Fix spurious </think> tags caused by empty string reasoning_content in streaming

[Copilot]

The streaming response from OpenAIChatCompletionClient.create_stream() was generating many spurious <think> and </think> tags when models returned empty string reasoning content chunks.

Problem

When using models like Qwen3-30B-A3B-Thinking that emit reasoning content during streaming, the API sometimes sends chunks with empty string reasoning_content values (""). The original logic incorrectly treated these empty strings as a signal to exit reasoning mode:

elif is_reasoning:
    # Exit reasoning mode - BUG: triggers for empty strings!
    reasoning_content = "</think>"
    # ...

This caused the streaming output to contain many unwanted tag pairs like:

<think>First part</think><think>Second part</think><think>Third part</think>

Instead of the expected:

<think>First partSecond partThird part</think>

Solution

Modified the exit condition to only trigger when reasoning_content is explicitly None:

elif reasoning_content is None and is_reasoning:
    # Exit reasoning mode only when reasoning_content is None (not when it's an empty string).
    reasoning_content = "</think>"
    # ...

Now empty string reasoning content is ignored (no action taken), and reasoning mode only exits when the API sends None as the proper termination signal.

Testing

Validated the fix with comprehensive tests showing:

• Before: Multiple spurious </think> tags generated for each empty string chunk
• After: Exactly one </think> tag generated only when reasoning actually ends

This minimal change preserves all existing functionality while fixing the reported streaming issue.

Fixes #7009.

---

💡 You can make Copilot smarter by setting up custom instructions, customizing its development environment and configuring Model Context Protocol (MCP) servers. Learn more Copilot coding agent tips in the docs.

[ekzhu]

@coach00 pls let us know if this fix works for you.

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 80.93%. Comparing base (5216bf3) to head (ea9fae1).
⚠️ Report is 1 commits behind head on main.

Additional details and impacted files

@@           Coverage Diff           @@
##             main    #7025   +/-   ##
=======================================
  Coverage   80.93%   80.93%           
=======================================
  Files         237      237           
  Lines       18240    18240           
=======================================
  Hits        14762    14762           
  Misses       3478     3478           

Flag	Coverage Δ
unittests	80.93% <100.00%> (ø)

Flags with carried forward coverage won't be shown. Click here to find out more.

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.
• 📦 JS Bundle Analysis: Save yourself from yourself by tracking and limiting bundle sizes in JS merges.

[coach00]

@ekzhu i have tried this fix and it works ! thanks very much !