phoenix-oss/llama-stack
1
0
Fork 0
forked from phoenix-oss/llama-stack-mirror
Code Pull requests Releases Packages 2 Activity Actions

fix(tools): do not index tools, only index toolgroups (#2261)

Browse source 
When registering a MCP endpoint, we cannot list tools (like we used to)
since the MCP endpoint may be behind an auth wall. Registration can
happen much sooner (via run.yaml).

Instead, we do listing only when the _user_ actually calls listing.
Furthermore, we cache the list in-memory in the server. Currently, the
cache is not invalidated -- we may want to periodically re-list for MCP
servers. Note that they must call `list_tools` before calling
`invoke_tool` -- we use this critically.

This will enable us to list MCP servers in run.yaml

## Test Plan

Existing tests, updated tests accordingly.
This commit is contained in:
Ashwin Bharambe 2025-05-25 13:27:52 -07:00 • committed by GitHub
parent 5a422e236c
commit ce33d02443
No known key found for this signature in database
GPG key ID: B5690EEEBB952194
19 changed files with 131 additions and 153 deletions
Download patch file Download diff file Expand all files Collapse all files

8
llama_stack/providers/datatypes.py
View file

@ -16,7 +16,7 @@ from llama_stack.apis.datatypes import Api
from llama_stack.apis.models import Model
from llama_stack.apis.scoring_functions import ScoringFn
from llama_stack.apis.shields import Shield
from llama_stack.apis.tools import Tool
from llama_stack.apis.tools import ToolGroup
from llama_stack.apis.vector_dbs import VectorDB
from llama_stack.schema_utils import json_schema_type
@ -74,10 +74,10 @@ class BenchmarksProtocolPrivate(Protocol):
async def register_benchmark(self, benchmark: Benchmark) -> None: ...
class ToolsProtocolPrivate(Protocol):
async def register_tool(self, tool: Tool) -> None: ...
class ToolGroupsProtocolPrivate(Protocol):
async def register_toolgroup(self, toolgroup: ToolGroup) -> None: ...
async def unregister_tool(self, tool_id: str) -> None: ...
async def unregister_toolgroup(self, toolgroup_id: str) -> None: ...
@json_schema_type

10
llama_stack/providers/inline/tool_runtime/rag/memory.py
View file

@ -25,14 +25,14 @@ from llama_stack.apis.tools import (
RAGQueryConfig,
RAGQueryResult,
RAGToolRuntime,
Tool,
ToolDef,
ToolGroup,
ToolInvocationResult,
ToolParameter,
ToolRuntime,
)
from llama_stack.apis.vector_io import QueryChunksResponse, VectorIO
from llama_stack.providers.datatypes import ToolsProtocolPrivate
from llama_stack.providers.datatypes import ToolGroupsProtocolPrivate
from llama_stack.providers.utils.inference.prompt_adapter import interleaved_content_as_str
from llama_stack.providers.utils.memory.vector_store import (
content_from_doc,
@ -49,7 +49,7 @@ def make_random_string(length: int = 8):
return "".join(secrets.choice(string.ascii_letters + string.digits) for _ in range(length))
class MemoryToolRuntimeImpl(ToolsProtocolPrivate, ToolRuntime, RAGToolRuntime):
class MemoryToolRuntimeImpl(ToolGroupsProtocolPrivate, ToolRuntime, RAGToolRuntime):
def __init__(
self,
config: RagToolRuntimeConfig,
@ -66,10 +66,10 @@ class MemoryToolRuntimeImpl(ToolsProtocolPrivate, ToolRuntime, RAGToolRuntime):
async def shutdown(self):
pass
async def register_tool(self, tool: Tool) -> None:
async def register_toolgroup(self, toolgroup: ToolGroup) -> None:
pass
async def unregister_tool(self, tool_id: str) -> None:
async def unregister_toolgroup(self, toolgroup_id: str) -> None:
return
async def insert(

10
llama_stack/providers/remote/tool_runtime/bing_search/bing_search.py
View file

@ -12,19 +12,19 @@ import httpx
from llama_stack.apis.common.content_types import URL
from llama_stack.apis.tools import (
ListToolDefsResponse,
Tool,
ToolDef,
ToolGroup,
ToolInvocationResult,
ToolParameter,
ToolRuntime,
)
from llama_stack.distribution.request_headers import NeedsRequestProviderData
from llama_stack.providers.datatypes import ToolsProtocolPrivate
from llama_stack.providers.datatypes import ToolGroupsProtocolPrivate
from .config import BingSearchToolConfig
class BingSearchToolRuntimeImpl(ToolsProtocolPrivate, ToolRuntime, NeedsRequestProviderData):
class BingSearchToolRuntimeImpl(ToolGroupsProtocolPrivate, ToolRuntime, NeedsRequestProviderData):
def __init__(self, config: BingSearchToolConfig):
self.config = config
self.url = "https://api.bing.microsoft.com/v7.0/search"
@ -32,10 +32,10 @@ class BingSearchToolRuntimeImpl(ToolsProtocolPrivate, ToolRuntime, NeedsRequestP
async def initialize(self):
pass
async def register_tool(self, tool: Tool) -> None:
async def register_toolgroup(self, toolgroup: ToolGroup) -> None:
pass
async def unregister_tool(self, tool_id: str) -> None:
async def unregister_toolgroup(self, toolgroup_id: str) -> None:
return
def _get_api_key(self) -> str:

10
llama_stack/providers/remote/tool_runtime/brave_search/brave_search.py
View file

@ -11,30 +11,30 @@ import httpx
from llama_stack.apis.common.content_types import URL
from llama_stack.apis.tools import (
ListToolDefsResponse,
Tool,
ToolDef,
ToolGroup,
ToolInvocationResult,
ToolParameter,
ToolRuntime,
)
from llama_stack.distribution.request_headers import NeedsRequestProviderData
from llama_stack.models.llama.datatypes import BuiltinTool
from llama_stack.providers.datatypes import ToolsProtocolPrivate
from llama_stack.providers.datatypes import ToolGroupsProtocolPrivate
from .config import BraveSearchToolConfig
class BraveSearchToolRuntimeImpl(ToolsProtocolPrivate, ToolRuntime, NeedsRequestProviderData):
class BraveSearchToolRuntimeImpl(ToolGroupsProtocolPrivate, ToolRuntime, NeedsRequestProviderData):
def __init__(self, config: BraveSearchToolConfig):
self.config = config
async def initialize(self):
pass
async def register_tool(self, tool: Tool) -> None:
async def register_toolgroup(self, toolgroup: ToolGroup) -> None:
pass
async def unregister_tool(self, tool_id: str) -> None:
async def unregister_toolgroup(self, toolgroup_id: str) -> None:
return
def _get_api_key(self) -> str:

11
llama_stack/providers/remote/tool_runtime/model_context_protocol/model_context_protocol.py
View file

@ -11,12 +11,13 @@ from llama_stack.apis.common.content_types import URL
from llama_stack.apis.datatypes import Api
from llama_stack.apis.tools import (
ListToolDefsResponse,
ToolGroup,
ToolInvocationResult,
ToolRuntime,
)
from llama_stack.distribution.request_headers import NeedsRequestProviderData
from llama_stack.log import get_logger
from llama_stack.providers.datatypes import ToolsProtocolPrivate
from llama_stack.providers.datatypes import ToolGroupsProtocolPrivate
from llama_stack.providers.utils.tools.mcp import convert_header_list_to_dict, invoke_mcp_tool, list_mcp_tools
from .config import MCPProviderConfig
@ -24,13 +25,19 @@ from .config import MCPProviderConfig
logger = get_logger(__name__, category="tools")
class ModelContextProtocolToolRuntimeImpl(ToolsProtocolPrivate, ToolRuntime, NeedsRequestProviderData):
class ModelContextProtocolToolRuntimeImpl(ToolGroupsProtocolPrivate, ToolRuntime, NeedsRequestProviderData):
def __init__(self, config: MCPProviderConfig, _deps: dict[Api, Any]):
self.config = config
async def initialize(self):
pass
async def register_toolgroup(self, toolgroup: ToolGroup) -> None:
pass
async def unregister_toolgroup(self, toolgroup_id: str) -> None:
return
async def list_runtime_tools(
self, tool_group_id: str | None = None, mcp_endpoint: URL | None = None
) -> ListToolDefsResponse:

10
llama_stack/providers/remote/tool_runtime/tavily_search/tavily_search.py
View file

@ -12,29 +12,29 @@ import httpx
from llama_stack.apis.common.content_types import URL
from llama_stack.apis.tools import (
ListToolDefsResponse,
Tool,
ToolDef,
ToolGroup,
ToolInvocationResult,
ToolParameter,
ToolRuntime,
)
from llama_stack.distribution.request_headers import NeedsRequestProviderData
from llama_stack.providers.datatypes import ToolsProtocolPrivate
from llama_stack.providers.datatypes import ToolGroupsProtocolPrivate
from .config import TavilySearchToolConfig
class TavilySearchToolRuntimeImpl(ToolsProtocolPrivate, ToolRuntime, NeedsRequestProviderData):
class TavilySearchToolRuntimeImpl(ToolGroupsProtocolPrivate, ToolRuntime, NeedsRequestProviderData):
def __init__(self, config: TavilySearchToolConfig):
self.config = config
async def initialize(self):
pass
async def register_tool(self, tool: Tool) -> None:
async def register_toolgroup(self, toolgroup: ToolGroup) -> None:
pass
async def unregister_tool(self, tool_id: str) -> None:
async def unregister_toolgroup(self, toolgroup_id: str) -> None:
return
def _get_api_key(self) -> str:

10
llama_stack/providers/remote/tool_runtime/wolfram_alpha/wolfram_alpha.py
View file

@ -12,19 +12,19 @@ import httpx
from llama_stack.apis.common.content_types import URL
from llama_stack.apis.tools import (
ListToolDefsResponse,
Tool,
ToolDef,
ToolGroup,
ToolInvocationResult,
ToolParameter,
ToolRuntime,
)
from llama_stack.distribution.request_headers import NeedsRequestProviderData
from llama_stack.providers.datatypes import ToolsProtocolPrivate
from llama_stack.providers.datatypes import ToolGroupsProtocolPrivate
from .config import WolframAlphaToolConfig
class WolframAlphaToolRuntimeImpl(ToolsProtocolPrivate, ToolRuntime, NeedsRequestProviderData):
class WolframAlphaToolRuntimeImpl(ToolGroupsProtocolPrivate, ToolRuntime, NeedsRequestProviderData):
def __init__(self, config: WolframAlphaToolConfig):
self.config = config
self.url = "https://api.wolframalpha.com/v2/query"
@ -32,10 +32,10 @@ class WolframAlphaToolRuntimeImpl(ToolsProtocolPrivate, ToolRuntime, NeedsReques
async def initialize(self):
pass
async def register_tool(self, tool: Tool) -> None:
async def register_toolgroup(self, toolgroup: ToolGroup) -> None:
pass
async def unregister_tool(self, tool_id: str) -> None:
async def unregister_toolgroup(self, toolgroup_id: str) -> None:
return
def _get_api_key(self) -> str: