CodeVector

CodeVector is a full-stack app for asking questions about a codebase using RAG (retrieval augmented generation). You can either upload a local .zip repo or import a GitHub repo URL, then chat with the app about the code.

Features

• Upload a local repository as a .zip
• Import a repository from a GitHub URL
• Chunk and embed source files with OpenAI embeddings
• Store vectors in Pinecone (namespaced by job ID)
• Ask questions and get grounded answers from retrieved code context
• Save query history to Supabase (for uploaded zip jobs)

Tech Stack

• Frontend: React 19, TypeScript, Vite, React Router
• Backend: Node.js, Express
• AI: OpenAI (text-embedding-3-small, gpt-4o-mini by default)
• Vector DB: Pinecone
• Data logging: Supabase

Repo Structure

CodeVector/
├── client/                # React + Vite frontend
├── server/                # Express API + RAG pipeline
├── README.md
├── package.json           # Placeholder/demo package file (root)
└── index.html             # Legacy/demo file at repo root

How It Works

1. A repo is uploaded as .zip (/api/repo/upload-local) or cloned/imported from GitHub (frontend supports this flow).
2. Server extracts and scans files (skipping node_modules, .git, dist, build).
3. Files are chunked and embedded with OpenAI.
4. Embeddings are upserted to Pinecone under a namespace (jobId).
5. User asks a question (/api/repo/query).
6. Relevant chunks are retrieved from Pinecone and passed to the LLM to generate an answer.

Getting Started

Prerequisites

• Node.js 18+ (reccomended)
• npm
• OpenAI API key
• Pinecone index/API key
• Supabase project (optional in concept, but current server code expects env vars for logging)

1. Install dependancies

Backend:

cd server
npm install

Frontend:

cd client
npm install

2. Configure enviroment variables (server)

Create server/.env with values similar to:

PORT=3001
OPENAI_API_KEY=your_openai_key
PINECONE_API_KEY=your_pinecone_key
PINECONE_INDEX=your_index_name
SUPABASE_URL=https://your-project.supabase.co
SUPABASE_SERVICE_KEY=your_service_key

# Optional model tuning
RAG_MODEL=gpt-4o-mini
RAG_TEMPERATURE=0.2

Notes:

• PINECONE_INDEX is required at startup.
• Current supabaseService initializes immediately, so missing Supabase vars may cause runtime issues.

3. Run the app

Backend (from server/):

npm run dev

Frontend (from client/):

npm run dev

Default local endpoints:

• Frontend: Vite dev server (usually http://localhost:5173)
• Backend: http://localhost:3001

Main Routes

Frontend routes

• / -> Upload page
• /chat/:projectId -> Chat for uploaded zip job
• /chat/repo/:repoName -> Chat for imported/cloned repo flow

Backend routes

Mounted under /api/repo:

• POST /upload-local -> upload .zip file (repoZip form field)
• POST /query -> ask a question about a previously indexed job
• GET /history/job/:jobId -> fetch query history for a job

Example API Usage

Upload a zip:

curl -X POST http://localhost:3001/api/repo/upload-local \
  -F "repoZip=@/path/to/repo.zip"

Query a job:

curl -X POST http://localhost:3001/api/repo/query \
  -H "Content-Type: application/json" \
  -d '{"jobId":"YOUR_JOB_ID","question":"What does the upload route do?","topK":5}'

Current Notes / Gaps

• Root package.json is mostly demo/placeholder metadata and not the main app runner.
• Some legacy files and alternate routes/controllers exist in server/ and are not all actively wired.
• The frontend GitHub import flow posts to a /repo endpoint and chat fallback uses /ask; depending on your backend wiring, you may need to align these endpoints.

Contributing

1. Fork the repo
2. Create a feature branch
3. Make changes
4. Test locally
5. Open a pull request

License

No explicit project license is defined for the main app folders yet (the root package.json lists MIT, but verify before publishing).