fix(azure): handle None values in streaming tool_calls

[VedantMadane]

Summary

Fixes TypeError when streaming tool calls via AzureAIChatCompletionClient when delta chunks have None values.

Problem

When using AzureAIChatCompletionClient with streaming enabled (model_client_stream=True) and tool calls, the code crashes with:

TypeError: can only concatenate str (not "NoneType") to str
  at autogen_ext/models/azure/_azure_ai_client.py line ~550

This occurs because tool_call_chunk.id, tool_call_chunk.function.name, or tool_call_chunk.function.arguments can be None in streamed deltas from Azure AI (GitHub Models / Azure AI Foundry).

Root Cause

The Azure client unconditionally concatenates these fields:

full_tool_calls[idx].id += tool_call_chunk.id
full_tool_calls[idx].name += tool_call_chunk.function.name
full_tool_calls[idx].arguments += tool_call_chunk.function.arguments

Solution

Add None guards before concatenation, matching the pattern already used in the OpenAI client:

if tool_call_chunk.id is not None:
    full_tool_calls[idx].id += tool_call_chunk.id
if tool_call_chunk.function is not None:
    if tool_call_chunk.function.name is not None:
        full_tool_calls[idx].name += tool_call_chunk.function.name
    if tool_call_chunk.function.arguments is not None:
        full_tool_calls[idx].arguments += tool_call_chunk.function.arguments

Testing

This fix follows the same pattern as the OpenAI client (_openai_client.py lines 986-993), which already handles None values correctly.

Environment

• Windows 11
• Python 3.13
• autogen-ext 0.7.5
• Using GitHub Models via Azure AI Foundry endpoint

Fixes #7157