• Framework: React (Vite) + TypeScript
• Styling: Tailwind CSS + shadcn/ui
• Linting / Formatting: Biome

• Runtime: Node.js
• API: Hono
  • Single codebase
  • Runs in web mode or desktop mode

• Web: PostgreSQL
• Desktop: PGlite (embedded Postgres, local file)
• ORM: Drizzle
  • Relational + app data
  • Embeddings stored via pgvector

• Embeddings (web + desktop): pgvector (Postgres extension)
  • Desktop: PGlite + pgvector extension
  • Web: Postgres + pgvector extension

• Web: WebLLM (in-browser via @browser-ai/web-llm, Vercel AI SDK)
• Desktop: llama.cpp (local runner / sidecar via @ai-sdk/openai)

all of these features should allow for "@"ing courses/documents/terms???

• Dashboard (Upload Docs, Manage Context, View Courses)
• Grounded Chat
• Quiz Generation
• Flashcard Deck Generation
• Voice to Note (.md)

• Allow users to dial how much performance of there computer they want to use (CPU/GPU)
• Find wattage Draw for the given thing they are doing. maybe a cool analytics page

• Open Source
• Point users to download the desktop app for more performance + full privacy
• Explain Deeply what is going on when you run a LLM on your machine. Animation?