fix: Agent-level knowledge sources with non-OpenAI embedders

- Remove OpenAI default from KnowledgeStorage
- Add proper embedder config inheritance from crew to agent
- Improve error messaging for missing embedder config
- Add tests for agent-level knowledge sources

Fixes #2164

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

src/crewai/agent.py

@@ -115,7 +115,7 @@ class Agent(BaseAgent):
         default="safe",
         description="Mode for code execution: 'safe' (using Docker) or 'unsafe' (direct execution).",
     )
-    embedder: Optional[Dict[str, Any]] = Field(
+    embedder_config: Optional[Dict[str, Any]] = Field(
         default=None,
         description="Embedder configuration for the agent.",
     )
@@ -150,9 +150,14 @@ class Agent(BaseAgent):
                 if isinstance(self.knowledge_sources, list) and all(
                     isinstance(k, BaseKnowledgeSource) for k in self.knowledge_sources
                 ):
+                    # Use agent's embedder config if provided, otherwise use crew's
+                    embedder_config = self.embedder_config
+                    if not embedder_config and self.crew:
+                        embedder_config = self.crew.embedder_config
+                    
                     self.knowledge = Knowledge(
                         sources=self.knowledge_sources,
-                        embedder=self.embedder,
+                        embedder_config=embedder_config,
                         collection_name=knowledge_agent_name,
                         storage=self.knowledge_storage or None,
                     )

src/crewai/crew.py

@@ -138,7 +138,7 @@ class Crew(BaseModel):
         default=None,
         description="An instance of the UserMemory to be used by the Crew to store/fetch memories of a specific user.",
     )
-    embedder: Optional[dict] = Field(
+    embedder_config: Optional[dict] = Field(
         default=None,
         description="Configuration for the embedder to be used for the crew.",
     )
@@ -308,7 +308,7 @@ class Crew(BaseModel):
                 ):
                     self.knowledge = Knowledge(
                         sources=self.knowledge_sources,
-                        embedder=self.embedder,
+                        embedder_config=self.embedder_config,
                         collection_name="crew",
                     )
 

src/crewai/knowledge/knowledge.py

@@ -15,29 +15,30 @@ class Knowledge(BaseModel):
     Args:
         sources: List[BaseKnowledgeSource] = Field(default_factory=list)
         storage: Optional[KnowledgeStorage] = Field(default=None)
-        embedder: Optional[Dict[str, Any]] = None
+        embedder_config: Optional[Dict[str, Any]] = None
     """
 
     sources: List[BaseKnowledgeSource] = Field(default_factory=list)
     model_config = ConfigDict(arbitrary_types_allowed=True)
     storage: Optional[KnowledgeStorage] = Field(default=None)
-    embedder: Optional[Dict[str, Any]] = None
+    embedder_config: Optional[Dict[str, Any]] = None
     collection_name: Optional[str] = None
 
     def __init__(
         self,
         collection_name: str,
         sources: List[BaseKnowledgeSource],
-        embedder: Optional[Dict[str, Any]] = None,
+        embedder_config: Optional[Dict[str, Any]] = None,
         storage: Optional[KnowledgeStorage] = None,
         **data,
     ):
         super().__init__(**data)
+        self.embedder_config = embedder_config
         if storage:
             self.storage = storage
         else:
             self.storage = KnowledgeStorage(
-                embedder=embedder, collection_name=collection_name
+                embedder_config=embedder_config, collection_name=collection_name
             )
         self.sources = sources
         self.storage.initialize_knowledge_storage()

src/crewai/knowledge/storage/knowledge_storage.py

@@ -48,11 +48,11 @@ class KnowledgeStorage(BaseKnowledgeStorage):
 
     def __init__(
         self,
-        embedder: Optional[Dict[str, Any]] = None,
+        embedder_config: Optional[Dict[str, Any]] = None,
         collection_name: Optional[str] = None,
     ):
         self.collection_name = collection_name
-        self._set_embedder_config(embedder)
+        self._set_embedder_config(embedder_config)
 
     def search(
         self,
@@ -179,15 +179,14 @@ class KnowledgeStorage(BaseKnowledgeStorage):
             raise
 
     def _create_default_embedding_function(self):
-        from chromadb.utils.embedding_functions.openai_embedding_function import (
+        raise ValueError(
-            OpenAIEmbeddingFunction,
+            "No embedder configuration provided. Please provide an embedder configuration "
+            "either at the crew level or agent level. You can configure embeddings using "
+            "the 'embedder_config' parameter with providers like 'openai', 'watson', etc. "
+            "Example: embedder_config={'provider': 'openai', 'config': {'api_key': 'your-key'}}"
         )
 
-        return OpenAIEmbeddingFunction(
+    def _set_embedder_config(self, embedder_config: Optional[Dict[str, Any]] = None) -> None:
-            api_key=os.getenv("OPENAI_API_KEY"), model_name="text-embedding-3-small"
-        )
-
-    def _set_embedder_config(self, embedder: Optional[Dict[str, Any]] = None) -> None:
         """Set the embedding configuration for the knowledge storage.
 
         Args:
@@ -195,7 +194,7 @@ class KnowledgeStorage(BaseKnowledgeStorage):
                 If None or empty, defaults to the default embedding function.
         """
         self.embedder = (
-            EmbeddingConfigurator().configure_embedder(embedder)
+            EmbeddingConfigurator().configure_embedder(embedder_config)
-            if embedder
+            if embedder_config
             else self._create_default_embedding_function()
         )

tests/test_agent_knowledge.py

@@ -0,0 +1,94 @@
+import pytest
+from unittest.mock import patch, MagicMock
+
+from chromadb.api.types import EmbeddingFunction
+from crewai import Agent, Crew, Task
+from crewai.knowledge.source.string_knowledge_source import StringKnowledgeSource
+from crewai.process import Process
+
+class MockEmbeddingFunction(EmbeddingFunction):
+    def __call__(self, texts):
+        return [[0.0] * 1536 for _ in texts]
+
+@pytest.fixture(autouse=True)
+def mock_vector_db():
+    """Mock vector database operations."""
+    with patch("crewai.knowledge.storage.knowledge_storage.KnowledgeStorage") as mock, \
+         patch("chromadb.PersistentClient") as mock_chroma:
+        # Mock ChromaDB client and collection
+        mock_collection = MagicMock()
+        mock_collection.query.return_value = {
+            "ids": [["1"]],
+            "distances": [[0.1]],
+            "metadatas": [[{"source": "test"}]],
+            "documents": [["Test content"]]
+        }
+        mock_chroma.return_value.get_or_create_collection.return_value = mock_collection
+        
+        # Mock the query method to return a predefined response
+        instance = mock.return_value
+        instance.query.return_value = [
+            {
+                "context": "Test content",
+                "score": 0.9,
+            }
+        ]
+        instance.reset.return_value = None
+        yield instance
+
+def test_agent_knowledge_with_custom_embedder(mock_vector_db):
+    agent = Agent(
+        role="test role",
+        goal="test goal",
+        backstory="test backstory",
+        knowledge_sources=[StringKnowledgeSource(content="test content")],
+        embedder_config={
+            "provider": "custom",
+            "config": {
+                "embedder": MockEmbeddingFunction()
+            }
+        }
+    )
+    assert agent.knowledge is not None
+    assert agent.knowledge.storage.embedder is not None
+
+def test_agent_inherits_crew_embedder(mock_vector_db):
+    test_agent = Agent(
+        role="test role",
+        goal="test goal",
+        backstory="test backstory"
+    )
+    test_task = Task(
+        description="test task",
+        expected_output="test output",
+        agent=test_agent
+    )
+    crew = Crew(
+        agents=[test_agent],
+        tasks=[test_task],
+        process=Process.sequential,
+        embedder_config={
+            "provider": "custom",
+            "config": {
+                "embedder": MockEmbeddingFunction()
+            }
+        }
+    )
+    agent = Agent(
+        role="test role",
+        goal="test goal",
+        backstory="test backstory",
+        knowledge_sources=[StringKnowledgeSource(content="test content")],
+        crew=crew
+    )
+    assert agent.knowledge is not None
+    assert agent.knowledge.storage.embedder is not None
+
+def test_agent_knowledge_without_embedder_raises_error(mock_vector_db):
+    with pytest.raises(ValueError, match="No embedder configuration provided"):
+        agent = Agent(
+            role="test role",
+            goal="test goal",
+            backstory="test backstory",
+            knowledge_sources=[StringKnowledgeSource(content="test content")]
+        )