phoenix-oss/llama-stack-mirror
0
0
Fork 1
mirror of https://github.com/meta-llama/llama-stack.git synced 2025-12-18 14:39:48 +00:00
Code Issues Projects Releases Packages Wiki Activity Actions

use OpenAIMessageParam for OpenAI message compatibility

Browse source
This commit is contained in:
Aakanksha Duggal 2025-08-13 13:45:04 -04:00
parent 172ef0c9c2
commit 279ac3bc51
2 changed files with 9 additions and 8 deletions
Download patch file Download diff file Expand all files Collapse all files

10
llama_stack/providers/inline/agents/meta_reference/agent_instance.py
View file

@ -50,6 +50,7 @@ from llama_stack.apis.inference import (
CompletionMessage,
Inference,
Message,
OpenAIMessageParam,
SamplingParams,
StopReason,
SystemMessage,
@ -516,18 +517,13 @@ class ChatAgent(ShieldRunnerMixin):
if self.agent_config.name:
span.set_attribute("agent_name", self.agent_config.name)
# Convert messages to OpenAI format
openai_messages = []
openai_messages: list[OpenAIMessageParam] = []
for message in input_messages:
openai_message = await convert_message_to_openai_dict_new(message)
openai_messages.append(openai_message)
# Convert tool definitions to OpenAI format
openai_tools = None
if self.tool_defs:
openai_tools = []
for tool_def in self.tool_defs:
openai_tool = convert_tooldef_to_openai_tool(tool_def)
openai_tools.append(openai_tool)
openai_tools = [convert_tooldef_to_openai_tool(x) for x in (self.tool_defs or [])]
# Extract tool_choice from tool_config for OpenAI compatibility
# Note: tool_choice can only be provided when tools are also provided

5
llama_stack/providers/utils/inference/openai_compat.py
View file

@ -1267,6 +1267,11 @@ async def prepare_openai_completion_params(**params):
elif isinstance(value, dict):
new_value = {k: await _prepare_value(v) for k, v in value.items()}
elif isinstance(value, BaseModel):
# Special handling for OpenAIMessageParam, preserve as Pydantic objects
if hasattr(value, 'role') and hasattr(value, 'content'):
new_value = value
else:
# Other BaseModel objects get converted to dicts
new_value = value.model_dump(exclude_none=True)
return new_value