app.py

import streamlit as st
import pandas as pd
import numpy as np
import torch
import ast
from sentence_transformers import SentenceTransformer
from sklearn.metrics.pairwise import cosine_similarity
import random
import asyncio
import os
os.environ["STREAMLIT_WATCHDOG"] = "0"

try:
    asyncio.get_running_loop()
except RuntimeError:
    asyncio.set_event_loop(asyncio.new_event_loop())


# Load pre-trained model
model = SentenceTransformer('all-MiniLM-L6-v2')

def get_embedding(text):
    return model.encode(text, convert_to_numpy=True)

# Load datasets
tweets_df = pd.read_csv("dataset_with_embeddings.csv")  # Contains 'tweet_id' and 'sentence_embedding'
original_tweets_df = pd.read_csv("dataset_with_actual_text.csv")  # Contains 'tweet_id' and 'original_text'

# Convert embeddings from string to list

def parse_embedding(embedding_str):
    try:
        return ast.literal_eval(embedding_str)
    except (ValueError, SyntaxError):
        return None  # Handle errors gracefully


tweets_df['sentence_embedding'] = tweets_df['sentence_embedding'].astype(str).apply(parse_embedding)

# Simulated user database
users = {"testuser": "password123"}  # Dictionary to store usernames and passwords

# Session state for login
if "logged_in" not in st.session_state:
    st.session_state["logged_in"] = False
    st.session_state["username"] = ""
    st.session_state["liked_tweets"] = []
    st.session_state["posted_tweets"] = []

# Login Page
if not st.session_state["logged_in"]:
    st.title("Tweet Recommendation System")
    username = st.text_input("Username")
    password = st.text_input("Password", type="password")
    if st.button("Login"):
        if username in users and users[username] == password:
            st.session_state["logged_in"] = True
            st.session_state["username"] = username
            st.success("Login successful!")
            st.experimental_rerun()
        else:
            st.error("Invalid credentials")
else:
    st.title(f"Welcome, {st.session_state['username']}!")
    
    # Randomly select 10 tweets for the Explore Page
    random_tweets = original_tweets_df.sample(10)
    
    st.subheader("Explore Tweets")
    for _, row in random_tweets.iterrows():
        tweet_id = row['tweet_id']
        text = row['original_text']
        
        with st.container():
            st.write(f"*Tweet:* {text}")
            if st.button(f"Like ❤ {tweet_id}", key=f"like_{tweet_id}"):
                if tweet_id not in st.session_state["liked_tweets"]:
                    st.session_state["liked_tweets"].append(tweet_id)
                st.success("Tweet Liked!")
    
    # Posting new tweets
    st.subheader("Post a Tweet")
    new_tweet = st.text_area("Write your tweet here:")
    if st.button("Post Tweet"):
        if new_tweet:
            new_tweet_embedding = get_embedding(new_tweet)
            st.session_state["posted_tweets"].append(new_tweet_embedding)
            st.success("Tweet posted successfully!")
    
    # Recommendation Logic
    if st.session_state["liked_tweets"] or st.session_state["posted_tweets"]:
        st.subheader("Recommended Tweets for You")
        
        # Get embeddings of liked tweets
        liked_embeddings = [tweets_df[tweets_df['tweet_id'] == tid]['sentence_embedding'].values[0] 
                            for tid in st.session_state["liked_tweets"]]
        
        # Get embeddings of posted tweets
        posted_embeddings = st.session_state["posted_tweets"]
        
        # Combine all embeddings
        user_profile_embedding = np.mean(liked_embeddings + posted_embeddings, axis=0)
        
        # Compute cosine similarity
        all_embeddings = np.vstack(tweets_df['sentence_embedding'].values)
        similarities = cosine_similarity([user_profile_embedding], all_embeddings)[0]
        
        # Get top 10 similar tweets
        top_indices = np.argsort(similarities)[-10:][::-1]
        recommended_tweets = tweets_df.iloc[top_indices]
        
        # Display recommended tweets
        for _, row in recommended_tweets.iterrows():
            tweet_id = row['tweet_id']
            text = original_tweets_df[original_tweets_df['tweet_id'] == tweet_id]['original_text'].values[0]
            with st.container():
                st.write(f"*Recommended Tweet:* {text}")