app.py

# import required packages
import google.generativeai as genai
import os
import PIL.Image
import gradio as gr
from gradio_multimodalchatbot import MultimodalChatbot
from gradio.data_classes import FileData

# Fetch an environment variable.
GG_API_KEY = os.environ.get('GG_API_KEY')
oaiusr = os.environ.get("OAI_USR")
oaipwd = os.environ.get("OAI_PWD")
genai.configure(api_key=GG_API_KEY)

model = genai.GenerativeModel('gemini-1.5-flash-latest')
modelvis = genai.GenerativeModel('gemini-1.5-flash-latest')

def gemini(input, file, chatbot=[]):
    messages = []
    print(chatbot)

    # Process previous chatbot messages if present
    if len(chatbot) != 0:
        for user, bot in chatbot:
            user, bot = user.text, bot.text
            messages.extend([
                {'role': 'user', 'parts': [user]},
                {'role': 'model', 'parts': [bot]}
            ])
        messages.append({'role': 'user', 'parts': [input]})
    else:
        messages.append({'role': 'user', 'parts': [input]})

    try:
        # Process image if file is provided
        if file is not None:
            with PIL.Image.open(file.name) as img:
                message = [{'role': 'user', 'parts': [input, img]}]
                response = modelvis.generate_content(message)
                gemini_video_resp = response.text
                messages.append({'role': 'model', 'parts': [gemini_video_resp]})

                # Construct list of messages in the required format
                user_msg = {"text": input, "files": [{"file": FileData(path=file.name)}]}
                bot_msg = {"text": gemini_video_resp, "files": []}
                chatbot.append([user_msg, bot_msg])
        else:
            response = model.generate_content(messages)
            gemini_resp = response.text

            # Construct list of messages in the required format
            user_msg = {"text": input, "files": []}
            bot_msg = {"text": gemini_resp, "files": []}
            chatbot.append([user_msg, bot_msg])
    except Exception as e:
        # Handling exceptions and raising error to the modal
        print(f"An error occurred: {e}")
        raise gr.Error(e)

    return chatbot, "", None

# Define the Gradio Blocks interface
with gr.Blocks() as demo:
    # Add a centered header using HTML
    gr.HTML("<center><h2>✨Tomoniai's Gemini Chat✨</h2></center>")

    # Initialize the MultimodalChatbot component
    multi = MultimodalChatbot(value=[], label="Gemini", height=400, show_copy_button=True, bubble_full_width=False,  likeable=True)
    
    with gr.Row():
        # Textbox for user input with increased scale for better visibility
        tb = gr.Textbox(show_label=False, scale=4)

        # Upload button for image files
        up = gr.UploadButton("Upload Image", file_types=["image"], scale=1)

    # Define the behavior on text submission
    tb.submit(gemini, [tb, up, multi], [multi, tb, up])

    # Define the behavior on image upload
    # Using chained then() calls to update the upload button's state
    up.upload(lambda: gr.UploadButton("Uploading Image..."), [], up) \
       .then(lambda: gr.UploadButton("Image Uploaded"), [], up) \
       .then(lambda: gr.UploadButton("Upload Image"), [], up)

demo.queue().launch(auth=(oaiusr, oaipwd),show_api=False)