Update app.py

app.py

@@ -89,14 +89,33 @@ def extract_images_from_pptx(pptx_path):
 def get_text_embedding(text):
     return text_model.encode(text).tolist()
 
+from transformers import CLIPProcessor, CLIPModel
+import torch
+import numpy as np
+from sklearn.decomposition import PCA
+
+# ✅ Load CLIP (512-dimensional output)
+model = CLIPModel.from_pretrained("openai/clip-vit-base-patch32")
+processor = CLIPProcessor.from_pretrained("openai/clip-vit-base-patch32")
 
-### Step 6: Convert Images to Embeddings ###
 def get_image_embedding(image_path):
+    """Extracts image embedding and reduces to 384 dimensions"""
+    from PIL import Image
+
     image = Image.open(image_path)
-    inputs = clip_processor(images=image, return_tensors="pt")
+    inputs = processor(images=image, return_tensors="pt")
+
     with torch.no_grad():
-        embedding = clip_model.get_image_features(**inputs)
-    return embedding.squeeze().tolist()
+        image_embedding = model.get_image_features(**inputs)  # Shape: (1, 512)
+
+    image_embedding = image_embedding.numpy().flatten()  # Convert to NumPy (512,)
+
+    # ✅ Reduce to 384 dimensions using PCA
+    pca = PCA(n_components=384)
+    image_embedding_384 = pca.fit_transform(image_embedding.reshape(1, -1))
+
+    return image_embedding_384.flatten().tolist()
+
 
 
 ### Step 7: Store Data in ChromaDB ###