AI assist (admin)

Settings → AI (/settings/ai). This page only connects a provider. What people click is documented in AI in the product.

Providers

Presets include OpenRouter (often the default free-tier path), Gemini, and OpenAI. Each card explains where to get a key. Anything that speaks OpenAI-compatible /v1 chat completions (including Azure OpenAI-style bases) can be set in advanced fields.

Required switches

• Enable AI — master flag (AI_ENABLED)
• API key — stored in env; UI shows a mask if a key already exists
• Test connection — verifies the key before you rely on it in production

Also:

• Semantic search — embeddings for smart search. After enabling: php artisan ai:embed-tasks
• Queue heavy AI jobs — insights, digests, notes-to-tasks run asynchronously (needs queue cron)

Demo mode blocks edits.

Advanced

Model, embedding model, API base URL, timeout, per-minute and per-day rate limits, embedding candidate limit.

Usage widgets: requests/tokens today, 7-day counts, remaining today, breakdown by workspace.

Roles

A user also needs Use AI features on their role. Global on + role off = no buttons.

Still failing?

Wrong base URL, model not allowed on that key, rate limit, or queue not running for heavy jobs. FAQ.