app.py

import gradio as gr
import tensorflow as tf
import numpy as np
from PIL import Image
import os
import google.generativeai as genai  # ✅ Gemini API

# ---------------- Load Model ----------------
MODEL_PATH = "final_model.h5"
if not os.path.exists(MODEL_PATH):
    raise FileNotFoundError(
        f"{MODEL_PATH} not found. Please upload your trained model (final_model.h5)."
    )

# Load Keras model
model = tf.keras.models.load_model(MODEL_PATH)

# ---------------- Gemini API ----------------
# 1. Get a free API key: https://aistudio.google.com/app/apikey
# 2. On Hugging Face, store your key as a secret:
#    Settings → Repository secrets → Add "GEMINI_API_KEY"
GEMINI_API_KEY = os.getenv("GEMINI_API_KEY", "AIzaSyDOhJk8UOVz20YYTd5_e685C68qaOp8nd0")  # ✅ safer than hardcoding

if GEMINI_API_KEY:
    genai.configure(api_key=GEMINI_API_KEY)
    gemini_model = genai.GenerativeModel("gemini-1.5-flash")  # ✅ Free, fast model
else:
    gemini_model = None
    print("⚠️ Warning: GEMINI_API_KEY not set. Explanations will be skipped.")

# ---------------- Prediction + Explanation ----------------
def predict_and_explain(image: Image.Image):
    # Preprocess image for model
    img = image.resize((224, 224))  # Adjust to your training input size
    img_array = np.array(img) / 255.0
    img_array = np.expand_dims(img_array, axis=0)

    # Model prediction
    prediction = model.predict(img_array, verbose=0)[0][0]

    if prediction > 0.5:
        result = f"🟥 Malignant (Cancer Detected) — Confidence: {prediction*100:.2f}%"
        prompt = "Explain in simple terms to a patient what it means that this skin lesion is malignant."
    else:
        result = f"🟩 Benign (No Cancer) — Confidence: {(1-prediction)*100:.2f}%"
        prompt = "Explain in simple terms to a patient what it means that this skin lesion is benign."

    # Generate explanation with Gemini
    explanation = "⚠️ Gemini explanation not available (API key missing)."
    if gemini_model:
        try:
            response = gemini_model.generate_content(prompt)
            explanation = response.text
        except Exception as e:
            explanation = f"⚠️ AI explanation failed: {str(e)}"

    return result, explanation

# ---------------- Gradio UI ----------------
demo = gr.Interface(
    fn=predict_and_explain,
    inputs=gr.Image(type="pil", label="📤 Upload Skin Lesion Image"),
    outputs=[
        gr.Textbox(label="🔎 Prediction", interactive=False),
        gr.Textbox(label="📖 Explanation (AI-powered)", interactive=False),
    ],
    title="🧬 Skin Cancer Detection with AI Explanation (Gemini)",
    description="Upload a skin lesion image. The model predicts if it is **Malignant** or **Benign** and Gemini explains the result in simple terms.",
)

# ---------------- Launch ----------------
if __name__ == "__main__":
    demo.launch(share=True)