from transformers import AutoTokenizer, AutoModelForSeq2SeqLM
import gradio as gr

model_id = "google/flan-t5-small"  # Extremely fast and CPU-friendly
tokenizer = AutoTokenizer.from_pretrained(model_id)
model = AutoModelForSeq2SeqLM.from_pretrained(model_id)

system_prompt = "You are 𝕴 𝖆𝖒 𝖍𝖎𝖒 — a fun, fast, emotionally tuned AI chatbot created by 𝕴 𝖆𝖒 𝖍𝖎𝖒. You reply quickly, like a chill and clever human friend."

def chat(history, message):
    prompt = f"{system_prompt}\nUser: {message}\nAI:"
    inputs = tokenizer(prompt, return_tensors="pt")
    outputs = model.generate(**inputs, max_new_tokens=100)
    reply = tokenizer.decode(outputs[0], skip_special_tokens=True).strip()

    history = history or []
    history.append((message, reply))
    return history, history

iface = gr.Interface(
    fn=chat,
    inputs=[gr.State(), gr.Textbox(placeholder="Talk to 𝕴 𝖆𝖒 𝖍𝖎𝖒...")],
    outputs=[gr.State(), gr.Chatbot(label="𝕴 𝖆𝖒 𝖍𝖎𝖒 AI Chatbot")],
    title="𝕴 𝖆𝖒 𝖍𝖎𝖒 — Superfast Chatbot",
    description="An extremely fast and chill AI chatbot, created by 𝕴 𝖆𝖒 𝖍𝖎𝖒. Running on Hugging Face Spaces (CPU only).",
    allow_flagging="never"
)

iface.launch()