🐛 Bug Report: GenerationSpanData missing response field causes _extract_response_attributes to never execute

[heyanlong]

Which component is this bug for?

OpenAI Instrumentation

📜 Description

GenerationSpanData missing response field causes _extract_response_attributes to never execute

@dont_throw
    def on_span_end(self, span):
        """Called when a span ends - finish OpenTelemetry span."""
        from agents import GenerationSpanData

        if not span or not hasattr(span, "span_data"):
            return

        if span in self._otel_spans:
            otel_span = self._otel_spans[span]
            span_data = getattr(span, "span_data", None)
            trace_content = should_send_prompts()
            if span_data and (
                type(span_data).__name__ == "ResponseSpanData"
                or isinstance(span_data, GenerationSpanData)
            ):
                # Extract prompt data from input
                input_data = getattr(span_data, "input", [])
                _extract_prompt_attributes(otel_span, input_data, trace_content)

                # Add function/tool specifications to the request using OpenAI semantic conventions
                response = getattr(span_data, "response", None) # if span_data is GenerationSpanData, so _extract_response_attributes to never execute
                if response:
                    model_settings = _extract_response_attributes(otel_span, response, trace_content)
                    self._last_model_settings = model_settings

👟 Reproduction steps

from opentelemetry.instrumentation.openai_agents import OpenAIAgentsInstrumentor

trace.set_tracer_provider(trace_provider)

OpenAIAgentsInstrumentor(
    replace_existing_processors=True
).instrument(
    agent_name="TestAgent",
    tracer_provider=trace_provider,
)

agent = Agent(
            name=name,
            instructions=f"{RECOMMENDED_PROMPT_PREFIX}\n\n{mark_instructions}",
            model=LitellmModel(model=model),
            model_settings=ModelSettings(
                include_usage=True
            ))
Runner.run_streamed(agent, messages, context=context, **kwargs)

👍 Expected behavior

There should be gen_ai.output.xxx

👎 Actual Behavior with Screenshots

🤖 Python Version

No response

📃 Provide any additional context for the Bug.

No response

👀 Have you spent some time to check if this bug has been raised before?

• I checked and didn't find similar issue

Are you willing to submit PR?

None

[RudraDudhat2509]

Hi, I'd like to fix this. Plan is to check what field GenerationSpanData carries for its output (vs .response on ResponseSpanData) and pass it correctly to _extract_response_attributes in _end_generation_span. Is a PR welcome?