

Usage

1. Register / Login using the button in the top-right corner (optional — anonymous use is supported for chat, but repo history requires an account).
2. Paste a GitHub URL (e.g. https://github.com/owner/repo) and click Ingest Repository.
3. Wait for indexing to complete — progress is shown in real time.
4. Ask questions about the codebase in the chat interface.
5. Previously ingested repos appear on the home screen for quick access.

API Endpoints

Method	Endpoint	Description
POST	/api/auth/register/	Create a new account
POST	/api/auth/token/	Obtain JWT tokens
POST	/api/auth/refresh/	Refresh access token
POST	/api/ingest/	Start repo ingestion
GET	/api/ingest/?job_id=<id>	Stream ingestion status (SSE)
GET	/api/chat/?repo_url=&query=&session_id=	Stream chat response (SSE)
GET	/api/chat/history/?session_id=<id>	Fetch chat history
GET	/api/repos/	List user's ingested repos

Environment Variables

Variable	Default	Description
GROQ_API_KEY	—	Required. Your Groq API key
DB_NAME	postgres	PostgreSQL database name
DB_USER	postgres	PostgreSQL user
DB_PASSWORD	postgres	PostgreSQL password
DB_HOST	localhost	PostgreSQL host
DB_PORT	5432	PostgreSQL port
REDIS_URL	redis://localhost:6379/0	Redis URL for Celery
PUBLIC_BASE_URL	—	Frontend: backend base URL

Project Structure

repochat/
├── backend/
│   ├── apps/
│   │   ├── accounts/     # User registration & JWT auth
│   │   ├── chat/         # Chat sessions & message history
│   │   ├── ingestion/    # Ingestion jobs & SSE status
│   │   └── repos/        # Repo registry
│   ├── services/
│   │   ├── github/       # Git clone & file parsing/chunking
│   │   ├── llm/          # LangChain + Groq streaming
│   │   └── vectorstore/  # ChromaDB embed & search
│   ├── tasks/
│   │   └── ingest.py     # Celery ingestion task
│   └── config/           # Django settings, URLs, Celery
└── frontend/
    └── src/
        ├── routes/       # SvelteKit pages (+page.svelte, chat/)
        └── lib/          # API client, stores, UI components

Notes

• Only public GitHub repositories are supported.
• Large repositories may take a minute or two to index.
• The LLM is instructed to answer strictly from the indexed code context and will say so if an answer isn't found.
• File types with language-aware chunking: Python, Go, Java, Rust, C#, Ruby, Kotlin, PHP, Scala, Swift, and more. JSON, Markdown, and HTML have dedicated splitters.