app.py

import gradio as gr
from transformers import pipeline

# ✅ Load FLAN-T5 model for polite, instruction-based replies
generator = pipeline("text2text-generation", model="google/flan-t5-small")

# 📌 Instruction template
TEMPLATE = (
    "You are a helpful, polite, and professional customer support agent. "
    "Reply to this customer message in a brand-consistent tone:\n\n"
    "{input}"
)

# 🔁 Reply Generator
def generate_reply(user_input):
    prompt = TEMPLATE.format(input=user_input)
    response = generator(prompt, max_length=150, do_sample=False)[0]["generated_text"]
    return response.strip()

# 🎛️ Gradio Interface
iface = gr.Interface(
    fn=generate_reply,
    inputs=gr.Textbox(lines=6, label="Customer Message", placeholder="Enter complaint or question..."),
    outputs=gr.Textbox(label="Auto-Generated Support Reply"),
    title="🤖 Auto-Reply Generator for Customer Support",
    description=(
        "Generate fast, polite, and professional replies to customer queries using Google's FLAN-T5 model. "
        "Perfect for CRM bots, helpdesk automation, and ticket response."
    ),
    examples=[
        ["I still haven't received my order and it's been 10 days."],
        ["My refund hasn't been processed yet."],
        ["Your app keeps crashing on my iPhone."],
        ["Great service, just wanted to say thanks!"]
    ]
)

iface.launch()