Update app.py

app.py

@@ -1,234 +1,171 @@
-import os
-from PIL import ImageDraw
-from PIL import ImageFont
-
-#https://huggingface.co/spaces/Galis/room_interior_quality/tree/main/FGoWx7peJuJ/secret_santa
-STABILITY_HOST = os.environ["STABILITY_HOST"]
-STABILITY_KEY = os.environ["STABILITY_KEY"]
-cohere_key = os.environ["cohere_key"]
-import cohere
-import random
-co = cohere.Client(cohere_key)
+import gradio as gr
 import io
 import os
 import warnings
-import math
-from math import sqrt
-from IPython.display import display
 from PIL import Image
 from stability_sdk import client
 import stability_sdk.interfaces.gooseai.generation.generation_pb2 as generation
-from PIL import Image
- 
+import google.generativeai as genai
+
+
+genai.configure(api_key=os.environ['genai_img'])
+
+# Replace with your API key
 stability_api = client.StabilityInference(
-    key=os.environ['STABILITY_KEY'], 
+    key=os.environ['STABILITY_KEY'],
     verbose=True,
+    engine="stable-diffusion-xl-1024-v1-0",  # You can experiment with different engines
 )
 
+def generate_image_from_text(prompt):
+    """Generates an image from a text prompt."""
+    try:
+        answers = stability_api.generate(
+            prompt=prompt,
+            seed=12345,  # You can adjust the seed for different results
+            steps=30,  # Adjust the number of steps for quality/speed trade-off
+            cfg_scale=8.0,
+            width=512,  # Adjust width and height as needed
+            height=512,
+            sampler=generation.SAMPLER_K_DPMPP_2M
+        )
+        for resp in answers:
+            for artifact in resp.artifacts:
+                if artifact.finish_reason == generation.FILTER:
+                    warnings.warn("Safety filter triggered. Please modify the prompt.")
+                    return None
+                if artifact.type == generation.ARTIFACT_IMAGE:
+                    img = Image.open(io.BytesIO(artifact.binary))
+                    return img
+    except Exception as e:
+        print(f"Error during image generation: {e}")
+        return None
+
+def generate_image_from_image(init_image, start_schedule, prompt):
+    """Generates an image using the provided initial image, start schedule, and prompt."""
+    try:
+        answers = stability_api.generate(
+            prompt=prompt,
+            init_image=init_image,
+            start_schedule=start_schedule,
+            seed=12345,  # You can adjust the seed for different results
+            steps=30,  # Adjust the number of steps for quality/speed trade-off
+            cfg_scale=8.0,
+            width=512,  # Adjust width and height as needed
+            height=512,
+            sampler=generation.SAMPLER_K_DPMPP_2M
+        )
+        for resp in answers:
+            for artifact in resp.artifacts:
+                if artifact.finish_reason == generation.FILTER:
+                    warnings.warn("Safety filter triggered. Please modify the prompt.")
+                    return None
+                if artifact.type == generation.ARTIFACT_IMAGE:
+                    img = Image.open(io.BytesIO(artifact.binary))
+                    return img
+    except Exception as e:
+        print(f"Error during image generation: {e}")
+        return None
+
+# Placeholder for model loading (explained later)
+def load_model():
+    # Replace with your Google Generative AI (GenAI) model loading logic
+    # Here's an example structure (assuming GenAI is available):
+    import pathlib
+
+    # Set up the model (replace with your actual API key and model name)
+    generation_config = {
+        "temperature": 1,
+        "top_p": 0.95,
+        "top_k": 64,
+        "max_output_tokens": 8192,
+    }
+
+    safety_settings = [
+        {
+            "category": "HARM_CATEGORY_HARASSMENT",
+            "threshold": "BLOCK_MEDIUM_AND_ABOVE"
+        },
+        {
+            "category": "HARM_CATEGORY_HATE_SPEECH",
+            "threshold": "BLOCK_MEDIUM_AND_ABOVE"
+        },
+        {
+            "category": "HARM_CATEGORY_SEXUALLY_EXPLICIT",
+            "threshold": "BLOCK_MEDIUM_AND_ABOVE"
+        },
+        {
+            "category": "HARM_CATEGORY_DANGEROUS_CONTENT",
+            "threshold": "BLOCK_MEDIUM_AND_ABOVE"
+        },
+    ]
+
+    model = genai.GenerativeModel(
+        model_name="gemini-1.5-pro-latest",
+        generation_config=generation_config,
+        safety_settings=safety_settings,
+    )
+    return model
+
+# Function to handle user input and model interaction
+def chat(image, query):
+    # Load the model on the first call
+    if not hasattr(chat, 'model'):
+        chat.model = load_model()
+
+    # Process the image (replace with your image processing logic)
+    # Here's a placeholder for potential image processing:
+    processed_image = image  # Assuming no processing needed for now
+
+    # Start or continue the conversation
+    convo = chat.model.start_chat(history=[
+        {
+            "role": "user",
+            "parts": [processed_image],
+        },
+    ])
+    response = convo.send_message(query)
+    # return str(response)
+    print(response)
+    return response.text
+
+
+    # Extract only the text content from the response
+
+
+# Gradio interface definition
+chat_interface = gr.Interface(
+    fn=chat,
+    inputs=[gr.Image(type="pil"), gr.Textbox(lines=4)],
+    outputs="text",
+    title="Sustainable Interior Design Chatbot",
+    description="Ask the AI for sustainable design suggestions based on an image of your room.",
+)
 
-def generate_caption_keywords(prompt, model='command-xlarge-20221108', max_tokens=200, temperature=random.uniform(0.1, 2), k=0, p=0.75, frequency_penalty=0, presence_penalty=0, stop_sequences=[]):
-    
-    response = co.generate(
-      model=model,
-      prompt=prompt,
-      max_tokens=max_tokens,
-      temperature=temperature,
-      k=k,
-      p=p,
-      frequency_penalty=frequency_penalty,
-      presence_penalty=presence_penalty,
-      stop_sequences=stop_sequences,
-      return_likelihoods='NONE')
-
-    def highlight_keywords(text):
-        keywords = []
-        text = text.lower()
-        text = re.sub(r'[^a-z\s]', '', text) # remove punctuation
-        text = re.sub(r'\b(the|and|of)\b', '', text) # remove stop words
-        words = text.split()
-        for word in words:
-            if word not in keywords:
-                keywords.append(word)
-        return keywords
-
-    caption = response.generations[0].text
-    keywords = highlight_keywords(caption)
-    keywords_string = ', '.join(keywords)
-
-    return caption, keywords_string
-
-
+with gr.Blocks() as demo:
+    gr.Markdown("**Baith-al-suroor بَیتُ الْسرور  🏡🤖**, Transform your space with the power of artificial intelligence." ) # Add title with emojis
+    gr.Markdown("Baith al suroor بَیتُ الْسرور  (house of happiness in Arabic)  🏡🤖  is a deeptech app that uses the power of artificial intelligence to transform your space. With the Cohere/GPT3 language   model, it can generate descriptions of your desired design, and the Stable Diffusion algorithm creates relevant images to bring your vision to your thoughts. Give Baith AI a try and see how it can elevate your interior design.--if you want to scale / reaserch / build mobile app / get secret key for research purpose on this space konnect me   @[Xhaheen](https://www.linkedin.com/in/sallu-mandya/)")
+    gr.Markdown("## Generate Images with Stability AI")
+
+    with gr.Accordion("Text-to-Image", open=False):
+        with gr.Row():
+            text_prompt = gr.Textbox(label="Prompt", lines=2,value='Zen-style (bedroom interior) With storage bench or ottoman and bed and accent chair and headboard and bedside table or night stand and night light and mirror and plant. . With natural light and serenity and harmony and clutter free and clean lines and mimimalist and Asian zen interior and Japanese minimalist interior and Japanese interior. . Cinematic photo, highly detailed, cinematic lighting, ultra-detailed, ultrarealistic, photorealism, 8k. Zen interior design style' ,placeholder="Enter your text prompt here...")
+            generate_text_button = gr.Button("Generate")
+        text_output = gr.Image(type="pil", label="Generated Image")
+
+    with gr.Accordion("Image-to-Image", open=False):
+        with gr.Row():
+            image_input = gr.Image(type="pil", label="Initial Image")
+            prompt_strength = gr.Slider(0.0, 1.0, value=0.85, label="Prompt Strength")
+        with gr.Row():
+            image_prompt = gr.Textbox(label="Prompt", lines=2,value='Zen-style (  interior) With storage bench or ottoman and bed and accent chair and headboard and bedside table or night stand and night light and mirror and plant. . With natural light and serenity and harmony and clutter free and clean lines and mimimalist and Asian zen interior and Japanese minimalist interior and Japanese interior. . Cinematic photo, highly detailed, cinematic lighting, ultra-detailed, ultrarealistic, photorealism, 8k. Zen interior design style' , placeholder="Enter your text prompt here...")
+            generate_image_button = gr.Button("Generate")
+        image_output = gr.Image(type="pil", label="Generated Image")
+
+    with gr.Accordion("Chat with AI", open=False):
+        chat_interface.render()
  
-def img2img( path ,secret_key,design,x_prompt,alt_prompt,strength,guidance_scale,steps):
-    # Read the size of the image
-############################
- 
-    img = Image.open(path)
-    width, height = img.size
-    num_pixels = width * height
-    
-    # Calculate the maximum number of pixels allowed
-    max_pixels = 1048576
-
-    # Calculate the new size of the image, making sure that the number of pixels does not exceed the maximum limit
-    if width * height > max_pixels:
-        # Calculate the new width and height of the image
-        ratio = width / height
-        new_width = int(math.sqrt(max_pixels * ratio))
-        new_height = int(math.sqrt(max_pixels / ratio))
-    else:
-        new_width = width
-        new_height = height
-
-    # Make sure that either the width or the height of the resized image is a multiple of 64
-    if new_width % 64 != 0:
-        new_width = ((new_width + 63) // 64) * 64
-    if new_height % 64 != 0:
-        new_height = ((new_height + 63) // 64) * 64
-
-    # Resize the image
-    img = img.resize((new_width, new_height), resample=Image.BILINEAR)
-
-    # Check if the number of pixels in the resized image is within the maximum limit
-    # If not, adjust the width and height of the image to bring the number of pixels within the maximum limit
-    if new_width * new_height > max_pixels:
-        while new_width * new_height > max_pixels:
-            new_width -= 1
-            new_height = int(max_pixels / new_width)
-
-
-
-    
-            # Calculate the closest multiple of 64 for each value
-            if new_width % 64 != 0:
-                new_width = (new_width // 64) * 64
-            if new_height % 64 != 0:
-                new_height = (new_height // 64) * 64
-
-            # Make sure that the final values are less than the original values
-            if new_width > 1407:
-                new_width -= 64
-            if new_height > 745:
-                new_height -= 64
-    
-    new_height ,new_width
-    # Initialize the values
-    widthz = new_width
-    heightz = new_height
-
-    # Calculate the closest multiple of 64 for each value
-    if widthz % 64 != 0:
-        widthz = (widthz // 64) * 64
-    if heightz % 64 != 0:
-        heightz = (heightz // 64) * 64
-
-    # Make sure that the final values are less than the original values
-    if widthz > 1407:
-        widthz -= 64
-    if heightz > 745:
-        heightz -= 64
-
-    
-
-    img = img.resize((widthz, heightz), resample=Image.BILINEAR)
- 
-
+    generate_text_button.click(generate_image_from_text, text_prompt, text_output)
+    generate_image_button.click(generate_image_from_image, [image_input, prompt_strength, image_prompt], image_output)
 
-######################################
-    max_attempts = 5 # maximum number of attempts before giving up
-    attempts = 0 # current number of attempts
-    while attempts < max_attempts:
-        try:
-            if x_prompt == True:
-                prompt = alt_prompt
-            else:
-                try:
-                    caption, keywords = generate_caption_keywords(design)
-                    prompt = keywords
-                except:
-                    prompt = design
-
-            # call the GRPC service to generate the image
-            answers = stability_api.generate(
-                prompt,
-                init_image=img,
-                seed=54321,
-                start_schedule=strength,
-            )
-            for resp in answers:
-                for artifact in resp.artifacts:
-                    if artifact.finish_reason == generation.FILTER:
-                        warnings.warn(
-                            "Your request activated the API's safety filters and could not be processed."
-                            "Please modify the prompt and try again.")
-                    if artifact.type == generation.ARTIFACT_IMAGE:
-                        img2 = Image.open(io.BytesIO(artifact.binary))
-                        img2 = img2.resize((new_width, new_height), resample=Image.BILINEAR)
-                        img2.save("new_image.jpg")
-                        print(type(img2))
-                         
-            # if the function reaches this point, it means it succeeded, so we can return the result
-                        if secret_key not in os.environ['secretz']:
-
-                          draw = ImageDraw.Draw(img2)
-
-                          # Set the font and text color
-                          font = ImageFont.truetype('arial.ttf', 32)
-                          text_color = (255, 255, 255)
-
-                          # Get the size of the image
-                          width, height = img2.size
-
-                          # Calculate the x and y coordinates for the text
-                          text_x = 10
-                          text_y = height - 100
-
-                          # Draw the text on the image
-                          draw.text((text_x, text_y), 'Please enter secret key to get HD image without \n  watermark', font=font, fill=text_color)
-
-                          # Draw the diagonal lines
-                          line_color = (0, 0, 0)
-                          draw.line((0, 0) + (width, height), fill=line_color, width=5)
-                          draw.line((0, height) + (width, 0), fill=line_color, width=5)
-
-                          # Save the image with the watermark
-                          img2.save('image_with_watermark.jpg')
-                          img2
-
-
-            return img2
-        except Exception as e:
-            # if an exception is thrown, we will increment the attempts counter and try again
-            attempts += 1
-            print("Attempt {} failed: {}".format(attempts, e))
-    # if the function reaches this point, it means the maximum number of attempts has been reached, so we will raise an exception
-    raise Exception("Maximum number of attempts reached, unable to generate image")
-
-
-
-import gradio as gr  
-
-gr.Interface(img2img,  [gr.Image(source="upload", type="filepath", label="Input Image"),
-                        gr.Textbox(label = 'enter secret key to get HD image without watermark , connect with Xhaheen to get key',value = 'secret_santa', type="password" ),
-    
-                        gr.Dropdown(['interior design of living room', 
-                                         'interior design of gaming room',
-                                         'interior design of kitchen',
-                                         'interior design of bedroom',
-                                         'interior design of bathroom',
-                                         'interior design of office',
-                                         'interior design of meeting room',
-                                         'interior design of personal room'],label="Click here to select your design by GPT-3/Cohere Language model",value = 'interior design'), 
-                        gr.Checkbox(label="Check Custom design if you already have prompt",value = False),
-
-                        gr.Textbox(label = ' Input custom Prompt Text'),
-                        gr.Slider(label='Strength , try with multiple value betweens 0.55 to 0.9 ', minimum = 0, maximum = 1, step = .01, value = .65),
-                        gr.Slider(2, 15, value = 7, label = 'Guidence Scale'),
-                        gr.Slider(10, 50, value = 50, step = 1, label = 'Number of Iterations')
-                        ], 
-                        gr.Image(), 
-             examples =[['1.png',"xxx",'interior design of living room','False','interior design',0.6,7,50],
-                  ['2.png',"xxx",'interior design of hall ','False','interior design',0.7,7,50],
-                  ['3.png',"xxx",'interior design of bedroom','False','interior design',0.6,7,50]],
-             title = "" +'**Baith-al-suroor بَیتُ الْسرور  🏡🤖**, Transform your space with the power of artificial intelligence. '+ "",
-                                    description="Baith al suroor بَیتُ الْسرور  (house of happiness in Arabic)  🏡🤖  is a deeptech app that uses the power of artificial intelligence to transform your space. With the Cohere/GPT3 language   model, it can generate descriptions of your desired design, and the Stable Diffusion algorithm creates relevant images to bring your vision to your thoughts. Give Baith AI a try and see how it can elevate your interior design.--if you want to scale / reaserch / build mobile app / get secret key for research purpose on this space konnect me   @[Xhaheen](https://www.linkedin.com/in/sallu-mandya/)").launch( show_api=False,debug = True)
+demo.launch(debug=True)