app.py

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

MODEL = "TheBloke/MythoMax-L2-13B-GGUF"  # GGUF quantized version (lighter)

tokenizer = AutoTokenizer.from_pretrained(MODEL)
pipeline_chat = pipeline(
    "text-generation",
    model=MODEL,
    tokenizer=tokenizer,
    device_map="auto",
    max_new_tokens=512,
    temperature=0.9,
    top_p=0.95,
)

def chat_fn(message, history):
    prompt = ""
    for user, bot in history:
        prompt += f"User: {user}\nAssistant: {bot}\n"
    prompt += f"User: {message}\nAssistant:"

    output = pipeline_chat(prompt)[0]["generated_text"]
    reply = output.split("Assistant:")[-1].strip()
    return reply

demo = gr.ChatInterface(chat_fn, title="💖 MythoMax Virtual GF")

demo.launch()