AI SaaS / RAG
Chaibook
A NotebookLM-style RAG notebook that answers only from your own material, with inline citations.
Overview
Chaibook lets you add your own PDFs, websites, YouTube videos, and text, then chat with an assistant that answers strictly from that material and cites its sources inline.
The Problem
General chat assistants answer from the open web and hallucinate. When you want answers grounded only in your own documents, you need retrieval that stays inside your material and shows where each claim came from.
The Solution
Chaibook ingests multiple content formats into per-workspace vector namespaces, then runs a streaming RAG chat that grounds every answer in your sources and links back to them with citations.
Key Features
- Multi-format ingestion — PDFs, websites, YouTube, and text are chunked, embedded, and indexed into per-workspace Pinecone namespaces via durable Inngest background jobs.
- Streaming RAG chat with [1], [2] source citations, plus an optional Tavily web-search tool.
- Generated study artifacts — summaries, flashcards, quizzes, and mind maps.
- Google auth (better-auth), Prisma on Neon PostgreSQL, OpenAI for chat and embeddings, and Mem0 long-term memory.
Technologies
- Next.js 16
- React 19
- TypeScript
- Vercel AI SDK
- TanStack Query
- shadcn/ui
- Express.js
- better-auth
- Prisma
- PostgreSQL (Neon)
- Pinecone
- OpenAI
- Mem0
- Tavily
- Firecrawl
- Inngest