AI providers
sudoterm ships day-1 support for four providers. Pick one (or several), paste your API key, swap models per conversation.
| Provider | Default model | Pricing | Notes |
|---|---|---|---|
| Anthropic | claude-sonnet-4-5 | $$ | Best reasoning + tool use |
| OpenAI | gpt-4o | $$ | Strong general purpose |
| Gemini | gemini-2.5-pro | $ | Huge context window |
| Deepseek | deepseek-chat | ¢ | Cheapest, surprisingly capable |
Add a key
Open the AI sidebar → click ⚙ → paste your key for each provider you want. Keys are AES-256-GCM-encrypted with the master key in ~/.sudoterm/master.key before being written to ~/.sudoterm/data.db. The plaintext key is held in memory only during a streaming request.
How streaming works
The daemon never proxies your conversation through us — it calls the provider directly with your key. We just decrypt → stream → forward over SSE to your browser. No logging, no caching, no telemetry.
Block → AI flow
Hover any command block in the terminal → click "ask AI". The composer pre-fills with the command + its output, ready to send to your chosen provider.
Swap models mid-conversation
The model selector in the AI sidebar drops down to the curated list for the active provider. Pick a cheaper model for boilerplate questions, escalate to a bigger one for hard ones. Provider + model preferences persist in localStorage.