CaviraOSS's PageLM Turns Any Document Into a Self-Hosted AI Tutor
An open-source project rebuilds NotebookLM's playbook into a self-hostable study platform with quizzes, flashcards, podcasts, and local model support.
Takeaways
- PageLM is an open-source NotebookLM alternative with 1.7k GitHub stars and Docker deployment.
- Turns PDFs, DOCX, and notes into quizzes, Cornell notes, flashcards, and AI podcasts.
- Supports Gemini, GPT, Claude, Grok, MiniMax, Ollama, and OpenRouter as LLM backends.
- TTS options include Edge TTS, ElevenLabs, Google TTS, Cartesia, Hume, and MiniMax.
- Built on Node, TypeScript, LangChain, Langraph, React, and Vite with WebSocket streaming.
- Custom CaviraOSS Community License permits personal and educational use, not commercial resale.
Google's NotebookLM popularized the idea of turning your own documents into a personal tutor, complete with AI-generated podcasts. PageLM from CaviraOSS is a community-driven take on that concept, packaging quizzes, flashcards, structured notes, and audio generation into a self-hostable stack that plugs into whichever LLM you already pay for or run locally.
The repo has crossed 1,700 stars and 230+ forks, and the pitch is simple: an open source AI powered education platform that transforms study materials into interactive learning experiences, slightly inspired by NotebookLM. Under the hood it is a full-stack TypeScript project built on Node, React, LangChain, and Langraph.
What you actually get
Rather than a single chat surface, PageLM ships as a suite of study-focused tools that operate on documents you upload (PDF, DOCX, Markdown, TXT). The feature list reads more like a study app than a raw RAG demo:
- Contextual Chat for asking questions grounded in uploaded documents
- SmartNotes that generate Cornell-style notes automatically from a topic or file
- Flashcards extracted as non-overlapping cards for spaced repetition
- Quizzes with hints, explanations, and scoring
- AI Podcast that turns notes into audio for learning on the go