Choose models and bring your own key

Personal AI model prefs and optional API key for Keeps, memory, and assistant steps.

Choose models and bring your own key

Pick models for Keeps, memory search, and personal assistant planning. Optionally store your own API key so usage bills to you.

Read settings

GET /v1/api/user/ai-settings
Authorization: Bearer <token>

You see model names, whether a key is stored, and the last four characters — never the raw key.

Change models

PATCH /v1/api/user/ai-settings
Authorization: Bearer <token>
Content-Type: application/json

{
  "keepsModel": "gpt-4.1-mini",
  "agiModel": "gpt-4o",
  "embedModel": "text-embedding-3-small"
}

No step-up required for models alone.

Set or clear your key

Needs a step-up token (x-step-up-token).

PATCH /v1/api/user/ai-settings
Authorization: Bearer <token>
x-step-up-token: <grant>
Content-Type: application/json

{ "cencoriApiKey": "your_key" }

DELETE /v1/api/user/ai-settings/api-key
Authorization: Bearer <token>
x-step-up-token: <grant>

How calls resolve

1. Your key, if stored
2. Platform key, if configured
3. Offline fallback (heuristics / local embeds)

Models follow the same cascade: your prefs → platform defaults → built-in defaults.