Added LLM and MongoDB Agent Scripts

llm/gemini_mongo.py

@@ -0,0 +1,351 @@
+#MONGO_URI=mongodb+srv://Admin:HelloKitty420@geobase.tyxsoir.mongodb.net/crashes
+
+import os
+import requests
+import time
+from datetime import datetime
+from pymongo import MongoClient
+from langchain_google_genai import ChatGoogleGenerativeAI
+from math import radians, sin, cos, sqrt, atan2
+
+# Configuration
+GEMINI_API_KEY = "AIzaSyBCbEOo4aK72507hqvpYkE9zXUe-z5aSXA"
+# OPENWEATHER_API_KEY = "19c6d988b89b040b603f4b3b1b1b304f"
+OPENWEATHER_API_KEY = "8754b3f387fc0f1d96a81f73e303e181"
+MONGO_URI = "mongodb+srv://Admin:HelloKitty420@geobase.tyxsoir.mongodb.net/crashes"
+
+llm = ChatGoogleGenerativeAI(model="gemini-2.5-flash-lite", api_key=GEMINI_API_KEY)
+
+def connect_to_mongodb():
+    """
+    Connect to MongoDB database and return the collection.
+    """
+    try:
+        print("Connecting to MongoDB...")
+        client = MongoClient(MONGO_URI)
+        # Test the connection
+        client.admin.command('ping')
+        print("✅ Successfully connected to MongoDB!")
+        
+        db = client.crashes  # Database name
+        collection = db.crashes  # Collection name - corrected to 'crashes'
+        
+        # Get collection stats
+        total_count = collection.estimated_document_count()
+        print(f"📊 Found {total_count:,} total crash records in database")
+        
+        # Check specifically for 2020+ data
+        filter_2020_plus = {"reportDate": {"$gte": datetime(2020, 1, 1)}}
+        count_2020_plus = collection.count_documents(filter_2020_plus)
+        print(f"📅 Found {count_2020_plus:,} crash records from 2020 onward")
+        
+        return collection
+        
+    except Exception as e:
+        print(f"❌ Failed to connect to MongoDB: {e}")
+        return None
+
+def get_crashes_within_radius_mongodb(collection, center_lat, center_lon, radius_km):
+    """
+    Query MongoDB for crashes within specified radius using geospatial query.
+    Filters for crashes from 2020 onward only.
+    
+    Args:
+        collection: MongoDB collection object
+        center_lat: Latitude of center point
+        center_lon: Longitude of center point
+        radius_km: Radius in kilometers
+    
+    Returns:
+        List of crash documents within radius from 2020 onward
+    """
+    try:
+        print(f"🔍 Querying crashes within {radius_km}km of ({center_lat:.6f}, {center_lon:.6f}) from 2020 onward...")
+        
+        # MongoDB geospatial query using $geoWithin and $centerSphere
+        # $centerSphere uses radians, so convert km to radians (divide by Earth's radius in km)
+        radius_radians = radius_km / 6371  # Earth's radius in km
+        
+        # Combined query: geospatial AND date filter for 2020+
+        query = {
+            "location": {
+                "$geoWithin": {
+                    "$centerSphere": [[center_lon, center_lat], radius_radians]
+                }
+            },
+            "reportDate": {
+                "$gte": datetime(2020, 1, 1)  # Only crashes from 2020 onward
+            }
+        }
+        
+        # Execute the query
+        cursor = collection.find(query)
+        crashes = list(cursor)
+        
+        print(f"📍 Found {len(crashes)} crashes within {radius_km}km radius (from 2020 onward)")
+        
+        # Add distance calculation to each crash for sorting
+        for crash in crashes:
+            if crash.get('location', {}).get('coordinates'):
+                crash_lon, crash_lat = crash['location']['coordinates']
+                distance = haversine_distance(center_lat, center_lon, crash_lat, crash_lon)
+                crash['distance_km'] = distance
+        
+        # Sort by distance
+        crashes.sort(key=lambda x: x.get('distance_km', float('inf')))
+        
+        return crashes
+        
+    except Exception as e:
+        print(f"❌ Error querying MongoDB: {e}")
+        return []
+
+def haversine_distance(lat1, lon1, lat2, lon2):
+    """
+    Calculate the great circle distance between two points 
+    on the earth (specified in decimal degrees)
+    Returns distance in kilometers
+    """
+    # Convert decimal degrees to radians
+    lat1, lon1, lat2, lon2 = map(radians, [lat1, lon1, lat2, lon2])
+    
+    # Haversine formula
+    dlat = lat2 - lat1
+    dlon = lon2 - lon1
+    a = sin(dlat/2)**2 + cos(lat1) * cos(lat2) * sin(dlon/2)**2
+    c = 2 * atan2(sqrt(a), sqrt(1-a))
+    
+    # Radius of earth in kilometers
+    r = 6371
+    
+    return c * r
+
+def get_current_weather(lat, lon, api_key):
+    """
+    Get basic weather data from OpenWeatherMap API.
+    """
+    try:
+        url = "https://api.openweathermap.org/data/2.5/weather"
+        response = requests.get(
+            url,
+            params={
+                "lat": lat,
+                "lon": lon,
+                "appid": api_key,
+                "units": "metric"
+            },
+            timeout=10
+        )
+        response.raise_for_status()
+        data = response.json()
+        
+        # Create summary from weather API
+        main = data.get("main", {})
+        weather = data.get("weather", [{}])[0]
+        wind = data.get("wind", {})
+        
+        summary_parts = []
+        
+        if main.get("temp") is not None:
+            summary_parts.append(f"Temperature: {main['temp']:.1f}°C")
+        
+        if weather.get("description"):
+            summary_parts.append(f"Conditions: {weather['description']}")
+        
+        if wind.get("speed") is not None:
+            summary_parts.append(f"Wind: {wind['speed']:.1f} m/s")
+        
+        if main.get("humidity") is not None:
+            summary_parts.append(f"Humidity: {main['humidity']}%")
+        
+        summary = " | ".join(summary_parts) if summary_parts else "Weather data available"
+        
+        return data, summary
+        
+    except Exception as e:
+        return None, f"Weather API failed: {str(e)}"
+
+def analyze_mongodb_crash_patterns(crashes, center_lat, center_lon, radius_km, weather_summary=None):
+    """
+    Analyze crash patterns from MongoDB data and generate safety assessment.
+    """
+    if not crashes:
+        return "No crash data available for the specified location and radius."
+    
+    total_crashes = len(crashes)
+    avg_distance = sum(crash.get('distance_km', 0) for crash in crashes) / total_crashes if crashes else 0
+    
+    # Analyze crash patterns from MongoDB structure
+    crash_analysis = {
+        'severity_counts': {},
+        'total_fatalities': 0,
+        'total_major_injuries': 0,
+        'total_minor_injuries': 0,
+        'speeding_involved': 0,
+        'impaired_involved': 0,
+        'pedestrian_crashes': 0,
+        'bicyclist_crashes': 0,
+        'vehicle_counts': {}
+    }
+    
+    # Analyze each crash
+    for crash in crashes:
+        # Severity analysis
+        severity = crash.get('severity', 'Unknown')
+        crash_analysis['severity_counts'][severity] = crash_analysis['severity_counts'].get(severity, 0) + 1
+        
+        # Casualty analysis
+        casualties = crash.get('casualties', {})
+        
+        # Count fatalities and injuries across all categories
+        for category in ['bicyclists', 'drivers', 'pedestrians', 'passengers']:
+            if category in casualties:
+                crash_analysis['total_fatalities'] += casualties[category].get('fatal', 0)
+                crash_analysis['total_major_injuries'] += casualties[category].get('major_injuries', 0)
+                crash_analysis['total_minor_injuries'] += casualties[category].get('minor_injuries', 0)
+        
+        # Count vulnerable road user involvement
+        if casualties.get('pedestrians', {}).get('total', 0) > 0:
+            crash_analysis['pedestrian_crashes'] += 1
+        if casualties.get('bicyclists', {}).get('total', 0) > 0:
+            crash_analysis['bicyclist_crashes'] += 1
+        
+        # Circumstances analysis
+        circumstances = crash.get('circumstances', {})
+        if circumstances.get('speeding_involved', False):
+            crash_analysis['speeding_involved'] += 1
+        
+        # Check for impairment
+        if (circumstances.get('pedestrians_impaired', False) or 
+            circumstances.get('bicyclists_impaired', False) or 
+            circumstances.get('drivers_impaired', False)):
+            crash_analysis['impaired_involved'] += 1
+        
+        # Vehicle analysis
+        vehicles = crash.get('vehicles', {})
+        total_vehicles = vehicles.get('total', 0)
+        crash_analysis['vehicle_counts'][str(total_vehicles)] = crash_analysis['vehicle_counts'].get(str(total_vehicles), 0) + 1
+    
+    # Create comprehensive summary for LLM
+    crash_summary = f"""
+SEVERITY BREAKDOWN: {dict(crash_analysis['severity_counts'])}
+CASUALTIES:
+- Fatal injuries: {crash_analysis['total_fatalities']}
+- Major injuries: {crash_analysis['total_major_injuries']}
+- Minor injuries: {crash_analysis['total_minor_injuries']}
+VULNERABLE ROAD USERS:
+- Crashes involving pedestrians: {crash_analysis['pedestrian_crashes']}
+- Crashes involving bicyclists: {crash_analysis['bicyclist_crashes']}
+RISK FACTORS:
+- Crashes involving speeding: {crash_analysis['speeding_involved']}
+- Crashes with impairment: {crash_analysis['impaired_involved']}
+VEHICLE INVOLVEMENT: {dict(crash_analysis['vehicle_counts'])}"""
+    
+    # Add current weather information if available
+    weather_info = ""
+    if weather_summary:
+        weather_info = f"""
+
+CURRENT WEATHER CONDITIONS:
+{weather_summary}"""
+    
+    # Create prompt for LLM
+    prompt = f"""You are a traffic safety expert analyzing recent crash data (2020 onward) and current conditions for location ({center_lat:.6f}, {center_lon:.6f}) within a {radius_km}km radius.
+    
+    CRASH STATISTICS (2020-Present):
+    - Total crashes in area: {total_crashes}
+    - Average distance from center: {avg_distance:.2f} km
+    - Search area: {radius_km}km radius (approximately {3.14159 * radius_km**2:.1f} km²)
+    
+    DETAILED CRASH ANALYSIS:{crash_summary}{weather_info}
+    
+    Based on this comprehensive recent MongoDB crash data (2020 onward), provide:
+    1. A danger level assessment (Low, Moderate, High, Very High)
+    2. Key safety concerns based on recent crash patterns AND current weather conditions
+    3. Specific recommendations for someone traveling to this location RIGHT NOW
+    4. Notable patterns in recent crash data (severity, vulnerable users, risk factors)
+    5. How current weather conditions may affect driving safety
+    
+    Focus on practical, actionable safety advice based on recent trends. Be specific about identified risks and provide clear recommendations."""
+    
+    try:
+        response = llm.invoke(prompt)
+        return response.content
+    except Exception as e:
+        return f"Error analyzing crash data with LLM: {e}"
+
+def main():
+    """
+    Main function to analyze crash danger using MongoDB geospatial queries.
+    """
+    print("🚗 MongoDB Traffic Crash Danger Analysis Tool (2020+ Data)")
+    print("=" * 65)
+    
+    # Connect to MongoDB
+    collection = connect_to_mongodb()
+    if collection is None:
+        print("❌ Could not connect to MongoDB. Exiting...")
+        return
+    
+    # Get user input for location and radius
+    try:
+        center_lat = float(input("Enter latitude: "))
+        center_lon = float(input("Enter longitude: "))
+        radius_km = float(input("Enter search radius in kilometers (default: 1.0): ") or "1.0")
+        
+        print(f"\n🔍 Analyzing recent crashes (2020+) within {radius_km}km of ({center_lat:.6f}, {center_lon:.6f})...")
+        
+        # Query MongoDB for nearby crashes using geospatial indexing
+        nearby_crashes = get_crashes_within_radius_mongodb(collection, center_lat, center_lon, radius_km)
+        
+        if len(nearby_crashes) > 0:
+            print(f"🔴 Closest crash: {nearby_crashes[0]['distance_km']:.3f}km away")
+            print(f"🔴 Furthest crash: {nearby_crashes[-1]['distance_km']:.3f}km away")
+            
+            # Display sample crash details from MongoDB structure
+            print("📊 Sample crash details from MongoDB:")
+            sample = nearby_crashes[0]
+            print(f"   - ID: {sample.get('crashId', 'N/A')}")
+            print(f"   - Severity: {sample.get('severity', 'N/A')}")
+            print(f"   - Address: {sample.get('address', 'N/A')}")
+            print(f"   - Ward: {sample.get('ward', 'N/A')}")
+            
+            casualties = sample.get('casualties', {})
+            total_casualties = 0
+            for cat in ['bicyclists', 'drivers', 'pedestrians', 'passengers']:
+                cat_data = casualties.get(cat, {})
+                total_casualties += (cat_data.get('fatal', 0) + 
+                                   cat_data.get('major_injuries', 0) + 
+                                   cat_data.get('minor_injuries', 0))
+            print(f"   - Total casualties: {total_casualties}")
+        else:
+            print("ℹ️  No crashes found within the specified radius.")
+        
+        # Get current weather conditions
+        print("\n🌤️ Fetching current weather conditions...")
+        weather_data, weather_summary = get_current_weather(center_lat, center_lon, OPENWEATHER_API_KEY)
+        
+        if weather_data is None:
+            print(f"⚠️  Weather data unavailable: {weather_summary}")
+            weather_summary = None
+        else:
+            print(f"🌤️  Current conditions: {weather_summary}")
+        
+        # Generate comprehensive safety analysis using LLM
+        print("\n🤖 Generating comprehensive safety assessment...")
+        analysis = analyze_mongodb_crash_patterns(nearby_crashes, center_lat, center_lon, radius_km, weather_summary)
+        
+        print("\n" + "="*65)
+        print("🚨 RECENT CRASH SAFETY ASSESSMENT REPORT (2020-Present)")
+        print("="*65)
+        print(analysis)
+        
+    except ValueError:
+        print("❌ Please enter valid numerical values for coordinates and radius.")
+    except KeyboardInterrupt:
+        print("\n⚠️ Analysis cancelled by user.")
+    except Exception as e:
+        print(f"❌ An error occurred: {e}")
+
+if __name__ == "__main__":
+    main()

llm/gemini_reroute.py

@@ -0,0 +1,547 @@
+import os
+import requests
+import json
+from datetime import datetime
+from pymongo import MongoClient
+from langchain_google_genai import ChatGoogleGenerativeAI
+from math import radians, sin, cos, sqrt, atan2, degrees, atan2
+from typing import List, Tuple, Dict, Optional
+
+
+
+
+# Configuration
+GEMINI_API_KEY = "AIzaSyBCbEOo4aK72507hqvpYkE9zXUe-z5aSXA"
+OPENWEATHER_API_KEY = "8754b3f387fc0f1d96a81f73e303e181"
+MONGO_URI = "mongodb+srv://Admin:HelloKitty420@geobase.tyxsoir.mongodb.net/crashes"
+MAPBOX_API_KEY = "pk.eyJ1IjoicGllbG9yZDc1NyIsImEiOiJjbWcxdTd6c3AwMXU1MmtxMDh6b2l5amVrIn0.5Es0azrah23GX1e9tmbjGw"
+
+llm = ChatGoogleGenerativeAI(model="gemini-2.5-flash-lite", api_key=GEMINI_API_KEY)
+
+class SafeRouteAnalyzer:
+    def __init__(self, mongo_uri: str):
+        """Initialize the safe route analyzer with MongoDB connection."""
+        try:
+            self.client = MongoClient(mongo_uri)
+            self.client.admin.command('ping')
+            self.db = self.client.crashes
+            self.collection = self.db.crashes
+            print("✅ Connected to MongoDB for route safety analysis")
+        except Exception as e:
+            print(f"❌ Failed to connect to MongoDB: {e}")
+            self.collection = None
+
+    def haversine_distance(self, lat1: float, lon1: float, lat2: float, lon2: float) -> float:
+        """Calculate distance between two points in kilometers."""
+        lat1, lon1, lat2, lon2 = map(radians, [lat1, lon1, lat2, lon2])
+        dlat = lat2 - lat1
+        dlon = lon2 - lon1
+        a = sin(dlat/2)**2 + cos(lat1) * cos(lat2) * sin(dlon/2)**2
+        c = 2 * atan2(sqrt(a), sqrt(1-a))
+        return 6371 * c  # Earth's radius in km
+
+    def get_route_from_mapbox(self, start_lat: float, start_lon: float, 
+                             end_lat: float, end_lon: float, profile: str = "driving") -> Dict:
+        """
+        Get route from Mapbox Directions API.
+        
+        Args:
+            start_lat, start_lon: Starting coordinates
+            end_lat, end_lon: Destination coordinates  
+            profile: 'driving', 'walking', or 'cycling'
+        
+        Returns:
+            Route data with coordinates, distance, duration
+        """
+        try:
+            url = f"https://api.mapbox.com/directions/v5/mapbox/{profile}/{start_lon},{start_lat};{end_lon},{end_lat}"
+            params = {
+                'access_token': MAPBOX_API_KEY,
+                'overview': 'full',
+                'geometries': 'geojson',
+                'steps': 'true'
+            }
+            
+            response = requests.get(url, params=params, timeout=15)
+            response.raise_for_status()
+            
+            data = response.json()
+            
+            if data.get('code') == 'Ok' and data.get('routes'):
+                route = data['routes'][0]
+                geometry = route['geometry']
+                
+                # Extract coordinates from GeoJSON format
+                coordinates = [[coord[1], coord[0]] for coord in geometry['coordinates']]  # Convert [lon,lat] to [lat,lon]
+                
+                return {
+                    'success': True,
+                    'coordinates': coordinates,  # List of [lat, lon] pairs
+                    'distance_km': route['distance'] / 1000,
+                    'duration_min': route['duration'] / 60,
+                    'geometry': geometry
+                }
+            else:
+                error_msg = data.get('message', 'No route found')
+                return {'success': False, 'error': error_msg}
+                
+        except Exception as e:
+            return {'success': False, 'error': str(e)}
+
+    def get_alternative_routes_mapbox(self, start_lat: float, start_lon: float,
+                                     end_lat: float, end_lon: float, num_alternatives: int = 3) -> List[Dict]:
+        """
+        Get multiple alternative routes using Mapbox Directions API.
+        """
+        try:
+            url = f"https://api.mapbox.com/directions/v5/mapbox/driving/{start_lon},{start_lat};{end_lon},{end_lat}"
+            params = {
+                'access_token': MAPBOX_API_KEY,
+                'alternatives': 'true',  # Request alternatives
+                'overview': 'full',
+                'geometries': 'geojson',
+                'steps': 'false'
+            }
+            
+            response = requests.get(url, params=params, timeout=15)
+            response.raise_for_status()
+            data = response.json()
+            
+            routes = []
+            if data.get('code') == 'Ok' and data.get('routes'):
+                for i, route in enumerate(data['routes'][:num_alternatives]):
+                    geometry = route['geometry']
+                    coordinates = [[coord[1], coord[0]] for coord in geometry['coordinates']]  # Convert [lon,lat] to [lat,lon]
+                    
+                    routes.append({
+                        'route_id': i,
+                        'coordinates': coordinates,
+                        'distance_km': route['distance'] / 1000,
+                        'duration_min': route['duration'] / 60,
+                        'geometry': geometry
+                    })
+            
+            return routes
+            
+        except Exception as e:
+            print(f"Error getting alternative routes: {e}")
+            return []
+
+    def analyze_route_safety(self, route_coordinates: List[Tuple[float, float]], 
+                           buffer_km: float = 0.2) -> Dict:
+        """
+        Analyze safety along a route by checking for crashes near route points.
+        
+        Args:
+            route_coordinates: List of (lat, lon) tuples along the route
+            buffer_km: How far to look for crashes around each route point
+            
+        Returns:
+            Safety analysis data
+        """
+        if self.collection is None:
+            return {'error': 'No database connection'}
+            
+        try:
+            all_nearby_crashes = []
+            safety_scores = []
+            
+            # Sample every Nth point to avoid too many queries (adjust based on route length)
+            sample_interval = max(1, len(route_coordinates) // 20)  # Max 20 sample points
+            sample_points = route_coordinates[::sample_interval]
+            
+            print(f"🔍 Analyzing safety at {len(sample_points)} points along route...")
+            
+            for i, (lat, lon) in enumerate(sample_points):
+                # Query crashes within buffer distance of this route point
+                radius_radians = buffer_km / 6371
+                
+                query = {
+                    "location": {
+                        "$geoWithin": {
+                            "$centerSphere": [[lon, lat], radius_radians]
+                        }
+                    },
+                    "reportDate": {
+                        "$gte": datetime(2020, 1, 1)
+                    }
+                }
+                
+                crashes_near_point = list(self.collection.find(query))
+                
+                # Calculate safety score for this point (lower = safer)
+                point_safety_score = self.calculate_point_safety_score(crashes_near_point)
+                safety_scores.append({
+                    'point_index': i * sample_interval,
+                    'coordinates': [lat, lon],
+                    'crashes_count': len(crashes_near_point),
+                    'safety_score': point_safety_score
+                })
+                
+                all_nearby_crashes.extend(crashes_near_point)
+            
+            # Remove duplicate crashes
+            unique_crashes = {}
+            for crash in all_nearby_crashes:
+                crash_id = crash.get('crashId', str(crash.get('_id')))
+                if crash_id not in unique_crashes:
+                    unique_crashes[crash_id] = crash
+            
+            unique_crashes_list = list(unique_crashes.values())
+            
+            # Calculate overall route safety metrics
+            total_crashes = len(unique_crashes_list)
+            avg_safety_score = sum(point['safety_score'] for point in safety_scores) / len(safety_scores) if safety_scores else 0
+            max_danger_score = max((point['safety_score'] for point in safety_scores), default=0)
+            
+            return {
+                'total_crashes_near_route': total_crashes,
+                'average_safety_score': avg_safety_score,
+                'max_danger_score': max_danger_score,
+                'safety_points': safety_scores,
+                'crashes_data': unique_crashes_list,
+                'route_length_points': len(route_coordinates)
+            }
+            
+        except Exception as e:
+            return {'error': str(e)}
+
+    def calculate_point_safety_score(self, crashes: List[Dict]) -> float:
+        """
+        Calculate a safety score for a point based on nearby crashes.
+        Higher score = more dangerous
+        """
+        if not crashes:
+            return 0.0
+            
+        score = 0.0
+        
+        for crash in crashes:
+            # Base score for any crash
+            base_score = 1.0
+            
+            # Weight by severity
+            severity = crash.get('severity', '').lower()
+            if 'fatal' in severity or 'major' in severity:
+                base_score *= 3.0
+            elif 'minor' in severity:
+                base_score *= 1.5
+            
+            # Weight by casualty count
+            casualties = crash.get('casualties', {})
+            total_casualties = 0
+            for category in ['bicyclists', 'drivers', 'pedestrians', 'passengers']:
+                if category in casualties:
+                    cat_data = casualties[category]
+                    total_casualties += (cat_data.get('fatal', 0) * 5 +
+                                       cat_data.get('major_injuries', 0) * 2 +
+                                       cat_data.get('minor_injuries', 0) * 1)
+            
+            base_score += total_casualties * 0.5
+            
+            # Weight by circumstances
+            circumstances = crash.get('circumstances', {})
+            if circumstances.get('speeding_involved', False):
+                base_score *= 1.3
+            if any([circumstances.get('pedestrians_impaired', False),
+                   circumstances.get('bicyclists_impaired', False), 
+                   circumstances.get('drivers_impaired', False)]):
+                base_score *= 1.4
+                
+            score += base_score
+        
+        return score
+
+    def generate_safety_report_with_llm(self, route_safety_data: Dict, 
+                                       route_info: Dict, weather_summary: str = None) -> str:
+        """
+        Use LLM to generate comprehensive safety report and route recommendations.
+        """
+        if 'error' in route_safety_data:
+            return f"Error analyzing route safety: {route_safety_data['error']}"
+            
+        crashes = route_safety_data.get('crashes_data', [])
+        safety_points = route_safety_data.get('safety_points', [])
+        
+        # Find most dangerous sections
+        dangerous_points = sorted(safety_points, key=lambda x: x['safety_score'], reverse=True)[:3]
+        
+        # Analyze crash patterns
+        severity_counts = {}
+        casualty_summary = {'fatal': 0, 'major': 0, 'minor': 0}
+        risk_factors = {'speeding': 0, 'impairment': 0, 'pedestrian': 0, 'bicyclist': 0}
+        
+        for crash in crashes:
+            severity = crash.get('severity', 'Unknown')
+            severity_counts[severity] = severity_counts.get(severity, 0) + 1
+            
+            # Count casualties
+            casualties = crash.get('casualties', {})
+            for category in ['bicyclists', 'drivers', 'pedestrians', 'passengers']:
+                if category in casualties:
+                    cat_data = casualties[category]
+                    casualty_summary['fatal'] += cat_data.get('fatal', 0)
+                    casualty_summary['major'] += cat_data.get('major_injuries', 0)
+                    casualty_summary['minor'] += cat_data.get('minor_injuries', 0)
+            
+            # Count risk factors
+            circumstances = crash.get('circumstances', {})
+            if circumstances.get('speeding_involved', False):
+                risk_factors['speeding'] += 1
+            if any([circumstances.get(f'{cat}_impaired', False) for cat in ['pedestrians', 'bicyclists', 'drivers']]):
+                risk_factors['impairment'] += 1
+            if casualties.get('pedestrians', {}).get('total', 0) > 0:
+                risk_factors['pedestrian'] += 1
+            if casualties.get('bicyclists', {}).get('total', 0) > 0:
+                risk_factors['bicyclist'] += 1
+        
+        weather_info = f"\n\nCURRENT WEATHER CONDITIONS:\n{weather_summary}" if weather_summary else ""
+        
+        prompt = f"""You are an expert traffic safety analyst and route planning specialist. Analyze this route's safety profile and provide recommendations.
+
+ROUTE INFORMATION:
+- Distance: {route_info.get('distance_km', 0):.1f} km
+- Estimated duration: {route_info.get('duration_min', 0):.0f} minutes
+- Analysis points along route: {len(safety_points)}
+
+SAFETY ANALYSIS (2020+ crash data):
+- Total crashes near route: {route_safety_data.get('total_crashes_near_route', 0)}
+- Average safety score: {route_safety_data.get('average_safety_score', 0):.2f}
+- Maximum danger score: {route_safety_data.get('max_danger_score', 0):.2f}
+
+CRASH BREAKDOWN:
+- Severity distribution: {severity_counts}
+- Casualties: {casualty_summary['fatal']} fatal, {casualty_summary['major']} major injuries, {casualty_summary['minor']} minor injuries
+- Risk factors: {risk_factors['speeding']} speeding-related, {risk_factors['impairment']} impairment-related
+- Vulnerable users: {risk_factors['pedestrian']} pedestrian crashes, {risk_factors['bicyclist']} bicyclist crashes
+
+MOST DANGEROUS SECTIONS:
+{chr(10).join([f"Point {p['point_index']}: {p['crashes_count']} crashes nearby, safety score {p['safety_score']:.1f}" for p in dangerous_points[:3]])}
+{weather_info}
+
+Please provide:
+1. Overall route safety assessment (SAFE/MODERATE RISK/HIGH RISK/DANGEROUS)
+2. Specific dangerous sections to watch out for
+3. Driving recommendations for this route considering current conditions
+4. Whether an alternative route should be recommended
+5. Time-of-day considerations if applicable
+6. Weather-specific precautions based on crash patterns
+
+Be specific and actionable in your recommendations."""
+
+        try:
+            response = llm.invoke(prompt)
+            return response.content
+        except Exception as e:
+            return f"Error generating safety analysis: {e}"
+
+    def find_safer_route(self, start_lat: float, start_lon: float,
+                        end_lat: float, end_lon: float) -> Dict:
+        """
+        Find the safest route among alternatives by analyzing crash data.
+        """
+        print("🗺️  Getting alternative routes...")
+        
+        # Get multiple route options
+        alternative_routes = self.get_alternative_routes_mapbox(start_lat, start_lon, end_lat, end_lon)
+        
+        if not alternative_routes:
+            print("❌ No routes found")
+            return {'error': 'No routes available'}
+        
+        print(f"📍 Analyzing {len(alternative_routes)} route options for safety...")
+        
+        # Analyze safety for each route
+        route_analyses = []
+        for i, route in enumerate(alternative_routes):
+            print(f"🔍 Analyzing route {i+1}/{len(alternative_routes)}...")
+            
+            safety_analysis = self.analyze_route_safety(route['coordinates'])
+            
+            if 'error' not in safety_analysis:
+                route_analyses.append({
+                    'route_id': i,
+                    'route_data': route,
+                    'safety_analysis': safety_analysis,
+                    'safety_score': safety_analysis.get('average_safety_score', float('inf'))
+                })
+        
+        if not route_analyses:
+            return {'error': 'Could not analyze any routes for safety'}
+        
+        # Sort routes by safety (lower score = safer)
+        route_analyses.sort(key=lambda x: x['safety_score'])
+        
+        # Get weather for additional context
+        weather_data, weather_summary = self.get_current_weather(start_lat, start_lon)
+        
+        # Generate safety reports for top routes
+        results = {
+            'recommended_route': route_analyses[0],
+            'alternative_routes': route_analyses[1:],
+            'weather_summary': weather_summary
+        }
+        
+        # Generate LLM analysis for the safest route
+        safest_route = route_analyses[0]
+        safety_report = self.generate_safety_report_with_llm(
+            safest_route['safety_analysis'], 
+            safest_route['route_data'],
+            weather_summary
+        )
+        
+        results['safety_report'] = safety_report
+        results['route_comparison'] = self.compare_routes_with_llm(route_analyses, weather_summary)
+        
+        return results
+
+    def compare_routes_with_llm(self, route_analyses: List[Dict], weather_summary: str = None) -> str:
+        """
+        Use LLM to compare multiple routes and explain why one is safer.
+        """
+        if len(route_analyses) < 2:
+            return "Only one route available for analysis."
+        
+        comparison_data = []
+        for i, analysis in enumerate(route_analyses):
+            route_data = analysis['route_data']
+            safety_data = analysis['safety_analysis']
+            
+            comparison_data.append({
+                'route_num': i + 1,
+                'distance_km': route_data.get('distance_km', 0),
+                'duration_min': route_data.get('duration_min', 0),
+                'crashes_near_route': safety_data.get('total_crashes_near_route', 0),
+                'safety_score': safety_data.get('average_safety_score', 0),
+                'max_danger_score': safety_data.get('max_danger_score', 0)
+            })
+        
+        weather_info = f"\nCurrent weather: {weather_summary}" if weather_summary else ""
+        
+        prompt = f"""Compare these route options for safety and provide a recommendation:
+
+ROUTE OPTIONS:
+{chr(10).join([f"Route {r['route_num']}: {r['distance_km']:.1f}km, {r['duration_min']:.0f}min, {r['crashes_near_route']} nearby crashes, safety score {r['safety_score']:.2f}" for r in comparison_data])}{weather_info}
+
+Provide:
+1. Which route is safest and why
+2. Trade-offs between routes (safety vs. time/distance)
+3. Clear recommendation with reasoning
+4. Any weather-related considerations
+
+Keep it concise and actionable."""
+
+        try:
+            response = llm.invoke(prompt)
+            return response.content
+        except Exception as e:
+            return f"Error comparing routes: {e}"
+
+    def get_current_weather(self, lat: float, lon: float) -> Tuple[Optional[Dict], str]:
+        """Get current weather conditions."""
+        try:
+            url = "https://api.openweathermap.org/data/2.5/weather"
+            response = requests.get(
+                url,
+                params={"lat": lat, "lon": lon, "appid": OPENWEATHER_API_KEY, "units": "metric"},
+                timeout=10
+            )
+            response.raise_for_status()
+            data = response.json()
+            
+            main = data.get("main", {})
+            weather = data.get("weather", [{}])[0]
+            wind = data.get("wind", {})
+            
+            summary = f"{main.get('temp', 'N/A')}°C, {weather.get('description', 'N/A')}, wind {wind.get('speed', 'N/A')} m/s"
+            return data, summary
+            
+        except Exception as e:
+            return None, f"Weather unavailable: {e}"
+
+
+def main():
+    """
+    Demo function showing how to use the SafeRouteAnalyzer.
+    """
+    print("🛣️  Safe Route Planning System")
+    print("=" * 50)
+    
+    analyzer = SafeRouteAnalyzer(MONGO_URI)
+    
+    if analyzer.collection is None:
+        print("❌ Cannot proceed without database connection")
+        return
+    
+    # Get input
+    try:
+        print("\n📍 Enter route details:")
+        start_lat = float(input("Starting latitude: "))
+        start_lon = float(input("Starting longitude: "))
+        end_lat = float(input("Destination latitude: "))
+        end_lon = float(input("Destination longitude: "))
+        
+        print(f"\n🚗 Planning safe route from ({start_lat:.4f}, {start_lon:.4f}) to ({end_lat:.4f}, {end_lon:.4f})")
+        
+        # Find the safest route
+        results = analyzer.find_safer_route(start_lat, start_lon, end_lat, end_lon)
+        
+        if 'error' in results:
+            print(f"❌ Error: {results['error']}")
+            return
+        
+        # Display results
+        recommended = results['recommended_route']
+        route_data = recommended['route_data']
+        safety_data = recommended['safety_analysis']
+        
+        print("\n" + "="*50)
+        print("🏆 RECOMMENDED SAFE ROUTE")
+        print("="*50)
+        print(f"📏 Distance: {route_data['distance_km']:.1f} km")
+        print(f"⏱️  Duration: {route_data['duration_min']:.0f} minutes")
+        print(f"🚨 Crashes nearby: {safety_data['total_crashes_near_route']}")
+        print(f"📊 Safety score: {safety_data['average_safety_score']:.2f} (lower is safer)")
+        
+        print(f"\n🌤️  Weather: {results.get('weather_summary', 'N/A')}")
+        
+        print("\n📋 SAFETY ANALYSIS:")
+        print("-" * 30)
+        print(results['safety_report'])
+        
+        if len(results['alternative_routes']) > 0:
+            print("\n🔄 ROUTE COMPARISON:")
+            print("-" * 30)
+            print(results['route_comparison'])
+        
+        # Output for Mapbox visualization
+        coordinates = recommended['route_data']['coordinates']
+        print(f"\n🗺️  Route coordinates for Mapbox ({len(coordinates)} points):")
+        print("First 5 points:", coordinates[:5])
+        print("Last 5 points:", coordinates[-5:])
+        
+        # You can save these coordinates to pass to your Mapbox visualization
+        route_data_to_save = {
+            'recommended_route': coordinates,
+            'route_info': route_data,
+            'safety_summary': {
+                'total_crashes': safety_data['total_crashes_near_route'],
+                'average_safety_score': safety_data['average_safety_score'],
+                'max_danger_score': safety_data['max_danger_score']
+            }
+        }
+        
+        with open('safe_route_coordinates.json', 'w') as f:
+            json.dump(route_data_to_save, f, indent=2)
+        print("📁 Route data saved to 'safe_route_coordinates.json'")
+        
+    except ValueError:
+        print("❌ Please enter valid numerical coordinates")
+    except KeyboardInterrupt:
+        print("\n⚠️  Route planning cancelled")
+    except Exception as e:
+        print(f"❌ Error: {e}")
+
+
+if __name__ == "__main__":
+    main()