app.py

import gradio as gr
from gradio_client import Client, handle_file
from huggingface_hub import InferenceClient

moondream_client = Client("vikhyatk/moondream2")

llama_client = InferenceClient("Qwen/QwQ-32B-Preview")

history = []

def describe_image(image, user_message):
    global history
    
    result = moondream_client.predict(
        img=handle_file(image),
        prompt="Describe this image.",
        api_name="/answer_question"
    )
    
    description = result  # Moondream2'nin cevabını alıyoruz

    history.append(f"User: {user_message}")
    history.append(f"Assistant: {description}")
    
    full_conversation = "\n".join(history)
    llama_result = llama_client.chat_completion(
        messages=[{"role": "user", "content": full_conversation}],
        max_tokens=512,
        temperature=0.7,
        top_p=0.95
    )
    
    return description + "\n\nAssistant: " + llama_result['choices'][0]['message']['content']

def chat_or_image(image, user_message):
    global history

    if image:
        return describe_image(image, user_message)
    else:
        history.append(f"User: {user_message}")
        full_conversation = "\n".join(history)
        llama_result = llama_client.chat_completion(
            messages=[{"role": "user", "content": full_conversation}],
            max_tokens=512,
            temperature=0.7,
            top_p=0.95
        )
        return llama_result['choices'][0]['message']['content']

demo = gr.Interface(
    fn=chat_or_image,
    inputs=[
        gr.Image(type="filepath", label="Resim Yükle (isteğe bağlı)"),
        gr.Textbox(label="Soru Sor ya da Konuş", placeholder="Soru sor...", lines=2)
    ],
    outputs="text",
)

if __name__ == "__main__":
    demo.launch(show_error=True)