phoenix-oss/llama-stack-mirror
0
0
Fork 1
mirror of https://github.com/meta-llama/llama-stack.git synced 2025-10-03 19:57:35 +00:00
Code Issues Projects Releases Packages Wiki Activity Actions

fix: don't pass default response format in Responses

Browse source 
# What does this PR do?


## Test Plan
This commit is contained in:
Eric Huang 2025-09-30 11:33:57 -07:00
parent 6cce553c93
commit f387e4023f
2 changed files with 10 additions and 9 deletions
Download patch file Download diff file Expand all files Collapse all files

5
llama_stack/providers/inline/agents/meta_reference/responses/streaming.py
View file

@ -127,13 +127,16 @@ class StreamingResponseOrchestrator:
messages = self.ctx.messages.copy()
while True:
# Text is the default response format for chat completion so don't need to pass it
# (some providers don't support non-empty response_format when tools are present)
response_format = None if self.ctx.response_format.type == "text" else self.ctx.response_format
completion_result = await self.inference_api.openai_chat_completion(
model=self.ctx.model,
messages=messages,
tools=self.ctx.chat_tools,
stream=True,
temperature=self.ctx.temperature,
response_format=self.ctx.response_format,
response_format=response_format,
)
# Process streaming chunks and build complete response

14
tests/unit/providers/agents/meta_reference/test_openai_responses.py
View file

@ -37,7 +37,6 @@ from llama_stack.apis.inference import (
OpenAIJSONSchema,
OpenAIResponseFormatJSONObject,
OpenAIResponseFormatJSONSchema,
OpenAIResponseFormatText,
OpenAIUserMessageParam,
)
from llama_stack.apis.tools.tools import Tool, ToolGroups, ToolInvocationResult, ToolParameter, ToolRuntime
@ -148,7 +147,7 @@ async def test_create_openai_response_with_string_input(openai_responses_impl, m
mock_inference_api.openai_chat_completion.assert_called_once_with(
model=model,
messages=[OpenAIUserMessageParam(role="user", content="What is the capital of Ireland?", name=None)],
response_format=OpenAIResponseFormatText(),
response_format=None,
tools=None,
stream=True,
temperature=0.1,
@ -823,16 +822,16 @@ async def test_store_response_uses_rehydrated_input_with_previous_response(
@pytest.mark.parametrize(
"text_format, response_format",
[
(OpenAIResponseText(format=OpenAIResponseTextFormat(type="text")), OpenAIResponseFormatText()),
(OpenAIResponseText(format=OpenAIResponseTextFormat(type="text")), None),
(
OpenAIResponseText(format=OpenAIResponseTextFormat(name="Test", schema={"foo": "bar"}, type="json_schema")),
OpenAIResponseFormatJSONSchema(json_schema=OpenAIJSONSchema(name="Test", schema={"foo": "bar"})),
),
(OpenAIResponseText(format=OpenAIResponseTextFormat(type="json_object")), OpenAIResponseFormatJSONObject()),
# ensure text param with no format specified defaults to text
(OpenAIResponseText(format=None), OpenAIResponseFormatText()),
# ensure text param of None defaults to text
(None, OpenAIResponseFormatText()),
# ensure text param with no format specified defaults to None
(OpenAIResponseText(format=None), None),
# ensure text param of None defaults to None
(None, None),
],
)
async def test_create_openai_response_with_text_format(
@ -855,7 +854,6 @@ async def test_create_openai_response_with_text_format(
# Verify
first_call = mock_inference_api.openai_chat_completion.call_args_list[0]
assert first_call.kwargs["messages"][0].content == input_text
assert first_call.kwargs["response_format"] is not None
assert first_call.kwargs["response_format"] == response_format