first commit to hf space

app.py

@@ -0,0 +1,236 @@
+import numpy as np
+import pandas as pd
+import os
+from PIL import Image
+import streamlit as st 
+from streamlit import components
+from datasets import Dataset, load_dataset, load_from_disk
+import faiss
+from scripts.preprocessing import preprocess
+
+# App config 
+icon = Image.open('./images/wine_icon.png')
+st.set_page_config(page_title="Sommeli-AI", 
+                   page_icon=icon,
+                   layout="wide")
+hide_default_format = """
+       <style>
+       #MainMenu {visibility: visible; }
+       footer {visibility: hidden;}
+       </style>
+       """
+st.markdown(hide_default_format, unsafe_allow_html=True)
+
+# App functions 
+@st.cache_data 
+def read_data(ds_path=None):
+
+    if ds_path is not None:
+        # Read in hf file 
+        embeddings_dataset = load_from_disk(ds_path)
+    else:
+        embeddings_dataset = load_dataset("pdjewell/sommeli_ai", split="train")
+
+    # Convert to pandas df
+    embeddings_dataset.set_format("pandas")
+    df = embeddings_dataset[:]
+
+    # preprocess data (add type col, remove dups)
+    df = preprocess(df)
+  
+    return df 
+
+
+def get_neighbours(df, query_embedding, k=6,
+                   metric='inner'):
+
+    # convert from pandas df to hf ds 
+    ds = Dataset.from_pandas(df)
+    ds.reset_format()
+    ds = ds.with_format("np")
+
+    # add faiss index
+    if metric == 'inner':
+        ds.add_faiss_index(column="embeddings",
+                       metric_type=faiss.METRIC_INNER_PRODUCT)
+    else: 
+        ds.add_faiss_index(column="embeddings",
+                       metric_type=faiss.METRIC_L2)
+        
+    scores, samples = ds.get_nearest_examples(
+        "embeddings", query_embedding, k=k)
+    
+    samples.pop('embeddings')
+    samples.pop('__index_level_0__')
+    
+    return scores, samples
+
+
+def filter_df_search(df: pd.DataFrame) -> pd.DataFrame:
+    
+    modify_search = st.checkbox("🔍 Further filter search selection")
+    
+    if not modify_search:
+        return df
+
+    df = df.copy()
+
+    modification_container_search = st.container()
+
+    with modification_container_search:
+        to_filter_columns = st.multiselect("Filter on:", 
+                                           ['Province', 'Region', 'Winery','Score', 'Price'],
+                                           key='search')
+
+        for column in to_filter_columns:
+            if column in ['Score', 'Price']: # Use slider for 'points' and 'price'
+                min_val = 0
+                max_val = int(df[column].max())
+                user_input = st.slider(f"Values for {column}", min_val, max_val, (min_val, max_val))
+                df = df[(df[column] >= user_input[0]) & (df[column] <= user_input[1])]
+            elif column in ['Country', 'Province', 'Region', 'Variety', 'Winery']: # Use multiselect for these columns
+                unique_values = df[column].dropna().unique()
+                default_values = [unique_values[0]] if len(unique_values) > 0 else [] # Select only the first unique value if it exists
+                user_input = st.multiselect(f"Values for {column}", unique_values, default_values)
+                df = df[df[column].isin(user_input)]
+
+    return df
+
+
+def filter_df_recs(df: pd.DataFrame) -> pd.DataFrame:
+    
+    modify_recs = st.checkbox("🔍 Filter recommendation results")
+    
+    if not modify_recs:
+        return df
+
+    df = df.copy()
+
+    modification_container_recs = st.container()
+
+    with modification_container_recs:
+
+        to_filter_columns2 = st.multiselect("Filter on:", 
+                                            ['Country','Province', 'Region', 'Variety', 'Winery',
+                                             'Score', 'Price'],
+                                            key='recs')
+
+        for column in to_filter_columns2:
+            if column in ['Score', 'Price']: # Use slider for 'points' and 'price'
+                min_val = 0
+                max_val = int(df[column].max())
+                user_input = st.slider(f"Values for {column}", min_val, max_val, (min_val, max_val))
+                df = df[(df[column] >= user_input[0]) & (df[column] <= user_input[1])]
+            elif column in ['Country', 'Province', 'Region', 'Variety', 'Winery']: # Use multiselect for these columns
+                unique_values = df[column].dropna().unique()
+                default_values = [unique_values[0]] if len(unique_values) > 0 else [] # Select only the first unique value if it exists
+                user_input = st.multiselect(f"Values for {column}", unique_values, default_values)
+                df = df[df[column].isin(user_input)]
+
+    return df
+
+
+if __name__ == "__main__":
+    st.title("🍷 Sommeli-AI")
+    col1, col2 = st.columns([0.6,0.4], gap="medium")
+ 
+    # Read in data 
+    ds_path = "./data/wine_ds.hf"
+    df = read_data(ds_path=None)
+
+    with col2:
+        st.header("Explore the world of wine  🌍")
+        wine_plot = st.radio('Select plot type:', ['2D','3D'],
+                            label_visibility = "hidden",
+                            horizontal=True)
+        st.text("Click the legend categories to filter")
+
+        # Load the HTML file
+        with open('./images/px_2d.html', 'r') as file:
+            plot2d_html = file.read()
+        # Load the HTML file
+        with open('./images/px_3d.html', 'r') as file:
+            plot3d_html = file.read()
+        # Display the HTML plot in the Streamlit app
+        if wine_plot == '2D':
+            components.v1.html(plot2d_html, width=512, height=512)
+        elif wine_plot == '3D': 
+            components.v1.html(plot3d_html, width=512, height=512)
+
+    with col1: 
+
+        # Select all wine types initially
+        st.header("Search for similar wines  🥂")
+        # Select wine type: default is all 
+        wine_types = df['Type'].unique()
+        selected_wine_types = st.multiselect("Select category 👇", wine_types, default=wine_types)
+        df = df[df['Type'].isin(selected_wine_types)]
+        subcol1, subcol2 = st.columns([0.5,0.5], gap="small")
+        with subcol1:
+            # Select wine variety: default is all 
+            wine_vars = df['Variety'].unique()
+            selected_wine_vars = st.multiselect("Narrow down the variety 🍇",['Select all'] + list(wine_vars),
+                                                default = 'Select all')
+            if "Select all" in selected_wine_vars:
+                df_search = df
+            else:
+                df_search = df[df['Variety'].isin(selected_wine_vars)]
+        
+        with subcol2:
+            # Select the country: default is all 
+            countries = df_search['Country'].unique()
+            selected_countries = st.multiselect("Narrow down the country 🌎",['Select all'] + list(countries),
+                                                default = 'Select all')
+            if "Select all" in selected_countries:
+                df_search = df_search
+            else:
+                df_search = df_search[df_search['Country'].isin(selected_countries)]
+
+        # Add additional filters 
+        df_search = filter_df_search(df_search)
+
+        # Create a search bar for the wine 'title'
+        selected_wine = st.selectbox("Search for and select a wine 👇", [''] + list(df_search["Title"].unique()))
+    
+        if selected_wine:
+            # Get the embedding for selected_wine
+            query_embedding = df.loc[df['Title']==selected_wine, 'embeddings'].iloc[0]
+
+            tasting_notes = df.loc[df['Title']==selected_wine, 'Tasting notes'].iloc[0]
+            st.write(f"Tasting notes: {tasting_notes}")
+
+            # CSS to inject contained in a string
+            hide_table_row_index = """
+                        <style>
+                        thead tr th:first-child {display:none}
+                        tbody th {display:none}
+                        </style>
+                        """
+            # Inject CSS with Markdown
+            st.markdown(hide_table_row_index, unsafe_allow_html=True)
+
+            # Display selected wine
+            st.header("	🍷 Your selected wine")
+            selected_cols = ['Title','Country','Province','Region','Winery',
+                            'Variety','Tasting notes','Score']
+            st.table(df.loc[df['Title']==selected_wine, selected_cols].fillna(""))
+
+            # Slider for results to show 
+            k = st.slider(f"Choose how many similar wines to show 👇", 1, 10, value=4)
+            
+            # Filter recommendation results 
+            df_results = filter_df_recs(df)
+            
+            # Display results as table 
+            if st.button("🔘 Press me to generate similar tasting wines"):
+                # Get neighbours
+                scores, samples = get_neighbours(df_results, query_embedding, 
+                                                 k=k+1, metric='l2')
+                recs_df = pd.DataFrame(samples).fillna("")
+                recs_df = recs_df.fillna(" ")
+                # Display results
+                st.header(f"🍾 Top {k} similar tasting wines")
+                st.table(recs_df.loc[1:,selected_cols])
+            
+        else:
+            print("Awaiting selection")

requirements.txt

@@ -0,0 +1,8 @@
+datasets==2.12.0
+faiss-cpu==1.7.4
+gitdb==4.0.10
+numpy==1.24.3
+pandas==2.0.2
+Pillow==9.5.0
+pyarrow==12.0.0
+streamlit==1.23.1

scripts/preprocessing.py

@@ -0,0 +1,162 @@
+import numpy as np
+import pandas as pd 
+import os
+from pathlib import Path
+
+def preprocess(df):
+    
+    df = df.copy()
+
+    White = ['White Blend', 'Pinot Gris', 'Riesling', 'Chardonnay', 'Chenin Blanc', 'Sauvignon Blanc', 
+            'Viognier-Chardonnay', 'Catarratto', 'Inzolia', 'Bordeaux-style White Blend', 'Grillo', 
+            'Albariño', 'Petit Manseng', 'Vernaccia', 'Grüner Veltliner', 'Viognier', 'Vermentino', 
+            'Grenache Blanc', 'Pinot Blanc', 'Alsace white blend', 'Portuguese White', 'Verdejo', 
+            'Fumé Blanc', 'Pinot Bianco', 'Ugni Blanc-Colombard', 'Friulano', 'Assyrtico', 'Vignoles', 
+            'Muscat', 'Muscadelle', 'Garganega', 'Pinot Grigio','Cortese', 'Melon', 'Vidal', 'Verdelho', 
+            'Marsanne', 'Vilana', 'Viura', 'Verduzzo', 'Verdicchio', 'Colombard', 'Sylvaner', 'Sémillon', 
+            'Antão Vaz', 'Verdejo-Viura', 'Chenin Blanc-Chardonnay', 'Insolia', 'Ribolla Gialla', 
+            'Weissburgunder', 'Traminer', 'Prié Blanc', 'Müller-Thurgau', 'Pansa Blanca', 'Muskat Ottonel',
+            'Sauvignon Blanc-Semillon', 'Semillon-Sauvignon Blanc', 'Bical', 'Viura-Chardonnay', 'Malvasia Bianca',
+            'Rhône-style White Blend', 'Scheurebe', 'Kerner', 'Carricante', 'Fiano', 'Früburgunder', 'Roussanne', 
+            'Avesso', 'Chinuri', 'Muscat Blanc à Petits Grains', 'Xarel-lo', 'Greco', 'Trebbiano', 'Prié Blanc',
+            'Falanghina', 'Bical', 'Gelber Muskateller', 'Turbiana', 'Refosco', 'Alvarinho', 'Manzoni', 'Assyrtiko', 
+            'Welschriesling', 'Rieslaner', 'Traminette', 'Marsanne-Viognier', 'Gewürztraminer-Riesling', 
+            'Austrian white blend', 'Tocai', 'Chardonnay-Viognier', 'Fernão Pires', 'Seyval Blanc', 'Muscat Canelli', 
+            'Arinto', 'Arneis', 'Malvasia', 'Altesse', 'Blanc du Bois', 'Provence white blend', 'Nosiola', 
+            'Roussanne-Viognier', 'Godello', 'Auxerrois', 'Albana', 'Muskat',  'Grechetto', 'Encruzado', 
+            'Garnacha Blanca', 'Pallagrello', 'Morava', 'Aleatico', 'Nascetta', 'Siria', 'Asprinio', 'Feteascǎ Regalǎ', 
+            'Tocai Friulano', 'Schiava', 'Chardonnay-Semillon', 'Palomino', 'Norton', 
+            'Loureiro-Arinto', 'Symphony', 'Edelzwicker', 'Madeira Blend', 'Gros and Petit Manseng', 'Jacquère', 
+            'Chenin Blanc-Sauvignon Blanc', 'Marzemino', 'Chardonnay-Sauvignon Blanc', 'Trebbiano Spoletino',
+            'Chasselas', 'Hárslevelü', 'Siegerrebe','Colombard-Sauvignon Blanc', 'Diamond',
+            'Gros Manseng', 'Muskateller', 'Aligoté', 'Muscat Blanc', 'Viognier-Roussanne', 'Pallagrello Bianco', 
+            'Veltliner', 'Chardonnay-Sauvignon', 'Chenin Blanc-Viognier', 'Vitovska', 'Grauburgunder', 'Macabeo', 
+            'Verdil', 'Treixadura', 'Coda di Volpe', 'Viura-Verdejo', 'Bombino Bianco', 'Pinot-Chardonnay', 
+            "Muscat d'Alexandrie", 'Chardonnay-Pinot Gris', 'Chardonnay-Pinot Blanc','Piquepoul Blanc', 'Orange Muscat',
+            'Ugni Blanc', 'Semillon-Chardonnay', 'Irsai Oliver', 'Greco Bianco', 'Viognier-Grenache Blanc', 'Pignoletto', 
+            'Muscatel', 'White Riesling', 'Hondarrabi Zuri', 'Nuragus', 'Xynisteri', 'Sauvignon Musqué', 'Roussanne-Marsanne', 
+            'Incrocio Manzoni', 'Terrantez', 'Bual', 'Verdejo-Sauvignon Blanc', 'Malvasia-Viura', 'Savatiano', 
+            'Macabeo-Chardonnay', 'Tamjanika', 'Macabeo-Moscatel', 'Códega do Larinho','Pinot Gris-Gewürztraminer',
+            'Viosinho', 'Paralleda', 'Malvar', 'Airen', 'Erbaluce', 'Verdosilla', 'Aidani', 'Vinhão', 'Rolle', 'Orangetraube', 
+            'Žilavka', 'Portuguiser', 'Gouveio', 'Bombino Nero', 'Malagouzia-Chardonnay', 'Elbling', 'Gragnano', 
+            'Pinot Blanc-Chardonnay', 'Petit Meslier', 'Chardonnay Weissburgunder', 'Robola', 'Folle Blanche', 'Malagouzia', 
+            'Rabigato', 'Sauvignonasse', 'Meseguera', 'Alvarinho-Chardonnay', 'Pinot Blanc-Viognier', 'Biancu Gentile', 
+            'Xinisteri','Moschofilero-Chardonnay','Sauvignon Blanc-Sauvignon Gris', 'Trebbiano di Lugana', 'Verdeca', 
+            'Chardonel', 'Silvaner-Traminer', 'Uvalino', 'Merseguera-Sauvignon Blanc', 'Cayuga', 
+            'Nasco', 'Vital', 'Apple', 'Pinot Grigio-Sauvignon Blanc', 'Valvin Muscat', 'Malvasia Fina', 
+            'Roditis-Moschofilero', 'Premsal', 'Jampal', 'Tokay Pinot Gris', 'Trajadura', 'Roscetto', 'Torontel', 
+            'Viognier-Valdiguié',
+            'Zierfandler', 'Marsanne-Roussanne', 'Pinot Meunier', 'Muskat Ottonel', 'Moscatel', 'Moschofilero', 'White Port', 
+            'Kisi', 'Kangoun', 'Posip', 'Uva di Troia', 'Zierfandler-Rotgipfler', 'Mauzac', 'Pinot Auxerrois', 'Neuburger', 
+            'Sämling', 'Rkatsiteli', 'Trousseau Gris', 'Malvasia Istriana', 'Morillon', 'Tokay', 'Gros Plant', 'Muscat Hamburg', 
+            'Emir', 'Tsolikouri', 'Narince', 'Grecanico', 'Madeleine Angevine', 'Doña Blanca', 'Graševina', 'Thrapsathiri', 
+            'Cococciola', 'Plyto', 'Azal', 'Moscatel Graúdo', 'Malvasia di Candia', 'Maria Gomes', 'Muscat of Alexandria', 
+            'Moscatel de Alejandría', 'Misket', 'Tamianka', 'Morio Muskat', 'Sauvignonasse', 
+            'Viognier-Marsanne', 'Ryzlink Rýnský', 'Muscadel', 'Roussanne-Grenache Blanc', 'Chancellor', 'Picapoll', 
+            'Blauburger', 'Athiri', 'Ondenc','Gewürztraminer', 'Torrontés', 'Furmint', 'Savagnin', 'Glera', 
+            'Roter Veltliner', 'Silvaner', 'Ruché', 'Pecorino', 'Sauvignon Gris', 'Vidal Blanc', 'Albanello', 
+            'Loureiro', 'Clairette', 'Verduzzo Friulano ', "Loin de l'Oeil", 'Timorasso', 'Pigato', 'Viognier-Gewürztraminer', 
+            'Sauvignon Blanc-Chenin Blanc', 'Colombard-Ugni Blanc', 'Mtsvane', 'Rivaner', 'Vespaiolo', 'Biancolella', 
+            'Riesling-Chardonnay', 'Maria Gomes-Bical', 'Gelber Traminer', 'Sercial', 'Grenache Gris', 'Chardonnay-Albariño',
+            'Roditis', 'Papaskarasi', 'Zibibbo', 'Malagousia', 'Rotgipfler', 'Durella', 'Cercial', 'Johannisberg Riesling', 
+            'Teran', 'Mantonico', 'Timorasso', 'Zlahtina', 'Shiraz-Roussanne', 'Tămâioasă Românească', 'Ansonica', 'Feteasca',
+            'Catalanesca', 'Moscato di Noto', 'Moscato Giallo','Sauvignon Blanc-Chardonnay', 'Sauvignon-Sémillon', "Cesanese d'Affile", 
+            'Sauvignon Blanc-Verdejo', 'Chardonnay-Riesling', 'Sauvignon Blanc-Assyrtiko','Zelen', 'Tempranillo Blanco', 
+            'Roter Traminer','Picpoul'
+    ]
+    Red = ['Portuguese Red', 'Pinot Noir', 'Tempranillo-Merlot', 'Frappato', 'Cabernet Sauvignon',
+            'Nerello Mascalese', 'Malbec', 'Tempranillo Blend', 'Meritage', 'Red Blend', 'Merlot', 
+            "Nero d'Avola", 'Gamay', 'Primitivo', 'Sangiovese', 'Cabernet Franc', 'Bordeaux-style Red Blend', 
+            'Aglianico', 'Petite Sirah', 'Touriga Nacional', 'Carmenère', 'Rosso', 'Shiraz-Cabernet Sauvignon', 
+            'Barbera', 'Rhône-style Red Blend', 'Graciano', 'Tannat-Cabernet', 'Sauvignon', 'Sangiovese Grosso', 
+            'Bonarda', 'Shiraz', 'Montepulciano', 'Grenache', 'Syrah', 'Nebbiolo', 'Blaufränkisch', 'Carignan-Grenache', 
+            'Sagrantino', 'Cabernet Sauvignon-Syrah', 'Tempranillo','Mencía', 'Zweigelt', 'Cannonau', 'Dolcetto', 
+            'Garnacha Tintorera', 'Pinot Nero', 'Pinotage', 'Syrah-Grenache', 'Antão Vaz', 'Cabernet Sauvignon-Carmenère', 
+            'Tinta Miúda', 'Monastrell', 'Merlot-Malbec', 'Cabernet Sauvignon-Merlot', 'Merlot-Argaman', 'Garnacha', 
+            'Negroamaro', 'Mourvèdre', 'Syrah-Cabernet', 'Tannat', 'Cabernet Sauvignon-Sangiovese', 'Austrian Red Blend', 
+            'Teroldego', 'Baga','Pinot Noir-Gamay', 'Cinsault', 'Corvina, Rondinella, Molinara', 'Tannat-Syrah', 'Charbono', 
+            'Provence red blend', 'Claret','Malbec-Merlot', 'Monastrell-Syrah', 'Malbec-Tannat', 'Malbec-Cabernet Franc', 
+            'Tinta de Toro', 'Cabernet Moravia', 'Chambourcin', 'Nero di Troia', 'Cesanese', 'Lagrein', 'Tinta Fina', 'St. Laurent', 
+            'Cabernet Sauvignon-Shiraz', 'Syrah-Cabernet Sauvignon', 'Pugnitello', 'Touriga Nacional Blend', 'Tinta Roriz', 
+            'Cabernet Franc-Cabernet Sauvignon', 'Grenache-Syrah', 'Tempranillo-Cabernet Sauvignon', 'Merlot-Cabernet Franc', 
+            'Syrah-Petite Sirah', 'Cabernet Blend', 'Maturana', 'Magliocco', 'Gamay Noir', 'Spätburgunder', 'Plavac Mali',
+            'Lemberger', 'Saperavi', 'Dornfelder', 'Ojaleshi', 'Mondeuse', 'Perricone', 'Syrah-Merlot', 'Cabernet Sauvignon-Malbec',
+            'Tinto Fino', 'Malbec-Cabernet Sauvignon','Carignano', 'Cabernet Franc-Merlot', 
+            'Syrah-Petit Verdot', 'Syrah-Mourvèdre', 'Shiraz-Grenache', 'Grenache-Carignan', 'Malbec-Syrah', 
+            'Cabernet Sauvignon-Tempranillo', 'Carignan', 'Cabernet-Syrah', 'Merlot-Cabernet Sauvignon', 
+            'Mourvèdre-Syrah', 'Negrette', 'Tinta Barroca', 'Merlot-Tannat','Castelão', 
+            'Grenache Blend', 'Sangiovese Cabernet', 'Touriga Nacional-Cabernet Sauvignon', 'Cabernet Sauvignon-Cabernet Franc', 
+            'Baco Noir', 'Tempranillo-Tannat', 'Touriga Franca', 'Barbera-Nebbiolo', 'Prieto Picudo', 'Gaglioppo', 'Carignane', 
+            'Tannat-Merlot', 'Nerello Cappuccio', 'Counoise', 'Mazuelo', 'Tinta del Pais', 'Vranec', 'Mavrud', 'Cabernet', 
+            'Grenache-Mourvèdre', 'Forcallà', 'Syrah-Tempranillo', 'Cabernet Sauvignon-Barbera', 'Merlot-Cabernet', 'Jaen', 
+            'Tinta del Toro', 'Prunelard', 'Garnacha-Syrah', 'Rufete', 'Tempranillo-Shiraz','Mansois',
+            'Mataro', 'Tinta Cao', 'Blauer Portugieser', 'Groppello', 'Poulsard', 'Grenache-Shiraz', 'Baga-Touriga Nacional', 
+            'Carineña', 'Ciliegiolo', 'Cabernet Sauvignon-Merlot-Shiraz', 'Sciaccerellu', 'Alicante', 'Rosenmuskateller', 
+            'Malbec-Cabernet', 'Touriga', 'Carmenère-Syrah', 'Mavroudi', 'Pinot Blanc-Pinot Noir', 'Tinto Velasco', 'Kadarka', 
+            'Sangiovese-Syrah', 'Tannat-Cabernet Franc', 'Fer Servadou', 'Mission', 'Kekfrankos', 'Blauburgunder', 'Marquette', 
+            'Romorantin', 'Braucol', 'Cabernet Franc-Malbec', 'Pallagrello Nero', 'Rebula', 'Vespolina', 'Shiraz-Malbec', 
+            'Rebo', 'Tempranillo-Malbec', 'Trousseau', 'Bacchus', 'Syrah-Malbec', 'Syrah-Cabernet Franc', 'Cariñena-Garnacha', 
+            'Sideritis','Rara Neagra', 'Molinara', 'Abouriou', 'Nielluciu', 'Malbec-Bonarda', 'Garnacha-Monastrell', 'Souzao', 
+            'Tinta Francisca', 'Malvasia Nera', 'Listán Negro', 'Pinotage-Merlot', 'Jacquez', 'Carignan-Syrah', 'Mavrotragano', 
+            'Bovale', 'Frankovka', 'Garnacha Blend', 'Merlot-Shiraz', 'Malbec Blend', 'Merlot-Syrah', 'Babić', 'Yapincak', 
+            'Mandilaria', 'Saperavi-Merlot', 'Teroldego Rotaliano', 'Garnacha-Tempranillo','Vermentino Nero',
+            'Albarossa', 'Cabernet Sauvignon Grenache', 'Black Monukka', 'Merlot-Grenache', 'Vranac', 'Tempranillo-Syrah', 
+            'Boğazkere', 'Tinta Amarela', 'Tinta Negra Mole', 'Chelois', 'Shiraz-Tempranillo', 'Biancale', 'Syrah-Bonarda', 
+            'Durif', 'Franconia', 'Malbec-Tempranillo', 'Monastrell-Petit Verdot', 'Sirica', 'Espadeiro', 'Blatina', 'Karalahna', 
+            'Garnacha-Cabernet', 'Garnacha-Cariñena', 'Cabernet Franc-Lemberger', 'Shiraz-Mourvèdre', 'Mavrokalavryta', 'Favorita', 
+            'Babosa Negro', 'Dafni', 'Petit Courbu', 'Kotsifali', 'Parraleta', 'Otskhanuri Sapere', 'Trollinger', 
+            'Tsapournakos', 'Francisa', 'Kuntra', 'Pignolo', 'Schwartzriesling','Sousão', 'Feteasca Neagra', 'Kinali Yapincak',
+            'Kalecik Karasi', 'Karasakiz', 'Raboso', 'Trepat', 'Freisa', 'Trincadeira', 'Melnik', 'Argaman', 'Piedirosso', 
+            'Marawi', 'Çalkarası', 'Tinta Francisca', 'Vidadillo', 'Other', 'Cabernet Pfeffer', 'Roviello', 'Colorino', 
+            'Tinta Madeira', 'Centesimino', 'Ramisco', 'Gamza', 'Bobal-Cabernet Sauvignon',
+            'Petit Verdot', 'Zinfandel', 'G-S-M', 'Monica', 'Cabernet Merlot', 'Cabernet Franc-Carmenère', 
+            'Grenache Noir', 'Xinomavro', 'Petite Verdot', 'Tempranillo-Garnacha', 'Carmenère-Cabernet Sauvignon', 
+            'Sangiovese-Cabernet Sauvignon', 'Shiraz-Cabernet', 'Syrah-Grenache-Viognier', 'Cabernet-Shiraz', 'Syrah-Carignan', 
+            'Cabernet-Malbec', 'Merlot-Petite Verdot', 'Duras', 'Aragonês', 'Agiorgitiko', 'Aragonez', 'Alfrocheiro', 'Corvina', 
+            'Alicante Bouschet', 'Tinto del Pais', 'Bobal', 'Susumaniello', 'Grolleau', 'Canaiolo', 'Bastardo', 'Tintilia', 
+            'St. Vincent', 'Caprettone','Black Muscat','Muscadine','Syrah-Viognier', 'Shiraz-Viognier', 'Carcajolu', 
+            'Marselan', 'Malbec-Petit Verdot', 'Grignolino', 'Pinot Noir-Syrah', 'Malbec-Carménère','País', 'Alvarelhão', 
+            'Okuzgozu', 'Tintilia','Mavrodaphne','Tintilia ', 
+    ] 
+
+    Rosé = ['Rosé', 'Rosato', 'Rosado','Portuguese Rosé', 'Prugnolo Gentile'] 
+    
+    Sparkling = ['Champagne Blend', 'Prosecco', 'Sparkling Blend','Portuguese Sparkling',
+                'Cerceal', 'Lambrusco','Lambrusco di Sorbara','Lambrusco Grasparossa',
+                'Torbato', 'Moscadello', 'Passerina', 'Brachetto', 'Ekigaïna', 'Picolit', 
+                'Sacy', 'Moscatel Roxo', 'Debit','Moscato', 'Valdiguié', 'Casavecchia', 
+                'Lambrusco Salamino', 'Moscato Rosa'] 
+    
+    Fortified = ['Sherry', 'Pedro Ximénez', 'White Port', 'Tokaji','Port']
+
+    red_dict = {variety: 'Red' for variety in Red}
+    white_dict = {variety: 'White' for variety in White}
+    rose_dict = {variety: 'Rosé' for variety in Rosé}
+    sparkling_dict = {variety: 'Sparkling' for variety in Sparkling}
+    fortified_dict = {variety: 'Fortified' for variety in Fortified}
+    wine_dict = {**red_dict, **white_dict, **rose_dict, **sparkling_dict, **fortified_dict}
+    
+    # Remove duplicates 
+    df = df.drop_duplicates(subset='description', keep="first")
+    
+    # Apply wine type dict map 
+    df['type'] = df['variety'].map(wine_dict)
+
+    # Fix one missing value:
+    df['type'].fillna('Red', inplace=True)
+        
+    # Rename cols
+    df = df.rename(columns={'country':'Country',
+        'description':'Tasting notes',
+        'designation':'Designation',
+        'points':'Score', 
+        'price': 'Price',
+        'province':'Province',
+        'region_1': 'Region', 
+        'title':'Title', 
+        'variety':'Variety',
+        'winery':'Winery',
+        'embeddings':'embeddings', 
+        'type':'Type'}) 
+    
+    return df