phoenix-oss/llama-stack-mirror
0
0
Fork 1
mirror of https://github.com/meta-llama/llama-stack.git synced 2025-08-02 08:44:44 +00:00
Code Issues Projects Releases Packages Wiki Activity Actions

fix tool calling by not relying on finish reason but tool_calls

Browse source
This commit is contained in:
Ashwin Bharambe 2025-04-28 13:39:50 -07:00
parent a1524390b9
commit 7323d8e86f
2 changed files with 6 additions and 3 deletions
Download patch file Download diff file Expand all files Collapse all files

6
llama_stack/providers/inline/agents/meta_reference/openai_responses.py
View file

@ -143,7 +143,7 @@ class OpenAIResponsesImpl:
stream=stream,
)
if isinstance(chat_response, AsyncIterator):
if stream:
# TODO: refactor this into a separate method that handles streaming
chat_response_id = ""
chat_response_content = []
@ -175,10 +175,10 @@ class OpenAIResponsesImpl:
)
else:
# dump and reload to map to our pydantic types
chat_response = OpenAIChatCompletion.model_validate_json(chat_response.model_dump_json())
chat_response = OpenAIChatCompletion(**chat_response.model_dump())
output_messages: List[OpenAIResponseOutput] = []
if chat_response.choices[0].finish_reason == "tool_calls":
if chat_response.choices[0].message.tool_calls:
output_messages.extend(
await self._execute_tool_and_return_final_output(model, stream, chat_response, messages)
)

3
tests/verifications/openai-api-verification-run.yaml
View file

@ -1,3 +1,6 @@
# This is a temporary run file because model names used by the verification tests
# are not quite consistent with various pre-existing distributions.
#
version: '2'
image_name: openai-api-verification
apis: