import gradio as gr
from PIL import Image
from inference.main import MultiModalPhi2

multimodal_phi2 = MultiModalPhi2(
    modelname_or_path="Navyabhat/Llava-Phi2",
    temperature=0.2,
    max_new_tokens=1024,
    device="cpu",
)

theme = gr.themes.Default(primary_hue="blue").set(
    loader_color="#FF0000",
    button_primary_background_fill="*primary_200",
    button_primary_background_fill_hover="*primary_300",
)

def add_content(chatbot, text, image, audio_upload, audio_mic) -> gr.Chatbot:
    textflag, imageflag, audioflag = False, False, False
    if text not in ["", None]:
        chatbot.append((text, None))
        textflag = True
    if image is not None:
        chatbot.append(((image,), None))
        imageflag = True
    if audio_mic is not None:
        chatbot.append(((audio_mic,), None))
        audioflag = True
    else:
        if audio_upload is not None:
            chatbot.append(((audio_upload,), None))
            audioflag = True
    if not any([textflag, imageflag, audioflag]):
        # Raise an error if neither text nor file is provided
        raise gr.Error("Enter a valid text, image or audio")
    return chatbot

def clear_data():
    return {prompt: None, image: None, audio_upload: None, audio_mic: None, chatbot: []}


def run(history, text, image, audio_upload, audio_mic):
    if text in [None, ""]:
        text = None

    if audio_upload is not None:
        audio = audio_upload
    elif audio_mic is not None:
        audio = audio_mic
    else:
        audio = None

    print("text", text)
    print("image", image)
    print("audio", audio)

    if image is not None:
        image = Image.open(image)
    outputs = multimodal_phi2(text, audio, image)
    # outputs = ""

    history.append((None, outputs.title()))
    return history, None, None, None, None
    
with gr.Blocks(theme=theme) as demo:
    gr.Markdown("## 🤖 Multi-modal LLM")
    gr.Markdown("This is a multi-modal LLM that takes text, image and audio as inputs.")

    with gr.Row():
        # with gr.Column(scale=4):
            
        #     with gr.Box():
        #         with gr.Row():
        #             # Adding image
        #             image = gr.Image(type="filepath", value=None)
                
        #         # with gr.Row():
        #             # Add audio
        #             audio_upload = gr.Audio(source="upload", type="filepath")
        #             audio_mic = gr.Audio(
        #                 source="microphone", type="filepath", format="mp3"
        #             )

        # with gr.Column(scale=8):
        #     with gr.Box():
        #         with gr.Row():
        chatbot = gr.Chatbot(
            avatar_images=("🧑", "🤖"),
            height=560,
        )

    with gr.Row():
        image = gr.Image(type="filepath", value=None)

        audio_upload = gr.Audio(source="upload", type="filepath")
        audio_mic = gr.Audio(
            source="microphone", type="filepath", format="mp3"
        )
        
    with gr.Row():
        prompt = gr.Textbox(
            placeholder="Ask anything", lines=2, label="Query", value=None, scale=4
        )
       
    with gr.Row():
        # Adding a Button
        submit = gr.Button(value = "Submit", variant="primary")
        clear = gr.Button(value="Clear")

    submit.click(
        add_content,
        inputs=[chatbot, prompt, image, audio_upload, audio_mic],
        outputs=[chatbot],
    ).success(
        run,
        inputs=[chatbot, prompt, image, audio_upload, audio_mic],
        outputs=[chatbot, prompt, image, audio_upload, audio_mic],
    )

    clear.click(
        clear_data,
        outputs=[prompt, image, audio_upload, audio_mic, chatbot],
    )

demo.launch()