AI Integrations¶

AI Integrations connect external AI models to your CxReports workspace. They are used by the AI Assistant and automatic dictionary translation.

As a workspace admin, navigate to the Admin application, then Connections → AI Integrations, to manage integrations.

Supported Engines¶

The model name is a free-form string, so any model identifier supported by the selected engine can be used. As new models are released, they can be used immediately by entering the model name.

Gemini models run with a low thinking level, whether they are picked from the list or typed in by hand. Gemini versions before 3 expect a different thinking setting and may reject requests. DeepSeek keeps its own default thinking behaviour. Existing histories without the reasoning DeepSeek requires cannot be continued with DeepSeek; start a new conversation or keep the previous provider.

Gemini requests use Google's own content filtering thresholds, the same as you would get calling the API directly.

Configuration¶

Each integration requires a Name, Engine, Model, and API Key. API keys are stored encrypted and only decrypted at the moment they are needed.

You can create multiple integrations -- for example, a faster model for simple tasks and a more capable model for complex work -- and switch between them per conversation in the AI Assistant.

Notes¶

• The AI permission in role configuration controls access to AI features.
• Dictionary translation uses the default language first, falling back to the entry key if needed.