app.py

import gradio as gr
import pandas as pd
from sentence_transformers import SentenceTransformer, util

# Load FAQ data
faq_df = pd.read_csv("lic_faq.csv")
model = SentenceTransformer('all-MiniLM-L6-v2')
faq_embeddings = model.encode(faq_df['question'].tolist(), convert_to_tensor=True)

def chatbot(query):
    query_embedding = model.encode(query, convert_to_tensor=True)
    scores = util.pytorch_cos_sim(query_embedding, faq_embeddings)[0]
    best_score = float(scores.max())  # Get max similarity score
    best_idx = int(scores.argmax())

    if best_score < 0.6:  # Threshold can be adjusted
        return (
            "🤖 I'm not confident I have the right answer for that.\n"
            "Please ask a question related to LIC policies, claims, onboarding, or commissions."
        )
    else:
        return faq_df.iloc[best_idx]['answer']


with gr.Blocks(title="LIC Agent Assistant") as demo:
    gr.Markdown(
        """
        <h1 style='text-align: center; color: #1a237e;'>🧑‍💼 LIC Agent Assistant Chatbot</h1>
        <p style='text-align: center;'>Ask questions about LIC policies, commissions, claims, onboarding, and more!</p>
        """,
        elem_id="header"
    )

    with gr.Row():
        with gr.Column(scale=1):
            user_input = gr.Textbox(
                label="Your Question",
                placeholder="E.g., How do I file a claim?",
                lines=2
            )
            submit_btn = gr.Button("Get Answer", variant="primary")

        with gr.Column(scale=1):
            output = gr.Textbox(
                label="Answer",
                placeholder="Response will appear here...",
                lines=6
            )

    submit_btn.click(fn=chatbot, inputs=user_input, outputs=output)

demo.launch()