Build a chat with document app called DocuRAG.
It will be a typical RAG setup where a user
can upload a document. It will be chunked, embedded,
and stored in a vector DB. Once done, a user can ask
questions about the document. The engine will retrieve
the relevant chunks after embedding the query. Finally,
it will generate a coherent response using GPT-4o based
on the query and the retrieved context. Add Google OAuth.
Use Supabase as the backend and LLMs/embedding models via
the OpenAI API. Build frontend in next.js.