AI Model Setup
2026-11-15
AI features run on models you configure yourself: built-in providers GLM, DeepSeek, Gemini and OpenAI, plus any OpenAI-compatible endpoint.
Desktop
- Open "Settings → AI services";
- Pick a provider or choose custom, then fill in the endpoint and model name;
- Enter the API key and save;
- Confirm with "Test connection".
Mobile
- Open "Settings → AI models";
- Enter an OpenAI-compatible endpoint and API key, then tap "Fetch model list" and pick a model.
API keys are stored only in the system keychain / keystore.
Network proxy
If you cannot reach the provider, set a proxy for AI requests (http / https / socks5) under "Settings → Network proxy". It applies to model requests only — sync and the built-in browser are unaffected.
Portal online tools
For the online tools (canvas, mind maps, PDF reading), fill in the "Model configuration" panel at the top; the configuration stays in your browser only.
Cost and privacy
Model requests go directly to your provider — the platform never touches them or adds a markup. API keys stay on your machine; usage and billing are settled in your provider account.