🧠 GitHub Repository Intelligence System

An advanced, AI-powered intelligence system designed for understanding, analyzing, and visualizing large codebases seamlessly. Leveraging LangGraph, CodeBERT, and RAG (Retrieval-Augmented Generation), it acts as your personal AI architect and developer for any repository.

🎯 Target Audience: Who is this for?

• Software Engineers & Developers: Quickly onboard to new or large codebases without spending days reading code.
• System Architects: Automatically generate architecture and component diagrams for existing code.
• Tech Leads & Code Reviewers: Use AI-powered static analysis to catch bugs and analyze dependency graphs.
• DevOps & Security Teams: Understand flow paths and codebase structures for infrastructure planning or vulnerability scanning.

🚀 Key Features

• 💬 Natural Language Q&A: Ask questions in plain English about any part of your codebase.
• 📖 Deep Code Explanations: Get detailed, context-aware explanations of functions, classes, and complex logic.
• 🏗️ Automated Architecture Diagrams: Auto-generate component, class, and dependency flow diagrams.
• 🐛 AI-Powered Bug Detection: Run static analysis to detect potential issues or code smells.
• 🔍 Semantic Search: CodeBERT-powered embedding search that understands the meaning, not just exact keywords.
• 📊 Graph-Based Dependency Analysis: Reason over code relationships visually and logically.

🏛️ Architecture Workflow

Here's how the intelligence system extracts and reasons over your code:

graph TD
    A[GitHub Repository] -->|Clone & Parse| B(Code Extractor)
    B -->|Functions/Classes| C(Code Chunker)
    C -->|Semantic Chunks| D[CodeBERT Embedding]
    D -->|Vectors| E[(Qdrant Vector DB)]
    
    F[User Query] --> G{LangGraph Agent Workflow}
    G -->|Retrieve| E
    G -->|Graph Reasoning| H(NetworkX / Graphviz)
    G -->|Context + Query| I[LLM: Gemini / GPT-4]
    I --> J[Response & Explanations]
    H --> K[Architecture Diagrams]

Loading

🛠️ Tech Stack Used

• Agent Orchestration: LangGraph
• LLM & Reasoning: Langchain, Google GenAI / OpenAI GPT-4
• Code Embeddings & Parsing: CodeBERT, SentenceTransformers, Tree-sitter (for Python, JS, TS, Go, Rust)
• Vector Database: Qdrant, ChromaDB
• Graph & Diagrams: Graphviz, NetworkX, PlantUML
• Backend & APIs: FastAPI, Uvicorn
• Web UI: Streamlit
• Version Control Integration: PyGithub, GitPython

⚙️ How to Use This Repository

1. Clone & Setup Environment

git clone https://github.com/Hemkumar247/Repo_intelligence.git
cd Repo_intelligence

# Create a virtual environment
python -m venv venv
source venv/bin/activate  # On Windows: venv\Scripts\activate

# Install dependencies
pip install -r requirements.txt

2. Configure API Keys

Copy the example environment variables and add your tokens:

cp .env.example .env
# Edit .env and include:
# - GITHUB_TOKEN
# - OPENAI_API_KEY / GEMINI_API_KEY
# - QDRANT_URL (If using managed Qdrant)

3. Start Vector Database (Local Qdrant)

docker run -p 6333:6333 qdrant/qdrant

4. Run the Application

You can run the intelligence system through three different interfaces:

Interactive Web UI:

streamlit run src/ui/app.py

REST API (FastAPI):

python -m src.api
# Access the interactive Swagger docs at http://localhost:8000/docs

Command Line (CLI):

python -m src.pipeline

💡 Quick API Examples

from src.pipeline import RepoIntelligencePipeline

pipeline = RepoIntelligencePipeline()

# 1. Index a Repo
pipeline.index_repository("https://github.com/torvalds/linux")

# 2. Ask Code Questions
answer = pipeline.ask("linux", "How does the scheduler work?")

# 3. Generate Architecture Diagram
arch = pipeline.generate_architecture("linux")

📜 License

This project is licensed under the MIT License.