import re import json import requests import traceback import time import os from typing import Dict, Any, List, Optional, Tuple from datetime import datetime, timedelta # Updated imports for pydantic from pydantic import BaseModel, Field # Updated imports for LangChain from langchain_core.prompts import PromptTemplate, ChatPromptTemplate from langchain_core.output_parsers import JsonOutputParser from langchain_ollama import OllamaLLM from langchain.chains import LLMChain from langchain.callbacks.manager import CallbackManager from langchain.callbacks.streaming_stdout import StreamingStdOutCallbackHandler from langchain_huggingface.embeddings import HuggingFaceEmbeddings # Enhanced HuggingFace imports for improved functionality from transformers import pipeline, AutoTokenizer, AutoModelForSequenceClassification import numpy as np # Import endpoints documentation from endpoints_documentation import endpoints_documentation # Set environment variables for HuggingFace os.environ["HF_HOME"] = "/tmp/huggingface" os.environ["HF_HUB_DISABLE_SYMLINKS_WARNING"] = "1" class ChatMessage(BaseModel): """Data model for chat messages""" message_id: str = Field(..., description="Unique identifier for the message") user_id: str = Field(..., description="User identifier") message: str = Field(..., description="The user's message") timestamp: datetime = Field(default_factory=datetime.now, description="When the message was sent") language: str = Field(default="english", description="Detected language of the message") class ChatResponse(BaseModel): """Data model for chatbot responses""" response_id: str = Field(..., description="Unique identifier for the response") response_type: str = Field(..., description="Type of response: 'conversation' or 'api_action'") message: str = Field(..., description="The chatbot's response message") api_call_made: bool = Field(default=False, description="Whether an API call was made") api_data: Optional[Dict[str, Any]] = Field(default=None, description="API response data if applicable") language: str = Field(default="english", description="Language of the response") timestamp: datetime = Field(default_factory=datetime.now, description="When the response was generated") class RouterResponse(BaseModel): """Data model for router chain response""" intent: str = Field(..., description="Either 'API_ACTION' or 'CONVERSATION'") confidence: float = Field(..., description="Confidence score between 0.0 and 1.0") reasoning: str = Field(..., description="Explanation of the decision") endpoint: Optional[str] = Field(default=None, description="API endpoint if intent is API_ACTION") method: Optional[str] = Field(default=None, description="HTTP method if intent is API_ACTION") params: Dict[str, Any] = Field(default_factory=dict, description="Parameters for API call") missing_required: List[str] = Field(default_factory=list, description="Missing required parameters") class HealthcareChatbot: def __init__(self): self.endpoints_documentation = endpoints_documentation self.ollama_base_url = "http://localhost:11434" self.model_name = "gemma3" self.BASE_URL = 'https://4646-197-54-54-66.ngrok-free.app' self.headers = {'Content-type': 'application/json'} self.user_id = '19cc0380-57e8-45db-94e4-6ad6e6802084' self.max_retries = 3 self.retry_delay = 2 # Store conversation history self.conversation_history = [] self.max_history_length = 10 # Keep last 10 exchanges # Initialize components self._initialize_language_tools() self._initialize_llm() self._initialize_parsers_and_chains() self._initialize_date_parser() print("Healthcare Chatbot initialized successfully!") self._print_welcome_message() def _print_welcome_message(self): """Print welcome message in both languages""" print("\n" + "="*60) print("🏥 HEALTHCARE CHATBOT READY") print("="*60) print("English: Hello! I'm your healthcare assistant. I can help you with:") print("• Booking and managing appointments") print("• Finding hospital information") print("• Viewing your medical records") print("• General healthcare questions") print() print("Arabic: مرحباً! أنا مساعدك الطبي. يمكنني مساعدتك في:") print("• حجز وإدارة المواعيد") print("• العثور على معلومات المستشفى") print("• عرض سجلاتك الطبية") print("• الأسئلة الطبية العامة") print("="*60) print("Type 'quit' or 'خروج' to exit\n") def _initialize_language_tools(self): """Initialize language processing tools""" try: self.embeddings = HuggingFaceEmbeddings(model_name="intfloat/multilingual-e5-large") self.language_classifier = pipeline( "text-classification", model="papluca/xlm-roberta-base-language-detection", top_k=1 ) self.sentiment_analyzer = pipeline( "sentiment-analysis", model="cardiffnlp/twitter-xlm-roberta-base-sentiment" ) print("✓ Language processing models loaded successfully") except Exception as e: print(f"⚠ Warning: Some language models failed to load: {e}") self.language_classifier = None self.sentiment_analyzer = None def _initialize_date_parser(self): """Initialize date parsing model""" try: self.date_parser = pipeline( "token-classification", model="Jean-Baptiste/roberta-large-ner-english", aggregation_strategy="simple" ) except Exception as e: print(f"⚠ Warning: Date parsing model failed to load: {e}") self.date_parser = None def _initialize_llm(self): """Initialize the LLM""" callbacks = [StreamingStdOutCallbackHandler()] self.llm = OllamaLLM( model=self.model_name, base_url=self.ollama_base_url, callbacks=callbacks, temperature=0.7, num_ctx=8192, top_p=0.9, request_timeout=60, ) def _initialize_parsers_and_chains(self): """Initialize all prompt templates and chains - REVAMPED to 3 chains only""" self.json_parser = JsonOutputParser(pydantic_object=RouterResponse) # UNIFIED ROUTER CHAIN - Handles both intent classification AND API routing self.router_prompt_template = PromptTemplate( template=""" You are a routing system. Your job is simple: 1. Understand what the user wants 2. Handle any dates/times in their request with PRECISE calculations 3. Check if any endpoint can do what they want 4. If yes = API_ACTION, if no = CONVERSATION ## Available API Endpoints Documentation {endpoints_documentation} ## User Query to Analyze Query: "{user_query}" Language: {detected_language} Current Context: - DateTime: {current_datetime} - Timezone: {timezone} - Current Day of Week: {current_day_name} ## Step-by-Step Analysis **STEP 1: What does the user want?** - If query is in Arabic, translate it to English first - Identify the exact action or information the user is requesting - Focus on understanding their underlying need, not just the words **STEP 2: Handle Date/Time Processing with PRECISE Calculations** IMPORTANT: Use the current datetime ({current_datetime}) and timezone ({timezone}) for ALL calculations. ### Current Date Reference Points: - Today is: {current_datetime} - Current day of week: {current_day_name} - Current timezone: {timezone} ### Arabic Date/Time Expressions Processing: **Basic Relative Dates:** - "اليوم" (today) = {current_datetime} date portion - "غدا" (tomorrow) = current date + 1 day - "أمس" (yesterday) = current date - 1 day - "بعد غد" (day after tomorrow) = current date + 2 days **Weekly Expressions - CALCULATE PRECISELY:** - "الأسبوع القادم" (next week) = current date + 7 days - "الأسبوع الماضي" (last week) = current date - 7 days **Specific Weekday Calculations - MOST IMPORTANT:** For expressions like "يوم [weekday] القادم" (next [weekday]): 1. Identify the target weekday from Arabic names: - الأحد (Sunday) = 0 - الاثنين (Monday) = 1 - الثلاثاء (Tuesday) = 2 - الأربعاء (Wednesday) = 3 - الخميس (Thursday) = 4 - الجمعة (Friday) = 5 - السبت (Saturday) = 6 2. Calculate days to add: - Get current weekday number (0=Sunday, 1=Monday, etc.) - Target weekday number - If target > current: days_to_add = target - current - If target <= current: days_to_add = 7 - (current - target) - Final date = current_date + days_to_add **Example Calculation:** If today is Sunday (June 1, 2025) and user says "يوم الاربع القادم" (next Wednesday): - Current weekday: 0 (Sunday) - Target weekday: 3 (Wednesday) - Days to add: 3 - 0 = 3 - Result: June 1 + 3 days = June 4, 2025 **Monthly/Yearly Expressions:** - "الشهر القادم" (next month) = add 1 month to current date - "الشهر الماضي" (last month) = subtract 1 month from current date - "السنة القادمة" (next year) = add 1 year to current date **Time Expressions:** - "صباحًا" (morning/AM) = 09:00 if no specific time given - "مساءً" (evening/PM) = 18:00 if no specific time given - "ظهرًا" (noon) = 12:00 - "منتصف الليل" (midnight) = 00:00 - "بعد ساعتين" (in 2 hours) = current time + 2 hours - "قبل ساعة" (1 hour ago) = current time - 1 hour **Date Format Output:** - Always convert final calculated date to ISO 8601 format: YYYY-MM-DDTHH:MM:SS - Include timezone offset if available - For date-only expressions, use 00:00:00 as default time **STEP 3: Find matching endpoint** - Read each endpoint description in the documentation - Check if any endpoint's purpose can fulfill what the user wants - Match based on functionality, not keywords **STEP 4: Decision** - Found matching endpoint = "API_ACTION" - No matching endpoint = "CONVERSATION" **STEP 5: Parameter Extraction (only if API_ACTION)** - Extract parameter values from user query - Use the CALCULATED dates/times from Step 2 - Convert all dates/times to ISO 8601 format (YYYY-MM-DDTHH:MM:SS) - List any missing required parameters - **CRITICAL: All parameters must be in English** - Translate any Arabic text to English - Convert names to English equivalents (e.g., "دكتور احمد" → "Dr. Ahmed") - Use standard English terms for all parameters ## Output Format {{ "intent": "CONVERSATION|API_ACTION", "confidence": 0.8, "reasoning": "User wants: [what user actually needs]. Date/time processing: [show exact calculation: current date + X days = final date]. Found endpoint: [endpoint path and why it matches] OR No endpoint matches this need", "endpoint": "/exact/endpoint/path", "method": "GET|POST|PUT|DELETE", "params": {{ // ALL VALUES MUST BE IN ENGLISH // Arabic terms must be translated to English equivalents }}, "missing_required": [], "calculated_datetime": "YYYY-MM-DDTHH:MM:SS (if date/time was processed)" }} ## CRITICAL REMINDERS: 1. ALWAYS use the provided current_datetime ({current_datetime}) as your base for calculations 2. For "next weekday" expressions, calculate the exact number of days to add 3. Show your calculation work in the reasoning field 4. Double-check weekday numbers: Sunday=0, Monday=1, Tuesday=2, Wednesday=3, Thursday=4, Friday=5, Saturday=6 5. **ALL PARAMETERS MUST BE IN ENGLISH** - translate any Arabic text before output **FINAL CHECK BEFORE OUTPUTTING:** 🔍 **MANDATORY LANGUAGE CHECK:** 1. Examine every value in the params object 2. If ANY value contains Arabic characters (ا-ي), you MUST: - Translate it to English - Convert names to English equivalents - Replace Arabic terms with English counterparts 3. Only output JSON when ALL parameters are in English Now analyze the user query step by step and give me the JSON response. """, input_variables=["user_query", "detected_language", "extracted_keywords", "sentiment_analysis", "endpoints_documentation", "current_datetime", "timezone", "current_day_name"] ) ''' Third Approach ''' # self.router_prompt_template = PromptTemplate( # template=""" # You are a routing system. Your job is simple: # 1. Understand what the user wants # 2. Handle any dates/times in their request with PRECISE calculations # 3. Check if any endpoint can do what they want # 4. If yes = API_ACTION, if no = CONVERSATION # ## Available API Endpoints Documentation # {endpoints_documentation} # ## User Query to Analyze # Query: "{user_query}" # Language: {detected_language} # Current Context: # - DateTime: {current_datetime} # - Timezone: {timezone} # - Current Day of Week: {current_day_name} # ## Step-by-Step Analysis # **STEP 1: What does the user want?** # - If query is in Arabic, translate it to English first # - Identify the exact action or information the user is requesting # - Focus on understanding their underlying need, not just the words # **STEP 2: Handle Date/Time Processing with PRECISE Calculations** # IMPORTANT: Use the current datetime ({current_datetime}) and timezone ({timezone}) for ALL calculations. # ### Current Date Reference Points: # - Today is: {current_datetime} # - Current day of week: {current_day_name} # - Current timezone: {timezone} # ### Arabic Date/Time Expressions Processing: # **Basic Relative Dates:** # - "اليوم" (today) = {current_datetime} date portion # - "غدا" (tomorrow) = current date + 1 day # - "أمس" (yesterday) = current date - 1 day # - "بعد غد" (day after tomorrow) = current date + 2 days # **Weekly Expressions - CALCULATE PRECISELY:** # - "الأسبوع القادم" (next week) = current date + 7 days # - "الأسبوع الماضي" (last week) = current date - 7 days # **Specific Weekday Calculations - MOST IMPORTANT:** # For expressions like "يوم [weekday] القادم" (next [weekday]): # 1. Identify the target weekday from Arabic names: # - الأحد (Sunday) = 0 # - الاثنين (Monday) = 1 # - الثلاثاء (Tuesday) = 2 # - الأربعاء (Wednesday) = 3 # - الخميس (Thursday) = 4 # - الجمعة (Friday) = 5 # - السبت (Saturday) = 6 # 2. Calculate days to add: # - Get current weekday number (0=Sunday, 1=Monday, etc.) # - Target weekday number # - If target > current: days_to_add = target - current # - If target <= current: days_to_add = 7 - (current - target) # - Final date = current_date + days_to_add # **Example Calculation:** # If today is Sunday (June 1, 2025) and user says "يوم الاربع القادم" (next Wednesday): # - Current weekday: 0 (Sunday) # - Target weekday: 3 (Wednesday) # - Days to add: 3 - 0 = 3 # - Result: June 1 + 3 days = June 4, 2025 # **Monthly/Yearly Expressions:** # - "الشهر القادم" (next month) = add 1 month to current date # - "الشهر الماضي" (last month) = subtract 1 month from current date # - "السنة القادمة" (next year) = add 1 year to current date # **Time Expressions:** # - "صباحًا" (morning/AM) = 09:00 if no specific time given # - "مساءً" (evening/PM) = 18:00 if no specific time given # - "ظهرًا" (noon) = 12:00 # - "منتصف الليل" (midnight) = 00:00 # - "بعد ساعتين" (in 2 hours) = current time + 2 hours # - "قبل ساعة" (1 hour ago) = current time - 1 hour # **Date Format Output:** # - Always convert final calculated date to ISO 8601 format: YYYY-MM-DDTHH:MM:SS # - Include timezone offset if available # - For date-only expressions, use 00:00:00 as default time # **STEP 3: Find matching endpoint** # - Read each endpoint description in the documentation # - Check if any endpoint's purpose can fulfill what the user wants # - Match based on functionality, not keywords # **STEP 4: Decision** # - Found matching endpoint = "API_ACTION" # - No matching endpoint = "CONVERSATION" # **STEP 5: Parameter Extraction (only if API_ACTION)** # - Extract parameter values from the user's query and translate the parameters value to English. # - Use the CALCULATED dates/times from Step 2 # - Convert all dates/times to ISO 8601 format (YYYY-MM-DDTHH:MM:SS) # - List any missing required parameters # ## Output Format # {{ # "intent": "CONVERSATION|API_ACTION", # "confidence": 0.8, # "reasoning": "User wants: [what user actually needs]. Date/time processing: [show exact calculation: current date + X days = final date]. Found endpoint: [endpoint path and why it matches] OR No endpoint matches this need", # "endpoint": "/exact/endpoint/path", # "method": "GET|POST|PUT|DELETE", # "params": {{}}, # "missing_required": [], # "calculated_datetime": "YYYY-MM-DDTHH:MM:SS (if date/time was processed)" # }} # ## CRITICAL REMINDERS: # 1. ALWAYS use the provided current_datetime ({current_datetime}) as your base for calculations # 2. For "next weekday" expressions, calculate the exact number of days to add # 3. Show your calculation work in the reasoning field # 4. Double-check weekday numbers: Sunday=0, Monday=1, Tuesday=2, Wednesday=3, Thursday=4, Friday=5, Saturday=6 # Now analyze the user query step by step and give me the JSON response. # """, # input_variables=["user_query", "detected_language", "extracted_keywords", # "sentiment_analysis", "endpoints_documentation", "current_datetime", # "timezone", "current_day_name"] # ) ''' Second Approach ''' # self.router_prompt_template = PromptTemplate( # template=""" # You are a routing system. Your job is simple: # 1. Understand what the user wants # 2. Handle any dates/times in their request # 3. Check if any endpoint can do what they want # 4. If yes = API_ACTION, if no = CONVERSATION # ## Available API Endpoints Documentation # {endpoints_documentation} # ## User Query to Analyze # Query: "{user_query}" # Language: {detected_language} # Current Context: # - DateTime: {current_datetime} # - Timezone: {timezone} # ## Step-by-Step Analysis # **STEP 1: What does the user want?** # - If query is in Arabic, translate it to English first # - Identify the exact action or information the user is requesting # - Focus on understanding their underlying need, not just the words # **STEP 2: Handle Date/Time Processing** # - Identify any temporal expressions in the user query # - Convert relative dates/times using the current context: # * "اليوم" (today) = current date # * "غدا" (tomorrow) = current date + 1 day # * "أمس" (yesterday) = current date - 1 day # * "الأسبوع القادم" (next week) = current date + 7 days # * "بعد ساعتين" (in 2 hours) = current time + 2 hours # * "صباحًا" (morning/AM), "مساءً" (evening/PM) # * "الشهر القادم" (next month) = current date + 1 month # * "الأسبوع الماضي" (last week) = current date - 7 days # - Handle different date formats and languages # - Account for timezone differences # - Convert to ISO 8601 format: YYYY-MM-DDTHH:MM:SS # **STEP 3: Find matching endpoint** # - Read each endpoint description in the documentation # - Check if any endpoint's purpose can fulfill what the user wants # - Match based on functionality, not keywords # **STEP 3: Decision** # - Found matching endpoint = "API_ACTION" # - No matching endpoint = "CONVERSATION" # **STEP 4: Parameter Extraction (only if API_ACTION)** # - Extract parameter values from user query # - Convert dates/times to ISO 8601 format (YYYY-MM-DDTHH:MM:SS) using the context above # - List any missing required parameters # ## Output Format # {{ # "intent": "CONVERSATION|API_ACTION", # "confidence": 0.8, # "reasoning": "User wants: [what user actually needs]. Date/time processing: [any date conversions made]. Found endpoint: [endpoint path and why it matches] OR No endpoint matches this need", # "endpoint": "/exact/endpoint/path", # "method": "GET|POST|PUT|DELETE", # "params": {{}}, # "missing_required": [] # }} # Now analyze the user query step by step and give me the JSON response. # """, # input_variables=["user_query", "detected_language", "extracted_keywords", # "sentiment_analysis", "endpoints_documentation", "current_datetime", "timezone"] # ) ''' First Approach ''' # self.router_prompt_template = PromptTemplate( # template=""" # You are a routing system. Your job is simple: # 1. Understand what the user wants # 2. Check if any endpoint can do what they want # 3. If yes = API_ACTION, if no = CONVERSATION # ## Available API Endpoints Documentation # {endpoints_documentation} # ## User Query to Analyze # Query: "{user_query}" # Language: {detected_language} # ## Step-by-Step Analysis # **STEP 1: What does the user want?** # - If query is in Arabic, translate it to English first # - Identify the exact action or information the user is requesting # - Focus on understanding their underlying need, not just the words # **STEP 2: Find matching endpoint** # - Read each endpoint description in the documentation # - Check if any endpoint's purpose can fulfill what the user wants # - Match based on functionality, not keywords # **STEP 3: Decision** # - Found matching endpoint = "API_ACTION" # - No matching endpoint = "CONVERSATION" # ## Output Format # {{ # "intent": "CONVERSATION|API_ACTION", # "confidence": 0.8, # "reasoning": "User wants: [what user actually needs]. Found endpoint: [endpoint path and why it matches] OR No endpoint matches this need", # "endpoint": "/exact/endpoint/path", # "method": "GET|POST|PUT|DELETE", # "params": {{}}, # "missing_required": [] # }} # Now analyze the user query step by step and give me the JSON response. # """, # input_variables=["user_query", "detected_language", "extracted_keywords", # "sentiment_analysis", "endpoints_documentation"] # ) # CONVERSATION CHAIN - Handles conversational responses self.conversation_template = PromptTemplate( template=""" You are a friendly and professional healthcare chatbot assistant. === RESPONSE GUIDELINES === - Respond ONLY in {detected_language} - Be helpful, empathetic, and professional - Keep responses concise but informative - Use appropriate medical terminology when needed - Maintain a caring and supportive tone === CONTEXT === User Message: {user_query} Language: {detected_language} Sentiment: {sentiment_analysis} Conversation History: {conversation_history} === LANGUAGE-SPECIFIC INSTRUCTIONS === FOR ARABIC RESPONSES: - Use Modern Standard Arabic (الفصحى) - Be respectful and formal as appropriate in Arabic culture - Use proper Arabic medical terminology - Keep sentences clear and grammatically correct FOR ENGLISH RESPONSES: - Use clear, professional English - Be warm and approachable - Use appropriate medical terminology === RESPONSE RULES === 1. Address the user's question or comment directly 2. Provide helpful information when possible 3. If you cannot help with something specific, explain what you CAN help with 4. Never provide specific medical advice - always recommend consulting healthcare professionals 5. Be encouraging and supportive 6. Do NOT mix languages in your response 7. End responses naturally without asking multiple questions Generate a helpful conversational response:""", input_variables=["user_query", "detected_language", "sentiment_analysis", "conversation_history"] ) # API RESPONSE CHAIN - Formats API responses for users # self.api_response_template = PromptTemplate( # template=""" # You are a professional healthcare assistant. Answer the user's question using the provided API data. # User Query: {user_query} # User Sentiment: {sentiment_analysis} # Response Language: {detected_language} # API Response Data: # {api_response} # === INSTRUCTIONS === # 1. Read and understand the API response data above # 2. Use ONLY the actual data from the API response - never make up information # 3. Respond in {detected_language} language only # 4. Write like you're talking to a friend or family member - warm, friendly, and caring # 5. Make it sound natural and conversational, not like a system message # 6. Convert technical data to simple, everyday language # === DATE AND TIME FORMATTING === # When you see date_time fields like '2025-05-30T10:28:10': # - For English: Convert to "May 30, 2025 at 10:28 AM" # - For Arabic: Convert to "٣٠ مايو ٢٠٢٥ في الساعة ١٠:٢٨ صباحاً" # === RESPONSE EXAMPLES === # For appointment confirmations: # - English: "Great! I've got your appointment set up for May 30, 2025 at 10:28 AM. Everything looks good!" # - Arabic: "ممتاز! موعدك محجوز يوم ٣٠ مايو ٢٠٢٥ الساعة ١٠:٢٨ صباحاً. كل شيء جاهز!" # For appointment info: # - English: "Your next appointment is on May 30, 2025 at 10:28 AM. See you then!" # - Arabic: "موعدك القادم يوم ٣٠ مايو ٢٠٢٥ الساعة ١٠:٢٨ صباحاً. نراك قريباً!" # === TONE GUIDELINES === # - Use friendly words like: "Great!", "Perfect!", "All set!", "ممتاز!", "رائع!", "تمام!" # - Add reassuring phrases: "Everything looks good", "You're all set", "كل شيء جاهز", "تم بنجاح" # - Sound helpful and caring, not robotic or formal # === LANGUAGE FORMATTING === # For Arabic responses: # - Use Arabic numerals: ٠١٢٣٤٥٦٧٨٩ # - Use Arabic month names: يناير، فبراير، مارس، أبريل، مايو، يونيو، يوليو، أغسطس، سبتمبر، أكتوبر، نوفمبر، ديسمبر # - Friendly, warm Arabic tone # For English responses: # - Use standard English numerals # - 12-hour time format with AM/PM # - Friendly, conversational English tone # === CRITICAL RULES === # - Extract dates and times exactly as they appear in the API response # - Never use example dates or placeholder information # - Respond only in the specified language # - Make your response sound like a helpful friend, not a computer # - Focus on answering the user's specific question with warmth and care # Generate a friendly, helpful response using the API data provided above. # """, # input_variables=["user_query", "api_response", "detected_language", "sentiment_analysis"] # ) self.api_response_template = PromptTemplate( template=""" You are a professional healthcare assistant. Generate a natural language response to the user's query using ONLY the provided API data. User Query: {user_query} User Sentiment: {sentiment_analysis} Response Language: {detected_language} API Response Data: {api_response} === CORE INSTRUCTIONS === 1. Analyze the API response structure and extract relevant data points 2. Cross-reference with the user's query to determine what information to include 3. Respond in {detected_language} using a warm, conversational tone 4. Convert technical data into natural language appropriate for healthcare communication === DATE/TIME HANDLING === 1. Identify all date/time fields in the API response (look for ISO 8601 format: YYYY-MM-DDTHH:MM:SS) 2. For English responses: - Format dates as "Month Day, Year at HH:MM AM/PM" - Convert times to 12-hour format with proper AM/PM 3. For Arabic responses: - Format dates as "Day Month Year الساعة HH:MM صباحاً/مساءً" - Use Arabic numerals (٠١٢٣٤٥٦٧٨٩) - Use Arabic month names 4. Preserve all original date/time values - only change the formatting === RESPONSE GUIDELINES === 1. Use ONLY data present in the API response 2. Maintain a professional yet friendly healthcare tone 3. Adapt to the user's sentiment: - Positive: reinforce with encouraging language - Neutral: provide clear, factual information - Negative: show empathy and offer assistance 4. Structure the response to directly answer the user's query 5. Include relevant details from the API response that address the user's needs === CRITICAL RULES === 1. Never invent or hallucinate information not present in the API response 2. If the API response doesn't contain requested information, say so politely 3. All dates/times must exactly match the API data 4. Maintain strict language consistency (respond only in {detected_language}) 5. Format all technical data (IDs, codes, etc.) for easy understanding Generate a helpful response that addresses the user's query using the API data. """, input_variables=["user_query", "api_response", "detected_language", "sentiment_analysis"] ) # Create the 3 chains self.router_chain = LLMChain(llm=self.llm, prompt=self.router_prompt_template) self.conversation_chain = LLMChain(llm=self.llm, prompt=self.conversation_template) self.api_response_chain = LLMChain(llm=self.llm, prompt=self.api_response_template) def detect_language(self, text): """Detect language of the input text""" if self.language_classifier and len(text.strip()) > 3: try: result = self.language_classifier(text) detected_lang = result[0][0]['label'] confidence = result[0][0]['score'] if detected_lang in ['ar', 'arabic']: return "arabic" elif detected_lang in ['en', 'english']: return "english" elif confidence > 0.8: return "english" # Default to English for unsupported languages except: pass # Fallback: Basic Arabic detection arabic_pattern = re.compile(r'[\u0600-\u06FF\u0750-\u077F\u08A0-\u08FF]+') if arabic_pattern.search(text): return "arabic" return "english" def analyze_sentiment(self, text): """Analyze sentiment of the text""" if self.sentiment_analyzer and len(text.strip()) > 3: try: result = self.sentiment_analyzer(text) return { "sentiment": result[0]['label'], "score": result[0]['score'] } except: pass return {"sentiment": "NEUTRAL", "score": 0.5} def extract_keywords(self, text): """Extract keywords from text""" # Simple keyword extraction words = re.findall(r'\b\w+\b', text.lower()) # Filter out common words and keep meaningful ones stopwords = {'the', 'and', 'or', 'but', 'in', 'on', 'at', 'to', 'for', 'of', 'with', 'by', 'is', 'are', 'was', 'were'} keywords = [w for w in words if len(w) > 3 and w not in stopwords] return list(set(keywords))[:5] # Return top 5 unique keywords def get_conversation_context(self): """Get recent conversation history as context""" if not self.conversation_history: return "No previous conversation" context = [] for item in self.conversation_history[-3:]: # Last 3 exchanges context.append(f"User: {item['user_message']}") context.append(f"Bot: {item['bot_response'][:100]}...") # Truncate long responses return " | ".join(context) def add_to_history(self, user_message, bot_response, response_type): """Add exchange to conversation history""" self.conversation_history.append({ 'timestamp': datetime.now(), 'user_message': user_message, 'bot_response': bot_response, 'response_type': response_type }) # Keep only recent history if len(self.conversation_history) > self.max_history_length: self.conversation_history = self.conversation_history[-self.max_history_length:] def parse_relative_date(self, text, detected_language): """Parse relative dates from text using a combination of methods""" today = datetime.now() # Handle common relative date patterns in English and Arabic tomorrow_patterns = { 'english': [r'\btomorrow\b', r'\bnext day\b'], 'arabic': [r'\bغدا\b', r'\bبكرة\b', r'\bغدًا\b', r'\bالغد\b'] } next_week_patterns = { 'english': [r'\bnext week\b'], 'arabic': [r'\bالأسبوع القادم\b', r'\bالأسبوع المقبل\b', r'\bالاسبوع الجاي\b'] } # Check for "tomorrow" patterns for pattern in tomorrow_patterns.get(detected_language, []) + tomorrow_patterns.get('english', []): if re.search(pattern, text, re.IGNORECASE): return (today + timedelta(days=1)).strftime('%Y-%m-%dT%H:%M:%S') # Check for "next week" patterns for pattern in next_week_patterns.get(detected_language, []) + next_week_patterns.get('english', []): if re.search(pattern, text, re.IGNORECASE): return (today + timedelta(days=7)).strftime('%Y-%m-%dT%H:%M:%S') # If NER model is available, use it to extract date entities if self.date_parser and detected_language == 'english': try: date_entities = self.date_parser(text) for entity in date_entities: if entity['entity_group'] == 'DATE': print(f"Found date entity: {entity['word']}") # Default to tomorrow if we detect any date return (today + timedelta(days=1)).strftime('%Y-%m-%dT%H:%M:%S') except Exception as e: print(f"Error in date parsing: {e}") # Default return None if no date pattern is recognized return None def parse_router_response(self, router_text): """Parse the router chain response into structured data""" try: # Clean the response text cleaned_response = router_text # Remove any comments (both single-line and multi-line) cleaned_response = re.sub(r'//.*?$', '', cleaned_response, flags=re.MULTILINE) cleaned_response = re.sub(r'/\*.*?\*/', '', cleaned_response, flags=re.DOTALL) # Remove any trailing commas cleaned_response = re.sub(r',(\s*[}\]])', r'\1', cleaned_response) # Try different methods to parse the JSON response try: # First attempt: direct JSON parsing of cleaned response parsed_response = json.loads(cleaned_response) except json.JSONDecodeError: try: # Second attempt: extract JSON from markdown code block json_match = re.search(r'```(?:json)?\s*(\{.*?\})\s*```', cleaned_response, re.DOTALL) if json_match: parsed_response = json.loads(json_match.group(1)) else: raise ValueError("No JSON found in code block") except (json.JSONDecodeError, ValueError): try: # Third attempt: find JSON-like content using regex json_pattern = r'\{\s*"intent"\s*:.*?\}' json_match = re.search(json_pattern, cleaned_response, re.DOTALL) if json_match: json_str = json_match.group(0) # Additional cleaning for the extracted JSON json_str = re.sub(r'//.*?$', '', json_str, flags=re.MULTILINE) json_str = re.sub(r',(\s*[}\]])', r'\1', json_str) parsed_response = json.loads(json_str) else: raise ValueError("Could not extract JSON using regex") except (json.JSONDecodeError, ValueError): print(f"Failed to parse JSON. Raw response: {router_text}") print(f"Cleaned response: {cleaned_response}") # Return default conversation response on parse failure return { "intent": "CONVERSATION", "confidence": 0.5, "reasoning": "Failed to parse router response - defaulting to conversation", "endpoint": None, "method": None, "params": {}, "missing_required": [] } # Validate required fields and set defaults validated_response = { "intent": parsed_response.get("intent", "CONVERSATION"), "confidence": parsed_response.get("confidence", 0.5), "reasoning": parsed_response.get("reasoning", "Router decision"), "endpoint": parsed_response.get("endpoint"), "method": parsed_response.get("method"), "params": parsed_response.get("params", {}), "missing_required": parsed_response.get("missing_required", []) } return validated_response except Exception as e: print(f"Error parsing router response: {e}") return { "intent": "CONVERSATION", "confidence": 0.5, "reasoning": f"Parse error: {str(e)}", "endpoint": None, "method": None, "params": {}, "missing_required": [] } def handle_conversation(self, user_query, detected_language, sentiment_result): """Handle conversational responses""" try: result = self.conversation_chain.invoke({ "user_query": user_query, "detected_language": detected_language, "sentiment_analysis": json.dumps(sentiment_result), "conversation_history": self.get_conversation_context() }) return result["text"].strip() except Exception as e: # Fallback response if detected_language == "arabic": return "أعتذر، واجهت مشكلة في المعالجة. كيف يمكنني مساعدتك؟" else: return "I apologize, I encountered a processing issue. How can I help you?" def backend_call(self, data: Dict[str, Any]) -> Dict[str, Any]: """Make API call to backend with retry logic""" endpoint_url = data.get('endpoint') endpoint_method = data.get('method') endpoint_params = data.get('params', {}).copy() print(f"🔗 Making API call to {endpoint_method} {self.BASE_URL + endpoint_url} with params: {endpoint_params}") # Inject patient_id if needed if 'patient_id' in endpoint_params: endpoint_params['patient_id'] = self.user_id retries = 0 response = None while retries < self.max_retries: try: if endpoint_method.upper() == 'GET': response = requests.get( self.BASE_URL + endpoint_url, params=endpoint_params, headers=self.headers, timeout=10 ) elif endpoint_method.upper() in ['POST', 'PUT', 'DELETE']: response = requests.request( endpoint_method.upper(), self.BASE_URL + endpoint_url, json=endpoint_params, headers=self.headers, timeout=10 ) response.raise_for_status() print('Backend Response:', response.json()) return response.json() except requests.exceptions.RequestException as e: retries += 1 if retries >= self.max_retries: return { "error": "Backend API call failed after multiple retries", "details": str(e), "status_code": getattr(e.response, 'status_code', None) if hasattr(e, 'response') else None } time.sleep(self.retry_delay) def handle_api_action(self, user_query, detected_language, sentiment_result, keywords, router_data): """Handle API-based actions using router data""" try: # Parse relative dates and inject into parameters # parsed_date = self.parse_relative_date(user_query, detected_language) # if parsed_date: # print(f"Parsed relative date: {parsed_date}") # # Inject parsed date if available and a date parameter exists # date_params = ['appointment_date', 'date', 'schedule_date', 'date_time', 'new_date_time'] # for param in date_params: # if param in router_data['params']: # router_data['params'][param] = parsed_date # Inject patient_id if needed if 'patient_id' in router_data['params']: router_data['params']['patient_id'] = self.user_id else: router_data['params']['patient_id'] = self.user_id print(f"🔍 Final API call data: {router_data}") # Make backend API call api_response = self.backend_call(router_data) print("🔗 API response received:", api_response) # Generate user-friendly response user_response_result = self.api_response_chain.invoke({ "user_query": user_query, "api_response": json.dumps(api_response, indent=2), "detected_language": detected_language, "sentiment_analysis": json.dumps(sentiment_result), }) print("🔗 Final user response:", user_response_result["text"].strip()) return { "response": user_response_result["text"].strip(), "api_data": api_response, "routing_info": router_data } except Exception as e: # Fallback error response if detected_language == "arabic": error_msg = "أعتذر، لم أتمكن من معالجة طلبك. يرجى المحاولة مرة أخرى أو صياغة السؤال بطريقة مختلفة." else: error_msg = "I apologize, I couldn't process your request. Please try again or rephrase your question." return { "response": error_msg, "api_data": {"error": str(e)}, "routing_info": None } def chat(self, user_message: str) -> ChatResponse: """Main chat method that handles user messages - REVAMPED to use 3 chains""" start_time = time.time() # Check for exit commands if user_message.lower().strip() in ['quit', 'exit', 'خروج', 'bye', 'goodbye']: if self.detect_language(user_message) == "arabic": return ChatResponse( response_id=str(time.time()), response_type="conversation", message="مع السلامة! أتمنى لك يوماً سعيداً. 👋", language="arabic" ) else: return ChatResponse( response_id=str(time.time()), response_type="conversation", message="Goodbye! Have a great day! 👋", language="english" ) try: print(f"\n{'='*50}") print(f"🔍 Processing: '{user_message}'") print(f"{'='*50}") # Step 1: Language and sentiment analysis detected_language = self.detect_language(user_message) sentiment_result = self.analyze_sentiment(user_message) keywords = self.extract_keywords(user_message) print(f"🌐 Detected Language: {detected_language}") print(f"😊 Sentiment: {sentiment_result}") print(f"🔑 Keywords: {keywords}") # Step 2: Router Chain - Determine intent and route appropriately print(f"\n🤖 Running Router Chain...") router_result = self.router_chain.invoke({ "user_query": user_message, "detected_language": detected_language, "extracted_keywords": json.dumps(keywords), "sentiment_analysis": json.dumps(sentiment_result), "conversation_history": self.get_conversation_context(), "endpoints_documentation": json.dumps(self.endpoints_documentation, indent=2), "current_datetime": datetime.now().strftime('%Y-%m-%dT%H:%M:%S'), "timezone": "UTC", "current_day_name": datetime.now().strftime('%A'), }) # Parse router response router_data = self.parse_router_response(router_result["text"]) print(f"🎯 Router Decision: {router_data}") # Step 3: Handle based on intent if router_data["intent"] == "CONVERSATION" and router_data['endpoint'] == '': print(f"\n💬 Handling as CONVERSATION") response_text = self.handle_conversation(user_message, detected_language, sentiment_result) # Add to conversation history self.add_to_history(user_message, response_text, "conversation") return ChatResponse( response_id=str(time.time()), response_type="conversation", message=response_text, api_call_made=False, language=detected_language, api_data=None ) elif router_data["intent"] == "API_ACTION": print(f"\n🔗 Handling as API_ACTION") # Check for missing required parameters # if router_data.get("missing_required"): # missing_params = router_data["missing_required"] # if detected_language == "arabic": # response_text = f"أحتاج إلى مزيد من المعلومات: {', '.join(missing_params)}" # else: # response_text = f"I need more information: {', '.join(missing_params)}" # return ChatResponse( # response_id=str(time.time()), # response_type="conversation", # message=response_text, # api_call_made=False, # language=detected_language # ) # Handle API action api_result = self.handle_api_action( user_message, detected_language, sentiment_result, keywords, router_data ) # Add to conversation history self.add_to_history(user_message, api_result["response"], "api_action") return ChatResponse( response_id=str(time.time()), response_type="api_action", message=api_result["response"], api_call_made=True, # api_data=api_result["api_data"] # api_data=json.dumps(api_result["api_data"]) if 'action_result' in api_result else None, language=detected_language ) else: # Fallback for unknown intent print(f"⚠️ Unknown intent: {router_data['intent']}") fallback_response = self.handle_conversation(user_message, detected_language, sentiment_result) return ChatResponse( response_id=str(time.time()), response_type="conversation", message=fallback_response, api_call_made=False, language=detected_language ) except Exception as e: print(f"❌ Error in chat method: {str(e)}") print(f"❌ Traceback: {traceback.format_exc()}") # Fallback error response if self.detect_language(user_message) == "arabic": error_message = "أعتذر، حدث خطأ في معالجة رسالتك. يرجى المحاولة مرة أخرى." else: error_message = "I apologize, there was an error processing your message. Please try again." return ChatResponse( response_id=str(time.time()), response_type="conversation", message=error_message, api_call_made=False, language=self.detect_language(user_message) ) finally: end_time = time.time() print(f"⏱️ Processing time: {end_time - start_time:.2f} seconds") def run_interactive_chat(self): """Run the interactive chat interface""" try: while True: try: # Get user input user_input = input("\n👤 You: ").strip() if not user_input: continue # Process the message response = self.chat(user_input) # Display the response print(f"\n🤖 Bot: {response.message}") # Check for exit if user_input.lower() in ['quit', 'exit', 'خروج', 'bye', 'goodbye']: break except KeyboardInterrupt: print("\n\n👋 Chat interrupted. Goodbye!") break except EOFError: print("\n\n👋 Chat ended. Goodbye!") break except Exception as e: print(f"\n❌ Error: {e}") continue except Exception as e: print(f"❌ Fatal error in chat interface: {e}") def clear_history(self): """Clear conversation history""" self.conversation_history = [] print("🗑️ Conversation history cleared.") # def main(): # """Main function to run the healthcare chatbot""" # try: # print("🚀 Starting Healthcare Chatbot...") # chatbot = HealthcareChatbot() # chatbot.run_interactive_chat() # except KeyboardInterrupt: # print("\n\n👋 Shutting down gracefully...") # except Exception as e: # print(f"❌ Fatal error: {e}") # print(f"❌ Traceback: {traceback.format_exc()}") # if __name__ == "__main__": # main() from fastapi import FastAPI, HTTPException from pydantic import BaseModel from typing import Dict, Any, Optional app = FastAPI( title="Healthcare AI Assistant", description="An AI-powered healthcare assistant that handles appointment booking and queries", version="1.0.0" ) # Initialize the AI agent agent = HealthcareChatbot() class QueryRequest(BaseModel): query: str @app.post("/query") async def process_query(request: QueryRequest): """ Process a user query and return a response """ try: response = agent.chat(request.query).message return response except Exception as e: raise HTTPException(status_code=500, detail=str(e)) @app.get("/health") async def health_check(): """ Health check endpoint """ return {"status": "healthy", "service": "healthcare-ai-assistant"} @app.get("/") async def root(): return {"message": "Hello World"} # if __name__ == "__main__": # import uvicorn # uvicorn.run(app, host="0.0.0.0", port=8000)