Fix Python frontend tools client sample

[Ran Shemtov (ranst91)]

Problem

The client sample in the Python zone of AG-UI > Frontend Tool Rendering cannot produce the output the page documents. Run against the server from the Getting Started tutorial, any prompt that requires a frontend tool prints only the run boundaries:

User (:q or quit to exit): What's the temperature reading from my sensors?

[Run Started]

[Run Finished]

Prompts that need no tool answer normally, so the gap is specific to frontend tool execution.

Cause

Four things in the sample:

1. Tool declarations carried no parameter schema. Only name and description were sent, so the model had no arguments it was allowed to fill and called the tool with {}.
2. The event loop ignored every tool-call event. It handled RUN_STARTED, TEXT_MESSAGE_CONTENT, RUN_FINISHED and RUN_ERROR only, so TOOL_CALL_START, TOOL_CALL_ARGS and TOOL_CALL_END fell through unhandled.
3. _handle_tool_call was never called, and read fields that do not exist on those events. The tool name arrives as toolCallName on TOOL_CALL_START, and the arguments arrive as JSON text fragments in TOOL_CALL_ARGS.delta, to be concatenated per tool call id and parsed at TOOL_CALL_END.
4. Tool results were never sent back. A frontend tool call ends the server's run, so the client has to record the call and its result and issue a follow-up run to get the agent's answer.

Fix

Only the client code block changes. The tool declarations now ship the JSON schema derived from each function's signature, the tool-call events are read and their argument fragments accumulated per call id, _handle_tool_call receives the tool name and the completed arguments, and the client records the call and result in the conversation before issuing the follow-up run. Run boundaries are reported once per user turn rather than once per run, matching the documented output.

The surrounding prose, the tool definitions section, the protocol description and the expected output are unchanged.

Verification

The sample was extracted from the page and run verbatim against a live AG-UI server built from the Getting Started tutorial:

User (:q or quit to exit): What's the temperature and humidity from my sensors?

[Run Started]

[Client Tool Call: read_climate_sensors]
  Arguments: {'include_temperature': True, 'include_humidity': True}
[Client Tool Result: {'temperature': 22.5, 'humidity': 45.0, 'air_quality_index': 75}]

Your sensors show:
- Temperature: 22.5 °C
- Humidity: 45%

[Run Finished]

Both sample tools were exercised, including get_user_location, which takes no parameters, and conversation history was confirmed to carry across turns.

[learn-build-service-prod]

Learn Build status updates of commit db0aab0:

✅ Validation status: passed

File	Status	Preview URL	Details
agent-framework/integrations/by-component/ui/ag-ui/frontend-tools.md	✅Succeeded

For more details, please refer to the build report.

[learn-build-service-prod]

Learn Build status updates of commit 0d21f2f:

✅ Validation status: passed

File	Status	Preview URL	Details
agent-framework/integrations/by-component/ui/ag-ui/frontend-tools.md	✅Succeeded

For more details, please refer to the build report.