termalTermalin’s AI assistant — the one that summarizes a session, explains terminal output and turns a request into a command — now runs on a model you choose, including one on your own machine.
Your own model, any provider
Open Settings → AI assistant → Model provider → Own endpoint and point it at any OpenAI-compatible API (OpenAI, Gemini, OpenRouter, Groq, DeepSeek) or a local model through Ollama or LM Studio. Requests go straight from your device to that endpoint with Termalin’s prompts; nothing passes through Termalin’s server, and no allowance applies. The key is optional for a local server. It works the same in the web cabinet, under Account → AI assistant.
If you’re running a model on your own hardware, giving a local LLM safe SSH access covers how to keep it to what you allow. The SSH for AI agents page has the wider picture.
Right-click in the web terminal
The web terminal’s right-click menu now matches the desktop: Copy, Paste, Explain the selection and Ask for a command.
It asks first, on iOS and Microsoft Store builds
The first time you use a summary, an explanation, a command request or “describe this server”, the app tells you where the text goes — the built-in provider or your own endpoint — and asks to allow it. Declining sends nothing, and you’re asked once per provider. Desktop builds from the site pick this up with their next update.
Also fixed
AI answers that came back empty on desktop (Explain, the command from a request, the session summary and the AI settings page read the server’s answer from the wrong place) now show up, and the assistant no longer slips an assumption onto the command line — the command and its note are kept apart.