app.py

import pandas as pd
import gradio as gr
from sentence_transformers import SentenceTransformer
import faiss
import numpy as np

# Step 1: Load train.csv directly
csv_path = "train.csv"  # file you uploaded directly
df = pd.read_csv(csv_path, header=None, names=["label", "title", "description"])

# Step 2: Combine title and description
df["content"] = df["title"].fillna("") + ". " + df["description"].fillna("")

# Step 3: Encode content using SentenceTransformer
model = SentenceTransformer('all-MiniLM-L6-v2')
corpus_embeddings = model.encode(df["content"].tolist(), show_progress_bar=True)

# Step 4: Build FAISS index
embedding_dim = corpus_embeddings.shape[1]
index = faiss.IndexFlatL2(embedding_dim)
index.add(corpus_embeddings)

# Step 5: Define retrieval function
def retrieve_and_respond(claim, k=5):
    query_embedding = model.encode([claim])
    D, I = index.search(np.array(query_embedding), k)
    
    results = []
    for idx in I[0]:
        row = df.iloc[idx]
        results.append(f"*Title:* {row['title']}\n*Description:* {row['description']}\n*Label:* {row['label']}\n")
    
    return "\n\n".join(results)

# Step 6: Gradio UI
iface = gr.Interface(
    fn=retrieve_and_respond,
    inputs=gr.Textbox(lines=2, placeholder="Enter a news-related claim here..."),
    outputs="markdown",
    title="Claim Verifier using AG News",
    description="Enter a claim and get the most relevant AG News articles to help verify or refute it."
)

iface.launch()