Skip to content
- Have an OpenAI-compatible model endpoint available (for example Ollama, LM Studio, or OpenRouter).
- Have the endpoint URL and API key (if required).
- Open Settings and choose Preferences, then Available AI Models.
- In server management, click Add server and enter the server URL.
- Add API key if required by your provider, then click Connect.
- Enable one or more models from the connected server.
- Select the enabled model in AI Chat.
- Open model settings and configure optional
system prompt, temperature, top_p, and top_k values.
- Send a test prompt to verify the model responds.
- For secure handling, note that API keys are stored in session storage and cleared on disconnect/session end.
- The custom server shows as connected in model settings.
- Enabled external models appear in the AI model selector.
- Test prompts return responses from the selected external model.
- Model parameter changes persist for that model profile.
- Connection fails
- Confirm URL format, server uptime, and CORS setup for browser access.
- Model list is empty
- Refresh server models and verify the endpoint exposes
/models.
- Agent is chat-only
- Check server mode/tool-calling support and use a model/server combination that supports agent actions.
- Auth errors persist
- Re-enter API key and reconnect from Models and AI settings.
