PageLM is a Node.js/TypeScript and React app, loosely modeled on NotebookLM. It takes uploaded PDF, DOCX, Markdown or TXT files and generates Cornell-style notes, flashcards, quizzes and TTS podcasts, with chat over the documents. It also has voice transcription, a homework planner, exam simulation and a debate mode. LLMs and embeddings are configurable across hosted providers and Ollama, with JSON file storage by default.
Who it is for: Students and educators wanting a self-hosted NotebookLM-style study tool
Strengths
Works with many LLM providers, including Ollama, OpenRouter and a self-hosted LiteLLM gateway
Produces notes, flashcards, quizzes and podcasts from the same uploaded document
Default JSON storage, so no database is needed to start
Several TTS options for podcasts, including Edge TTS, ElevenLabs and Google TTS
Weaknesses
Custom community license: commercial use or resale needs written permission
No tagged releases; setup is by cloning and running npm scripts
Podcast audio requires ffmpeg installed on the host
VoxCPM TTS integration uses the default voice only; no reference-audio cloning
What it needs
GPU optional
Compose
Needs Node.js 21.18+, ffmpeg, Docker (optional), LLM or TTS provider API keys
Models: Ollama, OpenAI, Anthropic Claude, Google Gemini, xAI Grok