Skip to content

[gpt-oss] use vLLM instead of openai types for streaming #25186

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
hmellor merged 6 commits into from
Sep 30, 2025
Merged

[gpt-oss] use vLLM instead of openai types for streaming #25186

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
6 commits
Select commit Hold shift + click to select a range
7b82dbf
initial commit, vllm ResponsesCompletedEvent type
qandrew Sep 18, 2025
64bb6d7
more types
qandrew Sep 20, 2025
a782f96
Merge branch 'main' into gpt-oss-streaming-3-wrap-types
qandrew Sep 26, 2025
f21a45d
ut
qandrew Sep 26, 2025
658048b
Merge branch 'main' into gpt-oss-streaming-3-wrap-types
qandrew Sep 27, 2025
68d99ec
read docs fix
Sep 30, 2025
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
8 changes: 8 additions & 0 deletions tests/entrypoints/openai/test_response_api_with_harmony.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -379,6 +379,14 @@ async def test_streaming(client: OpenAI, model_name: str, background: bool):
if event.type == "response.created":
resp_id = event.response.id

# test vllm custom types are in the response
if event.type in [
"response.completed", "response.in_progress",
"response.created"
]:
assert 'input_messages' in event.response.model_extra
assert 'output_messages' in event.response.model_extra

if current_event_mode != event.type:
current_event_mode = event.type
print(f"\n[{event.type}] ", end="", flush=True)
Expand Down
50 changes: 38 additions & 12 deletions vllm/entrypoints/openai/protocol.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -17,20 +17,32 @@
ChatCompletionAudio as OpenAIChatCompletionAudio)
from openai.types.chat.chat_completion_message import (
Annotation as OpenAIAnnotation)
# yapf: enable
from openai.types.responses import (
ResponseCodeInterpreterCallCodeDeltaEvent,
ResponseCodeInterpreterCallCodeDoneEvent,
ResponseCodeInterpreterCallCompletedEvent,
ResponseCodeInterpreterCallInProgressEvent,
ResponseCodeInterpreterCallInterpretingEvent, ResponseCompletedEvent,
ResponseContentPartAddedEvent, ResponseContentPartDoneEvent,
ResponseCreatedEvent, ResponseFunctionToolCall, ResponseInProgressEvent,
ResponseInputItemParam, ResponseOutputItem, ResponseOutputItemAddedEvent,
ResponseOutputItemDoneEvent, ResponsePrompt, ResponseReasoningItem,
ResponseReasoningTextDeltaEvent, ResponseReasoningTextDoneEvent,
ResponseStatus, ResponseWebSearchCallCompletedEvent,
ResponseWebSearchCallInProgressEvent, ResponseWebSearchCallSearchingEvent)
ResponseCodeInterpreterCallInterpretingEvent)
from openai.types.responses import (
ResponseCompletedEvent as OpenAIResponseCompletedEvent)
from openai.types.responses import (ResponseContentPartAddedEvent,
ResponseContentPartDoneEvent)
from openai.types.responses import (
ResponseCreatedEvent as OpenAIResponseCreatedEvent)
from openai.types.responses import ResponseFunctionToolCall
from openai.types.responses import (
ResponseInProgressEvent as OpenAIResponseInProgressEvent)
from openai.types.responses import (ResponseInputItemParam, ResponseOutputItem,
ResponseOutputItemAddedEvent,
ResponseOutputItemDoneEvent,
ResponsePrompt, ResponseReasoningItem,
ResponseReasoningTextDeltaEvent,
ResponseReasoningTextDoneEvent,
ResponseStatus,
ResponseWebSearchCallCompletedEvent,
ResponseWebSearchCallInProgressEvent,
ResponseWebSearchCallSearchingEvent)
# yapf: enable
Copy link
Copy Markdown
Collaborator

@yeqcharlotte yeqcharlotte Sep 24, 2025

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

is the format change intentional

Copy link
Copy Markdown
Contributor Author

@qandrew qandrew Sep 26, 2025

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

yep, yapf/isort had a conflict, it seems like it happens pretty often...

Copy link
Copy Markdown
Collaborator

@yeqcharlotte yeqcharlotte Sep 26, 2025

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

can we keep the original format?

Copy link
Copy Markdown
Contributor Author

@qandrew qandrew Sep 26, 2025

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

@yeqcharlotte we can't. The introduction of below makes the conflict, i tried to but there will be conflicts. You can see in serving_responses.py below # yapf conflicts with isort for this block happens often

from openai.types.responses import (
    ResponseInProgressEvent as OpenAIResponseInProgressEvent)

from openai.types.responses.response_reasoning_item import (
Content as ResponseReasoningTextContent)

Expand Down Expand Up @@ -2077,10 +2089,24 @@ class ResponseReasoningPartAddedEvent(OpenAIBaseModel):
"""The type of the event. Always `response.reasoning_part.added`."""


# vLLM Streaming Events
# Note: we override the response type with the vLLM ResponsesResponse type
class ResponseCompletedEvent(OpenAIResponseCompletedEvent):
response: ResponsesResponse # type: ignore[override]


class ResponseCreatedEvent(OpenAIResponseCreatedEvent):
response: ResponsesResponse # type: ignore[override]


class ResponseInProgressEvent(OpenAIResponseInProgressEvent):
response: ResponsesResponse # type: ignore[override]


StreamingResponsesResponse: TypeAlias = Union[
ResponseCreatedEvent,
ResponseInProgressEvent,
ResponseCompletedEvent,
"ResponseCreatedEvent",
"ResponseInProgressEvent",
"ResponseCompletedEvent",
ResponseOutputItemAddedEvent,
ResponseOutputItemDoneEvent,
ResponseContentPartAddedEvent,
Expand Down
23 changes: 13 additions & 10 deletions vllm/entrypoints/openai/serving_responses.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -22,16 +22,16 @@
ResponseCodeInterpreterCallCompletedEvent,
ResponseCodeInterpreterCallInProgressEvent,
ResponseCodeInterpreterCallInterpretingEvent,
ResponseCodeInterpreterToolCallParam, ResponseCompletedEvent,
ResponseContentPartAddedEvent, ResponseContentPartDoneEvent,
ResponseCreatedEvent, ResponseFunctionToolCall, ResponseFunctionWebSearch,
ResponseInProgressEvent, ResponseOutputItem, ResponseOutputItemAddedEvent,
ResponseOutputItemDoneEvent, ResponseOutputMessage, ResponseOutputText,
ResponseReasoningItem, ResponseReasoningTextDeltaEvent,
ResponseReasoningTextDoneEvent, ResponseStatus, ResponseTextDeltaEvent,
ResponseTextDoneEvent, ResponseWebSearchCallCompletedEvent,
ResponseWebSearchCallInProgressEvent, ResponseWebSearchCallSearchingEvent,
response_function_web_search, response_text_delta_event)
ResponseCodeInterpreterToolCallParam, ResponseContentPartAddedEvent,
ResponseContentPartDoneEvent, ResponseFunctionToolCall,
ResponseFunctionWebSearch, ResponseOutputItem,
ResponseOutputItemAddedEvent, ResponseOutputItemDoneEvent,
ResponseOutputMessage, ResponseOutputText, ResponseReasoningItem,
ResponseReasoningTextDeltaEvent, ResponseReasoningTextDoneEvent,
ResponseStatus, ResponseTextDeltaEvent, ResponseTextDoneEvent,
ResponseWebSearchCallCompletedEvent, ResponseWebSearchCallInProgressEvent,
ResponseWebSearchCallSearchingEvent, response_function_web_search,
response_text_delta_event)
from openai.types.responses.response_output_text import (Logprob,
LogprobTopLogprob)
# yapf: enable
Expand All @@ -58,6 +58,9 @@
InputTokensDetails,
OutputTokensDetails,
RequestResponseMetadata,
ResponseCompletedEvent,
ResponseCreatedEvent,
ResponseInProgressEvent,
ResponseReasoningPartAddedEvent,
ResponseReasoningPartDoneEvent,
ResponsesRequest,
Expand Down