openai/openai-python · error · RuntimeError

Expected to have received `response.created` before `{event.

Error message

Expected to have received `response.created` before `{event.type}`

What it means

Raised by the internal accumulation state when a non-`response.created` event arrives before `response.created`. The Responses streaming accumulator builds its initial ParsedResponseSnapshot from `response.created`, so any other event appearing first leaves it with nothing to accumulate into.

Source

Thrown at src/openai/lib/streaming/responses/_responses.py:370

                content = output.content[event.content_index]
                assert content.type == "output_text"
                content.text += event.delta
        elif event.type == "response.function_call_arguments.delta":
            output = snapshot.output[event.output_index]
            if output.type == "function_call":
                output.arguments += event.delta
        elif event.type == "response.completed":
            self._completed_response = parse_response(
                text_format=self._text_format,
                response=event.response,
                input_tools=self._input_tools,
            )

        return snapshot

    def _create_initial_response(self, event: RawResponseStreamEvent) -> ParsedResponseSnapshot:
        if event.type != "response.created":
            raise RuntimeError(f"Expected to have received `response.created` before `{event.type}`")

        return construct_type_unchecked(type_=ParsedResponseSnapshot, value=event.response.to_dict())

View on GitHub (pinned to 9917c6e28e)

Solutions

1. Fix the event source so `response.created` is the first event of the stream.
2. In tests, prepend a `response.created` event to your mock stream.
3. If using a proxy, verify it forwards the initial SSE event unchanged.

Example fix

// before
events = ["response.output_item.added", ...]

// after
events = ["response.created", "response.output_item.added", ..., "response.completed"]

def starts_with_created(events: list[str]) -> bool:
    return bool(events) and events[0] == "response.created"

try:
    stream = client.responses.stream(...)
except RuntimeError as e:
    if "response.created" in str(e):
        raise ValueError("event source violates Responses streaming protocol") from e
    raise

When it happens

Trigger: A stream whose first event is `response.in_progress`, `response.output_item.added`, or any delta event instead of `response.created`; typically from a custom/proxy server or a reordered mock event sequence.

Common situations: Hand-written SSE fixtures in tests that omit response.created; middleware or proxies dropping the first event; using a non-OpenAI backend that implements a different event order.

Related errors

• Didn't receive a `response.completed` event.
• Cannot provide both response_id/starting_after can't be prov
• input must be provided when creating a new response
• WebSocket connection closed with unsent messages
• Missing stream class

AI-assisted analysis of openai/openai-python@9917c6e28e (2026-08-28). Data as JSON: /api/errors/5773082ce2f16a33. Report an issue: GitHub.