StackRAG — Intelligent Code Assistant for VS Code

StackRAG is a Retrieval-Augmented Generation (RAG) powered VS Code extension that helps developers understand, debug, and improve their code using real-world knowledge from Stack Overflow and their own codebase.

It combines:

🔍 Codebase-aware retrieval
🧠 Stack Overflow knowledge
🤖 LLM-powered reasoning
⚡ Fast hybrid search

✨ Features

🧩 Codebase Understanding

Automatically indexes your workspace
Language-aware chunking for Python and JavaScript
Semantic code search across your project

🐛 Intelligent Bug Fixing

Detects issues in your code
Suggests fixes with explanations
Uses real Stack Overflow solutions as evidence

♻️ Smart Refactoring

Improves code quality and structure
Applies best-practice patterns
Maintains original functionality

🔎 Hybrid Retrieval

BM25 + Vector search ensemble
Context-aware ranking
High-quality relevant results

🛠️ Supported Languages

✅ Python
✅ JavaScript

(More languages coming soon)

📦 Installation

Install from the VS Code Marketplace
Reload VS Code
Configure your API key (see below)

🔑 Required Setup

StackRAG requires your own Mistral API key.

Step 1 — Open VS Code Settings

Search for:

StackRAG: Mistral Api Key
StackRAG: Qdrant Cluster Endpoint
StackRAG: Qdrant Api Key

Step 2 — Paste your key

Or add to settings.json:

{
  "stackrag.mistralApiKey": "your-api-key-here",
  "stackrag.qdrantClusterEndpoint": "provide the endpoint",
  "stackrag.qdrantApiKey": "your-qdrant-api-key"
}

⚠️ Your key is stored locally and never uploaded.

🚀 How to Use

1️⃣ Index Your Workspace

Open Command Palette
Run:

StackRAG: Index Workspace

This builds the semantic code index.

2️⃣ Ask Questions

Use the StackRAG panel to:

Explain code
Fix bugs
Refactor functions
Search relevant patterns

3️⃣ Bug Fixing

Provide:

Bug description
Code snippet
Optional related files

StackRAG will generate:

✅ Fixed code
✅ Explanation
✅ Supporting evidence

🧠 How It Works (High Level)

StackRAG uses a multi-stage RAG pipeline:

Workspace code chunking
Stack Overflow knowledge retrieval
Hybrid search (BM25 + vector)
LLM reasoning with context

This produces grounded, high-quality answers.

🔒 Privacy & Security

Your code stays local
No automatic uploads
API key stored in VS Code settings
You control all indexing

🐞 Known Limitations

Large repositories may take time to index
Currently optimized for Python and JavaScript
Requires user-provided API key

🗺️ Roadmap

More language support
Smarter repository summarization
Improved ranking
Streaming responses
UI enhancements

🤝 Contributing

Contributions are welcome! Feel free to open issues or submit pull requests.

⭐ Support

If you find StackRAG helpful, consider giving it a star on GitHub!

Built to make code understanding faster and smarter.

StackRAG

Shifa