fix(mistral): remove accidental tuple wrapping of CompletionUsage in metadata

[frankgoldfish]

Summary

In MistralAIChatCompletion._get_metadata_from_response, the CompletionUsage object is accidentally wrapped in a single-element tuple due to a trailing comma after the closing parenthesis of the constructor call.

Bug

# Before (buggy): metadata["usage"] is a tuple, not a CompletionUsage instance
metadata["usage"] = (
    CompletionUsage(
        prompt_tokens=response.usage.prompt_tokens,
        completion_tokens=response.usage.completion_tokens,
    ),   # <-- trailing comma creates a tuple
)

Any code that reads metadata["usage"].prompt_tokens or metadata["usage"].completion_tokens will get an AttributeError at runtime because tuples don't have those attributes.

Fix

# After (fixed): metadata["usage"] is a CompletionUsage instance
metadata["usage"] = CompletionUsage(
    prompt_tokens=response.usage.prompt_tokens,
    completion_tokens=response.usage.completion_tokens,
)

Files Changed

• python/semantic_kernel/connectors/ai/mistral_ai/services/mistral_ai_chat_completion.py

Test plan

• Verify that metadata["usage"] is a CompletionUsage instance (not a tuple) after a MistralAI chat completion response
• Confirm metadata["usage"].prompt_tokens and metadata["usage"].completion_tokens are accessible without error
• Run existing MistralAI unit tests

🤖 Generated with Claude Code

[moonbox3]

Any unit test that needs to be added/updated for this?

[Copilot]

Pull request overview

Fixes token usage metadata in the Mistral chat completion connector so metadata["usage"] is a CompletionUsage instance (not an accidental single element tuple), preventing runtime AttributeError when accessing token counts.

Changes:

• Remove trailing comma that wrapped CompletionUsage(...) in a single element tuple.
• Ensure metadata["usage"] is assigned directly to a CompletionUsage instance when response.usage is present.

---

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.

[Copilot]

This bug fix changes the shape of metadata["usage"] (tuple -> CompletionUsage), but there is no regression test asserting the type/attributes of metadata["usage"] on returned ChatMessageContent / streaming chunks. Add a unit test that verifies response.metadata["usage"] is a CompletionUsage instance and that prompt_tokens/completion_tokens are accessible to prevent the tuple-wrapping bug from reappearing.