app.py

import pandas as pd
from geopy.geocoders import Nominatim
from geopy.distance import geodesic
import requests
import gradio as gr
import os

# 🔐 Your OpenRouter API Key (keep this secret)
OPENROUTER_API_KEY = os.getenv('OPENROUTER_API_KEY')

# 📂 Load your accident dataset
df = pd.read_csv("final_merged.csv")  # Replace with your file

# 🌐 Get coordinates for user location
def geocode_location(location_name):
    geolocator = Nominatim(user_agent="accident-warning-system")
    location = geolocator.geocode(location_name)
    return (location.latitude, location.longitude) if location else None

# 📍 Find nearest accident locations
def find_nearest_accidents(user_latlon, df, top_n=10):
    df['DISTANCE'] = df.apply(
        lambda row: geodesic(user_latlon, (row['LATITUDE'], row['LONGITUDE'])).meters, axis=1)
    return df.sort_values('DISTANCE').head(top_n)

# 🧪 Filter by atmospheric and road condition
def filter_by_conditions(df, atmosphere, road_condition):
    return df[
        df['ATMOSPH_COND_DESC'].str.contains(atmosphere, case=False, na=False) &
        df['SURFACE_COND_DESC'].str.contains(road_condition, case=False, na=False)
    ]

# 📊 Summarize accident data
def summarize_data(df):
    return {
        "total_accidents": len(df),
        "severe_accidents": (df["SEVERITY"] == "Serious").sum(),
        "fatalities": int(df["NO_PERSONS_KILLED"].sum()),
        "common_types": df["ACCIDENT_TYPE_DESC"].value_counts().head(3).to_dict(),
        "age_groups": df["AGE_GROUP"].value_counts().head(2).to_dict(),
        "user_types": df["ROAD_USER_TYPE"].value_counts().head(2).to_dict()
    }

# 🧾 Format readable summary
def format_summary(summary, location, atmosphere, road_condition):
    return (
        f"📍 Location: {location}\n"
        f"🌦️ Atmosphere: {atmosphere}\n"
        f"🛣️ Road Condition: {road_condition}\n\n"
        f"🚧 Accidents Nearby: {summary['total_accidents']}\n"
        f"❗ Severe Accidents: {summary['severe_accidents']}\n"
        f"☠️ Fatalities: {summary['fatalities']}\n"
        f"🔁 Common Accident Types: {', '.join(summary['common_types'].keys())}\n"
        f"👥 Age Groups Involved: {', '.join(summary['age_groups'].keys())}\n"
        f"🚶‍♂️ User Types Affected: {', '.join(summary['user_types'].keys())}\n"
    )

# 💬 Generate LLM warning
def get_llm_warning(summary, location, atmosphere, road_condition):
    prompt = f"""
A driver is traveling to "{location}" during "{atmosphere}" weather and "{road_condition}" road conditions.

Nearby accident statistics:
- Total accidents: {summary['total_accidents']}
- Severe: {summary['severe_accidents']}
- Fatalities: {summary['fatalities']}
- Common types: {summary['common_types']}
- Age groups involved: {summary['age_groups']}
- Road users involved: {summary['user_types']}

Give a short safety warning with practical advice based on this data.
"""
    headers = {
        "Authorization": f"Bearer {OPENROUTER_API_KEY}",
        "Content-Type": "application/json"
    }
    payload = {
        "model": "openai/gpt-3.5-turbo",
        "messages": [{"role": "user", "content": prompt}]
    }
    try:
        response = requests.post("https://openrouter.ai/api/v1/chat/completions", headers=headers, json=payload)
        if response.status_code == 200:
            return response.json()['choices'][0]['message']['content']
        else:
            return f"⚠️ LLM Error: {response.status_code} - {response.text}"
    except Exception as e:
        return f"⚠️ Request failed: {e}"

# 🤖 Gradio function
def advisory_bot(location, atmosphere, road_condition):
    coords = geocode_location(location)
    if not coords:
        return "❌ Could not find the specified location.", ""

    nearest_df = find_nearest_accidents(coords, df)
    filtered_df = filter_by_conditions(nearest_df, atmosphere, road_condition)

    if filtered_df.empty:
        return f"⚠️ No recent accident data found near '{location}' for these conditions.", ""

    summary = summarize_data(filtered_df)
    summary_text = format_summary(summary, location, atmosphere, road_condition)
    llm_warning = get_llm_warning(summary, location, atmosphere, road_condition)

    return summary_text, llm_warning

# 🎛️ Gradio UI
gr.Interface(
    fn=advisory_bot,
    inputs=[
        gr.Textbox(label="📍 Location (e.g., Plenty Road, Melbourne)"),
        gr.Textbox(label="🌦️ Atmosphere (e.g., Rain, Fog, Clear)"),
        gr.Textbox(label="🛣️ Road Condition (e.g., Wet, Dry)")
    ],
    outputs=[
        gr.Textbox(label="📊 Accident Summary"),
        gr.Textbox(label="🛑 Safety Warning")
    ],
    title="🚧 Accident Risk Advisory",
    description="Input a location, weather, and road condition to get a data-based accident summary and safety warning."
).launch()