CodeMind AI

CodeMind AI is a local-first GitHub repository intelligence MVP. It clones a public repository, identifies semantic code sections, embeds them with sentence-transformers/all-MiniLM-L6-v2, persists them in FAISS, and answers questions with source citations.

Architecture

GitHub URL -> safe clone -> file scanner -> semantic parser/chunker
           -> embeddings -> persisted FAISS index -> retrieval
           -> grounded prompt -> Groq / OpenAI / Grok -> answer + citations

The backend uses small provider interfaces for embeddings and LLMs. FaissVectorStore, SentenceTransformerEmbeddings, and OpenAICompatibleProvider can be replaced later with Qdrant/Pinecone, another embedding model, or another model API without changing the API layer.

Run with Docker

1. Copy .env.example to .env and add GROQ_API_KEY, OPENAI_API_KEY, or XAI_API_KEY.
2. Run:

docker compose up --build

Open http://localhost:5173. The backend API and Swagger documentation are available at http://localhost:8000 and http://localhost:8000/docs.

An API key is optional. Without one, retrieval and citations still work, and CodeMind returns the most relevant code sections instead of a generated explanation.

To use Groq, set:

LLM_PROVIDER=groq
GROQ_API_KEY=your-groq-api-key
GROQ_MODEL=llama-3.3-70b-versatile

Run locally

Backend:

python -m venv .venv
.venv/Scripts/activate
pip install -r backend/requirements.txt
uvicorn backend.app:app --reload

Frontend:

cd frontend
npm install
npm run dev

The embedding model downloads on the first indexing request. Repository clones, indexes, model cache, and state are stored under backend/repositories, backend/indexes, and backend/data.

API

• POST /clone with {"repo_url":"https://github.com/owner/repository.git"}
• POST /index with {"repo_id":"..."}; omitting repo_id uses the active repository
• POST /ask with {"repo_id":"...","question":"Where is JWT implemented?","top_k":5}
• GET /health

Tests

pip install pytest
pytest backend/tests

The MVP deliberately leaves hybrid search, reranking, dependency graphs, visualizations, private repositories, and multi-repository retrieval as extension points rather than partially implementing them.

Production Deployment (AWS Elastic Beanstalk)

CodeMind AI is deployed live on AWS Elastic Beanstalk (ap-south-1 region) as a single container application using Docker:

• Live Web App: http://codemind-prod.ap-south-1.elasticbeanstalk.com/
• Health Check: http://codemind-prod.ap-south-1.elasticbeanstalk.com/health

See DEPLOYMENT.md for full step-by-step instructions, infrastructure architecture, environment variable configuration, and maintenance commands.