Prepare your account and request
The Edge Lab page is public. Live AI requires a THOX account with a confirmed email address. You do not need to bring an API key.
The live experience sends your consented request to hosted AI. Use only public or synthetic information, and keep the request between 1 and 2,000 characters. A sensitivity choice is not automatic detection of confidential content.
- Sign into your THOX account and confirm your email if requested.
- Choose a target profile and enter a public example such as a request to check device status.
- Review the current processing and model details, confirm the input is public, and give hosted-inference consent.
- Run the request and read the resulting proposal and its review requirements.
A proposal does not execute an action
Edge Lab classifies intent and returns a bounded proposal for human review. Target profiles represent examples, not enrolled devices. The demo does not run tools, shell commands, Android actions or device changes.
A successful response does not grant execution approval or connect a device. The separate browser examples are deterministic samples; they are not live AI responses and do not replace a failed live request.
Limits and common messages
Live requests are limited to 10 accepted attempts per account per UTC clock hour and 200 across Edge Lab per UTC day. An accepted attempt stays counted if the hosted request fails.
For a usage-limit message, follow the retry time. The hourly account window resets at the next UTC hour, while the shared daily window resets at 00:00 UTC.
If the processing or model configuration changes, refresh the page, inspect the new details and give consent again. A configured status does not guarantee that the hosted service is currently available.
If sign-in is rejected, confirm you are using the intended account and have completed email verification. For an unavailable service, try later rather than repeatedly submitting the same request.
Keep the example safe to share
The website adapter does not persist your prompt or generated receipt. Usage counters are stored, and hosting, authentication and AI processing services have their own operating policies. This is not a zero-retention or offline experience.
For device connectivity or a local installation, use the relevant setup guide. A hosted Edge Lab result does not verify your local hardware or network.