fix generating text embeddings from searhc query

src/services/embedding_service.py

@@ -4,37 +4,52 @@ import os
 from typing import List, Dict, Any, Union, Optional
 import numpy as np
 from PIL import Image
+import vertexai
+from vertexai.vision_models import MultiModalEmbeddingModel
 
 from src.config.config import settings
 
 logger = logging.getLogger(__name__)
 
 class EmbeddingService:
-    """Service for generating image and text embeddings"""
+    """Service for generating image and text embeddings using Vertex AI multimodal model"""
     
     def __init__(self):
         self.model = None
-        self.processor = None
-        self.model_name = "openai/clip-vit-base-patch32"
-        self.device = "cpu"  # Simplified without PyTorch
-        self.embedding_dim = 512  # Dimension of CLIP's embeddings
+        self.model_name = "multimodalembedding@001"
+        self.embedding_dim = 1408  # Vertex AI multimodal embedding dimensions
+        self._initialize_vertex_ai()
+    
+    def _initialize_vertex_ai(self):
+        """Initialize Vertex AI"""
+        try:
+            project_id = os.environ.get('GOOGLE_CLOUD_PROJECT') or settings.FIRESTORE_PROJECT_ID
+            location = os.environ.get('VERTEX_AI_LOCATION', 'us-central1')
+            
+            if project_id:
+                vertexai.init(project=project_id, location=location)
+                logger.info(f"Initialized Vertex AI with project {project_id} in location {location}")
+            else:
+                logger.error("PROJECT_ID not found in environment variables")
+                raise ValueError("Google Cloud Project ID not configured")
+        except Exception as e:
+            logger.error(f"Error initializing Vertex AI: {e}")
+            raise
     
     def _load_model(self):
-        """
-        Load the embedding model if not already loaded
-        """
+        """Load the Vertex AI multimodal embedding model"""
         if self.model is None:
             try:
-                logger.info(f"Loading embedding model on {self.device}")
-                # Placeholder for model loading logic
-                logger.info("Embedding model loaded successfully")
+                logger.info(f"Loading Vertex AI multimodal embedding model: {self.model_name}")
+                self.model = MultiModalEmbeddingModel.from_pretrained(self.model_name)
+                logger.info("Vertex AI multimodal embedding model loaded successfully")
             except Exception as e:
-                logger.error(f"Error loading embedding model: {e}")
+                logger.error(f"Error loading Vertex AI model: {e}")
                 raise
     
     def generate_image_embedding(self, image_data: bytes) -> List[float]:
         """
-        Generate embedding for an image
+        Generate embedding for an image using Vertex AI multimodal model
         
         Args:
             image_data: Binary image data
@@ -45,22 +60,38 @@ class EmbeddingService:
         try:
             self._load_model()
             
-            # Load the image
-            image = Image.open(io.BytesIO(image_data))
+            # Create Vertex AI image object
+            from vertexai.vision_models import Image as VertexImage
+            vertex_image = VertexImage(image_data)
             
-            # Placeholder for image embedding generation
-            # Returns a random normalized vector as placeholder
-            embedding = np.random.randn(self.embedding_dim).astype(np.float32)
-            embedding = embedding / np.linalg.norm(embedding)
+            # Generate image embedding using Vertex AI multimodal model
+            embeddings = self.model.get_embeddings(image=vertex_image)
+            
+            if embeddings is None or embeddings.image_embedding is None:
+                logger.error("Failed to generate image embeddings - no image embedding returned")
+                raise ValueError("Failed to generate image embeddings")
+            
+            # Get the image embedding vector
+            embedding_vector = embeddings.image_embedding
+            
+            # Convert to numpy array and normalize
+            embeddings_array = np.array(embedding_vector, dtype=np.float32)
+            
+            # Normalize the feature vector
+            norm = np.linalg.norm(embeddings_array)
+            if norm > 0:
+                embeddings_array = embeddings_array / norm
+            
+            logger.info(f"Generated image embeddings with shape: {embeddings_array.shape}")
+            return embeddings_array.tolist()
             
-            return embedding.tolist()
         except Exception as e:
             logger.error(f"Error generating image embedding: {e}")
             raise
     
     async def generate_text_embedding(self, text: str) -> Optional[List[float]]:
         """
-        Generate embedding for a text query
+        Generate embedding for a text query using Vertex AI multimodal model
         
         Args:
             text: Text query
@@ -71,12 +102,27 @@ class EmbeddingService:
         try:
             self._load_model()
             
-            # Placeholder for text embedding generation
-            # Returns a random normalized vector as placeholder
-            embedding = np.random.randn(self.embedding_dim).astype(np.float32)
-            embedding = embedding / np.linalg.norm(embedding)
+            # Generate text embedding using Vertex AI multimodal model
+            embeddings = self.model.get_embeddings(contextual_text=text)
+            
+            if embeddings is None or embeddings.text_embedding is None:
+                logger.error("Failed to generate text embeddings - no text embedding returned")
+                return None
+            
+            # Get the text embedding vector
+            embedding_vector = embeddings.text_embedding
+            
+            # Convert to numpy array and normalize
+            embeddings_array = np.array(embedding_vector, dtype=np.float32)
+            
+            # Normalize the feature vector
+            norm = np.linalg.norm(embeddings_array)
+            if norm > 0:
+                embeddings_array = embeddings_array / norm
+            
+            logger.info(f"Generated text embeddings with shape: {embeddings_array.shape}")
+            return embeddings_array.tolist()
             
-            return embedding.tolist()
         except Exception as e:
             logger.error(f"Error generating text embedding: {e}")
             return None