app.py

import os
import tempfile
from PIL import Image
import gradio as gr
from google import genai
from google.genai import types

def save_binary_file(file_name, data):
    with open(file_name, "wb") as f:
        f.write(data)

def generate(text, file_name, api_key, model="gemini-2.0-flash-exp"):
    client = genai.Client(api_key=(api_key.strip() if api_key and api_key.strip() != "" else os.environ.get("GEMINI_API_KEY")))
    files = [client.files.upload(file=file_name)]
    
    contents = [
        types.Content(
            role="user",
            parts=[
                types.Part.from_uri(file_uri=files[0].uri, mime_type=files[0].mime_type),
                types.Part.from_text(text=text),
            ],
        ),
    ]
    generate_content_config = types.GenerateContentConfig(
        temperature=1,
        top_p=0.95,
        top_k=40,
        max_output_tokens=8192,
        response_modalities=["image", "text"],
        response_mime_type="text/plain",
    )

    text_response = ""
    image_path = None
    with tempfile.NamedTemporaryFile(suffix=".png", delete=False) as tmp:
        temp_path = tmp.name
        for chunk in client.models.generate_content_stream(model=model, contents=contents, config=generate_content_config):
            if not chunk.candidates or not chunk.candidates[0].content or not chunk.candidates[0].content.parts:
                continue
            candidate = chunk.candidates[0].content.parts[0]
            if candidate.inline_data:
                save_binary_file(temp_path, candidate.inline_data.data)
                image_path = temp_path
                break
            else:
                text_response += chunk.text + "\n"
    
    del files
    return image_path, text_response

def process_image_and_prompt(composite_pil, prompt, gemini_api_key):
    try:
        if not composite_pil:
            raise gr.Error("Carregue uma imagem primeiro.", duration=5)
        if not prompt:
            raise gr.Error("Digite um prompt antes de gerar.", duration=5)
        if not gemini_api_key and not os.environ.get("GEMINI_API_KEY"):
            raise gr.Error("Insira uma chave API Gemini ou configure a variável GEMINI_API_KEY.", duration=10)
        
        with tempfile.NamedTemporaryFile(suffix=".png", delete=False) as tmp:
            composite_path = tmp.name
            composite_pil.save(composite_path)
        
        image_path, text_response = generate(text=prompt, file_name=composite_path, api_key=gemini_api_key)
        
        if image_path:
            result_img = Image.open(image_path)
            if result_img.mode == "RGBA":
                result_img = result_img.convert("RGB")
            return result_img, ""
        else:
            return None, text_response
    except Exception as e:
        raise gr.Error(f"Erro: {e}", duration=5)

# Interface Moderna
with gr.Blocks(css="style.css", theme=gr.themes.Soft(), title="Gemini Image Editor") as demo:
    gr.HTML(
    """
    <div class="header">
        <img src="https://www.gstatic.com/lamda/images/gemini_favicon_f069958c85030456e93de685481c559f160ea06b.png" alt="Logo" class="logo">
        <h1>Gemini Image Editor</h1>
        <p class="subtitle">Edite imagens com IA de forma simples e poderosa</p>
    </div>
    """
    )
    
    with gr.Row(elem_classes="main-container"):
        with gr.Column(scale=1, elem_classes="input-section"):
            image_input = gr.Image(type="pil", label="Carregar Imagem (PNG)", elem_classes="image-upload")
            prompt_input = gr.Textbox(placeholder="Digite o prompt (ex: 'add a hat')", label="Prompt", elem_classes="prompt-box")
            gemini_api_key = gr.Textbox(placeholder="Chave API Gemini", label="API Key", type="password", elem_classes="api-key-box")
            with gr.Row():
                submit_btn = gr.Button("Gerar", variant="primary", elem_classes="submit-btn")
                clear_btn = gr.Button("Limpar", variant="secondary", elem_classes="clear-btn")
        
        with gr.Column(scale=2, elem_classes="output-section"):
            output_image = gr.Image(label="Resultado", elem_classes="output-image")
            output_text = gr.Textbox(label="Mensagem", placeholder="Resultados de texto aparecem aqui", interactive=False, elem_classes="output-text")

    with gr.Accordion("ℹ️ Como Usar e Configurar", open=False, elem_classes="info-accordion"):
        gr.Markdown("""
        ### Como Usar
        1. Faça upload de uma imagem PNG.
        2. Digite um prompt em inglês (ex: "remove the background").
        3. Insira sua chave API Gemini ou configure a variável `GEMINI_API_KEY`.
        4. Clique em "Gerar" e veja o resultado!

        ### Configuração
        - Obtenha sua chave API em <a href="https://aistudio.google.com/apikey">Google AI Studio</a>.
        - Duplique este projeto em <a href="https://huggingface.co/spaces/ameerazam08/Gemini-Image-Edit?duplicate=true">Hugging Face</a>.
        - Contato: <a href="https://www.linkedin.com/in/dheiver-santos/">Dheiver Santos</a>.
        """)

    gr.Markdown("### Exemplos", elem_classes="examples-header")
    examples = [
        ["data/1.webp", "change text to 'AMEER'", ""],
        ["data/2.webp", "remove the spoon from hand only", ""],
        ["data/3.webp", "change text to 'Make it'", ""],
    ]
    gr.Examples(examples=examples, inputs=[image_input, prompt_input, gemini_api_key], outputs=[output_image, output_text],
                fn=process_image_and_prompt, cache_examples=False)

    # Eventos
    submit_btn.click(fn=process_image_and_prompt, inputs=[image_input, prompt_input, gemini_api_key], outputs=[output_image, output_text])
    clear_btn.click(fn=lambda: (None, "", "", None, ""), inputs=[], outputs=[image_input, prompt_input, gemini_api_key, output_image, output_text])

demo.queue(max_size=50).launch()