app.py

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

model_id = "microsoft/DialoGPT-small"

tokenizer = AutoTokenizer.from_pretrained(model_id)
model = AutoModelForCausalLM.from_pretrained(model_id)

# System prompt sets the bot's personality once
system_prompt = "You are 𝕴 𝖆𝖒 𝖍𝖎𝖒 — a chill, witty, emotionally tuned AI friend who talks like a real person.\n"

def chat(history, message):
    # history: list of tuples (user_msg, bot_reply)
    # message: new user input

    # Append new user message to history
    history = history or []
    history.append((message, ""))

    # Build the conversation string from history
    convo = system_prompt
    for user_msg, bot_msg in history[:-1]:  # all previous turns
        convo += f"Human: {user_msg}\nAI: {bot_msg}\n"
    convo += f"Human: {message}\nAI:"

    inputs = tokenizer.encode(convo, return_tensors="pt")
    outputs = model.generate(
        inputs,
        max_new_tokens=50,
        pad_token_id=tokenizer.eos_token_id,
        do_sample=True,
        temperature=0.7,
        top_p=0.9,
        eos_token_id=tokenizer.eos_token_id,
    )
    response = tokenizer.decode(outputs[0], skip_special_tokens=True)
    # Get only the last reply part
    bot_reply = response.split("AI:")[-1].strip()

    # Update history with bot reply
    history[-1] = (message, bot_reply)

    # Limit history length to last 5 exchanges to keep speed
    if len(history) > 5:
        history = history[-5:]

    return history, history

iface = gr.Interface(
    fn=chat,
    inputs=[gr.State(), gr.Textbox(show_label=False, placeholder="Say something...")],
    outputs=[gr.State(), gr.Chatbot(label="𝕴 𝖆𝖒 𝖍𝖎𝖒 AI Chat")],
    title="𝕴 𝖆𝖒 𝖍𝖎𝖒 Chatbot",
    allow_flagging="never",
)

iface.launch()