Menu

Using OpenTodo GuidesVersion 0.6.0

Set up the assistant

Settings, Integrations, the Assistant card as the owner sees it: whether the assistant is on, the provider and model everyone uses, the monthly limit and the Configure button.

The assistant is off until the server owner turns it on here. Members' own settings are described in Assistant settings; the owner's further choices in Assistant owner options.

Turn it on#

Settings → Integrations → Assistant shows the owner whether the assistant is on, the provider and model everyone uses, the providers allowed (flagging one without a key), the monthly limit and whether members may bring their own key. Select Configure (owner only) to change them:

  1. Tick Allow this provider for one or more providers and enter the model and key.
  2. Choose the Provider everyone uses.
  3. Tick Turn on the assistant for this server and select Save assistant settings. The dialog closes and the section shows the new state.
  4. In Configure, select Test connection under a provider: it sends one tiny request with the model, base URL and key you see on screen, saved or not, and reports success or the provider's error. Testing never saves anything; a passing test of unsaved values says so, and you still select Save. It is the only request the server makes outside someone's turn.
Provider Model Key Notes
Anthropic (Claude) claude-opus-5-5 by default; claude-sonnet-5-5 and claude-haiku-5-5 are offered, any model id can be typed An Anthropic API key Uses the official Anthropic Go SDK and the Messages API
OpenAI Any chat model id from your account An OpenAI API key Chat Completions at api.openai.com
Google Gemini Any Gemini model id A Gemini API key Google's OpenAI-compatible endpoint on generativelanguage.googleapis.com
Compatible endpoint The model name your server serves Optional Ollama, LM Studio, OpenRouter, Mistral and others; enter the base URL, for example http://192.168.1.20:11434/v1

The model must support tool use. Test connection reports an endpoint that rejects tool definitions as no_tool_support. A model on your own network is set up as a compatible endpoint; see A local model.

Changing the provider asks everyone to accept the disclosure again. Where keys are stored, and the switches for members' own keys and the monthly allowance, are in Assistant owner options.

Network details for operators are in Running with Docker; the routes are in the API reference.

Edit this page on GitHub