Update app.py

app.py

@@ -5,7 +5,9 @@ import numpy as np
 import os
 from huggingface_hub import InferenceClient
 from sklearn.metrics.pairwise import cosine_similarity
-from recipe_pipeline import RecipeDigitalizerPipeline # Ensure your pipeline file is named recipe_pipeline.py
+
+# --- FIX: Import from YOUR file name (IO_pipeline) ---
+from IO_pipeline import RecipeDigitalizerPipeline 
 
 # ==========================================
 # 1. SETUP API CLIENT
@@ -17,14 +19,23 @@ client = InferenceClient(token=os.getenv("HF_TOKEN"))
 print("⏳ Loading Datasets...")
 
 # Load the Pre-computed Embeddings
-# We use standard pickle loading. Since we saved numpy arrays, we don't need torch.
-with open('recipe_embeddings.pkl', 'rb') as f:
-    data_bundle = pickle.load(f)
-    stored_embeddings = data_bundle['embeddings'] # This is a numpy matrix
-
-# Load the CSV Dataset
-df_recipes = pd.read_csv('RecipeData_10K.csv')
-print("✅ Resources Loaded Successfully!")
+# We use standard pickle loading. 
+try:
+    with open('recipe_embeddings.pkl', 'rb') as f:
+        data_bundle = pickle.load(f)
+        # Ensure we get the matrix (handle both dict and list formats)
+        if isinstance(data_bundle, dict):
+            stored_embeddings = data_bundle['embeddings']
+        else:
+            stored_embeddings = data_bundle
+            
+    # Load the CSV Dataset
+    df_recipes = pd.read_csv('RecipeData_10K.csv')
+    print("✅ Resources Loaded Successfully!")
+except FileNotFoundError as e:
+    print(f"❌ CRITICAL ERROR: Missing file {e.filename}")
+    stored_embeddings = None
+    df_recipes = pd.DataFrame()
 
 
 # ==========================================
@@ -36,8 +47,7 @@ def get_embedding_via_api(text):
     try:
         # We use the feature_extraction task
         response = client.feature_extraction(text, model=API_MODEL)
-        # The API returns a list of floats (or list of list). We convert to numpy.
-        # Note: BGE-Small is 384 dimensions.
+        # The API returns a list of floats. We convert to numpy.
         return np.array(response)
     except Exception as e:
         print(f"API Error: {e}")
@@ -70,7 +80,9 @@ def format_recipe_text(json_data):
 
 def find_similar_recipes(user_query_text):
     """Finds recipes using API embeddings + Scikit-Learn (No Torch)."""
-    
+    if stored_embeddings is None:
+        return "❌ Error: Embeddings file not loaded."
+
     # 1. Get Embedding from API
     instruction = "Represent this recipe for retrieving similar dishes: "
     query_vec = get_embedding_via_api(instruction + user_query_text)
@@ -83,7 +95,6 @@ def find_similar_recipes(user_query_text):
         query_vec = query_vec.reshape(1, -1)
 
     # 2. Calculate Cosine Similarity (using Numpy/Scikit, very fast)
-    # stored_embeddings is (10000, 384)
     scores = cosine_similarity(query_vec, stored_embeddings)[0]
     
     # 3. Get Top 3 Indices using Numpy
@@ -111,8 +122,11 @@ def find_similar_recipes(user_query_text):
 
 def magic_pipeline(image_path):
     # Step 1: Image -> Text (API)
-    digitizer = RecipeDigitalizerPipeline()
-    json_result = digitizer.run_pipeline(image_path)
+    try:
+        digitizer = RecipeDigitalizerPipeline()
+        json_result = digitizer.run_pipeline(image_path)
+    except Exception as e:
+        return f"Error in IO_pipeline: {e}", ""
     
     # Step 2: Format
     readable_text, query_text = format_recipe_text(json_result)