Choose AI models, connect provider keys, and tune how CookieShift calls language models.
AI Settings controls which models power Assistant, banner generation, translations, and compliance tools for your workspace.
Model selection
Primary model
The default model for high-quality tasks (assistant replies, compliance narratives, banner generation). Common options include Gemini, GPT-4 class, and Claude models—availability depends on your plan and server configuration.
Secondary / economy model
Used for lighter tasks or when you want lower cost per token. Pair a strong primary model with an efficient secondary model for routine classification hints.
Changes apply to new requests after you save.
Provider configuration
Depending on your deployment, you may configure:
| Provider | Typical use |
|---|---|
| Google (Gemini) | Default for many CookieShift-hosted flows |
| OpenAI | GPT-family models |
| Anthropic | Claude models |
| OpenRouter | Route to multiple backends with one key |
API keys (BYOK)
Enter keys in the secure fields provided. Keys are stored for your workspace only and used when BYOK mode is enabled.
Safety and behavior toggles
Settings may include options such as:
- Temperature / creativity — lower for compliance copy, higher for marketing variants
- Max tokens — cap response length for cost control
- Feature flags — enable or disable specific AI endpoints per workspace
Exact toggles vary by plan; unavailable options are hidden or disabled.
Testing changes
After saving:
- Open AI Assistant and send a short test prompt
- Check AI Usage for a successful request
- If errors persist, verify the key, model name, and provider status page
Who can edit settings
Workspace owners and admins should manage AI Settings. Editors can consume AI features but may not see key fields.
Related
- AI Usage — verify requests succeed
- Billing — hosted token allowances vs BYOK
- Translations — locale drafts using the configured model