Instrument Responses.parse and AsyncResponses.parse

What problem do you want to solve?

opentelemetry-instrumentation-genai-openai instruments Responses.create() and AsyncResponses.create(), but not Responses.parse() or AsyncResponses.parse().

As a result, applications using the Responses API structured-output helper do not emit a GenAI span:

from openai import AsyncOpenAI
from pydantic import BaseModel

class CalendarEvent(BaseModel):
    name: str
    date: str
    participants: list[str]

client = AsyncOpenAI()

response = await client.responses.parse(
    model="gpt-4o-mini",
    input="Alice and Bob are going to a science fair on Friday.",
    text_format=CalendarEvent,
)

Responses.parse() calls the SDK's request path directly rather than delegating to the instrumented Responses.create() method. Therefore, instrumenting create() does not cover this API.

This differs from chat.completions.parse(), which is already instrumented by #18. That PR explicitly did not change the Responses API path.

I could not find an existing issue or pull request covering Responses.parse() or AsyncResponses.parse().

Describe the solution you'd like

Instrument both methods when they are available in the installed OpenAI SDK:

• openai.resources.responses.responses.Responses.parse
• openai.resources.responses.responses.AsyncResponses.parse

The implementation could reuse the existing Responses API invocation and response extraction logic used by Responses.create().

Describe alternatives you've considered

No response

Additional Context

• openai==3.0.0
• opentelemetry-instrumentation-genai-openai==1.0b0

The current instrumentation source also registers Responses API create, stream, and retrieve, but does not register Responses.parse or AsyncResponses.parse.

Would you like to implement it?

None

Tip

_{React with 👍 to help prioritize this issue. Please use comments to provide useful context, avoiding +1 or me too, to help us triage it. Learn more here.}

Start with the instrumentation source that registers Responses API create, stream, and retrieve, then compare its existing invocation and response extraction logic with openai.resources.responses.responses.Responses.parse and AsyncResponses.parse. Add coverage for both parse methods when available in the installed OpenAI SDK, reusing the existing Responses behavior, and verify that structured-output calls emit a GenAI span.