fix: coerce non-string metadata values instead of dropping them

[ChunkyTortoise]

Summary

• Coerce non-string metadata values via str() in _propagate_attributes instead of dropping them with a warning
• Matches the documented behavior in the v3→v4 migration guide: "Non-string values are automatically coerced to strings"
• Fixes noisy warnings when using LangGraph, which injects langgraph_step (int), langgraph_triggers (list), langgraph_path (tuple) into metadata on every graph step

Changes

• langfuse/_client/propagation.py: Convert non-string metadata values to strings before validation (3-line change)
• tests/test_propagate_attributes.py: Add test for non-string metadata coercion with LangGraph-style metadata types

Test plan

• New test test_non_string_metadata_values_coerced verifies int/list/tuple metadata is coerced to strings
• Existing tests should continue to pass (string values unchanged, over-200-char values still dropped)

Fixes #1571

Disclaimer: Experimental PR review

Greptile Summary

This PR fixes a noisy-warning regression introduced in the v3→v4 migration by coercing non-string metadata values to strings instead of dropping them. The change is minimal and well-targeted: a single str(value) if not isinstance(value, str) else value guard in _propagate_attributes ensures LangGraph-injected metadata (langgraph_step: int, langgraph_triggers: list, langgraph_path: tuple) flows through rather than being silently discarded.

Key changes:

• langfuse/_client/propagation.py: Coerce non-string metadata values before calling _validate_string_value, matching the v3→v4 migration guide's "Non-string values are automatically coerced to strings" promise.
• tests/test_propagate_attributes.py: New test test_non_string_metadata_values_coerced validates all three LangGraph metadata types (int, list, tuple) as well as string pass-through.

Minor follow-ups worth addressing:

• The public propagate_attributes signature still declares metadata: Optional[Dict[str, str]]; updating it to Optional[Dict[str, Any]] would prevent spurious type-checker errors for callers passing LangGraph metadata.
• The docstring Note still says metadata values "must be strings", which contradicts the new behavior.

Confidence Score: 5/5

• Safe to merge; the fix is a targeted 3-line change with a clear motivation, correct logic, and adequate test coverage.
• The core logic change is minimal and correct — coercing before validation rather than dropping. The two remaining items (type annotation and docstring) are non-blocking style/doc cleanups that don't affect runtime behavior. No existing tests are broken and the new test covers the intended cases.
• No files require special attention beyond the minor type annotation and docstring updates noted above.

Important Files Changed

Filename	Overview
langfuse/_client/propagation.py	Core fix coerces non-string metadata values via str() before validation. Logic is correct and minimal. Two minor follow-ups: the public function's type annotation (Dict[str, str]) and the docstring Note no longer match the new behavior.
tests/test_propagate_attributes.py	New test covers int, list, and tuple coercion with realistic LangGraph metadata keys, and verifies string values pass through unchanged. Adequate coverage for the change.

Flowchart

%%{init: {'theme': 'neutral'}}%%
flowchart TD
    A["propagate_attributes(metadata=...)"] --> B{"metadata is not None?"}
    B -- No --> G["Skip metadata block"]
    B -- Yes --> C["Iterate key, value in metadata.items()"]
    C --> D{"isinstance(value, str)"}
    D -- Yes --> E["str_value = value"]
    D -- No --> F["str_value = str(value) — NEW: coerce"]
    E --> H{"len(str_value) <= 200"}
    F --> H
    H -- Pass --> I["validated_metadata[key] = str_value"]
    H -- Fail --> J["warn + drop value"]
    I --> K["_set_propagated_attribute with validated_metadata"]

Loading

_{Reviews (1): Last reviewed commit: "fix: coerce non-string metadata values i..." | Re-trigger Greptile}

[claude]

Claude Code Review

This pull request is from a fork — automated review is disabled. A repository maintainer can comment @claude review to run a one-time review.

[CLAassistant]

Thank you for your submission! We really appreciate it. Like many open source projects, we ask that you sign our Contributor License Agreement before we can accept your contribution.
_{You have signed the CLA already but the status is still pending? Let us recheck it.}

[chatgpt-codex-connector]

💡 Codex Review

Here are some automated review suggestions for this pull request.

Reviewed commit: 2ac10d8244

ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

• Open a pull request for review
• Mark a draft as ready
• Comment "@codex review".

If Codex has suggestions, it will comment; otherwise it will react with 👍.

Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".

[chatgpt-codex-connector]

Handle str failures when coercing metadata

Calling str(value) unconditionally for non-string metadata can now raise if a user-supplied object has a failing __str__ implementation (e.g., proxy/mock/domain objects), which causes propagate_attributes to throw and break request/span execution. Before this change, non-string metadata was dropped without raising, so this is a regression in fault tolerance for malformed metadata inputs.

Useful? React with 👍 / 👎.