Skip to content

AI Configuration

The AI assistant is configured in App Settings → AI Configuration. All settings are stored in nextcp2Config.json and applied immediately after saving — no restart is required, the AI client is rebuilt on the fly.

ProviderIdentifierNotes
Google GeminigoogleOfficial Google Gemini API. Requires an API key.
OpenAI-compatibleopenaiWorks with OpenAI itself and any OpenAI-compatible endpoint such as OpenWebUI, Ollama (OpenAI layer), vLLM or LiteLLM.
SettingDescription
AI enabledMaster switch for the AI feature.
AI providerSelects the active provider (google or openai). Switching providers restores the last used settings of that provider from its profile.
Base URLOnly for OpenAI-compatible endpoints; disabled for Google. Enter the API root of your endpoint, e.g. http://localhost:3000/api for OpenWebUI. Do not append /v1 or /chat/completions — this is added automatically.
API keyProvider API key. For Google it is required. For OpenAI-compatible endpoints it is sent as Authorization: Bearer header and may be optional, depending on your server.
ModelThe model to use. The dropdown is filled live from the provider (Google model API or {baseUrl}/models); you can filter by typing.
Send nextCP toolsWhen enabled (default), nextCP/2 sends its built-in tools (device selection, radio playback, language) to the model. Disable this if your endpoint provides its own server-side tools.
Conversation memoryWhen enabled, recent messages of the conversation are sent as context with each request. Default: off.
Tool IDsOpenWebUI only: comma-separated list of server-side tool IDs to attach to each request, or * to use all tools registered on the endpoint.
Server-side toolsOpenWebUI only: if the endpoint exposes tools, they are listed as checkboxes for granular selection.

Use Save AI config to persist your changes, or Reset AI config to discard unsaved edits.

nextCP/2 stores a profile per provider. When you switch the provider in the dropdown, the API key, base URL, model and tool settings of the previous provider are saved, and the settings of the newly selected provider are restored. This makes it easy to alternate between, for example, Google Gemini and a local OpenWebUI instance without re-entering credentials.

  1. Create an API key in Google AI Studio.
  2. Select provider google and paste the API key.
  3. Pick a model from the dropdown (e.g. gemini-2.5-flash).
  4. Save the AI config and open the chat window.
  1. Select provider openai.
  2. Set the base URL to your OpenWebUI instance including the API path, e.g. http://nas.local:3000/api.
  3. Enter your OpenWebUI API key (Settings → Account in OpenWebUI).
  4. Pick a model from the dropdown.
  5. Optional: if you use tools registered in OpenWebUI, set Tool IDs (or * for all) and consider disabling Send nextCP tools to avoid duplicate tool sets.
  6. Save the AI config.
SymptomCause / solution
”ChatClient not initialized! Please check the AI configuration …”Provider, model, base URL or API key is missing or invalid. Verify the settings and save again.
Model dropdown is emptyThe model list could not be fetched. For Google check the API key; for OpenAI-compatible endpoints check that {baseUrl}/models is reachable.
Assistant answers but doesn’t control devicesSend nextCP tools is disabled, or the selected model does not support tool calling.
Answers lack context from previous messagesEnable Conversation memory.