Update app.py

app.py

@@ -1,6 +1,3 @@
-# !pip install flask-cors
-# !pip install Flask pyngrok requests cloudinary SpeechRecognition pydub happytransformer transformers torch faiss-cpu sentence-transformers pandas unsloth bitsandbytes webrtcvad
-# !ngrok config add-authtoken 2nFD4jJkAN642UzGI86nDsSC4qs_2cDEGBUFVpbQ5KaDuu4ys
 import os
 import faiss
 import torch
@@ -8,6 +5,7 @@ import pandas as pd
 from sentence_transformers import SentenceTransformer
 from flask import Flask, request, jsonify, render_template
 from flask_cors import CORS
+from pyngrok import ngrok
 import requests
 import cloudinary
 import cloudinary.uploader
@@ -27,51 +25,39 @@ from werkzeug.utils import secure_filename
 from geopy.geocoders import Nominatim
 import pickle
 import numpy as np
-from huggingface_hub import snapshot_download
 
 # Configure logging
 logging.basicConfig(level=logging.INFO)
 
-# Initialize Flask app
-app = Flask(__name__, template_folder="templates")
-CORS(app)
-
 # Load environment variables
-API_KEY = os.getenv("API_KEY", "default_key") 
-CSE_ID = os.getenv("CSE_ID", "default_cse")
-CLOUDINARY_CLOUD_NAME = os.getenv("CLOUDINARY_CLOUD_NAME", "default_cloud")
-CLOUDINARY_API_KEY = os.getenv("CLOUDINARY_API_KEY", "default_key")
-CLOUDINARY_API_SECRET = os.getenv("CLOUDINARY_API_SECRET", "default_secret")
-
-# Define paths for models and data
-MODEL_PATH = os.path.join("models", "new_rag_model", "model_state_dict.pth")
-FAISS_INDEX_PATH = os.path.join("models", "new_rag_model", "property_faiss.index")
-DATASET_PATH = os.path.join("models", "new_rag_model", "property_data.csv")
-MODEL_DIR = os.path.join("models", "llm_model")
+API_KEY = os.getenv("AIzaSyC5FSchUVhKWetUIYPMe92B_1oRqhGplqI")
+CSE_ID = os.getenv("c03c5384c2c5d424b")
+CLOUDINARY_CLOUD_NAME = os.getenv("dn4rackei")
+CLOUDINARY_API_KEY = os.getenv("599266248716888")
+CLOUDINARY_API_SECRET = os.getenv("DRAaasqskCvfAhJhcKB6AKxrD7U")
+
+# Define paths
+load_dir = "./models/new_rag_model/"
+model_path = os.path.join(load_dir, "model_state_dict.pth")
+faiss_index_path = os.path.join(load_dir, "property_faiss.index")
+dataset_path = os.path.join(load_dir, "property_data.csv")
+model_dir = "./models/llm_model"
+# model_dir = "/content/drive/MyDrive/newllmmodel/final_model"
+# model_dir = "/content/drive/MyDrive/real_estate_model/final_model"
+# model_dir = "/content/drive/MyDrive/rag"
 
 # Check device
 device = "cuda" if torch.cuda.is_available() else "cpu"
 print(f"Using device: {device}")
 
-# Initialize conversation context
-conversation_context = {}
-
 # Load SentenceTransformer model
 def load_sentence_transformer():
     print("Loading SentenceTransformer model...")
     try:
-        # Set environment variable to auto-accept trust_remote_code
-        os.environ['TRUST_REMOTE_CODE'] = "1"
-
-        from sentence_transformers import SentenceTransformer
-        model_embedding = SentenceTransformer(
-            'jinaai/jina-embeddings-v3',
-            trust_remote_code=True
-        ).to(device)
+        model_embedding = SentenceTransformer("jinaai/jina-embeddings-v3", trust_remote_code=True).to(device)
 
-        if not os.path.exists(MODEL_PATH):
-            raise FileNotFoundError(f"Model state dict not found at {MODEL_PATH}. Please ensure the file exists.")
-        state_dict = torch.load(MODEL_PATH, map_location=device)
+        # Load and optimize model state dict
+        state_dict = torch.load(model_path, map_location=device)
 
         # Dequantize if needed
         for key, tensor in state_dict.items():
@@ -90,18 +76,14 @@ def load_sentence_transformer():
 # Load FAISS index
 def load_faiss_index():
     print("Loading FAISS index...")
-    if not os.path.exists(FAISS_INDEX_PATH):
-        raise FileNotFoundError(f"FAISS index not found at {FAISS_INDEX_PATH}. Please ensure the file exists.")
-    index = faiss.read_index(FAISS_INDEX_PATH)
+    index = faiss.read_index(faiss_index_path)
     print("FAISS index loaded successfully.")
     return index
 
 # Load dataset
 def load_dataset():
     print("Loading dataset...")
-    if not os.path.exists(DATASET_PATH):
-        raise FileNotFoundError(f"Dataset file not found at {DATASET_PATH}. Please ensure the file exists.")
-    df = pd.read_csv(DATASET_PATH)
+    df = pd.read_csv(dataset_path)
     print("Dataset loaded successfully.")
     return df
 
@@ -112,7 +94,7 @@ class CustomRagRetriever:
         self.model = model
         self.pca = None
         # Load PCA if it exists
-        pca_path = os.path.join(os.path.dirname(MODEL_PATH), "pca_model.pkl")
+        pca_path = os.path.join(os.path.dirname(model_path), "pca_model.pkl")
         if os.path.exists(pca_path):
             with open(pca_path, 'rb') as f:
                 self.pca = pickle.load(f)
@@ -159,11 +141,11 @@ retriever = CustomRagRetriever(index, model_embedding)
 # Load tokenizer and LLM model
 def load_tokenizer_and_model():
     print("Loading tokenizer...")
-    tokenizer = AutoTokenizer.from_pretrained(MODEL_DIR)
+    tokenizer = AutoTokenizer.from_pretrained(model_dir)
     print("Tokenizer loaded successfully.")
 
     print("Loading LLM model...")
-    model_llm = AutoModelForCausalLM.from_pretrained(MODEL_DIR).to(device)
+    model_llm = AutoModelForCausalLM.from_pretrained(model_dir).to(device)
     print("LLM model loaded successfully.")
     return tokenizer, model_llm
 
@@ -381,8 +363,30 @@ def transcribe_with_vad(audio_file):
                 print(f"Could not request results from Google Speech Recognition service; {e}")
     return ""
 
+# Flask app
+app = Flask(__name__, template_folder="sample_data/templates")
+conversation_context = {}
+
+# Configure CORS
+CORS(app, resources={
+    r"/*": {
+        "origins": ["http://localhost:4200", "https://localhost:4200"],
+        "methods": ["GET", "POST", "OPTIONS"],
+        "allow_headers": ["Content-Type", "X-Session-ID"]
+    }
+})
+
+@app.before_request
+def handle_preflight():
+    if request.method == 'OPTIONS':
+        response = app.make_default_options_response()
+        response.headers.add('Access-Control-Allow-Headers', 'Content-Type, X-Session-ID')
+        response.headers.add('Access-Control-Allow-Methods', 'GET, POST, OPTIONS')
+        return response
+
 @app.route('/')
 def index():
+    print("Rendering index page")
     return render_template('index.html')
 
 @app.route('/search', methods=['POST'])
@@ -642,5 +646,9 @@ def set_location():
         return jsonify({"error": f"Error processing location: {str(e)}"}), 500
 
 if __name__ == '__main__':
-    # For Hugging Face Spaces, we need to listen on 0.0.0.0:7860
+    # Remove ngrok configuration
+    # public_url = ngrok.connect(5000)
+    # print(f' * ngrok tunnel "http://127.0.0.1:5000" -> "{public_url}"')
+    
+    # Update to use port 7860 (standard for Spaces)
     app.run(host='0.0.0.0', port=7860)