Fix/anthropic colud not end with trailing whitespace at assistant content

[SongChiYoung]

Why are these changes needed?

This PR fixes a 400 - invalid_request_error that occurs when using Anthropic models and the final message is from the assistant and ends with trailing whitespace.

Example error:

Error code: 400 - {'error': {'code': 'invalid_request_error', 'message': 'messages: final assistant content cannot end with trailing whitespace', ...}}

To unblock ongoing internal usage, this patch introduces an ad-hoc fix that strips trailing whitespace if the model is Anthropic and the last message is from the assistant.

Related issue number

Ad-hoc fix for issue discussed here:
#6167

Follow-up structural proposal here:
#6167
#6167 (comment)

Checks

• I've included any doc changes needed for https://microsoft.github.io/autogen/. See https://github.com/microsoft/autogen/blob/main/CONTRIBUTING.md to build and test documentation locally.
• I've added tests (if relevant) corresponding to the changes introduced in this PR.
• I've made sure all auto checks have passed.

[ekzhu]

Why not use the message transformation?

[SongChiYoung]

@ekzhu

Thanks for the question! (I wrote it quickly at first, but just updated the comment now.)

Why not use the message transformation?

The reason I didn't use the transformer is because this issue specifically applies to the last AssistantMessage, and the current message transformation logic works on a per-message basis — it doesn't have visibility into the full message sequence.

If we apply .rstrip() blindly to all AssistantMessages in the transformer stage, it could have unintended side effects, such as:

["show under message\n", "blah, blah"] → ["show under message", "blah, blah"]

That’s why I applied the fix after the message list is fully constructed — to avoid over-sanitizing content that’s meant to preserve formatting.

Let me know if you’d prefer a different way to organize this logic!

[SongChiYoung]

(Aside: Why did we originally design the transformer to work per-message?)

That was intentional — based on the Single Responsibility Principle.

Message transformers are meant to map the structure of each LLMMessage into SDK-specific args (e.g., role, name, content) in isolation. They're responsible for how each message "looks" structurally, not how the message sequence should be interpreted as a whole.

In contrast, message stream-level rules — like "strip trailing space from the last AssistantMessage" — depend on list-level context and vendor quirks. So they should live in a separate layer like MessageCleaner, which operates on the full message sequence.

This keeps responsibilities clear:

• Transformer → adapts message structure
• Cleaner → sanitizes message list

[ekzhu]

@ekzhu

Thanks for the question! (I wrote it quickly at first, but just updated the comment now.)

Why not use the message transformation?

The reason I didn't use the transformer is because this issue specifically applies to the last AssistantMessage, and the current message transformation logic works on a per-message basis — it doesn't have visibility into the full message sequence.

If we apply .rstrip() blindly to all AssistantMessages in the transformer stage, it could have unintended side effects, such as:

["show under message\n", "blah, blah"] → ["show under message", "blah, blah"]

That’s why I applied the fix after the message list is fully constructed — to avoid over-sanitizing content that’s meant to preserve formatting.

Let me know if you’d prefer a different way to organize this logic!

Understood, let's starts with the design in this PR and wait for usages and feedbacks.

[SongChiYoung]

Sounds good - thanks!
Let’s see how it goes.

[codecov]

Codecov Report

Attention: Patch coverage is 78.57143% with 3 lines in your changes missing coverage. Please review.

Project coverage is 77.31%. Comparing base (aec04e7) to head (3fd2fab).
Report is 1 commits behind head on main.

Files with missing lines	Patch %	Lines
.../autogen_ext/models/anthropic/_anthropic_client.py	71.42%	2 Missing ⚠️
...xt/src/autogen_ext/models/openai/_openai_client.py	85.71%	1 Missing ⚠️

Additional details and impacted files

@@            Coverage Diff             @@
##             main    #6168      +/-   ##
==========================================
+ Coverage   77.26%   77.31%   +0.04%     
==========================================
  Files         197      197              
  Lines       13783    13797      +14     
==========================================
+ Hits        10650    10667      +17     
+ Misses       3133     3130       -3     

Flag	Coverage Δ
unittests	77.31% <78.57%> (+0.04%)	⬆️

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.