phoenix-oss/llama-stack-mirror
0
0
Fork 1
mirror of https://github.com/meta-llama/llama-stack.git synced 2025-12-12 12:06:04 +00:00
Code Issues Projects Releases Packages Wiki Activity Actions

fix: add traces for tool calls and mcp tool listing

Browse source 
Signed-off-by: Gordon Sim <gsim@redhat.com>
This commit is contained in:
Gordon Sim 2025-10-07 16:47:40 +01:00
parent 696fefbf17
commit 5659c764dd
2 changed files with 40 additions and 19 deletions
Download patch file Download diff file Expand all files Collapse all files

19
llama_stack/providers/inline/agents/meta_reference/responses/streaming.py
View file

@ -46,6 +46,7 @@ from llama_stack.apis.inference import (
OpenAIMessageParam,
)
from llama_stack.log import get_logger
from llama_stack.providers.utils.telemetry import tracing
from .types import ChatCompletionContext, ChatCompletionResult
from .utils import convert_chat_choice_to_response_message, is_function_tool_call
@ -586,14 +587,22 @@ class StreamingResponseOrchestrator:
never_allowed = mcp_tool.allowed_tools.never
# Call list_mcp_tools
tool_defs = await list_mcp_tools(
endpoint=mcp_tool.server_url,
headers=mcp_tool.headers or {},
)
tool_defs = None
list_id = f"mcp_list_{uuid.uuid4()}"
attributes = {
"server_label": mcp_tool.server_label,
"server_url": mcp_tool.server_url,
"mcp_list_tools_id": list_id,
}
async with tracing.span("list_mcp_tools", attributes):
tool_defs = await list_mcp_tools(
endpoint=mcp_tool.server_url,
headers=mcp_tool.headers or {},
)
# Create the MCP list tools message
mcp_list_message = OpenAIResponseOutputMessageMCPListTools(
id=f"mcp_list_{uuid.uuid4()}",
id=list_id,
server_label=mcp_tool.server_label,
tools=[],
)

40
llama_stack/providers/inline/agents/meta_reference/responses/tool_executor.py
View file

@ -35,6 +35,7 @@ from llama_stack.apis.inference import (
from llama_stack.apis.tools import ToolGroups, ToolInvocationResult, ToolRuntime
from llama_stack.apis.vector_io import VectorIO
from llama_stack.log import get_logger
from llama_stack.providers.utils.telemetry import tracing
from .types import ChatCompletionContext, ToolExecutionResult
@ -219,12 +220,18 @@ class ToolExecutor:
from llama_stack.providers.utils.tools.mcp import invoke_mcp_tool
mcp_tool = mcp_tool_to_server[function_name]
result = await invoke_mcp_tool(
endpoint=mcp_tool.server_url,
headers=mcp_tool.headers or {},
tool_name=function_name,
kwargs=tool_kwargs,
)
attributes = {
"server_label": mcp_tool.server_label,
"server_url": mcp_tool.server_url,
"tool_name": function_name,
}
async with tracing.span("invoke_mcp_tool", attributes):
result = await invoke_mcp_tool(
endpoint=mcp_tool.server_url,
headers=mcp_tool.headers or {},
tool_name=function_name,
kwargs=tool_kwargs,
)
elif function_name == "knowledge_search":
response_file_search_tool = next(
(t for t in ctx.response_tools if isinstance(t, OpenAIResponseInputToolFileSearch)),
@ -234,15 +241,20 @@ class ToolExecutor:
# Use vector_stores.search API instead of knowledge_search tool
# to support filters and ranking_options
query = tool_kwargs.get("query", "")
result = await self._execute_knowledge_search_via_vector_store(
query=query,
response_file_search_tool=response_file_search_tool,
)
async with tracing.span("knowledge_search", {}):
result = await self._execute_knowledge_search_via_vector_store(
query=query,
response_file_search_tool=response_file_search_tool,
)
else:
result = await self.tool_runtime_api.invoke_tool(
tool_name=function_name,
kwargs=tool_kwargs,
)
attributes = {
"tool_name": function_name,
}
async with tracing.span("invoke_tool", attributes):
result = await self.tool_runtime_api.invoke_tool(
tool_name=function_name,
kwargs=tool_kwargs,
)
except Exception as e:
error_exc = e