[WIP] Add improved RAG system for document uploading and categorization

[Copilot]

• Explore repository (empty - only README.md exists)
• Set up monorepo root with Bun workspaces, Biome config, .gitignore, .env.example
• Create shared types package (packages/shared)
• Build backend app (apps/server) with Bun + Elysia.js
  • Database schema (PostgreSQL + pgvector)
  • Document upload & processing routes
  • Embedding generation service (OpenAI)
  • LLM chat route with RAG
  • Medical specialists route
  • PowerSync backend configuration
• Build frontend app (apps/web) with React + Vite + TanStack
  • PowerSync client setup + schema
  • TanStack Router setup
  • Document upload UI (drag & drop, multi-format)
  • Medical specialist selector
  • RAG chat interface with source citations
  • Document list/management
• Update README with setup instructions
• Code review and security scan

Original prompt

I want to create a better RAG system for easier uploading of documents, all types of docs, images, voice etc. this gets categorized and builds a RAG/vector db that can be used directly for an LLM to be "specialized". The theory is to create an LLM that can pick a medical specialist role and produce information and tips with sources. Since a lot of hospitals have a lot of PMs and routines in different forms, this could help aggregate it. I want to do the powersync hackathon so powersync needs to be used in the project. I like tanstack and bun, biome.

---

📱 Kick off Copilot coding agent tasks wherever you are with GitHub Mobile, available on iOS and Android.