Open source Notebook LM alternative

By /

Star Count: 8.8K+ Privacy-first multimodal AI note-taking and podcast generation tool open-notebook is an open-source project developed by lfnovo that offers a privacy-oriented alternative to Notebook LM with self-hosted deployment and high customization. It integrates multimodal content processing, AI-powered intelligent interactions, and advanced podcast generation for research note-taking, knowledge management, and content creation. The project’s MIT license, emphasizing data control and flexible scalability, has become a popular choice for privacy-sensitive users.

1. Deployment tutorial

The following is explained by the three levels of “Quick Start”, “Typical Deployment (Docker)”, and “Precautions”.

1. Quick Startup (Easiest)

  • Execute from the command line:git clone https://github.com/lfnovo/open-notebook.git cd open-notebook cp .env.example docker.env docker compose -f docker-compose.full.yml up -d This process is mentioned in the user test article to run in WSL2 or Linux environment.
  • Once launched, the browser access http://localhost:8502/notebooks is sufficient.
  • System requirements (official) include: Docker engine, at least 4 GB RAM, 2 GB free disk space.

2. Typical deployment (using Docker image + environment variables)

The following is a more complete process for you to deploy at home/school on a single machine (such as PC or cloud host):

  • Clone Warehouse:git clone https://github.com/lfnovo/open-notebook.git cd open-notebook
  • Copy environment configuration:cp .env.example docker.env
  •  docker.envEdit (or .env) set key environment variables, such as your AI model API KEY, database connection, and so on.
  • Run with mirror (example):docker run -d --name open-notebook -p 8502:8502 -p 5055:5055 -v ./notebook_data:/app/data -v ./surreal_data:/mydata -e OPENAI_API_KEY=你的_key -e SURREAL_URL="ws://localhost:8000/rpc" -e SURREAL_USER="root" -e SURREAL_PASSWORD="root" -e SURREAL_NAMESPACE="open_notebook" -e SURREAL_DATABASE="production" lfnovo/open_notebook:v1-latest-single
  • or use docker-compose: Provided docker-compose.full.ymlin the project (or similar) for easy management.
  • Access the interface:http://<你的主机>:8502 Log in/register.
  • Next, you need to go to the “Model Provider” settings interface to configure your chosen LLM interface for the system.

3. Notes & Tips

  • If you’re on Windows, it’s recommended to use WSL2 + Docker to run it.
  • You need to prepare at least one API KEY for the AI model (such as OpenAI, Anthropic, etc.) or prepare a local model. Otherwise, the AI assistance function will not work.
  • Database: SurrealDB is used as the background data store by default. You must configure SURREAL_URLyour username and password.
  • Resource Consumption: While the foundation is operational, large volumes of documents, multi-user, or local LLMs may require higher configurations.
  • Note that if you upgrade from an older version to v1.0 or higher, the image label/port may change.
  • Image tag description: Recommended or v1-latest-singlev1-latest  ; New users can ignore the “Breaking Changes” reminder.

2. AI model support

Summarize which models/providers are supported by Open Notebook, and when on-premises models are supported.

Supported models/providers

According to the official documentation, the system supports a wide range of LLM providers:

  • OpenAI (GPT Series)
  • Anthropic (Claude Series)
  • Google Gemini / Vertex AI and others
  • OpenRouter (Proxy Other Models)
  • On-premises models are also supported through Ollama or other local-hosted deployments.

On-premises model & privacy control

  • The project emphasizes the self-custody capability of “data is completely under your control”.
  • If you choose a local LLM or run it in an internal network, you can theoretically not pass data to external services.
  • Supports vector search, full-text search, and chat context control for enhanced AI assistance.

Usage suggestions (based on your learning needs)

  • Since you are taking interdisciplinary notes on deep learning/physics and mathematics/philosophy, it is recommended to choose a relatively strong LLM provider (such as OpenAI’s GPT-4 series) as the main model for generating summaries, parsing documents, and forming a “language system” thinking structure.
  • If you’re concerned about cost or data privacy, consider deploying an on-premises model (if you have machine resources) – it may not be as accurate as the cloud model, but it has more control.
  • You can think of model providers as “replaceable modules”: projects support multiple providers, so you can switch them in the future without being locked in.
  • For your use where you intend to integrate multimedia materials (PDFs, videos, articles) and build a “mental structure”, local control + vector search + multi-model support is an advantage.

GitHub：https://github.com/lfnovo/open-notebook
Tubing:

Scroll to Top