app.py

import gradio as gr
import io
import os
import warnings
from PIL import Image
from stability_sdk import client
import stability_sdk.interfaces.gooseai.generation.generation_pb2 as generation
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'],
    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.",
)

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, leveraging diffusion models and powerful Gemini 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()
 
    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)

demo.launch(debug=True)