Upload 3 files

.gitattributes

@@ -33,3 +33,4 @@ saved_model/**/* filter=lfs diff=lfs merge=lfs -text
 *.zip filter=lfs diff=lfs merge=lfs -text
 *.zst filter=lfs diff=lfs merge=lfs -text
 *tfevents* filter=lfs diff=lfs merge=lfs -text
+src/fragrance_faiss.index filter=lfs diff=lfs merge=lfs -text

src/fragrance_faiss.index

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:a235f0b4596acedbea9331f0fdc4f7c354e2cbbf3631adca4f5cf83b8778b988
+size 73921581

src/fragrance_metadata.pkl

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:1a3dbbdd17c47cd65492a0ff104e4be008879e7862dfa5d26b7f896fa9831a4d
+size 5195729

src/streamlit_app.py

@@ -1,40 +1,179 @@
-import altair as alt
-import numpy as np
-import pandas as pd
-import streamlit as st
-
-"""
-# Welcome to Streamlit!
-
-Edit `/streamlit_app.py` to customize this app to your heart's desire :heart:.
-If you have any questions, checkout our [documentation](https://docs.streamlit.io) and [community
-forums](https://discuss.streamlit.io).
-
-In the meantime, below is an example of what you can do with just a few lines of code:
-"""
-
-num_points = st.slider("Number of points in spiral", 1, 10000, 1100)
-num_turns = st.slider("Number of turns in spiral", 1, 300, 31)
-
-indices = np.linspace(0, 1, num_points)
-theta = 2 * np.pi * num_turns * indices
-radius = indices
-
-x = radius * np.cos(theta)
-y = radius * np.sin(theta)
-
-df = pd.DataFrame({
-    "x": x,
-    "y": y,
-    "idx": indices,
-    "rand": np.random.randn(num_points),
-})
-
-st.altair_chart(alt.Chart(df, height=700, width=700)
-    .mark_point(filled=True)
-    .encode(
-        x=alt.X("x", axis=None),
-        y=alt.Y("y", axis=None),
-        color=alt.Color("idx", legend=None, scale=alt.Scale()),
-        size=alt.Size("rand", legend=None, scale=alt.Scale(range=[1, 150])),
-    ))
+
+import streamlit as st
+import faiss
+import numpy as np
+import pandas as pd
+from sentence_transformers import SentenceTransformer
+import pickle
+from langchain_ollama import ChatOllama
+
+# Fragrance card function
+def create_fragrance_card(name, rating, brand, perfumer_text, top_notes, middle_notes, base_notes, accords_text, explanation):
+    # Create fragrance card HTML
+    card_html = f"""
+        <div style="border: 1px solid #ddd; padding: 15px; margin: 10px; border-radius: 15px; 
+                    background: linear-gradient(to bottom right, #ffffff, #f2f6fc); 
+                    width: 400px; color: #222; box-shadow: 0 4px 8px rgba(0,0,0,0.1);">
+            <h3 style="color: #3a3a3a; text-align: center;">{name} ⭐{rating}</h3>
+            <p><strong>🏷️ Brand:</strong> {brand}</p>
+            <p><strong>👃 Perfumer(s):</strong> {perfumer_text}</p>
+            <p><strong>🌿 Top Notes:</strong> {top_notes}</p>
+            <p><strong>💖 Heart Notes:</strong> {middle_notes}</p>
+            <p><strong>🌲 Base Notes:</strong> {base_notes}</p>
+            <p><strong>🎼 Main Accords:</strong> {accords_text}</p>
+            <p><strong>💡 AI Explanation:</strong> {explanation}</p>
+        </div>
+    """
+    
+    return card_html
+
+# Load FAISS database, metadata, and encoder with cache
+@st.cache_resource
+def load_resources():
+    index = faiss.read_index('fragrance_faiss.index')
+    with open('fragrance_metadata.pkl', 'rb') as f:
+        metadata = pickle.load(f)
+    encoder = SentenceTransformer('paraphrase-mpnet-base-v2')
+    return index, metadata, encoder
+
+# Gets a brief explanation from Ollama for why this fragrance matches the user's query
+def get_ollama_explanation(query, description):
+    prompt = f"""
+            A user is searching for a fragrance with this description: "{query}"
+            
+            One recommendation is:
+            {description}
+            
+            Explain in 1-2 sentences, in plain English, why this fragrance matches the user's query.
+            """
+    response = llm.invoke(prompt)
+    return response.content.strip()
+
+# Load Ollama
+llm = ChatOllama(model="llama3.2")
+
+# Initialize app
+st.set_page_config(page_title="Fragrance Recommendation System", layout="wide")
+
+# Add title to top of app interface
+st.title("Fragrance Recommendation System")
+
+# Sidebar filters
+st.sidebar.header("Filters")
+query = st.text_input("Describe your ideal fragrance:")
+
+col1, col2 = st.columns(2)
+with col1:
+    k = st.slider("Number of recommendations:", 1, 10, 5)
+with col2:
+    min_rating = st.slider("Minimum rating:", 1.0, 5.0, 3.5)
+
+gender_filter = st.sidebar.selectbox("Gender:", ["All", "Male", "Female", "Unisex"])
+brand_filter = st.sidebar.text_input("Brand (leave empty for all):", "").title()
+note_filter = st.sidebar.text_input("Notes (comma-separated):", "").lower()
+
+# Load resources
+index, metadata, encoder = load_resources()
+
+# Convert rating_values to numeric
+if 'rating_value' in metadata.columns:
+    metadata['rating_value'] = pd.to_numeric(
+        metadata['rating_value'], 
+        errors='coerce')
+
+# Press button and start recommendations
+if st.button('Get Recommendations'):
+    with st.spinner('Finding your fragrance recs...'):
+        if query == "":
+            st.warning("No query entered.")
+        else:
+            # Apply filters sequentially
+            current_df = metadata.copy()
+            
+            # Gender filter
+            if gender_filter != "All":
+                current_df = current_df[current_df['gender'].str.lower() == gender_filter.lower()]
+            
+            # Brand filter
+            if brand_filter:
+                current_df = current_df[current_df['brand'].str.contains(brand_filter, case=False, na=False)]
+            
+            # Rating filter (with NaN handling)
+            if 'rating_value' in current_df.columns:
+                current_df = current_df[current_df['rating_value'].ge(min_rating)]
+            
+            # Note filter
+            if note_filter:
+                notes = [n.strip().lower() for n in note_filter.split(",")]
+                def note_check(row):
+                    note_fields = [
+                        str(row['top']).lower() if pd.notna(row['top']) else "",
+                        str(row['middle']).lower() if pd.notna(row['middle']) else "",
+                        str(row['base']).lower() if pd.notna(row['base']) else ""
+                    ]
+                    return any(note in field for note in notes for field in note_fields)
+                
+                current_df = current_df[current_df.apply(note_check, axis=1)]
+            
+            valid_indices = current_df.index.tolist()
+            
+            # Check if any fragrances remain
+            if not valid_indices:
+                st.warning("No fragrances match all your filters. Try relaxing some criteria.")
+                st.stop()
+
+            # Grab the vectors for fragrances still present after the filters
+            filtered_vectors = np.vstack([index.reconstruct(int(idx)) for idx in valid_indices])
+            temp_index = faiss.IndexFlatIP(filtered_vectors.shape[1])
+            temp_index.add(filtered_vectors)
+        
+            # Encode the query and normalize it for cosine similarity
+            query_vector = encoder.encode([query])
+            faiss.normalize_L2(query_vector)
+            
+            # Perform the search and returns indices of the most similar vectors and their similarity scores
+            sim_score, I = temp_index.search(query_vector, min(k, len(valid_indices)))
+
+            # Get the recommened fragrance's indices and similarity score
+            results = [(valid_indices[i], sim_score[0][j]) for j, i in enumerate(I[0])]
+    
+            # Display results
+            st.subheader(f"Recommended Fragrances ({len(results)} results)")
+            cols = st.columns(3)
+            
+            for idx, (result_idx, sim_score) in enumerate(results):
+                rec = metadata.loc[result_idx]
+    
+                # Extract data with fallbacks
+                name = rec.get('perfume', 'Unknown')
+                brand = rec.get('brand', 'Unknown')
+                perfumer_text = rec.get('perfumer', 'Unknown')
+                top_notes = rec.get('top', 'Unknown')
+                middle_notes = rec.get('middle', 'Unknown')
+                base_notes = rec.get('base', 'Unknown')
+                accords_text = rec.get('accord', 'Unknown')
+                rating = rec.get('rating_value', '?')
+    
+                 # Create natural language fragrance description
+                description = (
+                    f"The fragrance is called {name}. It is by {brand}. "
+                    f"The perfumer is {perfumer_text}. The top notes are {top_notes}, "
+                    f"the heart notes are {middle_notes}, and the base notes are {base_notes}. "
+                    f"The main accords are {accords_text}."
+                )
+    
+                explanation = get_ollama_explanation(query, description)
+                
+                # Add rating to card
+                card = create_fragrance_card(
+                        name,
+                        rating,
+                        brand, 
+                        perfumer_text, 
+                        top_notes,
+                        middle_notes, 
+                        base_notes, 
+                        accords_text,
+                        explanation
+                    )
+                cols[idx % 3].markdown(card, unsafe_allow_html=True)