Update app.py

app.py

@@ -1,6 +1,3 @@
-# Complete Malayalam Hospital Booking Chatbot using Llama 3.1-8B-Instruct
-# with HuggingFace Transformers Library in Google Colab
-
 import os
 import json
 import gradio as gr
@@ -11,23 +8,17 @@ import uuid
 import re
 import time
 import random
+import pandas as pd
+import numpy as np
 from transformers import AutoModelForCausalLM, AutoTokenizer
-from google.colab import auth
-from googleapiclient.discovery import build
 
 # Set up timezone for India
 IST = pytz.timezone('Asia/Kolkata')
 
 # ===== CONFIGURATION =====
 
-# Path to store the model locally (to avoid re-downloading)
-MODEL_PATH = "/content/llama-3.1-8b-instruct"
-
-# Replace with your actual Hugging Face token
-HF_TOKEN = ""  # Will be set via Colab input
-
-# Google Calendar API scopes
-SCOPES = ['https://www.googleapis.com/auth/calendar']
+# Model ID on Hugging Face
+MODEL_ID = "meta-llama/Meta-Llama-3.1-8B-Instruct"
 
 # Available doctors and departments for booking
 available_doctors = {
@@ -78,7 +69,7 @@ function_definitions = [
   },
   {
     "name": "book_appointment",
-    "description": "Book an appointment with a doctor and add it to Google Calendar",
+    "description": "Book an appointment with a doctor",
     "parameters": {
       "type": "object",
       "properties": {
@@ -171,8 +162,8 @@ def check_time_slots(doctor_name, date):
         "available_slots": available_slots
     }
 
-def book_appointment(appointment_details, calendar_service):
-    """Book an appointment with a doctor and add it to Google Calendar"""
+def book_appointment(appointment_details):
+    """Book an appointment with a doctor"""
     try:
         # Validate the appointment details first
         doctor_exists = False
@@ -205,44 +196,8 @@ def book_appointment(appointment_details, calendar_service):
         start_datetime = datetime.datetime(year, month, day, hours, minutes, 0, tzinfo=IST)
         end_datetime = start_datetime + datetime.timedelta(minutes=30)  # 30 minutes appointment
 
-        # Create the calendar event
-        event = {
-            'summary': f"Medical appointment with {appointment_details['doctor_name']}",
-            'location': 'City Hospital, Kochi, Kerala',
-            'description': appointment_details.get('description', 'Regular checkup'),
-            'start': {
-                'dateTime': start_datetime.isoformat(),
-                'timeZone': 'Asia/Kolkata',
-            },
-            'end': {
-                'dateTime': end_datetime.isoformat(),
-                'timeZone': 'Asia/Kolkata',
-            },
-            'attendees': [
-                {'email': 'doctor@cityhospital.com'},
-                {'email': 'patient@example.com'}  # In a real app, use actual email
-            ],
-            'reminders': {
-                'useDefault': False,
-                'overrides': [
-                    {'method': 'email', 'minutes': 24 * 60},
-                    {'method': 'popup', 'minutes': 60},
-                ],
-            },
-        }
-
-        # Add to Google Calendar
-        if calendar_service:
-            try:
-                event = calendar_service.events().insert(calendarId='primary', body=event).execute()
-                appointment_id = event['id']
-            except Exception as e:
-                print(f"Calendar service error: {e}")
-                # Generate a mock ID if calendar service fails
-                appointment_id = str(uuid.uuid4())
-        else:
-            # If no calendar service, generate a mock ID
-            appointment_id = str(uuid.uuid4())
+        # Generate a unique appointment ID
+        appointment_id = str(uuid.uuid4())
 
         # Store in local database
         appointments_db[appointment_id] = {
@@ -253,6 +208,8 @@ def book_appointment(appointment_details, calendar_service):
             "date": appointment_details["date"],
             "time": appointment_details["time"],
             "description": appointment_details.get("description", ""),
+            "start_datetime": start_datetime,
+            "end_datetime": end_datetime
         }
 
         return {
@@ -274,7 +231,7 @@ def book_appointment(appointment_details, calendar_service):
             "message": f"Failed to book appointment: {str(e)}"
         }
 
-def cancel_appointment(appointment_id, patient_phone, calendar_service):
+def cancel_appointment(appointment_id, patient_phone):
     """Cancel an existing appointment"""
     try:
         # Check if appointment exists in our database
@@ -291,14 +248,6 @@ def cancel_appointment(appointment_id, patient_phone, calendar_service):
                 "message": "Patient phone number does not match our records"
             }
 
-        # Delete from Google Calendar
-        if calendar_service:
-            try:
-                calendar_service.events().delete(calendarId='primary', eventId=appointment_id).execute()
-            except Exception as e:
-                print(f"Error deleting from calendar: {e}")
-                # Continue anyway to delete from local database
-
         # Remove from local database
         del appointments_db[appointment_id]
 
@@ -312,62 +261,21 @@ def cancel_appointment(appointment_id, patient_phone, calendar_service):
             "message": f"Failed to cancel appointment: {str(e)}"
         }
 
-# ===== GOOGLE CALENDAR AUTHENTICATION =====
-
-def get_calendar_service():
-    """Authenticate and return the Google Calendar service"""
-    creds = None
-
-    try:
-        # Authenticate using Colab's auth helper
-        auth.authenticate_user()
-
-        # Get credentials from the authenticated Colab user
-        from google.auth import default
-        creds, _ = default()
-
-        # Build and return the service
-        service = build('calendar', 'v3', credentials=creds)
-        return service
-    except Exception as e:
-        print(f"Error authenticating with Google Calendar: {e}")
-        print("Continuing without Google Calendar integration.")
-        return None
-
-# ===== LLAMA 3.1 MODEL SETUP =====
+# ===== LLAMA MODEL SETUP =====
 
 def load_llama_model():
     """Load the Llama 3.1 model and tokenizer"""
-    model_name = "meta-llama/Meta-Llama-3.1-8B-Instruct"
-
     print("Loading Llama 3.1 model and tokenizer...")
 
     try:
-        # Check if model is already downloaded
-        if os.path.exists(MODEL_PATH):
-            print(f"Loading model from local path: {MODEL_PATH}")
-            tokenizer = AutoTokenizer.from_pretrained(MODEL_PATH)
-            model = AutoModelForCausalLM.from_pretrained(
-                MODEL_PATH,
-                torch_dtype=torch.bfloat16,
-                device_map="auto",
-                low_cpu_mem_usage=True
-            )
-        else:
-            print(f"Downloading model from Hugging Face Hub")
-            tokenizer = AutoTokenizer.from_pretrained(model_name, token=HF_TOKEN)
-            model = AutoModelForCausalLM.from_pretrained(
-                model_name,
-                torch_dtype=torch.bfloat16,
-                device_map="auto",
-                low_cpu_mem_usage=True,
-                token=HF_TOKEN
-            )
-
-            # Save model locally to avoid re-downloading
-            print(f"Saving model to: {MODEL_PATH}")
-            tokenizer.save_pretrained(MODEL_PATH)
-            model.save_pretrained(MODEL_PATH)
+        # Use 4-bit quantization for smaller memory footprint
+        tokenizer = AutoTokenizer.from_pretrained(MODEL_ID)
+        model = AutoModelForCausalLM.from_pretrained(
+            MODEL_ID,
+            torch_dtype=torch.bfloat16,
+            device_map="auto",
+            load_in_4bit=True
+        )
 
         print("Model loaded successfully!")
         return model, tokenizer
@@ -440,9 +348,9 @@ def extract_function_call(response_text):
 
     return None
 
-def process_chat(message, chat_history, language, model_tokenizer_calendar):
+def process_chat(message, chat_history, language, model_tokenizer):
     """Process a chat message, calling functions when necessary"""
-    model, tokenizer, calendar_service = model_tokenizer_calendar
+    model, tokenizer = model_tokenizer
 
     try:
         # Create system prompt based on language preference
@@ -505,9 +413,9 @@ Current language preference: {language}"""
             elif function_name == "check_time_slots" and "doctor_name" in function_args and "date" in function_args:
                 function_result = check_time_slots(function_args["doctor_name"], function_args["date"])
             elif function_name == "book_appointment":
-                function_result = book_appointment(function_args, calendar_service)
+                function_result = book_appointment(function_args)
             elif function_name == "cancel_appointment" and "appointment_id" in function_args and "patient_phone" in function_args:
-                function_result = cancel_appointment(function_args["appointment_id"], function_args["patient_phone"], calendar_service)
+                function_result = cancel_appointment(function_args["appointment_id"], function_args["patient_phone"])
             else:
                 function_result = {"error": "Invalid function call or missing parameters"}
 
@@ -562,7 +470,7 @@ Current language preference: {language}"""
 
 # ===== GRADIO INTERFACE =====
 
-def create_gradio_interface(model, tokenizer, calendar_service):
+def create_gradio_interface(model, tokenizer):
     """Create the Gradio interface for the chatbot"""
 
     with gr.Blocks(css="""
@@ -627,7 +535,7 @@ def create_gradio_interface(model, tokenizer, calendar_service):
         # Set up event handlers
         submit.click(
             process_chat,
-            inputs=[msg, chat_history, language, gr.State((model, tokenizer, calendar_service))],
+            inputs=[msg, chat_history, language, gr.State((model, tokenizer))],
             outputs=[chatbot, chat_history]
         ).then(
             lambda: "",
@@ -637,7 +545,7 @@ def create_gradio_interface(model, tokenizer, calendar_service):
 
         msg.submit(
             process_chat,
-            inputs=[msg, chat_history, language, gr.State((model, tokenizer, calendar_service))],
+            inputs=[msg, chat_history, language, gr.State((model, tokenizer))],
             outputs=[chatbot, chat_history]
         ).then(
             lambda: "",
@@ -674,8 +582,8 @@ def create_gradio_interface(model, tokenizer, calendar_service):
 
         # Initial welcome message
         demo.load(
-            lambda: ([("", "Hello! I'm the online assistant for City Hospital. How can I help you today?")],
-                    [("", "Hello! I'm the online assistant for City Hospital. How can I help you today?")]),
+            lambda: ([("", "Hello! I'm the online assistant for City Hospital. How can I help you today?")], 
+                     [("", "Hello! I'm the online assistant for City Hospital. How can I help you today?")]),
             inputs=None,
             outputs=[chatbot, chat_history]
         )
@@ -684,44 +592,15 @@ def create_gradio_interface(model, tokenizer, calendar_service):
 
 # ===== MAIN EXECUTION =====
 
-def main():
-    global HF_TOKEN
+print("===== Malayalam Hospital Booking Chatbot =====")
+print("Using Llama 3.1-8B-Instruct for Hugging Face Spaces")
 
-    print("===== Malayalam Hospital Booking Chatbot =====")
-    print("Using Llama 3.1-8B-Instruct with Google Calendar integration")
+# Load the Llama model and tokenizer
+model, tokenizer = load_llama_model()
 
-    # Install required packages in Colab
-    try:
-        import IPython
-        print("Installing required packages...")
-        IPython.get_ipython().system('pip install transformers>=4.37.0')
-        IPython.get_ipython().system('pip install accelerate>=0.25.0')
-        IPython.get_ipython().system('pip install bitsandbytes>=0.41.0')
-        IPython.get_ipython().system('pip install sentencepiece>=0.1.99')
-        IPython.get_ipython().system('pip install gradio==3.50.2')
-        IPython.get_ipython().system('pip install google-auth google-auth-oauthlib google-auth-httplib2')
-        IPython.get_ipython().system('pip install google-api-python-client')
-        IPython.get_ipython().system('pip install pytz')
-        print("All packages installed successfully!")
-    except:
-        print("Not running in IPython environment or packages already installed.")
-
-    # Get HF token from user input
-    HF_TOKEN = input("Enter your Hugging Face token with access to meta-llama models: ")
-
-    # Load the Llama model and tokenizer
-    model, tokenizer = load_llama_model()
-
-    if model is None or tokenizer is None:
-        print("Failed to load the model. Please check your Hugging Face token and try again.")
-        return
-
-    # Get calendar service
-    calendar_service = get_calendar_service()
-
-    # Create and launch the Gradio interface
-    demo = create_gradio_interface(model, tokenizer, calendar_service)
-    demo.launch(share=True, debug=True)
-
-if __name__ == "__main__":
-    main()
+if model is None or tokenizer is None:
+    raise ValueError("Failed to load the model. Please check configuration and try again.")
+
+# Create and launch the Gradio interface
+demo = create_gradio_interface(model, tokenizer)
+demo.launch()