Remove unused client parameter from prepare_tools

Author: Jazzcort

src/utils/responses.py

@@ -81,7 +81,6 @@
 from llama_stack_client import APIConnectionError, APIStatusError, AsyncLlamaStackClient
 
 import constants
-from client import AsyncLlamaStackClientHolder
 from configuration import configuration
 from constants import DEFAULT_RAG_TOOL
 from log import get_logger
@@ -219,8 +218,7 @@ async def maybe_get_topic_summary(
     return await get_topic_summary(input_text, client, model_id)
 
 
-async def prepare_tools(  # pylint: disable=too-many-arguments,too-many-positional-arguments,unused-argument
-    client: AsyncLlamaStackClient,
+async def prepare_tools(  # pylint: disable=too-many-arguments,too-many-positional-arguments
     vector_store_ids: Optional[list[str]],
     no_tools: Optional[bool],
     token: str,
@@ -230,7 +228,6 @@ async def prepare_tools(  # pylint: disable=too-many-arguments,too-many-position
     """Prepare tools for Responses API including RAG and MCP tools.
 
     Args:
-        client: The Llama Stack client instance
         vector_store_ids: The list of vector store IDs to use for RAG tools
             or None to fall back to rag.tool configuration
         no_tools: Whether to skip tool preparation
@@ -360,7 +357,6 @@ async def prepare_responses_params(  # pylint: disable=too-many-arguments,too-ma
 
     # Prepare tools for responses API
     tools = await prepare_tools(
-        client,
         query_request.vector_store_ids,
         query_request.no_tools,
         token,
@@ -1739,9 +1735,7 @@ async def _resolve_client_tools(
 
     # Optionally merge server-configured tools (RAG, MCP) with client tools
     if merge_server_tools:
-        client = AsyncLlamaStackClientHolder().get_client()
         server_tools = await prepare_tools(
-            client=client,
             vector_store_ids=vector_store_ids,
             no_tools=False,
             token=token,
@@ -1769,9 +1763,7 @@ async def _resolve_server_tools(
     Returns:
         List of server-configured tools, or None if none are configured.
     """
-    client = AsyncLlamaStackClientHolder().get_client()
     return await prepare_tools(
-        client=client,
         vector_store_ids=None,  # allow all vector stores configured
         no_tools=False,
         token=token,
@@ -1815,9 +1807,7 @@ async def resolve_tool_choice(
 
     if tools is None:
         # Register all tools configured in LCORE configuration
-        client = AsyncLlamaStackClientHolder().get_client()
         prepared_tools = await prepare_tools(
-            client=client,
             vector_store_ids=None,  # allow all vector stores configured
             no_tools=False,
             token=token,

tests/integration/endpoints/test_query_integration.py

@@ -4,7 +4,6 @@
 # pylint: disable=too-many-arguments  # Integration tests need many fixtures
 # pylint: disable=too-many-positional-arguments  # Integration tests need many fixtures
 
-
 import pytest
 from fastapi import HTTPException, Request, status
 from llama_stack_client import APIConnectionError
@@ -707,18 +706,10 @@ async def test_query_v2_endpoint_uses_tools_when_available(
     -------
         None
     """
-    _ = test_config
+    # prepare_tools does not require llama-stack client anymore so the way to enable RAG tools is through config
+    test_config.rag.tool = ["vs-test-123"]
     _ = patch_db_session
 
-    # Mock vector stores to be available (simulating RAG tools)
-    mock_vector_store = mocker.MagicMock()
-    mock_vector_store.id = "vs-test-123"
-
-    mock_list_result = mocker.MagicMock()
-    mock_list_result.data = [mock_vector_store]
-
-    mock_llama_stack_client.vector_stores.list.return_value = mock_list_result
-
     query_request = QueryRequest(query="What is Ansible?", no_tools=False)
 
     response = await query_endpoint_handler(

tests/integration/endpoints/test_responses_byok_integration.py

@@ -79,7 +79,6 @@ def _patch_all_client_holders(mocker: MockerFixture, mock_client: Any) -> None:
     for module in (
         "app.endpoints.responses",
         "utils.endpoints",
-        "utils.responses",
     ):
         holder = mocker.patch(f"{module}.AsyncLlamaStackClientHolder")
         holder.return_value.get_client.return_value = mock_client

tests/unit/app/endpoints/test_responses.py

@@ -122,12 +122,6 @@ def _patch_base(mocker: MockerFixture, config: AppConfig) -> None:
     mocker.patch(f"{MODULE}.check_configuration_loaded")
     mocker.patch(f"{MODULE}.check_tokens_available")
     mocker.patch(f"{MODULE}.validate_model_provider_override")
-    mock_holder = mocker.Mock()
-    mock_holder.get_client.return_value = mocker.Mock()
-    mocker.patch(
-        f"{UTILS_RESPONSES_MODULE}.AsyncLlamaStackClientHolder",
-        return_value=mock_holder,
-    )
     mocker.patch(
         f"{UTILS_RESPONSES_MODULE}.prepare_tools",
         new=mocker.AsyncMock(return_value=None),