Skip to main content
Edition: Community Model Configuration is where you set up AI providers once for the active project. The AI Chatbot and AI Search widgets — and source training — use the providers configured here.

Per provider

Open a provider (OpenAI, Azure OpenAI, Anthropic, Mistral, Google Gemini, or Custom LLM / Ollama) and set: Azure OpenAI also needs your resource Endpoint (required) and an optional API version. Instead of picking models, enter your chat deployment (required) and embedding deployment names, or use Refresh to load them from Azure. See Providers › Azure OpenAI. A temperature of 0.2–0.5 suits most chatbots and search: lower values keep answers focused and close to your content. Each provider shows Configured, Not configured, or Key rejected. If a key is rejected, enter a valid key — until then, widgets cannot use that provider and the widget settings offer to switch to another configured provider.

How widgets and sources use it

  • In the AI Chatbot and AI Search model settings, choose one of the configured providers. Chat model, embedding model, API key, and tuning come from Model Configuration.
  • When adding a website source, AI model for training lists the configured providers with a working key. If that provider’s model changes here, the next training uses the new model.
  • Remove configuration deletes the saved key and model settings. Widgets already using the provider keep working until you change their provider.