AI Chat
Ask the box, the model answers from the cloud
A set of preset questions sends a prompt to the self-hosted server with a single tap, and the model answer is streamed back to the screen; you can also ask directly from the console with ai <question>. The device only sends intents and renders results — model keys and orchestration stay on the server, with no plaintext credentials in the firmware.
Turn counting supports multi-round exchanges, answers wrap to the screen width automatically, and in-flight requests are clearly indicated to prevent double submission.
On-screen interface
A recreation of this page on the 2.41-inch AMOLED display (600 × 450). Actual interaction and layout follow the latest firmware.
Key capabilities
Preset prompts
Common questions in one tap, no typing required.
Keys never on device
All model calls happen server-side; the firmware holds no credentials.
Multi-turn context
Turn counting keeps follow-up questions coherent.
Console access
ai <question> for scripts and remote debugging.
Technical specs
Typical scenarios
On-the-spot answers
Ask what a parameter means or how to debug, without picking up a phone.
Scripted queries
Combine with remote commands for scheduled Q&A and summaries.
Related commands
page 27 Jump to AI Chat ai <question> e.g. ai how many watts is 5V 1A ai status Show in-flight state, turn count and errors
- Requires network access; only one request may be in flight at a time; answer length is bounded by the screen and the transport frame.