Open Notebook is an open-source, privacy-first platform designed as a flexible alternative to Google’s Notebook LM. It empowers researchers, students, and knowledge workers to manage multi-modal content—such as PDFs, videos, audio, and web pages—while maintaining complete control over their data. Unlike cloud-dependent tools, Open Notebook runs locally or on your own infrastructure using Docker or from source, ensuring no sensitive research is sent to third-party servers. Built with Next.js, React, and SurrealDB, it integrates with 16+ AI providers including OpenAI, Anthropic, Ollama, and LM Studio, enabling users to choose models based on cost, performance, or privacy needs. The platform also introduces advanced features like multi-speaker podcast generation and a full REST API, making it suitable for both individual researchers and teams seeking customizable, vendor-agnostic AI-assisted note-taking and knowledge management.
What You Get
- Multi-model AI support - Connect to 16+ LLM providers including OpenAI, Anthropic, Ollama, LM Studio, Google GenAI, and Azure OpenAI; choose the best model per task for cost or privacy optimization
- Privacy-first, self-hosted deployment - Run entirely locally with Docker Compose or from source; no data leaves your environment, ensuring full control over sensitive research
- Professional multi-speaker podcast generation - Create AI-powered podcasts with up to 4 customizable speakers, full script control, and voice profiles using ElevenLabs or other TTS providers
- Universal content ingestion - Import and process PDFs, videos, audio files, web pages, Word docs, and more; all content is indexed for search and AI analysis
- Intelligent full-text and vector search - Quickly find information across all your notes and sources using hybrid search combining keyword matching and semantic embeddings
- Context-aware AI chat - Ask questions about your research materials with full context from uploaded documents, ensuring accurate and source-backed responses
- Custom content transformations - Apply summarization, extraction, or rewriting actions to your documents using configurable AI pipelines
- Comprehensive REST API - Programmatically manage notebooks, sources, and AI interactions via a documented API endpoint at
http://localhost:5055/docs
- Multi-language UI - Fully localized interface in English, Portuguese, Chinese (Simplified & Traditional), Japanese, Spanish, French, Korean, and Russian
- Optional password protection - Secure public deployments with authentication to restrict access without compromising self-hosting benefits
Common Use Cases
- Building a private research repository - A graduate student ingests 50+ academic papers and lecture videos into Open Notebook, uses vector search to find related concepts, and generates AI-assisted summaries for their thesis without exposing data to cloud services
- Creating a multi-speaker educational podcast - A university professor uses Open Notebook to convert a recorded lecture into a professional 3-speaker podcast with distinct voices and scripts, exporting directly to audio platforms
- Problem → Solution flow: Vendor lock-in with Notebook LM → Open Notebook - A user is restricted by Google’s limited AI model options and data policies; they migrate to Open Notebook, switch to Ollama for local inference, and reduce costs by 70% while gaining podcast features
- Team knowledge management in remote R&D - A startup’s AI research team uses Open Notebook to centralize experiment notes, source code snippets, and video demos; they leverage the REST API to automate content ingestion from their CI/CD pipeline
Under The Hood
The lfnovo-open-notebook project is a full-stack AI-powered note-taking and podcast processing platform that integrates RAG-based chat, structured note management, and audio transcription workflows into a unified system. It serves as a research assistant tool that leverages modern AI and data processing capabilities to support both textual and audio-based knowledge workflows.
Architecture
The project follows a monolithic architecture with clear separation of concerns between frontend, API, and domain services, enabling a cohesive development experience.
- The codebase adopts a layered architecture with distinct modules for API routing, business logic, and data handling, ensuring clear separation between presentation, application, and domain layers.
- Services such as chat, embedding, and podcast processing are modularized into dedicated components with well-defined responsibilities, promoting reusability and maintainability.
- Key design patterns include middleware for authentication, strategy-based command execution, and service-oriented architecture that allows for extensibility and integration with various AI providers.
- Component interactions are managed through well-defined API endpoints and internal service calls, supporting both synchronous and asynchronous workflows with clear data flow.
Tech Stack
The project is a full-stack application built with Python and TypeScript, integrating modern web and AI technologies to support its research assistant functionality.
- The backend is implemented in Python 3.11-3.12, using FastAPI for the REST API and LangChain for AI integrations, while the frontend is built with TypeScript and Next.js.
- It heavily relies on LangChain and LangGraph for AI workflows, SurrealDB for database operations, and includes support for multiple AI providers like OpenAI, Anthropic, and Ollama.
- The project uses uv for dependency management, Ruff and MyPy for linting and type checking, and Docker for containerization.
- It integrates pytest with async support for unit and integration testing, alongside Vitest for frontend testing.
Code Quality
The codebase demonstrates a balanced approach to quality with comprehensive test coverage and consistent error handling patterns.
- Testing is well-distributed across multiple domains including notebooks, embeddings, and API interactions, with a focus on structured test coverage.
- Error handling is well-distributed with appropriate use of try/catch blocks and HTTP exceptions, though some areas show potential for more granular exception management.
- Code consistency is maintained through established patterns and clear separation of concerns, though some technical debt indicators suggest opportunities for refactoring.
What Makes It Unique
The project distinguishes itself as a unified AI-powered note-taking and podcast processing platform that blends RAG-based chat, structured note management, and audio transcription workflows into a single cohesive system.
- It intertwines multiple AI services and data pipelines to support both textual and audio-based knowledge workflows in a single integrated platform.
- The system enables researchers to manage notes, process podcasts, and engage in AI-assisted chat within a unified interface.
- Its modular architecture allows for extensibility and integration with various AI providers, making it adaptable to evolving research needs.