import gradio as gr
from llama_cpp import Llama
from huggingface_hub import hf_hub_download

# GGUF 다운로드
model_path = hf_hub_download(
    repo_id="HaGPT/news-intelligence-chatbot",
    filename="meta-llama-3.1-8b.Q4_K_M.gguf",
)

llm = Llama(
    model_path=model_path,
    n_ctx=4096,
    n_threads=6,
    n_gpu_layers=0,
    verbose=False
)

def chat_fn(message, history):
    messages = []
    for user, bot in history:
        messages.append({"role": "user", "content": user})
        messages.append({"role": "assistant", "content": bot})

    messages.append({"role": "user", "content": message})

    response = llm.create_chat_completion(
        messages=messages,
        max_tokens=512,
        temperature=0.7
    )

    answer = response["choices"][0]["message"]["content"]
    return answer

interface = gr.ChatInterface(
    fn=chat_fn,
    title="📰 News Intelligence Bot",
    description="Your local GGUF chatbot running on llama.cpp"
)

interface.launch()