chatbot

src/spatial/README.md

@@ -133,5 +133,23 @@ Ensure latitude and longitude have high precision (up to six decimal places) for
 
 ---
 
+### **Chatbot Tool**
+
+The Chatbot Tool provides a conversational interface for users to interact with the AirQo platform and receive information about air quality and related topics.
+
+```http
+POST http://127.0.0.1:5000/api/v2/spatial/chat
+```
+
+##### Example Request Body
+```json
+{
+    "grid_id": "659d036497e611001236cd1b",
+    "start_time": "2024-12-01T00:00",
+    "end_time": "2025-03-06T00:00",
+    "prompt":"with 300 words, write the air quality report "
+}
+```
+
 This README provides an overview of the setup, API endpoints, and example requests. For further details, consult the official AirQo API documentation.
 

src/spatial/configure.py

@@ -26,6 +26,11 @@ class Config:
     )
     ANALTICS_URL = os.getenv("ANALTICS_URL")
     GOOGLE_API_KEY = os.getenv("GOOGLE_API_KEY")
+    # Redis configuration
+    REDIS_HOST = os.getenv("REDIS_HOST", "localhost")  # Default to 'localhost' if not set
+    REDIS_PORT = int(os.getenv("REDIS_PORT", 6379))    # Convert to int, default to 6379
+    REDIS_DB = int(os.getenv("REDIS_DB", 0))           # Convert to int, default to 0
+    REDIS_PASSWORD = os.getenv("REDIS_PASSWORD", None) # Default to None if not set
 class ProductionConfig(Config):
     DEBUG = False
     TESTING = False

src/spatial/controllers/controllers.py

@@ -1,5 +1,6 @@
 # controller/controller.py
 from flask import Blueprint, request, jsonify
+from views.chatbot_views import ChatbotView
 from views.getis_services import SpatialDataHandler
 from views.getis_confidence_services import SpatialDataHandler_confidence
 from views.localmoran_services import SpatialDataHandler_moran
@@ -12,7 +13,7 @@
 from views.satellite_predictions import SatellitePredictionView
 from views.site_category_view import SiteCategorizationView
 from views.site_selection_views import SiteSelectionView
-from views.report_view import ReportView
+from views.report_view import ReportView 
 
 
 controller_bp = Blueprint("controller", __name__)
@@ -78,4 +79,9 @@ def fetch_air_quality_without_llm():
 
 @controller_bp.route("/air_quality_report_with_customised_prompt", methods=["POST"])
 def fetch_air_quality_with_customised_prompt():
-    return ReportView.generate_air_quality_report_with_customised_prompt_gemini()
+    return ReportView.generate_air_quality_report_with_customised_prompt_gemini()
+
+@controller_bp.route("/chatbot", methods=["POST"])
+def Chatbot_Views():
+    return ChatbotView.chat_endpoint()
+

src/spatial/models/chatbot_model.py

@@ -0,0 +1,191 @@
+import requests
+import redis
+import json
+import google.generativeai as genai
+import logging
+import re
+import threading
+from flask import Flask
+from urllib.parse import urlencode
+from configure import Config
+
+# Configure logging
+logging.basicConfig(level=logging.INFO, format='%(asctime)s - %(levelname)s - %(message)s')
+
+# Configure API keys
+GOOGLE_API_KEY = Config.GOOGLE_API_KEY
+genai.configure(api_key=GOOGLE_API_KEY)
+
+# Initialize Redis client
+try:
+    redis_client = redis.StrictRedis(
+        host=Config.REDIS_HOST or 'localhost',
+        port=Config.REDIS_PORT or 6379,
+        db=Config.REDIS_DB or 0,
+        password=Config.REDIS_PASSWORD or None,
+        decode_responses=True
+    )
+    # Test the connection 
+    redis_client.ping()
+    logging.info("Connected to Redis")
+except Exception as e:
+    logging.error(f"Error connecting to Redis: {e}")
+    redis_client = None
+# lock for thread safety to prevent race conditions when multiple users request the same data.
+data_fetch_lock = threading.Lock()
+
+class DataFetcher:
+    @staticmethod
+    def fetch_air_quality_data(grid_id, start_time, end_time):
+        """Fetch air quality data and cache it in Redis to avoid redundant API calls."""
+        cache_key = f"air_quality:{grid_id}:{start_time}:{end_time}"
+
+        # Check if data is cached in Redis
+        cached_data = None
+        if redis_client:
+            try:
+                cached_data = redis_client.get(cache_key)
+            except Exception as e:
+                logging.error(f"Error retrieving data from Redis: {e}")
+        else:
+            logging.error("Redis client not available, skipping cache check")
+
+        if cached_data:
+            logging.info(f"Retrieved cached data for {cache_key}")
+            return json.loads(cached_data)
+
+        token = Config.AIRQO_API_TOKEN
+        analytics_url = Config.ANALTICS_URL
+        if not token:
+            logging.error("AIRQO_API_TOKEN is not set.")
+            return None
+
+        query_params = {'token': token}
+        url = f"{analytics_url}?{urlencode(query_params)}"
+        payload = {"grid_id": grid_id, "start_time": start_time, "end_time": end_time}
+        logging.info(f"Fetching air quality data with payload: {payload}")
+
+        try:
+            response = requests.post(url, json=payload, timeout=5)
+            response.raise_for_status()
+            data = response.json()
+            # Cache response in Redis for 1 hour 
+            # setex is redis command to set a key with an expiration time
+            with data_fetch_lock:
+                if redis_client:
+                    redis_client.setex(cache_key, 3600, json.dumps(data))
+            logging.info(f"Data fetched and cached for grid_id: {grid_id}")
+            return data
+        except requests.exceptions.HTTPError as http_err:
+            logging.error(f"HTTP error: {http_err}")
+        except requests.exceptions.RequestException as req_err:
+            logging.error(f"Request error: {req_err}")
+        except ValueError as json_err:
+            logging.error(f"JSON error: {json_err}")
+        return None
+
+class AirQualityChatbot:
+    def __init__(self, air_quality_data):
+        self.data = air_quality_data or {} 
+        self.grid_name = self.data.get('airquality', {}).get('sites', {}).get('grid name', ['Unknown'])[0]
+        self.annual_data = self.data.get('airquality', {}).get('annual_pm', [{}])[0] or {}
+        self.daily_mean_data = self.data.get('airquality', {}).get('daily_mean_pm', []) or []
+        self.diurnal = self.data.get('airquality', {}).get('diurnal', []) or []
+        self.monthly_data = self.data.get('airquality', {}).get('site_monthly_mean_pm', []) or []
+        self.site_names = [item.get('site_name', 'Unknown') for item in self.data.get('airquality', {}).get('site_annual_mean_pm', [])] or ['Unknown']
+        self.num_sites = self.data.get('airquality', {}).get('sites', {}).get('number_of_sites', 'Unknown')
+        self.starttime = self.data.get('airquality', {}).get('period', {}).get('startTime', '')[:10] or 'N/A'
+        self.endtime = self.data.get('airquality', {}).get('period', {}).get('endTime', '')[:10] or 'N/A'
+        self.annual_pm2_5 = self.annual_data.get("pm2_5_calibrated_value", 'N/A')
+
+        # Sort daily_mean_data to get the most recent measurement
+        if self.daily_mean_data:
+            sorted_daily = sorted(self.daily_mean_data, key=lambda x: x.get('date', ''), reverse=True)
+            self.today_pm2_5 = sorted_daily[0].get('pm2_5_calibrated_value', 'N/A') if sorted_daily else 'N/A'
+            self.today_date = sorted_daily[0].get('date', 'N/A') if sorted_daily else 'N/A'
+        else:
+            self.today_pm2_5 = 'N/A'
+            self.today_date = 'N/A'
+
+        self.peak_diurnal = max(self.diurnal, key=lambda x: x.get('pm2_5_calibrated_value', 0)) if self.diurnal else {}
+
+        try:
+            # Gemini model
+            self.gemini_model = genai.GenerativeModel('gemini-2.0-flash')
+        except Exception as e:
+            logging.error(f"Failed to initialize Gemini model: {e}")
+            self.gemini_model = None
+        self.lock = threading.Lock()
+
+    def _prepare_data_context(self):
+        """Prepare a concise data context for the LLM."""
+        return (
+            f"AirQo data for {self.grid_name} ({self.starttime}-{self.endtime}): "
+            f"Annual PM2.5={self.annual_pm2_5} µg/m³, Sites={self.num_sites}, "
+            f"Most recent daily PM2.5={self.today_pm2_5} µg/m³ on {self.today_date}, "
+            f"Diurnal peak={self.peak_diurnal.get('pm2_5_calibrated_value', 'N/A')} µg/m³ at {self.peak_diurnal.get('hour', 'N/A')}:00, "
+            f"Site names={self.site_names}."
+        )
+
+    def _rule_based_response(self, user_prompt):
+        """Handle common queries with precomputed responses."""
+        prompt = user_prompt.lower()
+
+        if re.search(r"(today|now).*air.*quality", prompt):
+            if self.today_pm2_5 != 'N/A':
+                return f"The most recent PM2.5 in {self.grid_name} is {self.today_pm2_5} µg/m³ on {self.today_date}."
+            return "No recent air quality data available."
+
+        if re.search(r"(worst|highest|peak).*time", prompt):
+            if self.peak_diurnal:
+                return f"Pollution peaks at {self.peak_diurnal.get('hour', 'N/A')}:00 with {self.peak_diurnal.get('pm2_5_calibrated_value', 'N/A')} µg/m³."
+            return "No diurnal data available."
+
+        if re.search(r"how.*many.*(site|sites|monitors)", prompt):
+            if self.num_sites != 'Unknown':
+                return f"There are {self.num_sites} monitoring sites in {self.grid_name}."
+            return "Number of sites is not available."
+
+        if re.search(r"(year|annual).*average", prompt):
+            if self.annual_pm2_5 != 'N/A':
+                return f"The annual PM2.5 average in {self.grid_name} is {self.annual_pm2_5} µg/m³."
+            return "Annual air quality data is not available."
+
+        if re.search(r"(where|which|list).*site|sites|locations", prompt):
+            if self.site_names != ['Unknown']:
+                return f"Monitoring sites in {self.grid_name}: {', '.join(self.site_names)}."
+            return "Site information is not available."
+
+        return None
+
+    def _llm_response(self, user_prompt):
+        """Generate a response using the Gemini model for complex queries."""
+        if not self.gemini_model:
+            return "Language model is not available."
+
+        full_prompt = (
+            f"Data: {self._prepare_data_context()}\n"
+            f"User: {user_prompt}\n"
+            "Respond concisely and accurately based on the data."
+        )
+
+        try:
+            response = self.gemini_model.generate_content(full_prompt)
+            return response.text
+        except Exception as e:
+            logging.error(f"LLM error: {e}")
+            return "Sorry, I couldn't generate a response."
+
+    def chat(self, user_prompt):
+        """Process user queries and return appropriate responses."""
+        if not self.data:
+            return "Air quality data is not available for the specified grid and time period."
+        if not user_prompt or not isinstance(user_prompt, str):
+            return "Please provide a valid question about air quality."
+        if len(user_prompt) > 500:
+            return "Your question is too long. Please keep it under 500 characters."
+
+        rule_response = self._rule_based_response(user_prompt)
+        if rule_response:
+            return rule_response
+        return self._llm_response(user_prompt)

src/spatial/models/report_datafetcher.py

@@ -77,7 +77,7 @@ def __init__(self, data):
 
 
         # Initialize models once in the constructor
-        self.gemini_model = genai.GenerativeModel('gemini-pro')
+        self.gemini_model = genai.GenerativeModel('gemini-2.0-flash')
 
     def _prepare_base_info(self):
         return (

src/spatial/requirements.txt

@@ -26,3 +26,4 @@ joblib~=1.4.2
 lightgbm~=4.1.0 
 numpy
 google-generativeai 
+redis

src/spatial/views/chatbot_views.py

@@ -0,0 +1,80 @@
+from flask import request, jsonify
+# Assuming these are in models/chatbot_model.py based on the import
+from models.chatbot_model import AirQualityChatbot, DataFetcher  
+import numpy as np
+import logging
+
+#logging.basicConfig(filename="report_log.log", level=logging.INFO, filemode="w")
+logger = logging.getLogger(__name__)
+
+class ChatbotView:
+    @staticmethod
+    def chat_endpoint():
+        """
+        Handles chatbot API requests for air quality information
+        Expects JSON payload with grid_id, start_time, end_time, and prompt
+        Returns JSON response with chatbot's answer or error message
+        """
+        # Validate request payload
+        payload = request.json
+        if not payload or not all(key in payload for key in ["grid_id", "start_time", "end_time", "prompt"]):
+            logger.error("Invalid payload: missing required fields")
+            return jsonify({
+                "error": "Missing required fields: grid_id, start_time, end_time, prompt",
+                "status": "failure"
+            }), 400
+
+        # Extract parameters
+        grid_id = payload["grid_id"]
+        start_time = payload["start_time"]
+        end_time = payload["end_time"]
+        user_prompt = payload["prompt"]
+
+        # Validate prompt
+        if not user_prompt or not isinstance(user_prompt, str):
+            logger.error(f"Invalid prompt received: {user_prompt}")
+            return jsonify({
+                "error": "No valid prompt provided",
+                "status": "failure"
+            }), 400
+
+        try:
+            # Fetch air quality data with logging
+            logger.info(f"Fetching data for grid_id: {grid_id}, {start_time} to {end_time}")
+            air_quality_data = DataFetcher.fetch_air_quality_data(grid_id, start_time, end_time)
+
+            if not air_quality_data or 'airquality' not in air_quality_data:
+                logger.error(f"No valid air quality data returned for grid_id: {grid_id}")
+                return jsonify({
+                    "error": "Failed to fetch air quality data",
+                    "status": "failure"
+                }), 500
+
+            # Initialize chatbot and get response
+            chatbot = AirQualityChatbot(air_quality_data)
+            response = chatbot.chat(user_prompt)
+
+            if not response:
+                logger.warning(f"Empty response generated for prompt: {user_prompt}")
+                return jsonify({
+                    "error": "No response generated",
+                    "status": "failure"
+                }), 500
+
+            logger.info(f"Successfully processed request for {grid_id}")
+            return jsonify({
+                "response": response,
+                "status": "success",
+                "grid_id": grid_id,
+                "period": {
+                    "start_time": start_time,
+                    "end_time": end_time
+                }
+            }), 200
+
+        except Exception as e:
+            logger.error(f"Unhandled exception: {str(e)}")
+            return jsonify({
+                "error": "Internal server error", 
+                "status": "failure"
+            }), 500