UI: task expiration not handled gracefully

[jezekra1]

Describe the bug
When the user does not respond in time, the task will expire and the error is not nicely explained to the user.

To Reproduce
Steps to reproduce the behavior:

1. Write agent that uses input_required (e.g. request form)
2. (optional): set task_timeout=timedelta(seconds=1) in server.run to speed this up
3. Run agent in UI, wait for expiration (default 10 minutes or configured in 2.)
4. Send the form
5. See error

Expected behavior
Nice message to the user: "You failed to respond in time, why are you so slow?"

Logs / Screenshots / Code snippets

Additional context
The expiration might only work correctly after the latest SDK refactor on main.

[dosubot]

Currently, when a task expires due to user inactivity, the UI surfaces a generic error toast and marks the agent message as failed, but does not provide a user-friendly explanation for the timeout. The error code (-32602) is propagated from the backend as a JSON-RPC InvalidRequestError, and the message shown is either a default or extracted from the backend event, which isn't specific to expiration scenarios. There is no logic in the UI to detect task expiration and display a custom message like "You failed to respond in time" AgentRunProvider error handling.

To improve this, you could update the error handling in the UI (for example, in the AgentRunProvider or useHandleError hook) to check for error code -32602 or specific error messages from the backend, and display a tailored message to the user when a task expires. This would make the experience much clearer and friendlier error toast logic.

If needed, you could also adjust backend error messages to make it easier for the UI to distinguish task expiration from other errors, but the main improvement can be made in the UI error handler.

To reply, just mention @dosu.

---

^{How did I do? Good | Irrelevant | Incorrect | Verbose | Hallucination | Report 🐛 | Other}