import gradio as gr from gradio_client import Client import os import logging import requests from huggingface_hub import login, whoami, HfFolder, HfApi, SpaceHardware import torch webhook_server = os.getenv('webhook_server') API_TOKEN = os.getenv('ZeroGPU') #BASE_URL = os.getenv('SpaceURL') # Get user information #ser_info = whoami(token=API_TOKEN) # Set your token login(token=API_TOKEN, add_to_git_credential=True) # Initialize API client api = HfApi(token=API_TOKEN) # Set up logging logging.basicConfig(level=logging.INFO) logger = logging.getLogger(__name__) # Initialize the client for image generation "prithivMLmods/IMAGINEO-4K" "mukaist/DALLE-4K" client_image = Client("measmonysuon/DALLE-4K") # Get user information to confirm authentication user_info = whoami(token=API_TOKEN) print("User Info:", user_info) # Define resolutions and default style resolutions = { "896x1152": (896, 1152), "1024x1024": (1024, 1024), "1216x832": (1216, 832) } DEFAULT_STYLE = "3840 x 2160" def generate_image(prompt, resolution_key, style=DEFAULT_STYLE): resolution = resolutions.get(resolution_key, (1024, 1024)) width, height = resolution full_prompt = f"{prompt}, Canon EOS R5, 4K, Photo-Realistic, appearing photorealistic with super fine details, high resolution, natural look, hyper realistic photography, cinematic lighting, --ar 64:37, --v 6.0, --style raw, --stylize 750" try: result = client_image.predict( prompt=full_prompt, negative_prompt="(deformed, distorted, disfigured:1.3), poorly drawn, bad anatomy, wrong anatomy, extra limb, missing limb, floating limbs, (mutated hands and fingers:1.4), disconnected limbs, mutation, mutated, ugly, disgusting, blurry, amputation", use_negative_prompt=True, style=style, seed=0, width=width, height=height, guidance_scale=5, randomize_seed=True, api_name="/run" ) logger.info("Image generation successful.") return result except Exception as e: logger.error(f"Error generating image: {e}") return None def gradio_interface(prompt, resolution_key, user_chat_id): if not user_chat_id.strip(): return None, "User Chat ID cannot be blank. Please provide a valid User Chat ID or register with @supportBot to get UID." # Request large hardware for image generation api.request_space_hardware(repo_id='measmonysuon/DALLE-4K', hardware='a10g-large') # Get GPU info gpu_info = get_gpu_info() try: # Generate the image result = generate_image(prompt, resolution_key) if result and result[0]: file_path = result[0][0].get('image') if file_path: # Combine base URL with file path to create full URL base_url = "https://measmonysuon-imagen.hf.space/file=" full_file_url = base_url + file_path if os.path.exists(file_path): # Update user points and notify webhook update_user_points(user_chat_id, -5) notify_webhook_with_file(user_chat_id, full_file_url, gpu_info) # Notify webhook with full file URL return full_file_url, f"Check your telegram, The image was generated successfully. {gpu_info}" return None, "The image file is not available. Please try again later." return None, "The image file is not available. Please try again later." return None, "There was an error processing your photo. Please try again later." finally: # Log GPU info before resetting hardware logger.info(f"GPU info before reset: {gpu_info}") # Always request to switch back hardware api.request_space_hardware(repo_id='measmonysuon/DALLE-4K', hardware='zero-a10g') logger.info("Hardware reset to Zero-A10G") def handle_generate_image(prompt, resolution_key, user_chat_id): points = get_user_points(user_chat_id) if isinstance(points, str): if points == "User not found": return None, "Please register with https://t.me/myexpsupportBot" return None, "Error retrieving points. Please try again later." try: points = int(points) except ValueError: return None, "Error processing points. Please try again later." if points >= 5: result = gradio_interface(prompt, resolution_key, user_chat_id) if result[0]: # Include the GPU info in the success message gpu_info = result[1] # Assuming gradio_interface returns GPU info in the second part of the tuple return result[0], f"Check your telegram, The image was generated successfully. {gpu_info}" return None, "There was an error processing your photo. Please try again later." return None, "Insufficient points. Please get more points before generating an image." def get_gpu_info(): if torch.cuda.is_available(): device_name = torch.cuda.get_device_name(0) device_properties = torch.cuda.get_device_properties(0) total_memory = f"{device_properties.total_memory // (1024 ** 2)} MB" return { 'is_cuda_available': True, 'device_name': device_name, 'total_memory': total_memory } else: return { 'is_cuda_available': False, 'device_name': 'N/A', 'total_memory': 'N/A' } def request_otp(user_chat_id): try: response = requests.post(f"{webhook_server}/request_otp", json={"user_chat_id": user_chat_id}) response.raise_for_status() logger.debug(f"Response status code: {response.status_code}") logger.debug(f"Response content: {response.text}") response_data = response.json() if 'error' in response_data: logger.error(f"Error response from /request_otp: {response_data['error']}") return {"success": False, "error": response_data.get("error", "Unknown error")} return {"success": response_data.get("success", False)} except requests.RequestException as e: logger.error(f"Request exception: {e}") return {"success": False, "error": "Request exception occurred"} except ValueError as e: logger.error(f"JSON decode error: {e}") return {"success": False, "error": "Invalid JSON response"} def verify_otp(user_chat_id, otp): try: response = requests.post(f"{webhook_server}/verify_otp", json={"user_chat_id": user_chat_id, "otp": otp}) response.raise_for_status() return response.json().get("valid", False) except requests.RequestException as e: logger.error(f"Request exception: {e}") return False except ValueError as e: logger.error(f"JSON decode error: {e}") return False def retrieve_user_points(user_chat_id, otp): if not verify_otp(user_chat_id, otp): return None, "Invalid or expired OTP. Please request a new OTP." response = requests.get(f"{webhook_server}/get_points", params={"user_chat_id": user_chat_id}) if response.status_code == 200: data = response.json() points = data.get("points", 0) return points, "OTP verification and points retrieved successfully. Now, enter pormpt and generate image" return None, "Error retrieving points. Please try again later." def update_user_points(user_chat_id, points): webhook_url = f"{webhook_server}/update_user_points" payload = {'user_chat_id': user_chat_id, 'points': points} try: response = requests.post(webhook_url, json=payload) response.raise_for_status() logger.info(f"User with ID {user_chat_id} had their points updated by {points}. Webhook response: {response.status_code}") except requests.RequestException as e: logger.error(f"Error updating user points via webhook: {e}") def get_user_points(user_chat_id): try: response = requests.get(f"{webhook_server}/get_points", params={'user_chat_id': user_chat_id}) response.raise_for_status() data = response.json() if 'error' in data: return "User not found" return data.get('points', 0) except requests.RequestException as e: logger.error(f"Error retrieving user points: {e}") return "Error retrieving points" def notify_webhook(user_chat_id): webhook_url = f"{webhook_server}/notify_update" payload = {'user_chat_id': user_chat_id} try: response = requests.post(webhook_url, json=payload) response.raise_for_status() logger.info(f"Webhook notification sent successfully. Response: {response.status_code}") except requests.RequestException as e: logger.error(f"Error sending webhook notification: {e}") def notify_webhook_with_file(user_chat_id, full_file_url, gpu_info): webhook_url = f"{webhook_server}/notify_image" payload = { 'user_chat_id': user_chat_id, 'file_path': full_file_url, 'gpu_info': gpu_info } try: response = requests.post(webhook_url, json=payload) response.raise_for_status() logger.info(f"Webhook notification sent successfully for user {user_chat_id}. Response: {response.status_code}") except requests.RequestException as e: logger.error(f"Error sending webhook notification with file URL: {e}") ''' def gradio_interface(prompt, resolution_key, user_chat_id): if not user_chat_id.strip(): return None, "User Chat ID cannot be blank. Please provide a valid User Chat ID." result = generate_image(prompt, resolution_key) if result and result[0]: file_path = result[0][0].get('image') if file_path and os.path.exists(file_path): update_user_points(user_chat_id, -5) notify_webhook_with_file(user_chat_id, file_path) # Notify webhook with file path return file_path, "The image was generated successfully." return None, "The image file is not available. Please try again later." return None, "There was an error processing your photo. Please try again later." ''' def request_otp_and_update_ui(user_chat_id): # Validate user_chat_id if not user_chat_id.strip(): return "User Chat ID cannot be blank. Please provide a valid User Chat ID or register with @supportBot to get UID." otp_response = request_otp(user_chat_id) if otp_response.get("success"): return "OTP has been sent to your Telegram" return "Error requesting OTP. Please try again later." # Define HTML content directly PRIVACY_POLICY_HTML = """ Privacy Policy

Privacy Policy

Effective Date:15 August, 2024

1. Introduction

Welcome to @supportBot and Its Mini App. We value your privacy and are committed to protecting your personal data. This Privacy Policy explains how we collect, use, and secure your information when you interact with our Telegram bot and mini app.

2. Information We Collect

We collect the following types of information:

3. How We Use Your Information

We use your information for the following purposes:

4. Data Security

We implement appropriate technical and organizational measures to protect your data from unauthorized access, alteration, disclosure, or destruction. These measures include:

5. Data Retention

We retain your data only as long as necessary to fulfill the purposes outlined in this Privacy Policy or as required by law. Once the data is no longer needed, it will be securely deleted or anonymized.

6. Third-Party Services

We may use third-party services to assist in providing our Telegram bot and mini app. These third parties are contractually obligated to protect your data and use it only for the purposes we specify.

7. Your Rights

You have the right to:

To exercise these rights, please contact us at Email:measmonysuon@gmail.com.

8. Changes to This Privacy Policy

We may update this Privacy Policy from time to time. Any changes will be posted on this page, and we will notify you of significant updates. Your continued use of our services after changes indicates your acceptance of the updated policy.

9. Contact Us

If you have any questions or concerns about this Privacy Policy or our data practices, please contact us at:

Thank you for using @supportBot and Its Mini App. We are committed to protecting your privacy and providing a safe and secure experience.

""" def create_gradio_interface(): with gr.Blocks() as interface: gr.HTML("""

AI ImaGen DALLĀ·E 4K

Welcome! Please enter your UID and OTP to continue or Register with @SupportBot to get UID

Read more about our PRIVACY POLICY below

""") with gr.Row(): user_chat_id_input = gr.Textbox(label="Your UID", placeholder="Enter your UID") user_otp_input = gr.Textbox(label="OTP", type="password", placeholder="Enter your OTP") points_output = gr.Textbox(label="Your Balance", placeholder="Your Points", interactive=False) with gr.Row(): request_otp_button = gr.Button("Request OTP") get_points_button = gr.Button("Verify OTP") message_output = gr.Textbox(label="Message", placeholder="Results and notifications will be shown here", interactive=False) prompt_input = gr.Textbox(label="Prompt", placeholder="Enter your prompt here...") resolution_dropdown = gr.Dropdown(choices=list(resolutions.keys()), label="Resolution", value="1024x1024") generate_button = gr.Button("Generate") result_output = gr.Image(label="Generated Image", type="pil") generate_button.click( fn=handle_generate_image, inputs=[prompt_input, resolution_dropdown, user_chat_id_input], outputs=[result_output, message_output] ) get_points_button.click( fn=retrieve_user_points, inputs=[user_chat_id_input, user_otp_input], outputs=[points_output, message_output] ) request_otp_button.click( fn=request_otp_and_update_ui, inputs=[user_chat_id_input], outputs=[message_output] ) # Display Privacy Policy with gr.Row(): gr.HTML(""" """.format(PRIVACY_POLICY_HTML)) gr.HTML(""" """) return interface def launch_gradio(): interface = create_gradio_interface() interface.queue(max_size=20, api_open=False) # Configure queue and API accessibility interface.launch(show_error=True) # Enable verbose error reporting if __name__ == '__main__': launch_gradio()