fix(openai): support Langfuse arguments in stream helpers

[gigio1023]

What does this PR do?

Allows the native Chat Completions and Responses stream() helpers to accept Langfuse-specific arguments. The inner create(stream=True) call remains the single owner of the generation observation.

• Covers sync and async Chat helpers from OpenAI 1.40, using the legacy beta registration before 1.92, and Responses helpers from 1.66.
• Preserves native stream managers, structured-output and final-result helpers, user-provided extra_body, and sync manager reuse.
• Finalizes each observation once on exhaustion, explicit close, or context-manager exit after partial or no consumption.
• Treats OpenAI unset defaults as absent so inputs, metadata, and model parameters remain clean. Actual Responses retrieval streams remain uninstrumented.

Fixes langfuse/langfuse#15018

Type of change

• Bug fix
• New feature
• Breaking change
• Refactor
• Documentation update
• Tooling, CI, or repo maintenance

Verification

uv run --frozen pytest -q tests/unit/test_openai.py tests/unit/test_openai_stream_compatibility.py
for version in 1.40.0 1.66.0 1.92.0 2.29.0 2.30.0; do uv run --frozen --with "openai==$version" pytest -q tests/unit/test_openai_stream_compatibility.py; done
LANGFUSE_PUBLIC_KEY=test-public-key LANGFUSE_SECRET_KEY=test-secret-key LANGFUSE_BASE_URL=http://localhost:3000 bash scripts/codex/quick-check.sh

Checklist

• I self-reviewed the diff using code_review.md.
• I added or updated tests for behavior changes.
• I updated docs, examples, or .env.template if needed.
• I did not hand-edit generated files; if generated files changed, I used the upstream regeneration path.
• I did not commit secrets or credentials.

Greptile Summary

This PR extends Langfuse instrumentation to OpenAI’s native Chat Completions and Responses stream helpers while preserving their native managers and keeping the inner create(stream=True) call responsible for the observation.

• Registers synchronous and asynchronous stream helpers across the supported OpenAI SDK version layouts.
• Tunnels Langfuse-only arguments through stream managers and removes them before constructing provider requests.
• Finalizes stream observations on exhaustion, explicit close, or context-manager exit.
• Adds focused lifecycle, argument-forwarding, manager-reuse, structured-output, and multi-version compatibility tests.

Confidence Score: 5/5

The PR appears safe to merge with no concrete blocking or independently actionable non-blocking issues identified.

The stream wrappers preserve provider request arguments and native stream behavior while tests cover supported version boundaries, synchronous and asynchronous lifecycle paths, argument forwarding, and exactly-once observation finalization.

_{Reviews (1): Last reviewed commit: "fix(openai): support Langfuse arguments ..." | Re-trigger Greptile}

Context used:

• Knowledge Base — OpenAI instrumentation
• Knowledge Base — Prompts and framework integrations

[CLAassistant]

All committers have signed the CLA.

[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.