app.py

import pandas as pd
import gradio as gr
from datetime import datetime
from sentence_transformers import SentenceTransformer
from sklearn.metrics.pairwise import cosine_similarity
import re


# Load dataset 
df = pd.read_csv("analyticsvidhyacomplete.csv", parse_dates=["Date"])

# Preprocessing
df['Date'] = pd.to_datetime(df['Date'], format='mixed', dayfirst=True, errors='coerce')
df["combined_text"] = df["Title"].astype(str) + " " + df["Description"].astype(str)  + " " + df["Content"].astype(str)

# Loading query CSV
query_df = pd.read_csv("query.csv")
query_df.dropna(subset=["Topic", "Subtopic", "TopN"], inplace=True)

# dropdown options
query_df["QueryOption"] = query_df.apply(
    lambda row: f"{row['Topic']} - {row['Subtopic']} (TopN: {int(row['TopN'])})", axis=1
)

query_options = query_df["QueryOption"].tolist()

# Load model
model = SentenceTransformer("all-MiniLM-L6-v2")

text_embeddings = model.encode(df["combined_text"].tolist(), convert_to_tensor=False)



def retrieve_records(selected_query):

    match = re.match(r"(.+?) - (.+?) \(TopN: (\d+)\)", selected_query)
    if not match:
        return "Invalid query format selected."
    
    topic, subtopic, top_n = match.groups()
    top_n = int(top_n)
    full_query = f"{topic} {subtopic}"
    
    query_embedding = model.encode([full_query], convert_to_tensor=False)
    scores = cosine_similarity(query_embedding, text_embeddings).flatten()
    df["similarity"] = scores

    top_results = df.sort_values(by=["similarity", "Date"], ascending=[False, False]).head(top_n)


    
    # Format markdown output
    markdown_output = ""
    for _, row in top_results.iterrows():
        markdown_output += f"### [{row['Title']}]({row['Link']})\n"
        markdown_output += f"**Date**: {row['Date'].strftime('%Y-%m-%d')}\n\n"
        markdown_output += f"{row['Description']}\n\n---\n"

    return markdown_output




iface = gr.Interface(
    fn=retrieve_records,
    inputs=[
        gr.Dropdown(choices=query_options, label="Select a query"),
    ],
    outputs=gr.Markdown(label="Top Similar Records"),
    title="Top-N Article Retriever"
)


if __name__ == "__main__":
    iface.launch()