app.py

import os

# ✅ Use temp dir for safe model caching in Spaces/Docker
os.environ["TRANSFORMERS_CACHE"] = "/tmp/hf_cache"

import streamlit as st
from PIL import Image
import random
import torch
from transformers import AutoImageProcessor, SiglipForImageClassification, logging

# Optional: show more debug info if something fails
logging.set_verbosity_error()

# Constants
MODEL_NAME = "prithivMLmods/Recycling-Net-11"

# Daily sustainability tips
TIPS = [
    "Rinse containers before recycling to avoid contamination.",
    "Avoid using plastic bags for recyclables – use bins or boxes.",
    "Compost your kitchen scraps instead of tossing them.",
    "Recycle electronics only at designated e-waste centers.",
    "Buy products made from recycled materials to close the loop.",
    "Don’t recycle greasy pizza boxes – compost or trash them.",
    "Learn your local recycling rules – they vary by region.",
    "Use reusable bags, bottles, and containers to reduce waste.",
    "Donate old clothes and furniture instead of throwing them away.",
    "Avoid single-use plastics whenever possible.",
]

# Government recycling links
GOVERNMENT_LINKS = {
    "Pakistan": "https://environment.gov.pk/",
    "India": "https://www.cpcb.nic.in/",
    "China": "http://english.mee.gov.cn/",
    "Japan": "https://www.env.go.jp/en/",
    "USA": "https://www.epa.gov/recycle",
    "UK": "https://www.gov.uk/recycling-collections",
    "Canada": "https://www.canada.ca/en/services/environment/conservation/recycling.html",
    "Germany": "https://www.bmu.de/en/topics/water-waste-soil/waste-management",
}

# Load model and processor
@st.cache_resource(show_spinner="🔄 Loading AI model...")
def load_model():
    try:
        processor = AutoImageProcessor.from_pretrained(MODEL_NAME, revision="main")
        model = SiglipForImageClassification.from_pretrained(MODEL_NAME, revision="main")
        model.eval()
        return processor, model
    except Exception as e:
        st.error("❌ Failed to load the model. Please check the model name or your connection.")
        st.exception(e)
        raise e

# Prediction function
def predict(image: Image.Image, processor, model):
    inputs = processor(images=image, return_tensors="pt")
    with torch.no_grad():
        outputs = model(**inputs)
    logits = outputs.logits
    probs = torch.nn.functional.softmax(logits, dim=-1)
    conf, idx = torch.max(probs, dim=-1)
    class_name = model.config.id2label[idx.item()]
    confidence = conf.item()
    return class_name, confidence

# Recycling tip per label
def get_suggestion(label: str) -> str:
    suggestions = {
        "aluminium": "Rinse and recycle aluminum cans. They are infinitely recyclable.",
        "batteries": "Do not throw in the trash. Use proper e-waste collection centers.",
        "cardboard": "Flatten and keep dry. Avoid greasy pizza boxes.",
        "glass": "Rinse and remove lids. Separate by color if required.",
        "hard plastic": "Check recycling codes. Clean before recycling.",
        "paper": "Do not recycle shredded paper in curbside bins. Reuse or compost instead.",
        "paper towel": "Compost if clean. Trash if soiled.",
        "polystyrene": "Rarely accepted in curbside. Reuse or bring to special centers.",
        "soft plastics": "Often require store drop-off. Don’t mix with other recyclables.",
        "takeaway cups": "Check local rules. Many are lined and not recyclable curbside.",
    }
    return suggestions.get(label, "Please check your local rules for proper disposal of this item.")

# Main app
def main():
    st.set_page_config(page_title="♻️ Recycling Helper AI", layout="centered")
    st.title("♻️ Recycling Helper AI")
    st.subheader("An AI-powered app to identify recyclable materials and promote sustainability.")
    st.markdown("---")

    # Sidebar
    with st.sidebar:
        st.header("📘 About This App")
        st.markdown(
            "This open-source app helps you identify recyclable materials from waste images "
            "using a machine learning model. It promotes proper disposal and reduces contamination "
            "in the recycling stream. Built for hackathons using Hugging Face + Streamlit."
        )

        st.markdown("---")
        st.header("🌐 Recycling Resources")
        st.markdown("For proper recycling and disposal of waste, refer to the following resources:")
        for country, url in GOVERNMENT_LINKS.items():
            st.markdown(f"- [{country}]({url})", unsafe_allow_html=True)

        st.markdown("---")
        st.header("🌱 Daily Sustainability Tip")
        tip = random.choice(TIPS)
        st.success(tip)

    # Load model
    processor, model = load_model()

    # Upload image
    st.markdown("### 📤 Upload Waste Image")
    uploaded_file = st.file_uploader("Upload an image of a recyclable item", type=["png", "jpg", "jpeg"])

    if uploaded_file is not None:
        try:
            image = Image.open(uploaded_file).convert("RGB")
            st.image(image, caption="Uploaded Image", use_column_width=True)

            with st.spinner("🔍 Classifying image..."):
                label, confidence = predict(image, processor, model)

            st.success(f"**Predicted Material:** `{label}`  \n**Confidence:** `{confidence:.2%}`")
            st.info(f"**Disposal Tip:** {get_suggestion(label)}")

        except Exception as e:
            st.error("An error occurred during prediction.")
            st.exception(e)

    with st.expander("🔍 Show All Recognizable Materials"):
        st.write(model.config.id2label)

    st.markdown("---")
    st.caption("Made with 💚 for a sustainable future | Hackathon 2025")

# Run
if __name__ == "__main__":
    main()