app.py

import os
import gradio as gr
import torch
import torchvision.transforms as transforms
from torchvision import models
from PIL import Image
import openai

# --- SETUP ---
openai.api_key = os.getenv("")

# Load pretrained ResNet50 model
model = models.resnet50(pretrained=True)
model.eval()

# Image preprocessing
transform = transforms.Compose([
    transforms.Resize((224, 224)),
    transforms.ToTensor(),
    transforms.Normalize(mean=[0.485, 0.456, 0.406], std=[0.229, 0.224, 0.225])
])

# Fallback label list for environments where downloading fails
fallback_labels = [
    "tench", "goldfish", "great white shark", "tiger shark", "hammerhead",  # ... truncated
    "fire engine", "garbage truck", "pickup", "tow truck", "trailer truck"
]

try:
    LABELS_URL = "https://raw.githubusercontent.com/pytorch/hub/master/imagenet_classes.txt"
    local_path = torch.hub.download_url_to_file(LABELS_URL, 'imagenet_classes.txt')
    with open(local_path) as f:
        imagenet_labels = [line.strip() for line in f]
except Exception as e:
    print("Warning: Failed to download imagenet labels. Using fallback list.")
    imagenet_labels = fallback_labels

# Mock SDG mapping (you can expand this)
sdg_mapping = {
    "garbage truck": "SDG 11 - Sustainable Cities",
    "water": "SDG 6 - Clean Water",
    "tree": "SDG 13 - Climate Action",
    "smoke": "SDG 3 - Good Health"
}

# --- FUNCTIONS ---
def classify_image(img):
    img = transform(img).unsqueeze(0)
    with torch.no_grad():
        outputs = model(img)
        _, predicted = outputs.max(1)
        label = imagenet_labels[predicted.item() % len(imagenet_labels)]  # Fixed bracket error
    matched_sdg = next((v for k, v in sdg_mapping.items() if k in label.lower()), "SDG 12 - Responsible Consumption")
    return label, matched_sdg

def get_indicators(sdg):
    indicators = {
        "SDG 11 - Sustainable Cities": ["Access to transport", "Waste collection"],
        "SDG 6 - Clean Water": ["Water quality", "Wastewater treatment"],
        "SDG 13 - Climate Action": ["Carbon emissions", "Reforestation"],
        "SDG 3 - Good Health": ["Air quality", "Pollution levels"],
        "SDG 12 - Responsible Consumption": ["Recycling", "Resource usage"]
    }
    return indicators.get(sdg, ["No indicators found"])

def get_environmental_impact(sdg):
    impacts = {
        "SDG 11 - Sustainable Cities": "Reduces urban waste and improves living conditions.",
        "SDG 6 - Clean Water": "Improves access to clean and safe water.",
        "SDG 13 - Climate Action": "Combats climate change through ecosystem health.",
        "SDG 3 - Good Health": "Promotes healthier communities by reducing pollutants.",
        "SDG 12 - Responsible Consumption": "Encourages sustainable use of resources."
    }
    return impacts.get(sdg, "Helps achieve sustainable development.")

def problems_resolved(sdg):
    problems = {
        "SDG 11 - Sustainable Cities": ["Garbage overflow", "Unplanned development"],
        "SDG 6 - Clean Water": ["Dirty water", "No filtration"],
        "SDG 13 - Climate Action": ["Deforestation", "Pollution"],
        "SDG 3 - Good Health": ["Air pollution", "Toxic waste"],
        "SDG 12 - Responsible Consumption": ["Overuse of materials", "Lack of recycling"]
    }
    return problems.get(sdg, ["General sustainability issues"])

def chat_with_user(message):
    try:
        response = openai.ChatCompletion.create(
            model="gpt-3.5-turbo",
            messages=[
                {"role": "system", "content": "You are a helpful assistant for environmental reporting."},
                {"role": "user", "content": message}
            ]
        )
        return response.choices[0].message["content"]
    except Exception as e:
        return f"[Error contacting OpenAI API: {str(e)}]"

# --- UI COMPONENTS ---
countries = ["Pakistan", "India", "Bangladesh", "Nepal"]
cities = ["Islamabad", "Lahore", "Karachi", "Peshawar", "Quetta", "Delhi", "Dhaka", "Kathmandu"]

with gr.Blocks() as demo:
    with gr.Tab("Public Reporter"):
        country = gr.Dropdown(choices=countries, label="Select Country")
        city = gr.Dropdown(choices=cities, label="Select City")
        area = gr.Textbox(label="Area")
        desc = gr.Textbox(label="Issue Description")
        image = gr.Image(type="pil")
        submit_btn = gr.Button("Submit Report")

        output_label = gr.Textbox(label="Detected Issue")
        output_sdg = gr.Textbox(label="Relevant SDG")
        output_indicators = gr.Dropdown(label="Indicators")
        output_impact = gr.Textbox(label="Environmental Impact")
        output_problems = gr.Textbox(label="Problems Resolved")

        def process_report(img):
            label, sdg = classify_image(img)
            return label, sdg, get_indicators(sdg), get_environmental_impact(sdg), ", ".join(problems_resolved(sdg))

        submit_btn.click(fn=process_report, inputs=[image], 
                         outputs=[output_label, output_sdg, output_indicators, output_impact, output_problems])

        chatbot = gr.ChatInterface(fn=chat_with_user, chatbot=gr.Chatbot(), title="Ask About Reporting")

    with gr.Tab("Government Officials"):
        gr.Markdown("## Submitted Reports (Simulated)")
        gr.Dataframe(headers=["Country", "City", "Area", "Issue", "SDG", "Status"],
                     value=[["Pakistan", "Lahore", "Gulberg", "Garbage", "SDG 11", "Pending"]])
        gr.Markdown("### Suggested Resources: Garbage truck, Cleanup crew")
        gr.Markdown("### View by SDG, City, or Status coming soon...")

# --- LAUNCH ---
demo.launch()