app.py

import gradio as gr
import requests
import io

# API endpoint (local FastAPI server)
API_URL = "http://127.0.0.1:8000/predict_image"

# Function to send image and process response
def classify_trash(img):
    try:
        # Convert PIL image to in-memory file
        img_bytes = io.BytesIO()
        img.save(img_bytes, format="PNG")  # format-agnostic
        img_bytes.seek(0)

        files = {"file": ("uploaded.png", img_bytes, "image/png")}
        response = requests.post(API_URL, files=files)

        if response.status_code == 200:
            data = response.json()
            result = f"🧠 **Predicted Class**: `{data['predicted_class']}`\n"
            result += f"📊 **Confidence**: `{data['confidence']*100:.2f}%`\n\n"
            result += "📋 **Class Probabilities:**\n"
            for item in data["all_probabilities"]:
                result += f"- {item['class']}: `{float(item['confidence'])*100:.2f}%`\n"
            return result
        else:
            return f"❌ Error: {response.status_code} - {response.text}"
    except Exception as e:
        return f"❌ Exception occurred: {str(e)}"

# Gradio UI definition
def launch_ui():
    gr.Interface(
        fn=classify_trash,
        inputs=gr.Image(type="pil", label="📤 Upload a Trash Image"),
        outputs=gr.Markdown(label="🧠 Classification Result"),
        title="EcoVision - Trash Classifier",
        description="Upload an image to detect the type of trash (e.g., cardboard, metal, plastic, etc.).",
        allow_flagging="never",
    ).launch()

# Run the UI only when executed directly
if __name__ == "__main__":
    print("🚀 Launching Gradio UI...")
    launch_ui()