AI Overview
AI is optional. Connect your own API key or have an admin set a server-wide key. Everything else works without it.
The AI features are split across these pages:
- Photo Analysis — auto-fill bin name, items, tags, and notes from photos
- AI Assistant — natural-language commands and "where is X" search
- AI Reorganization — bulk-restructure bins, areas, and tags
- Task Routing — route different tasks to different providers/models
- Custom Prompts & Tuning — override prompts, temperature, token limits
Supported Providers
| Provider | Recommended Model |
|---|---|
| OpenAI | gpt-5-mini |
| Anthropic | claude-sonnet-4-6 |
| Google Gemini | gemini-3-flash-preview |
| OpenAI-compatible | — |
Getting an API key
Create an API key from your provider's developer platform:
- OpenAI — platform.openai.com → API Keys. Pricing →
- Anthropic — console.anthropic.com → API Keys. Pricing →
- Google Gemini — aistudio.google.com → Get API key. Free tier available. Pricing →
Per-User Setup
Configure your provider, API key, and model in Settings → AI. AI features activate immediately after saving.
Server-Wide Setup (Admin)
Admins can configure a shared AI key via environment variables so all users get AI features without individual setup:
ini
AI_PROVIDER=openai
AI_API_KEY=sk-...
AI_MODEL=gpt-5-mini
# For OpenAI-compatible endpoints:
AI_ENDPOINT_URL=https://your-endpoint.example.com/v1Users can still configure their own keys in Settings → AI, which takes precedence over the server-wide key.
INFO
Set AI_ENCRYPTION_KEY to an AES-256-GCM key to encrypt stored API keys at rest. Without this, keys are stored as plaintext in the database.