Keep your data local and choose your AI models for personal research notes
Gather PDFs, videos, and web pages in one place to chat with AI and even create podcasts. It offers features similar to Google's Notebook LM, but all data stays on your computer or server, so you do not have to worry about privacy. It is an open-source tool that lets you connect various AI models freely to adjust cost and performance.
Data sovereignty and model choice
The core of this tool is that your research materials do not go to an external cloud. If you install it locally with Docker, all processing happens in your environment, and connecting a local model server like Ollama allows you to use AI features without an internet connection. If you prefer the cloud, you can choose from more than 20 providers such as OpenAI, Anthropic, and Google, so you are not locked into a specific vendor. API keys are entered directly in the user interface rather than in a configuration file, making management simple.
Conversing across research materials and generating content
Add materials in various formats, such as PDFs, videos, audio, and web pages, to your notes, and you can have contextual conversations with AI based on these materials. Answers include citations so you can verify the reliability of the information, and full-text and vector search help you quickly find necessary information within vast amounts of material. In particular, a podcast generation feature allows for 1 to 4 speakers, enabling you to convert research content into audio for sharing. You can customize each speaker's profile to produce a professional podcast.
Installation and initial setup
You need Docker Desktop or Docker Engine with the Compose plugin. After downloading the docker-compose.yml file from the repository, you must set a secret string to encrypt API keys. Start the service with the docker compose up -d command, and you can access it in your browser after about 30 seconds. On first run, select your desired AI provider from the Models menu in the sidebar, enter the API key, and specify the language and embedding models to finish setup. To use local models, run Ollama alongside and enter the Base URL.
Security and limitations to check before use
By default, the UI and API ports are open to all network interfaces and there is no password, so if you use it on a public network, you must limit port binding to localhost or set a password. Some providers, such as Anthropic or Groq, do not support embedding models, so you must use them together with another provider that supports embedding features. The current version is optimized for a single-user environment, and multi-user support or agent features are only planned directions and have not yet been implemented. It is distributed under the MIT license, allowing you to modify and use it freely.