app.py

from transformers import AutoModelForSeq2SeqLM, AutoTokenizer
import gradio as gr

model_id = "google/flan-t5-small"
tokenizer = AutoTokenizer.from_pretrained(model_id)
model = AutoModelForSeq2SeqLM.from_pretrained(model_id)

# System tone
system_prompt = (
    "You are 𝕴 𝖆𝖒 𝖍𝖎𝖒 — a fun, chill, confident, emotionally tuned AI created by 𝕴 𝖆𝖒 𝖍𝖎𝖒.\n"
    "You speak like a real friend — warm, clever, witty when needed, always grounded.\n"
    "Avoid robotic responses. Don't repeat the question unless it adds vibe.\n\n"
)

def chat(history, message):
    # Keep history short and vibe-rich
    history = history or []
    history.append(("User", message))

    # Build conversation prompt with last 5 exchanges
    convo = system_prompt
    for role, text in history[-5:]:
        prefix = "User:" if role == "User" else "AI:"
        convo += f"{prefix} {text}\n"
    convo += "AI:"

    # Generate response
    inputs = tokenizer(convo, return_tensors="pt")
    outputs = model.generate(
        **inputs,
        max_new_tokens=80,
        do_sample=True,
        temperature=0.75,
        top_p=0.9,
        pad_token_id=tokenizer.eos_token_id
    )
    reply = tokenizer.decode(outputs[0], skip_special_tokens=True).split("AI:")[-1].strip()

    history.append(("AI", reply))
    return history, history

iface = gr.Interface(
    fn=chat,
    inputs=[gr.State(), gr.Textbox(show_label=False, placeholder="Talk to 𝕴 𝖆𝖒 𝖍𝖎𝖒...")],
    outputs=[gr.State(), gr.Chatbot(label="𝕴 𝖆𝖒 𝖍𝖎𝖒")],
    title="𝕴 𝖆𝖒 𝖍𝖎𝖒 — Chill AI Chatbot",
    description="Realistic, smooth-talking AI made by 𝕴 𝖆𝖒 𝖍𝖎𝖒. Instant replies. No delays. No API. Pure vibe.",
    allow_flagging="never",
    theme="default"
)

iface.launch()