app.py

import base64
import os
import mimetypes
from google import genai
from google.genai import types
import gradio as gr
import io
from PIL import Image

def save_binary_file(file_name, data):
    f = open(file_name, "wb")
    f.write(data)
    f.close()

def generate_image(prompt, image=None, output_filename="generated_image"):
    # Initialize client with the API key
    client = genai.Client(
        api_key="AIzaSyAQcy3LfrkMy6DqS_8MqftAXu1Bx_ov_E8",
    )

    model = "gemini-2.0-flash-exp-image-generation"
    parts = [types.Part.from_text(text=prompt)]
    
    # If an image is provided, add it to the content
    if image:
        # Convert PIL Image to bytes
        img_byte_arr = io.BytesIO()
        image.save(img_byte_arr, format="PNG")
        img_bytes = img_byte_arr.getvalue()
        # Add the image as a Part with inline_data
        parts.append({
            "inline_data": {
                "mime_type": "image/png",
                "data": img_bytes
            }
        })

    contents = [
        types.Content(
            role="user",
            parts=parts,
        ),
    ]
    generate_content_config = types.GenerateContentConfig(
        temperature=1,
        top_p=0.95,
        top_k=40,
        max_output_tokens=8192,
        response_modalities=[
            "image",
            "text",
        ],
        safety_settings=[
            types.SafetySetting(
                category="HARM_CATEGORY_CIVIC_INTEGRITY",
                threshold="OFF",
            ),
        ],
        response_mime_type="text/plain",
    )

    # Generate the content
    response = client.models.generate_content_stream(
        model=model,
        contents=contents,
        config=generate_content_config,
    )

    # Process the response
    for chunk in response:
        if not chunk.candidates or not chunk.candidates[0].content or not chunk.candidates[0].content.parts:
            continue
        if chunk.candidates[0].content.parts[0].inline_data:
            inline_data = chunk.candidates[0].content.parts[0].inline_data
            file_extension = mimetypes.guess_extension(inline_data.mime_type)
            filename = f"{output_filename}{file_extension}"
            save_binary_file(filename, inline_data.data)
            
            # Convert binary data to PIL Image for Gradio display
            img = Image.open(io.BytesIO(inline_data.data))
            return img, f"Image saved as {filename}"
        else:
            return None, chunk.text

    return None, "No image generated"

# Function to handle chat interaction
def chat_handler(prompt, user_image, chat_history, output_filename="generated_image"):
    # Add the prompt to the chat history
    if prompt:
        chat_history.append({"role": "user", "content": prompt})
    
    # If no input, return early
    if not prompt and not user_image:
        chat_history.append({"role": "assistant", "content": "Please provide a prompt or an image."})
        return chat_history, user_image, None, ""

    # Generate image based on user input
    img, status = generate_image(prompt or "Generate an image", user_image, output_filename)
    
    # Add the status message to the chat history
    chat_history.append({"role": "assistant", "content": status})
    
    return chat_history, user_image, img, ""

# Create Gradio interface
with gr.Blocks(title="Image Editing Chatbot") as demo:
    gr.Markdown("# Image Editing Chatbot")
    gr.Markdown("Upload an image and/or type a prompt to generate or edit an image using Google's Gemini model")
    
    # Chatbot display area for text messages
    chatbot = gr.Chatbot(
        label="Chat",
        height=300,
        type="messages",
        avatar_images=(None, None)
    )
    
    # Separate image outputs
    with gr.Row():
        uploaded_image_output = gr.Image(label="Uploaded Image")
        generated_image_output = gr.Image(label="Generated Image")
    
    # Input area
    with gr.Row():
        with gr.Column():
            image_input = gr.Image(
                label="Upload Image",
                type="pil",
                scale=1,
                height=100
            )
            prompt_input = gr.Textbox(
                label="Prompt",
                placeholder="Enter your image description here...",
                lines=3
            )
            filename_input = gr.Textbox(
                label="Output Filename",
                value="generated_image",
                placeholder="Enter desired filename (without extension)"
            )
            generate_btn = gr.Button("Generate Image")
    
    # State to maintain chat history
    chat_state = gr.State([])

    # Connect the button to the chat handler
    generate_btn.click(
        fn=chat_handler,
        inputs=[prompt_input, image_input, chat_state, filename_input],
        outputs=[chatbot, uploaded_image_output, generated_image_output, prompt_input]
    )

    # Also allow Enter key to submit
    prompt_input.submit(
        fn=chat_handler,
        inputs=[prompt_input, image_input, chat_state, filename_input],
        outputs=[chatbot, uploaded_image_output, generated_image_output, prompt_input]
    )

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