set_model_response escape latin characters in final response

[amenegola]

When defining output_schema in a LlmAgent with other tools, output schema processor executes json.dumps with ensure_ascii = True, making the agent respond with escaped latin characters, and therefore presenting responses not human readable:

adk-python/src/google/adk/flows/llm_flows/_output_schema_processor.py

Line 110 in 168c724

return json.dumps(func_response.response)

This line could be changed with ensure_ascii=False:

json.dumps(func_response.response, ensure_ascii=False)

[amenegola]

Hello @seanzhou1023

Following up on this issue, I've opened a Pull Request (#2937) with the proposed fix. The changes involve explicitly setting ensure_ascii=False in the json.dumps call, which I've tested locally to correctly handle Latin characters.

Please feel free to review it when you have a moment, and let me know if there's anything else I can do to help move this forward.

Thanks for your time!