fix: fix Chinese character escaping issue

[RiviaAzusa]

Add the ensure_ascii=False parameter to all json.dumps calls to ensure non-ASCII characters (such as Chinese) are not escaped into \uXXXX format.

Files to modify:

langfuse/_client/attributes.py: Core serialization functions
langfuse/_utils/request.py: API request data serialization
langfuse/_client/utils.py: Debug output formatting
langfuse/_task_manager/score_ingestion_consumer.py: Score processing serialization

Before Fix & After Fix:

---

Important

Add ensure_ascii=False to json.dumps calls to prevent non-ASCII character escaping in multiple files.

• Behavior:
  • Add ensure_ascii=False to json.dumps in attributes.py for core serialization functions.
  • Add ensure_ascii=False to json.dumps in request.py for API request data serialization.
  • Add ensure_ascii=False to json.dumps in utils.py for debug output formatting.
  • Add ensure_ascii=False to json.dumps in score_ingestion_consumer.py for score processing serialization.

^{This description was created by for 843ed96. You can customize this summary. It will automatically update as commits are pushed.}

Disclaimer: Experimental PR review

Greptile Summary

Updated On: 2025-09-09 06:09:40 UTC

This PR addresses a localization issue where non-ASCII characters (specifically Chinese characters) were being escaped to \uXXXX format in JSON serialization throughout the Langfuse Python SDK. The fix adds the ensure_ascii=False parameter to all json.dumps() calls across four key files to preserve Unicode characters in their original readable form.

The changes affect multiple serialization touchpoints:

• langfuse/_client/attributes.py: Core serialization for OpenTelemetry span attributes
• langfuse/_utils/request.py: API request data serialization sent to Langfuse servers
• langfuse/_client/utils.py: Debug output formatting for span data
• langfuse/_task_manager/score_ingestion_consumer.py: Score processing serialization and size calculations

All modified files use the existing EventSerializer class but now explicitly disable ASCII-only encoding. This ensures that when users trace LLM applications containing Chinese text (or other Unicode content), the data remains human-readable in the Langfuse UI, logs, and debug output. The change maintains JSON validity while improving the user experience for international developers.

Confidence score: 5/5

• This PR is safe to merge with minimal risk as it only affects JSON output formatting without changing data structures or API compatibility
• Score reflects that the changes are straightforward, well-targeted, and follow a consistent pattern across all affected files
• No files require special attention as all changes follow the same simple pattern of adding the ensure_ascii=False parameter

[CLAassistant]

All committers have signed the CLA.

[greptile-apps]

_{4 files reviewed, no comments}

_{Edit Code Review Bot Settings | Greptile}

[nimarb]

Hi @RiviaAzusa , thank you for the contribution! If possible, could you add at least one test for this behavior in order to ensure not running into this problem again? Thanks!

[gfx]

I really want this PR is merged.

[chenyong]

I ran into the same problem, and Kimi K2 discovered the same patch. Any update on this?

And CJK full text search in traces list breaks due to this issue.

[johnmai-dev]

I really want this PR is merged.