Open Notebook

Open Notebook is a self-hosted research notebook: sources, notes, transformations and chat over your material. Its AI runs on providers you set in the UI.

Upstream: source.

Plugboard options

Give Open Notebook its own key. Docker needs a reachable address (Serving other machines).

{
  "options": {
    "listen": { "host": "0.0.0.0", "port": 4010 },
    "keys": { "open-notebook": "{env:PLUGBOARD_KEY_OPEN_NOTEBOOK}" },
    "models": ["opencode-go/glm-5.3-flash"],
    "aliases": { "chat": "opencode-go/glm-5.3-flash" },
    "apps": { "open-notebook": { "allowedModels": ["chat"] } }
  }
}

Connect

The OpenAI Compatible provider takes one base URL per configuration. In the UI:

1. Add a configuration under OpenAI Compatible. Set a name, the Base URL http://host.docker.internal:4010/v1, and the open-notebook key as the API key.
2. Click Test. It calls GET <Base URL>/models, and plugboard answers with the models this app may use.
3. If chat is missing from the list, type its id in Discover Models and add it.
4. Add chat as a language model.

Set the provider in the database configuration. The OPENAI_COMPATIBLE_BASE_URL and OPENAI_COMPATIBLE_API_KEY env vars are a deprecated fallback.

What works

Language models for chat and transformations, through the chat alias.

Limits

• Adding sources needs an embedding model, which plugboard does not serve. Add a second configuration for an embedding provider (Ollama or another OpenAI-compatible server). Each model remembers which configuration it came from.
• Speech to text and text to speech are not served. Open Notebook's docs point to Speaches for local speech.
• Model calls time out after ESPERANTO_LLM_TIMEOUT (180 seconds by default). Raise it for a slow model.