GitSurfer

GitSurfer is an intelligent, multi-provider codebase analysis and research assistant for GitHub repositories. It leverages advanced LLMs (Gemini, OpenAI, Anthropic, Cohere) and vector databases to dynamically fetch, summarize, embed, and answer questions about any public GitHub repository, providing deep insights and research capabilities for developers and researchers.

---

Features

• Fetches and analyzes GitHub repositories (tree structure, file contents)
• Summarizes repository structure using LLMs
• Embeds code and documentation into a vector database (ChromaDB)
• Supports multiple LLM and embedding providers: Gemini, OpenAI, Anthropic, Cohere
• Interactive research assistant: Ask questions about the codebase and get detailed, contextual answers
• Extensible modular architecture using LangGraph and LangChain
• Rich logging and error handling

---

Graphs

Embedder sub-Graph

Fetcher sub-Graph

GitSurfer main Graph

---

Project Structure

GitSurfer/
├── app/
│   ├── core/           # Core utilities, LLM/embedding logic
│   ├── graphs/         # Main assistant, fetcher, embedder, researcher graphs
│   ├── retriever/      # Data ingestion and retriever logic
├── config/             # Settings and environment variable loader
├── DATA/               # Persisted vector DBs
├── temp/               # Temporary files (chunks, summaries)
├── logs/               # Log files
├── logger.py           # Logging configuration
├── requirements.txt    # Python dependencies
├── .env                # Environment variables (not committed)

---

Installation

1. Clone the repository

   git clone <your-fork-or-repo-url>
   cd GitSurfer

2. Install dependencies

   pip install -r requirements.txt

3. Set up environment variables

   • Copy .env.example to .env and fill in your API keys:
     • GOOGLE_API_KEY (for Gemini)
     • OPENAI_API_KEY (for OpenAI)
     • ANTHROPIC_API_KEY (for Anthropic)
     • COHERE_API_KEY (for Cohere)
     • GITHUB_TOKEN (for increased GitHub API limits)
   • You can also specify model names and other settings in .env.

---

Usage

The main entry point is the app/graphs/git_assistant.py script. It runs an interactive CLI assistant:

python app/graphs/git_assistant.py

Workflow:

1. Enter a GitHub repository URL when prompted.
2. GitSurfer fetches the repo, summarizes its structure, and creates a vector DB.
3. Ask any question about the codebase (design, functions, usage, etc.).
4. Interactively continue the research session or exit.

Example:

🔄 Processing repository...
👤 Input required: Enter GitHub repo URL
🤖 Assistant: Repository fetched and analyzed. Ask your question!
👤 You: What does the main.py file do?
🤖 Assistant: [detailed answer]

---

Configuration

• All settings (provider selection, model names, directories) are managed in config/settings.py and via environment variables.
• Supports switching between providers for both LLM and embeddings.
• Vector DBs are persisted under DATA/.

---

Prerequisites

• Python 3.9+
• API keys for at least one supported LLM/embedding provider
• (Optional) GitHub Personal Access Token for higher API rate limits

---

Environment Variables

Variable	Description
GOOGLE_API_KEY	Gemini API key
OPENAI_API_KEY	OpenAI API key
ANTHROPIC_API_KEY	Anthropic API key
COHERE_API_KEY	Cohere API key
GITHUB_TOKEN	GitHub token for API calls
GEMINI_LLM_MODEL	Gemini model name (default set)
OPENAI_LLM_MODEL	OpenAI model name (default set)
...	See config/settings.py for all

---

Testing

Run tests using:

pytest

---

Credits

• Author: Lalan Kumar (kumar8074)
• Built with LangChain, LangGraph, and ChromaDB

---

License

MIT License