app.py

import streamlit as st
import pandas as pd
import os
from dotenv import load_dotenv
from langchain_openai import OpenAIEmbeddings
from langchain.chains import RetrievalQA
from langchain_openai import OpenAI
import lancedb
from langchain_community.vectorstores import LanceDB
from langchain_core.documents import Document

# Page configuration
st.set_page_config(page_title="Film Buff", page_icon="🎬", layout="wide")
st.header('Film Buff 🎬 - Share your favorite movie and get personalized recommendations', divider=True)

# Load environment variables and initialize embeddings
load_dotenv()
openai_api_key = os.environ['OPENAI_API_KEY']
embeddings = OpenAIEmbeddings()

# Load movie dataset
md = pd.read_pickle('movies.pkl')

# Initialize LanceDB connection
db = lancedb.connect("data/sample-lancedb")
# Check if table exists before creating
if "movies" not in db.table_names():
    table_name = db.create_table("movies", md)
table = db.open_table('movies')

# First, create a new DataFrame with the required schema
search_df = pd.DataFrame({
    'Genres': md['genres'],
    'title': md['title'],
    'Overview': md['overview'],
    'Rating': md['weighted_rate'],
    'text': md['text'],
    'adult': md['adult'],
    'vector': md['vector'],  # The pre-computed embeddings
    'metadata': [{'title': title, 'genres': genres} 
            for title, genres in zip(md['title'], md['genres'])]
})

# Drop the existing table if it exists
if "movies_search" in db.table_names():
    db.drop_table("movies_search")

# Create a new table with the correct schema
table = db.create_table("movies_search", data=search_df)

# Create the vector store
docsearch = LanceDB(
    connection=db,
    embedding=embeddings,
    table_name="movies_search"
)
# Sidebar for user preferences
st.sidebar.title("Film Buff Recommendation System")
st.sidebar.markdown("Please enter your details and preferences below:")

# User input collection
age = st.sidebar.slider("What is your age?", 1, 100, 25)
gender = st.sidebar.radio("What is your gender?", ["Male", "Female", "Other"])
unique_genres = sorted(md.explode('genres')["genres"].unique())
genre = st.sidebar.selectbox("What is your favourite film genre?", unique_genres)

# Filter movies based on selected genre
df_filtered = md[md['genres'].apply(lambda x: genre in x)]

# Prompt template for recommendation system
PROMPT_TEMPLATE = """
You are a film recommender system helping users find films matching their preferences.
Context: {context}

User Profile:
Age: {age}
Gender: {gender}

Question: {question}
Your response:"""

# Initialize QA chain
qa = RetrievalQA.from_chain_type(
    llm=OpenAI(),
    chain_type="stuff",
    retriever=docsearch.as_retriever(search_kwargs={'data': df_filtered}),
    return_source_documents=True
)

# User query interface
query = st.text_input(
    'Enter your question:',
    placeholder='What action films do you suggest?'
)

# Add a search button
search_button = st.button('Get Recommendations 🎬')

# Only process query when button is clicked
if search_button and query:
    with st.spinner('Searching for the perfect film...'):
        result = qa.invoke({
            "query": query,
            "context": PROMPT_TEMPLATE.format(
                context="",
                age=age,
                gender=gender,
                question=query
            )
        })
        st.success('Recommendations found!')
        st.write(result['result'])
    
# Optional: Add a message when no query is entered
elif search_button and not query:
    st.warning('Please enter a film-related question first!')