Update app.py

app.py

@@ -8,7 +8,6 @@ from transformers import CLIPProcessor, CLIPModel
 from PIL import Image
 import chromadb
 import numpy as np
-from sklearn.decomposition import PCA
 
 app = FastAPI()
 
@@ -21,7 +20,7 @@ pdf_file = "Sutures and Suturing techniques.pdf"
 pptx_file = "impalnt 1.pptx"
 
 # Initialize Embedding Models
-text_model = SentenceTransformer('all-MiniLM-L6-v2')
+text_model = SentenceTransformer('paraphrase-MiniLM-L12-v2')  # 512D embeddings
 model = CLIPModel.from_pretrained("openai/clip-vit-base-patch32")
 processor = CLIPProcessor.from_pretrained("openai/clip-vit-base-patch32")
 
@@ -86,23 +85,17 @@ def extract_images_from_pptx(pptx_path):
         print(f"Error extracting images from PPTX: {e}")
         return []
 
-# Convert Text to Embeddings
+# Convert Text to Embeddings (512D)
 def get_text_embedding(text):
     return text_model.encode(text).tolist()
 
-# Extract Image Embeddings and Reduce to 384 Dimensions
+# Extract Image Embeddings (512D)
 def get_image_embedding(image_path):
     try:
         image = Image.open(image_path)
         inputs = processor(images=image, return_tensors="pt")
         with torch.no_grad():
             image_embedding = model.get_image_features(**inputs).numpy().flatten()
-        
-        # Ensure embedding is 384-dimensional
-        if len(image_embedding) != 384:
-            pca = PCA(n_components=384)
-            image_embedding = pca.fit_transform(image_embedding.reshape(1, -1)).flatten()
-        
         return image_embedding.tolist()
     except Exception as e:
         print(f"Error generating image embedding: {e}")
@@ -113,21 +106,12 @@ def store_data(texts, image_paths):
     for i, text in enumerate(texts):
         if text:
             text_embedding = get_text_embedding(text)
-            if len(text_embedding) == 384:
-                collection.add(ids=[f"text_{i}"], embeddings=[text_embedding], documents=[text])
+            collection.add(ids=[f"text_{i}"], embeddings=[text_embedding], documents=[text])
     
-    all_embeddings = [get_image_embedding(img_path) for img_path in image_paths if get_image_embedding(img_path) is not None]
-    
-    if all_embeddings:
-        all_embeddings = np.array(all_embeddings)
-        
-        # Apply PCA only if necessary
-        if all_embeddings.shape[1] != 384:
-            pca = PCA(n_components=384)
-            all_embeddings = pca.fit_transform(all_embeddings)
-        
-        for j, img_path in enumerate(image_paths):
-            collection.add(ids=[f"image_{j}"], embeddings=[all_embeddings[j].tolist()], documents=[img_path])
+    for j, img_path in enumerate(image_paths):
+        img_embedding = get_image_embedding(img_path)
+        if img_embedding:
+            collection.add(ids=[f"image_{j}"], embeddings=[img_embedding], documents=[img_path])
     
     print("Data stored successfully!")
 
@@ -148,7 +132,7 @@ def process_and_store(pdf_path=None, pptx_path=None):
 
 # FastAPI Endpoints
 @app.get("/")
-def create_vector():
+def greet_json():
     # Run Data Processing
     process_and_store(pdf_path=pdf_file, pptx_path=pptx_file)
     return {"Document store": "created!"}