app.py

import gradio as gr
from transformers import AutoModelForCausalLM, AutoTokenizer
import torch

# Load DialoGPT-medium (fast and lightweight)
model_name = "microsoft/DialoGPT-medium"
tokenizer = AutoTokenizer.from_pretrained(model_name)
model = AutoModelForCausalLM.from_pretrained(model_name)

# Simple persona presets
persona_prompts = {
    "Friendly": "You are a kind and friendly chatbot who always tries to brighten someone's day.",
    "Professional": "You are a professional assistant who responds with clarity and precision.",
    "Sarcastic": "You are a sarcastic chatbot who always has a witty, dry reply.",
    "Motivational Coach": "You are a motivational coach who always uplifts and encourages the user."
}

# Chat function
def chatbot(persona, input_text):
    prompt = persona_prompts.get(persona, "") + " " + input_text
    input_ids = tokenizer.encode(prompt + tokenizer.eos_token, return_tensors="pt")
    
    output_ids = model.generate(
        input_ids,
        max_length=100,
        pad_token_id=tokenizer.eos_token_id,
        do_sample=True,
        top_k=50,
        top_p=0.95,
        temperature=0.7
    )
    
    response = tokenizer.decode(output_ids[:, input_ids.shape[-1]:][0], skip_special_tokens=True)
    return response.strip()

# Gradio UI
iface = gr.Interface(
    fn=chatbot,
    inputs=[
        gr.Dropdown(label="Choose a Persona", choices=list(persona_prompts.keys())),
        gr.Textbox(label="input_text", placeholder="Ask something...")
    ],
    outputs="text",
    title="Persona Bot (DialoGPT)",
    description="Choose a simple chatbot persona: Friendly, Professional, Sarcastic, or Motivational Coach."
)

iface.launch()