Update app.py

app.py

@@ -6,7 +6,7 @@ import numpy as np
 from transformers import pipeline
 import time
 
-# --- 1. DATA LOADING AND PREPROCESSING ---
+# --- 1. DATA LOADING AND PREPROCESSING (FINAL FIX) ---
 print("===== Application Startup =====")
 start_time = time.time()
 
@@ -20,29 +20,33 @@ def extract_title_and_ingredients(sample):
 
 def extract_each_feature(sample):
     """
-    FIXED: Extract each feature of a recipe from a sample and clean up potential duplications.
+    FIXED: More robustly extracts recipe features and cleans duplicated content.
     """
     full_text = sample['input']
     
+    # Extract Title
     title = full_text[:full_text.find("\\n")]
     
-    # Find the start and end of the ingredients block
-    ingredients_start = full_text.find("Ingredients:")
-    directions_start = full_text.find("Directions:")
-    
-    # Extract ingredients cleanly
-    if ingredients_start != -1 and directions_start != -1:
-        ingredients = full_text[ingredients_start + len("Ingredients:"):directions_start].strip()
-    else:
-        ingredients = "Not available"
-        
-    # Extract directions and clean them
-    if directions_start != -1:
-        directions_raw = full_text[directions_start + len("Directions:"):].strip()
-        # FIX: Split by "Ingredients:" to remove any duplicated content
-        directions = directions_raw.split("Ingredients:")[0].strip()
-    else:
-        directions = "Not available"
+    # Extract Ingredients
+    ingredients = "Not available"
+    if "Ingredients:" in full_text:
+        start_ing = full_text.find("Ingredients:") + len("Ingredients:")
+        end_ing = full_text.find("Directions:")
+        if end_ing > start_ing:
+            ingredients = full_text[start_ing:end_ing].strip()
+
+    # Extract and clean Directions
+    directions = "Not available"
+    if "Directions:" in full_text:
+        start_dir = full_text.find("Directions:") + len("Directions:")
+        # Take the text after the first "Directions:"
+        temp_directions = full_text[start_dir:].strip()
+        # If "Ingredients:" appears again, it's a duplicate. Cut it off.
+        if "Ingredients:" in temp_directions:
+            end_dir = temp_directions.find("Ingredients:")
+            directions = temp_directions[:end_dir].strip()
+        else:
+            directions = temp_directions
 
     return {
         "title": title,
@@ -70,9 +74,10 @@ index = faiss.IndexFlatL2(dimension)
 index.add(embeddings)
 print(f"Index is ready. Total vectors in index: {index.ntotal}")
 
-# --- 3. SYNTHETIC GENERATION ---
+# --- 3. SYNTHETIC GENERATION (IMPROVED MODEL AND PROMPT) ---
 print("Loading generative model...")
-generator = pipeline('text-generation', model='distilgpt2')
+# Switched back to gpt2 for better instruction following
+generator = pipeline('text-generation', model='gpt2')
 
 def get_recommendations_and_generate(query_ingredients, k=3):
     # 1. Get Recommendations
@@ -90,43 +95,54 @@ def get_recommendations_and_generate(query_ingredients, k=3):
         }
         results.append(recipe)
         
-    # 2. Generate a new recipe with a structured "few-shot" prompt
-    prompt = f"""Create a full recipe including a title, ingredients, and directions based on the following items: {query_ingredients}.
-
-### Title:
-[Recipe Title]
-
-### Ingredients:
-- [Ingredient 1]
-- [Ingredient 2]
-- [Ingredient 3]
-
-### Directions:
-1. [Step 1]
-2. [Step 2]
-3. [Step 3]
----
-Recipe:
-### Title:
+    # 2. Generate a new recipe with a better "few-shot" prompt
+    prompt = f"""Given the ingredients: {query_ingredients}, create a complete recipe.
+
+### Title ###
+[A creative recipe title]
+
+### Ingredients ###
+- [List of ingredients]
+
+### Directions ###
+1. [First step]
+2. [Second step]
+3. [And so on...]
+
+### Title ###
 """
     # Generate the recipe text
-    generated_outputs = generator(prompt, max_new_tokens=200, num_return_sequences=1)
+    generated_outputs = generator(prompt, max_new_tokens=250, num_return_sequences=1, pad_token_id=50256)
     generated_text = generated_outputs[0]['generated_text'].replace(prompt, "").strip()
 
-    # 3. Parse the generated text into a structured format
+    # 3. More robustly parse the generated text
     try:
-        title_part, rest = generated_text.split("### Ingredients:", 1)
-        ingredients_part, directions_part = rest.split("### Directions:", 1)
-        
+        title = "AI Generated Recipe"
+        ingredients = ""
+        directions = ""
+
+        if "### Ingredients ###" in generated_text:
+            parts = generated_text.split("### Ingredients ###")
+            title = parts[0].strip()
+            rest = parts[1]
+            if "### Directions ###" in rest:
+                dir_parts = rest.split("### Directions ###")
+                ingredients = dir_parts[0].strip()
+                directions = dir_parts[1].strip()
+            else:
+                ingredients = rest.strip()
+        else:
+            directions = generated_text # Fallback
+
         generated_recipe = {
-            "title": title_part.strip(),
-            "ingredients": ingredients_part.strip(),
-            "directions": directions_part.strip()
+            "title": title,
+            "ingredients": ingredients,
+            "directions": directions
         }
-    except ValueError:
-        # Fallback if the model doesn't follow the format perfectly
+    except Exception as e:
+        print(f"Error parsing generated text: {e}")
         generated_recipe = {
-            "title": "AI Generated Recipe",
+            "title": "AI Generated Recipe (Parsing Error)",
             "ingredients": "Could not determine ingredients.",
             "directions": generated_text
         }