Python: Bug: Python: Incompatibility between yaml response_format and AzureAIInference

[mathrb]

Describe the bug
Declaring the response_format in a yaml file does not work with AzureAIInference.

To Reproduce
Steps to reproduce the behavior:
Create a yaml function:

name: DummyPrompt
description: hello
template_format: jinja2
execution_settings:
  azure_ai_inference:
    max_tokens: 100
    temperature: 0.0
    response_format: 
      type: json_schema
      json_schema:
        name: dummy_format
        schema:
          type: object
          properties:
           message: {type: string}
template: |
  <message role="system">
    Your goal is to return the message of the user as JSON.
  </message>
  <message role="user">
    {{ message }}
  </message>

Try to invoke the function with AzureAIInferenceChatCompletion.

SemanticKernel, and more specificaly AzureAIInferenceChatPromptExecutionSettings expects that response_format to be {"type": "json_schema", "json_schema": {}} link
The problem lies in the fact that the response_format is passed directly to AzureAIInferenceChatCompletion
Therefore, the JsonSchemaFormat cannot be created because it contains {"type": "json_schema", "json_schema": {}} instead of the content of json_schema

Expected behavior
The json_schema is correctly passed when read from dict.

Platform

• Language: Python
• Source: semantic-kernel==1.32.0 azure-ai-inference==1.0.0b9
• AI model: N/A, the problem arise before actually calling the model

Additional context

In the code:
# Case 4: response_format is a dictionary (legacy), create JsonSchemaFormat from dict
Does it mean that using YAML isn't intended anymore, or am I missing something?

[moonbox3]

It's a bug, @mathrb. I'll work on fixing it.