app.py

import streamlit as st
import pandas as pd
# import CosineDistanceClustering
import PU
import bert_embeddings
import oneclass

# Function to recommend papers
def recommend_papers(positive_df, unlabelled_df, model="tfidf"):
    pass
    
# Streamlit app
def main():
    st.title("ArXiv Feed Recommendations")

    # User input for model selection
    model = st.selectbox("Select Model", ["CosineDistanceClustering", "PU","oneclass"])

    # Upload CSV files
    st.subheader("Upload Positive Labeled CSV")
    positive_file = st.file_uploader("Upload CSV", type=['csv'], key="positive")
    st.subheader("Upload Unlabelled Data CSV")
    unlabelled_file = st.file_uploader("Upload CSV", type=['csv'], key="unlabelled")


    if positive_file is not None and unlabelled_file is not None:
        # Read CSV files
        positive_df = pd.read_csv(positive_file)
        unlabelled_df = pd.read_csv(unlabelled_file)

        # Show uploaded data
        st.subheader("Positive Labeled Data")
        st.write(positive_df)
        st.subheader("Unlabelled Data")
        st.write(unlabelled_df)

        # Button to trigger recommendation
        if st.button("Recommend"):
            # Call recommend_papers function
            recommended_indices = recommend_papers(positive_df, unlabelled_df, model.lower())
            st.write(recommended_indices)

            # Download CSV
            st.markdown(get_csv_download_link(recommended_indices), unsafe_allow_html=True)

# Function to generate a download link for CSV
def get_csv_download_link(data):
    csv = data.to_csv(index=False)
    b64 = base64.b64encode(csv.encode()).decode()  
    href = f'<a href="data:file/csv;base64,{b64}" download="recommendations.csv">Download CSV</a>'
    return href

if __name__ == "__main__":
    main()