AI Configuration
Configure AI providers and assign models to specific actions from the AI tab in Super Admin. All AI features in Clarive require at least one configured provider.
Add a provider
Click Add Provider and enter:
- Name — a label for the provider (e.g., "OpenAI", "Anthropic", "Ollama")
- Endpoint URL — the API base URL. Any OpenAI-compatible endpoint works.
- API Key — your provider's API key (stored encrypted)
After adding a provider, click Fetch Models to retrieve the list of available models from the endpoint.
Assign models to actions
Each AI-powered feature in Clarive uses a specific action type. Assign a model to each action to control which model handles that task:
| Action | What it does |
|---|---|
| Generation | Main prompt generation in the AI Wizard |
| Evaluation | Quality scoring after generation (Clarity, Effectiveness, Completeness, Faithfulness) |
| Clarification | Follow-up questions during the review step |
| System Message | Auto-generate system messages for entries |
| Decomposition | Split a single prompt into a multi-step chain |
| Fill Template Fields | Generate example values for template variables |
| Playground Judge | Score playground/test matrix responses (Accuracy, Helpfulness, Relevance, Coherence, Safety) |
| Polish Description | Rewrite rough descriptions in the AI Wizard |
| Merge Conflict | AI-assisted resolution when two users save conflicting changes |
Use Quick Setup to assign the same model to all actions at once — useful when getting started with a single provider.
Model capabilities
When fetching models, Clarive detects whether each model supports reasoning (extended thinking). Reasoning models show additional settings in the Playground like Reasoning Effort and Show Thinking, and hide the temperature slider.