RAG-based AI Coding Assistant

Context

Developers working with an internal React component library need access to its documentation as they write code. This extension brings a retrieval-based AI coding assistant into Visual Studio Code.

My contribution

I came up with the idea and owned the project through implementation and production deployment.

• Built the Visual Studio Code extension with Python, OpenAI models, LangChain and FAISS.
• Connected internal React component-library documentation to the assistant.
• Used FAISS to search the documentation knowledge base.
• Used LangChain for retrieval and orchestration with OpenAI models.

Result

The assistant is used in production by 30+ internal developers. Users reported less manual coding, and it now answers library questions that previously came directly to me. That reduced routine support demands and freed me to focus on other engineering work. The assistant retrieves relevant internal documentation directly in Visual Studio Code.