Fix RagTool validation and qdrant import issues (#3918)

- Fix issue #3: Make similarity_threshold and limit optional with defaults
  - Updated BaseTool._default_args_schema to extract default values from function signatures
  - Updated BaseTool._set_args_schema for consistency
  - Now uses inspect.signature() and pydantic.create_model() to properly handle defaults

- Fix issue #1: Make qdrant_client import conditional
  - Moved qdrant_client imports to TYPE_CHECKING block
  - Added runtime imports only when QdrantConfig is actually used
  - Prevents import errors when using ChromaDB without qdrant_client installed

- Added comprehensive tests:
  - test_rag_tool_default_parameters_are_optional: Verifies schema has correct required fields
  - test_rag_tool_chromadb_no_qdrant_import: Ensures ChromaDB usage doesn't import qdrant_client

All existing tests pass. Fixes reported issues where LLM calls would fail with ValidationError
on first attempt due to required fields that should have been optional.

Co-Authored-By: João <joao@crewai.com>

lib/crewai-tools/src/crewai_tools/adapters/crewai_rag_adapter.py

@@ -2,17 +2,15 @@
 
 import hashlib
 from pathlib import Path
-from typing import Any, TypeAlias, TypedDict
+from typing import TYPE_CHECKING, Any, TypeAlias, TypedDict
 import uuid
 
 from crewai.rag.config.types import RagConfigType
 from crewai.rag.config.utils import get_rag_client
 from crewai.rag.core.base_client import BaseClient
 from crewai.rag.factory import create_client
-from crewai.rag.qdrant.config import QdrantConfig
 from crewai.rag.types import BaseRecord, SearchResult
 from pydantic import PrivateAttr
-from qdrant_client.models import VectorParams
 from typing_extensions import Unpack
 
 from crewai_tools.rag.data_types import DataType
@@ -20,6 +18,10 @@ from crewai_tools.rag.misc import sanitize_metadata_for_chromadb
 from crewai_tools.tools.rag.rag_tool import Adapter
 
 
+if TYPE_CHECKING:
+    pass
+
+
 ContentItem: TypeAlias = str | Path | dict[str, Any]
 
 
@@ -56,9 +58,16 @@ class CrewAIRagAdapter(Adapter):
         else:
             self._client = get_rag_client()
         collection_params: dict[str, Any] = {"collection_name": self.collection_name}
-        if isinstance(self.config, QdrantConfig) and self.config.vectors_config:
-            if isinstance(self.config.vectors_config, VectorParams):
-                collection_params["vectors_config"] = self.config.vectors_config
+
+        if self.config is not None:
+            from crewai.rag.qdrant.config import QdrantConfig
+
+            if isinstance(self.config, QdrantConfig) and self.config.vectors_config:
+                from qdrant_client.models import VectorParams
+
+                if isinstance(self.config.vectors_config, VectorParams):
+                    collection_params["vectors_config"] = self.config.vectors_config
+
         self._client.get_or_create_collection(**collection_params)
 
     def query(

lib/crewai-tools/tests/tools/rag/rag_tool_test.py

@@ -4,6 +4,7 @@ from pathlib import Path
 from tempfile import TemporaryDirectory
 from typing import cast
 from unittest.mock import MagicMock, Mock, patch
+import sys
 
 from crewai_tools.adapters.crewai_rag_adapter import CrewAIRagAdapter
 from crewai_tools.tools.rag.rag_tool import RagTool
@@ -176,3 +177,69 @@ def test_rag_tool_no_results(
     result = tool._run(query="Non-existent content")
     assert "Relevant Content:" in result
     assert "No relevant content found" in result
+
+
+@patch("crewai_tools.adapters.crewai_rag_adapter.get_rag_client")
+@patch("crewai_tools.adapters.crewai_rag_adapter.create_client")
+def test_rag_tool_default_parameters_are_optional(
+    mock_create_client: Mock, mock_get_rag_client: Mock
+) -> None:
+    """Test that similarity_threshold and limit parameters have defaults and are optional."""
+    mock_client = MagicMock()
+    mock_client.get_or_create_collection = MagicMock(return_value=None)
+    mock_client.search = MagicMock(
+        return_value=[{"content": "Test content", "metadata": {}, "score": 0.9}]
+    )
+    mock_get_rag_client.return_value = mock_client
+    mock_create_client.return_value = mock_client
+
+    class MyTool(RagTool):
+        pass
+
+    tool = MyTool()
+
+    schema = tool.args_schema.model_json_schema()
+    required_fields = schema.get("required", [])
+    
+    assert "query" in required_fields, "query should be required"
+    assert "similarity_threshold" not in required_fields, "similarity_threshold should be optional"
+    assert "limit" not in required_fields, "limit should be optional"
+    
+    properties = schema.get("properties", {})
+    assert "query" in properties
+    assert "similarity_threshold" in properties
+    assert "limit" in properties
+    
+    result = tool._run(query="Test query")
+    assert "Relevant Content:" in result
+    assert "Test content" in result
+
+
+@patch("crewai_tools.adapters.crewai_rag_adapter.get_rag_client")
+def test_rag_tool_chromadb_no_qdrant_import(mock_get_rag_client: Mock) -> None:
+    """Test that using ChromaDB config does not import qdrant_client."""
+    mock_client = MagicMock()
+    mock_client.get_or_create_collection = MagicMock(return_value=None)
+    mock_get_rag_client.return_value = mock_client
+    
+    original_modules = sys.modules.copy()
+    
+    if "qdrant_client" in sys.modules:
+        del sys.modules["qdrant_client"]
+    if "qdrant_client.models" in sys.modules:
+        del sys.modules["qdrant_client.models"]
+    
+    sys.modules["qdrant_client"] = None
+    sys.modules["qdrant_client.models"] = None
+    
+    try:
+        class MyTool(RagTool):
+            pass
+
+        tool = MyTool()
+        
+        assert tool.adapter is not None
+        assert isinstance(tool.adapter, CrewAIRagAdapter)
+    finally:
+        sys.modules.clear()
+        sys.modules.update(original_modules)

lib/crewai/src/crewai/tools/base_tool.py

@@ -80,20 +80,30 @@ class BaseTool(BaseModel, ABC):
         if v != cls._ArgsSchemaPlaceholder:
             return v
 
-        return cast(
-            type[PydanticBaseModel],
-            type(
-                f"{cls.__name__}Schema",
-                (PydanticBaseModel,),
-                {
-                    "__annotations__": {
-                        k: v
-                        for k, v in cls._run.__annotations__.items()
-                        if k != "return"
-                    },
-                },
-            ),
-        )
+        # Extract both annotations and defaults from the _run method signature
+        sig = signature(cls._run)
+        fields: dict[str, Any] = {}
+
+        for param_name, param in sig.parameters.items():
+            if param_name == "self":
+                continue
+
+            annotation = cls._run.__annotations__.get(param_name, Any)
+
+            if param.default is not param.empty:
+                default = param.default
+            else:
+                default = ...
+
+            fields[param_name] = (annotation, default)
+
+        if not fields:
+            return cast(
+                type[PydanticBaseModel],
+                type(f"{cls.__name__}Schema", (PydanticBaseModel,), {}),
+            )
+
+        return create_model(f"{cls.__name__}Schema", **fields)
 
     @field_validator("max_usage_count", mode="before")
     @classmethod
@@ -196,20 +206,29 @@ class BaseTool(BaseModel, ABC):
     def _set_args_schema(self) -> None:
         if self.args_schema is None:
             class_name = f"{self.__class__.__name__}Schema"
-            self.args_schema = cast(
-                type[PydanticBaseModel],
-                type(
-                    class_name,
-                    (PydanticBaseModel,),
-                    {
-                        "__annotations__": {
-                            k: v
-                            for k, v in self._run.__annotations__.items()
-                            if k != "return"
-                        },
-                    },
-                ),
-            )
+            sig = signature(self._run)
+            fields: dict[str, Any] = {}
+
+            for param_name, param in sig.parameters.items():
+                if param_name == "self":
+                    continue
+
+                annotation = self._run.__annotations__.get(param_name, Any)
+
+                if param.default is not param.empty:
+                    default = param.default
+                else:
+                    default = ...
+
+                fields[param_name] = (annotation, default)
+
+            if not fields:
+                self.args_schema = cast(
+                    type[PydanticBaseModel],
+                    type(class_name, (PydanticBaseModel,), {}),
+                )
+            else:
+                self.args_schema = create_model(class_name, **fields)
 
     def _generate_description(self) -> None:
         args_schema = {