xinyun/vllm
1
0
Fork 0
mirror of https://git.datalinker.icu/vllm-project/vllm.git synced 2026-05-12 00:04:44 +08:00
Code Issues Packages Projects Releases Wiki Activity

[gpt-oss] Streaming Output for Python Tool (#23409)

Browse Source 
Signed-off-by: zjy0516 <riverclouds.zhu@qq.com>
This commit is contained in:
Jiangyun Zhu 2025-08-24 12:42:38 +08:00 committed by GitHub
parent 65197a5fb3
commit c55c028998
No known key found for this signature in database
GPG Key ID: B5690EEEBB952194
1 changed files with 44 additions and 26 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

70
vllm/entrypoints/openai/serving_responses.py
View File

@ -1069,7 +1069,48 @@ class OpenAIServingResponses(OpenAIServing):
delta=ctx.parser.last_content_delta, delta=ctx.parser.last_content_delta,
sequence_number=-1, sequence_number=-1,
)) ))
# built-in tools will be triggered on the analysis channel
# However, occasionally built-in tools will
# still be output to commentary.
elif (ctx.parser.current_channel == "commentary"
or ctx.parser.current_channel == "analysis"
) and ctx.parser.current_recipient == "python":
if not sent_output_item_added:
sent_output_item_added = True
yield _send_event(
openai_responses_types.
ResponseOutputItemAddedEvent(
type="response.output_item.added",
sequence_number=-1,
output_index=current_output_index,
item=openai_responses_types.
ResponseCodeInterpreterToolCallParam(
type="code_interpreter_call",
id=current_item_id,
code=None,
container_id="auto",
outputs=None,
status="in_progress",
),
))
yield _send_event(
openai_responses_types.
ResponseCodeInterpreterCallInProgressEvent(
type=
"response.code_interpreter_call.in_progress",
sequence_number=-1,
output_index=current_output_index,
item_id=current_item_id,
))
yield _send_event(
openai_responses_types.
ResponseCodeInterpreterCallCodeDeltaEvent(
type="response.code_interpreter_call_code.delta",
sequence_number=-1,
output_index=current_output_index,
item_id=current_item_id,
delta=ctx.parser.last_content_delta,
))
if ctx.is_assistant_action_turn() and len(ctx.parser.messages) > 0: if ctx.is_assistant_action_turn() and len(ctx.parser.messages) > 0:
previous_item = ctx.parser.messages[-1] previous_item = ctx.parser.messages[-1]
if (self.tool_server is not None if (self.tool_server is not None
@ -1165,30 +1206,6 @@ class OpenAIServingResponses(OpenAIServing):
and self.tool_server.has_tool("python") and self.tool_server.has_tool("python")
and previous_item.recipient is not None and previous_item.recipient is not None
and previous_item.recipient.startswith("python")): and previous_item.recipient.startswith("python")):
yield _send_event(
openai_responses_types.ResponseOutputItemAddedEvent(
type="response.output_item.added",
sequence_number=-1,
output_index=current_output_index,
item=openai_responses_types.
ResponseCodeInterpreterToolCallParam(
type="code_interpreter_call",
id=current_item_id,
code="",
container_id="auto",
outputs=[],
status="in_progress",
),
))
yield _send_event(
openai_responses_types.
ResponseCodeInterpreterCallInProgressEvent(
type="response.code_interpreter_call.in_progress",
sequence_number=-1,
output_index=current_output_index,
item_id=current_item_id,
))
# TODO: do we need to add delta event here?
yield _send_event( yield _send_event(
openai_responses_types. openai_responses_types.
ResponseCodeInterpreterCallCodeDoneEvent( ResponseCodeInterpreterCallCodeDoneEvent(
@ -1196,7 +1213,8 @@ class OpenAIServingResponses(OpenAIServing):
sequence_number=-1, sequence_number=-1,
output_index=current_output_index, output_index=current_output_index,
item_id=current_item_id, item_id=current_item_id,
code=previous_item.content[0].text)) code=previous_item.content[0].text,
))
yield _send_event( yield _send_event(
openai_responses_types. openai_responses_types.
ResponseCodeInterpreterCallInterpretingEvent( ResponseCodeInterpreterCallInterpretingEvent(