Update src/streamlit_app.py

src/streamlit_app.py

@@ -3,6 +3,10 @@ import requests
 import pandas as pd
 from together import Together
 import os
+import numpy as np
+from sentence_transformers import SentenceTransformer
+from sklearn.metrics.pairwise import cosine_similarity
+import pickle
 
 # =============================================================================
 # CONFIGURATION - Using Secrets Management
@@ -231,6 +235,51 @@ def extract_unique_names(content_list, field):
                     unique_names.add(cleaned_name)
     return sorted(list(unique_names))
 
+def generate_content_embeddings(content_list, model_name='all-MiniLM-L6-v2'):
+    """Generate embeddings for content using sentence transformers"""
+    # Check if embeddings already exist
+    if os.path.exists('content_embeddings.pkl'):
+        with open('content_embeddings.pkl', 'rb') as f:
+            return pickle.load(f)
+    
+    # Initialize the model
+    model = SentenceTransformer(model_name)
+    
+    # Prepare content descriptions
+    descriptions = []
+    for content in content_list:
+        # Combine relevant text fields for better embedding
+        text = f"{content.get('title', '')} {content.get('description', '')} {content.get('listed_in', '')} {content.get('cast', '')} {content.get('director', '')}"
+        descriptions.append(text)
+    
+    # Generate embeddings
+    embeddings = model.encode(descriptions, show_progress_bar=True)
+    
+    # Save embeddings
+    with open('content_embeddings.pkl', 'wb') as f:
+        pickle.dump(embeddings, f)
+    
+    return embeddings
+
+def get_content_recommendations(content_list, selected_content_idx, embeddings, n_recommendations=5):
+    """Get content recommendations based on similarity"""
+    # Calculate similarity scores
+    similarity_scores = cosine_similarity([embeddings[selected_content_idx]], embeddings)[0]
+    
+    # Get indices of top similar items (excluding self)
+    similar_indices = similarity_scores.argsort()[::-1][1:n_recommendations+1]
+    
+    # Get similarity scores for recommended items
+    scores = similarity_scores[similar_indices]
+    
+    # Get recommended content items with their similarity scores
+    recommendations = [
+        (content_list[idx], score) 
+        for idx, score in zip(similar_indices, scores)
+    ]
+    
+    return recommendations
+
 # =============================================================================
 # MAIN APP
 # =============================================================================
@@ -276,6 +325,10 @@ def main():
         st.error("Could not load Netflix content. Please check your NocoDB connection.")
         st.stop()
     
+    # Generate content embeddings
+    with st.spinner("Generating content embeddings (this may take a while the first time)..."):
+        embeddings = generate_content_embeddings(all_content)
+    
     # Extract unique values for filters
     all_ratings = sorted(list(set([c.get('rating') for c in all_content if c.get('rating')])))
     all_genres = sorted(list(set(
@@ -413,7 +466,7 @@ def main():
                 st.write(f"- Cast: {filters['cast']}")
             st.write("---")
             
-            # Show first 10 items
+            # Show first 10 items with recommendation buttons
             for i, content in enumerate(st.session_state.filtered_content[:10]):
                 with st.expander(f"{content.get('title', 'N/A')} ({content.get('release_year', 'N/A')})"):
                     # Content details in columns
@@ -434,6 +487,19 @@ def main():
                     # Description
                     st.write(f"**📖 Description:**")
                     st.write(content.get('description', 'N/A'))
+                    
+                    # Find content index in original list for recommendations
+                    try:
+                        content_idx = all_content.index(content)
+                        if st.button(f"🎯 Get Similar Content", key=f"rec_{i}"):
+                            recommendations = get_content_recommendations(all_content, content_idx, embeddings)
+                            st.write("**🎬 Similar Content You Might Like:**")
+                            for rec, score in recommendations:
+                                score_percentage = int(score * 100)
+                                st.write(f"- **{rec.get('title')}** ({rec.get('release_year')}) - {score_percentage}% match")
+                                st.write(f"  *{rec.get('type')} | {rec.get('listed_in')}*")
+                    except ValueError:
+                        st.error("Unable to generate recommendations for this title.")
             
             if filtered_count > 10:
                 st.info(f"Showing first 10 of {filtered_count:,} titles. Adjust filters to narrow results.")