app.py

import gradio as gr
import io
import os
from PIL import Image, ImageDraw
from anthropic import Anthropic
from anthropic.types import TextBlock
from anthropic.types.beta import BetaMessage, BetaTextBlock, BetaToolUseBlock
max_tokens = 4096
import base64
model = 'claude-3-5-sonnet-20241022'
system = """<SYSTEM_CAPABILITY>
* You are utilizing a Windows system with internet access.
* The current date is Monday, November 18, 2024.
</SYSTEM_CAPABILITY>"""

def save_image_or_get_url(image, filename="processed_image.png"):
    if not os.path.isdir("static"):
        os.mkdir("static")
    filepath = os.path.join("static", filename)
    image.save(filepath)
    return filepath

def draw_circle_on_image(image, center, radius=30):
    """
    Draws a circle on the given image using a center point and radius.

    Parameters:
        image (PIL.Image): The image to draw on.
        center (tuple): A tuple (x, y) representing the center of the circle.
        radius (int): The radius of the circle.

    Returns:
        PIL.Image: The image with the circle drawn.
    """
    if not isinstance(center, tuple) or len(center) != 2:
        raise ValueError("Center must be a tuple of two values (x, y).")
    if not isinstance(radius, (int, float)) or radius <= 0:
        raise ValueError("Radius must be a positive number.")

    # Calculate the bounding box for the circle
    bbox = [
        center[0] - radius, center[1] - radius,  # Top-left corner
        center[0] + radius, center[1] + radius  # Bottom-right corner
    ]

    # Create a drawing context
    draw = ImageDraw.Draw(image)

    # Draw the circle
    draw.ellipse(bbox, outline="red", width=15)  # Change outline color and width as needed

    return image


def pil_image_to_base64(pil_image):
    # Save the PIL image to an in-memory buffer as a file-like object
    buffered = io.BytesIO()
    pil_image.save(buffered, format="PNG")  # Specify format (e.g., PNG, JPEG)
    buffered.seek(0)  # Rewind the buffer to the beginning

    # Encode the bytes from the buffer to Base64
    image_data = base64.b64encode(buffered.getvalue()).decode("utf-8")
    return image_data







# Function to simulate chatbot responses
def chatbot_response(input_text, image, key, chat_history):

    if not key:
        return chat_history + [[input_text, "Please enter a valid key."]]
    if image is None:
        return chat_history + [[input_text, "Please upload an image."]]
    new_size = (1512, 982)  # For example, resizing to 800x600 pixels
    image = image.resize(new_size)
    api_key =key
    client = Anthropic(api_key=api_key)



    messages = [{'role': 'user', 'content': [TextBlock(text=f'Look at my screenshot, {input_text}', type='text')]},
                {'role': 'assistant', 'content': [BetaTextBlock(
                    text="I'll help you check your screen, but first I need to take a screenshot to see what you're looking at.",
                    type='text'), BetaToolUseBlock(id='toolu_01PSTVtavFgmx6ctaiSvacCB',
                                                   input={'action': 'screenshot'}, name='computer',
                                                   type='tool_use')]}]
    image_data = pil_image_to_base64(image)

    tool_res = {'role': 'user', 'content': [{'type': 'tool_result', 'tool_use_id': 'toolu_01PSTVtavFgmx6ctaiSvacCB',
                                             'is_error': False,
                                             'content': [{'type': 'image',
                                                          'source': {'type': 'base64', 'media_type': 'image/png',
                                                                     'data': image_data}}]}]}
    messages.append(tool_res)
    params = [{'name': 'computer', 'type': 'computer_20241022', 'display_width_px': 1512, 'display_height_px': 982,
               'display_number': None}, {'type': 'bash_20241022', 'name': 'bash'},
              {'name': 'str_replace_editor', 'type': 'text_editor_20241022'}]
    raw_response = client.beta.messages.with_raw_response.create(
        max_tokens=max_tokens,
        messages=messages,
        model=model,
        system=system,
        tools=params,
        betas=["computer-use-2024-10-22"],
        temperature=0.0,
    )
    response = raw_response.parse()
    scale_x = image.width / 1512
    scale_y = image.height / 982
    for r in response.content:
        if hasattr(r, 'text'):
            chat_history = chat_history + [[input_text, r.text]]

        if hasattr(r, 'input') and 'coordinate' in r.input:
            coordinate = r.input['coordinate']
            new_image = draw_circle_on_image(image, (int(coordinate[0] * scale_x), int(coordinate[1] * scale_y)))

            # Save the image or encode it as a base64 string if needed
            image_url = save_image_or_get_url(
                new_image)  # Define this function to save or generate the URL for the image

            # Include the image as part of the chat history
            image_html = f'<img src="{image_url}" alt="Processed Image" style="max-width: 100%; max-height: 200px;">'
            chat_history = chat_history + [[None, (image_url,)]]
    return chat_history

    # Read the image and encode it in base64





    # Simulated response
    response = f"Received input: {input_text}\nKey: {key}\nImage uploaded successfully!"
    return chat_history + [[input_text, response]]


# Create the Gradio interface
with gr.Blocks() as demo:
    with gr.Row():
        with gr.Column():
            image_input = gr.Image(label="Upload Image", type="pil", interactive=True)
        with gr.Column():
            chatbot = gr.Chatbot(label="Chatbot Interaction", height=400)

    with gr.Row():
        user_input = gr.Textbox(label="Type your message here", placeholder="Enter your message...")
        key_input = gr.Textbox(label="API Key", placeholder="Enter your key...", type="password")

    # Button to submit
    submit_button = gr.Button("Submit")

    # Initialize chat history
    chat_history = gr.State(value=[])

    # Set interactions
    submit_button.click(
        fn=chatbot_response,
        inputs=[user_input, image_input, key_input, chat_history],
        outputs=[chatbot],
    )

# Launch the app
demo.launch()