app.py

import streamlit as st
from transformers import pipeline, AutoModelForImageClassification, AutoFeatureExtractor
from PIL import Image
import openai
import os
from dotenv import load_dotenv

# =======================
# Load Environment Variables from .env File
# =======================
load_dotenv()  # Explicitly load the .env file

# Set OpenAI API key
openai.api_key = os.getenv("OPENAI_API_KEY")

# Debugging: Check if API key is loaded
if not openai.api_key or not openai.api_key.startswith("sk-"):
    st.error("OpenAI API key is not set or is invalid. Please check the `.env` file or your environment variable setup.")
    st.stop()

# =======================
# Streamlit Page Config
# =======================
st.set_page_config(
    page_title="AI-Powered Skin Cancer Detection",
    page_icon="🩺",
    layout="wide",
    initial_sidebar_state="expanded"
)

# =======================
# Load Skin Cancer Model (PyTorch)
# =======================
@st.cache_resource
def load_model():
    """
    Load the pre-trained skin cancer classification model using PyTorch.
    Use the AutoModelForImageClassification and AutoFeatureExtractor for explicit local caching.
    """
    try:
        extractor = AutoFeatureExtractor.from_pretrained("Anwarkh1/Skin_Cancer-Image_Classification")
        model = AutoModelForImageClassification.from_pretrained("Anwarkh1/Skin_Cancer-Image_Classification")
        return pipeline("image-classification", model=model, feature_extractor=extractor, framework="pt")
    except Exception as e:
        st.error(f"Error loading the model: {e}")
        return None

model = load_model()

# =======================
# Generate OpenAI Explanation
# =======================
def generate_openai_explanation(label, confidence):
    """
    Generate a detailed explanation for the classification result using OpenAI's GPT model.
    """
    prompt = (
        f"The AI model has classified an image of a skin lesion as **{label}** with a confidence of **{confidence:.2%}**.\n"
        f"Explain what this classification means, including potential characteristics of this lesion type, "
        f"what steps a patient should take next, and how the AI might have arrived at this conclusion. "
        f"Use language that is easy for a non-medical audience to understand."
    )
    try:
        response = openai.Completion.create(
            model="text-davinci-003",  # Replace with "gpt-4" if available
            prompt=prompt,
            max_tokens=300,
            temperature=0.7
        )
        return response.choices[0].text.strip()
    except Exception as e:
        return f"Error generating explanation: {e}"

# =======================
# Streamlit App Title and Sidebar
# =======================
st.title("🔍 AI-Powered Skin Cancer Classification and Explanation")
st.write("Upload an image of a skin lesion, and the AI model will classify it and provide a detailed explanation.")

st.sidebar.info("""
**AI Cancer Detection Platform**  
This application uses AI to classify skin lesions and generate detailed explanations for informational purposes.  
It is not intended for medical diagnosis. Always consult a healthcare professional for medical advice.
""")

# =======================
# File Upload and Prediction
# =======================
uploaded_image = st.file_uploader("Upload a skin lesion image (PNG, JPG, JPEG)", type=["png", "jpg", "jpeg"])

if uploaded_image:
    # Display uploaded image
    image = Image.open(uploaded_image).convert("RGB")
    st.image(image, caption="Uploaded Image", use_column_width=True)

    # Perform classification
    if model is None:
        st.error("Model could not be loaded. Please try again later.")
    else:
        with st.spinner("Classifying the image..."):
            try:
                results = model(image)
                label = results[0]['label']
                confidence = results[0]['score']

                # Display prediction results
                st.markdown(f"### Prediction: **{label}**")
                st.markdown(f"### Confidence: **{confidence:.2%}**")

                # Provide confidence-based insights
                if confidence >= 0.8:
                    st.success("High confidence in the prediction.")
                elif confidence >= 0.5:
                    st.warning("Moderate confidence in the prediction. Consider additional verification.")
                else:
                    st.error("Low confidence in the prediction. Results should be interpreted with caution.")

                # Generate explanation
                with st.spinner("Generating a detailed explanation..."):
                    explanation = generate_openai_explanation(label, confidence)

                st.markdown("### Explanation")
                st.write(explanation)

            except Exception as e:
                st.error(f"Error during classification: {e}")