Open Notebook

What is Open Notebook?

Open Notebook is a self-hosted, open source research notebook that lets you collect sources — PDFs, web pages, YouTube videos, audio, and raw text — and then chat with them, summarize them, and generate podcasts from them using AI. It is an independent, privacy-focused alternative to Google’s NotebookLM that keeps your documents on your own infrastructure and lets you pick which AI provider processes them.

The project is MIT-licensed and built by developer lfnovo. It grounds every AI answer in the material you upload (retrieval-augmented generation), so responses cite your sources rather than the model’s training data.

What is Open Notebook best for?

Open Notebook is best for researchers, students, and privacy-conscious builders who want a NotebookLM-style workflow but on their own server and with their own model choices. It fits people who care where their documents live, who want to swap in cheaper or local models, and who need programmatic access through an API. It is a single-user research tool, not a team knowledge base.

What can Open Notebook do?

• Ingest many source types — PDFs, web URLs, YouTube links, Office documents, audio, video, and plain text into notebooks.
• Chat with your sources using RAG, so answers stay grounded in the material you added.
• Choose from 16+ AI providers — OpenAI, Anthropic, Google, Mistral, plus local models via Ollama or LM Studio — and mix providers per task (chat, summaries, embeddings) to cut cost.
• Generate multi-speaker podcasts with 1–4 speakers, per-speaker voices and personalities, and custom intro/outro templates — versus NotebookLM’s fixed two-host format.
• Search across everything with combined full-text and vector search.
• Automate via a full REST API (OpenAPI), so you can wire ingestion and generation into your own pipelines.
• Back up simply — SurrealDB stores everything in one container and data directory you can archive.

Where does Open Notebook fall short?

• Citations are basic. Answers reference sources, but there is no inline highlighting back to the exact PDF page or video timestamp the way NotebookLM does — the maintainers flag this as an area still improving.
• Single-user only. There is no built-in multi-user support or role-based access control, so sharing it with a team means putting authentication in front of it with a reverse proxy.
• Audio processing needs external providers. Speech-to-text and text-to-speech call out to third-party services by default; a fully offline audio pipeline takes extra setup, and embeddings/large libraries want 4GB+ of RAM to avoid running out of memory.

Is Open Notebook free?

Yes — Open Notebook is free and open source under the MIT license, with no subscription or seat fees. You self-host it and only pay for what the AI models actually cost you. Point it at your own OpenAI or Anthropic keys, or run local models through Ollama or LM Studio for no per-token cost at all.

What does Open Notebook replace?

Open Notebook is a self-hosted stand-in for Google’s NotebookLM — the same source-grounded chat, summaries, and audio overviews, but private and provider-agnostic. It also overlaps with Notion AI for asking questions and summarizing across your own documents, though Open Notebook is a dedicated research tool rather than a writing assistant inside a broader workspace. If you mainly need a document-grounded RAG engine to embed in an app, a purpose-built engine like RAGFlow may fit better.

FAQ

Is Open Notebook open source? Yes. It is released under the MIT license, one of the most permissive open source licenses, so you can self-host, modify, and extend it freely. The full source is on GitHub.

Can I self-host Open Notebook for free? Yes. It runs via Docker on your own machine or server at no license cost. Your only spend is the AI usage — and that can be zero if you run local models through Ollama or LM Studio.

Is Open Notebook a good NotebookLM alternative? For privacy and flexibility, yes — you keep your documents local, choose any of 16+ AI providers, get a full REST API, and generate 1–4 speaker podcasts. NotebookLM still has more polished citations and a smoother hosted UI, so the trade is control and openness versus convenience.

What do I need to run Open Notebook? Docker (Docker Desktop is enough), an API key for at least one supported AI provider (or a local model runner like Ollama), and around 4GB+ of RAM for comfortable use with larger source libraries.