Fix type error and test issues

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

src/crewai/utilities/embedding_configurator.py

@@ -104,16 +104,25 @@ class EmbeddingConfigurator:
 
     @staticmethod
     def _configure_vertexai(config, model_name):
-        from chromadb.utils.embedding_functions.google_embedding_function import (
-            GoogleVertexEmbeddingFunction,
-        )
+        try:
+            from chromadb.utils.embedding_functions.google_embedding_function import (
+                GoogleVertexEmbeddingFunction,
+            )
 
-        return GoogleVertexEmbeddingFunction(
-            model_name=model_name,
-            api_key=config.get("api_key"),
-            project_id=config.get("project_id"),
-            region=config.get("region"),
-        )
+            from crewai.utilities.embedding_functions import (
+                FixedGoogleVertexEmbeddingFunction,
+            )
+            
+            return FixedGoogleVertexEmbeddingFunction(
+                model_name=model_name,
+                api_key=config.get("api_key"),
+                project_id=config.get("project_id"),
+                region=config.get("region"),
+            )
+        except ImportError as e:
+            raise ImportError(
+                "Google Vertex dependencies are not installed. Please install them to use Vertex embedding."
+            ) from e
 
     @staticmethod
     def _configure_google(config, model_name):

src/crewai/utilities/embedding_functions.py

@@ -0,0 +1,40 @@
+from typing import Any, List, Optional
+from urllib.parse import parse_qs, urlencode, urlparse, urlunparse
+
+import requests
+from chromadb import Documents, Embeddings
+from chromadb.utils.embedding_functions.google_embedding_function import (
+    GoogleVertexEmbeddingFunction,
+)
+
+
+class FixedGoogleVertexEmbeddingFunction(GoogleVertexEmbeddingFunction):
+    """
+    A wrapper around ChromaDB's GoogleVertexEmbeddingFunction that fixes the URL typo
+    where 'publishers/goole' is incorrectly used instead of 'publishers/google'.
+    
+    Issue reference: https://github.com/crewaiinc/crewai/issues/2690
+    """
+    
+    def __init__(self, 
+                model_name: str = "textembedding-gecko", 
+                api_key: Optional[str] = None,
+                **kwargs: Any):
+        api_key_str = "" if api_key is None else api_key
+        super().__init__(model_name=model_name, api_key=api_key_str, **kwargs)
+        
+        self._original_post = requests.post
+        requests.post = self._patched_post
+    
+    def __del__(self):
+        if hasattr(self, '_original_post'):
+            requests.post = self._original_post
+    
+    def _patched_post(self, url, *args, **kwargs):
+        if 'publishers/goole' in url:
+            url = url.replace('publishers/goole', 'publishers/google')
+        
+        return self._original_post(url, *args, **kwargs)
+    
+    def __call__(self, input: Documents) -> Embeddings:
+        return super().__call__(input)

tests/utilities/test_embedding_configurator.py

@@ -0,0 +1,37 @@
+from unittest.mock import MagicMock, patch
+
+import pytest
+
+from crewai.utilities.embedding_configurator import EmbeddingConfigurator
+from crewai.utilities.embedding_functions import FixedGoogleVertexEmbeddingFunction
+
+
+class TestEmbeddingConfigurator:
+    @pytest.fixture
+    def embedding_configurator(self):
+        return EmbeddingConfigurator()
+    
+    def test_configure_vertexai(self, embedding_configurator):
+        with patch('crewai.utilities.embedding_functions.FixedGoogleVertexEmbeddingFunction') as mock_class:
+            mock_instance = MagicMock()
+            mock_class.return_value = mock_instance
+            
+            config = {
+                "provider": "vertexai",
+                "config": {
+                    "api_key": "test-key",
+                    "model": "test-model",
+                    "project_id": "test-project",
+                    "region": "test-region"
+                }
+            }
+            
+            result = embedding_configurator.configure_embedder(config)
+            
+            mock_class.assert_called_once_with(
+                model_name="test-model",
+                api_key="test-key",
+                project_id="test-project",
+                region="test-region"
+            )
+            assert result == mock_instance

tests/utilities/test_embedding_functions.py

@@ -0,0 +1,57 @@
+from unittest.mock import MagicMock, patch
+
+import pytest
+import requests
+
+from crewai.utilities.embedding_functions import FixedGoogleVertexEmbeddingFunction
+
+
+class TestFixedGoogleVertexEmbeddingFunction:
+    @pytest.fixture
+    def embedding_function(self):
+        with patch('requests.post') as mock_post:
+            mock_response = MagicMock()
+            mock_response.json.return_value = {"predictions": [[0.1, 0.2, 0.3]]}
+            mock_post.return_value = mock_response
+            
+            function = FixedGoogleVertexEmbeddingFunction(
+                model_name="test-model",
+                api_key="test-key"
+            )
+            
+            yield function, mock_post
+            
+            if hasattr(function, '_original_post'):
+                requests.post = function._original_post
+            
+    def test_url_correction(self, embedding_function):
+        function, mock_post = embedding_function
+        
+        typo_url = "https://us-central1-aiplatform.googleapis.com/v1/projects/test-project/locations/us-central1/publishers/goole/models/test-model:predict"
+        
+        expected_url = "https://us-central1-aiplatform.googleapis.com/v1/projects/test-project/locations/us-central1/publishers/google/models/test-model:predict"
+        
+        with patch.object(function, '_original_post') as mock_original_post:
+            mock_response = MagicMock()
+            mock_response.json.return_value = {"predictions": [[0.1, 0.2, 0.3]]}
+            mock_original_post.return_value = mock_response
+            
+            response = function._patched_post(typo_url, json={})
+            
+            mock_original_post.assert_called_once()
+            call_args = mock_original_post.call_args
+            assert call_args[0][0] == expected_url
+            
+    def test_embedding_call(self, embedding_function):
+        function, mock_post = embedding_function
+        
+        mock_response = MagicMock()
+        mock_response.json.return_value = {"predictions": [[0.1, 0.2, 0.3]]}
+        mock_post.return_value = mock_response
+        
+        embeddings = function(["test text"])
+        
+        mock_post.assert_called_once()
+        
+        assert isinstance(embeddings, list)
+        assert len(embeddings) > 0