fixed main.py

main.py

@@ -1,5 +1,3 @@
-# app.py
-
 import os
 from fastapi import FastAPI
 from pydantic import BaseModel
@@ -10,7 +8,6 @@ from sentence_transformers import SentenceTransformer
 from typing import List, Optional
 
 # --- Configuration ---
-# Using environment variables is best practice for production.
 MONGO_URI = os.getenv("MONGO_URI", "mongodb+srv://saisunil22ecs:9m2ajd0GxVn43Fbu@majorproject.g0g1as0.mongodb.net/?retryWrites=true&w=majority&appName=MajorProject")
 DB_NAME = os.getenv("MONGO_DB", "legal_chatbot_db")
 COLLECTION_NAME = os.getenv("MONGO_COLLECTION", "datasets")
@@ -19,7 +16,6 @@ EMBED_MODEL = os.getenv("EMBED_MODEL", "sentence-transformers/all-MiniLM-L6-v2")
 
 # --- Resource Loading ---
 def load_resources():
-    """A single function to load all models and the database connection at startup."""
     try:
         print("🔄 Loading intent classification model...")
         tokenizer = AutoTokenizer.from_pretrained(MODEL_PATH)
@@ -27,7 +23,13 @@ def load_resources():
         print("✅ Intent model loaded.")
 
         print("🔄 Loading embedding model for vector search...")
-        embedding_model = SentenceTransformer(EMBED_MODEL)
+        
+        # --- THIS IS THE FIX ---
+        # We specify a local cache directory where the app has write permissions.
+        cache_dir = "./model_cache"
+        embedding_model = SentenceTransformer(EMBED_MODEL, cache_folder=cache_dir)
+        # ---
+        
         print("✅ Embedding model loaded.")
 
         print("🔄 Connecting to MongoDB Atlas...")
@@ -43,7 +45,6 @@ def load_resources():
         return None, None, None, None
 
 tokenizer, intent_model, embedding_model, collection = load_resources()
-
 # --- FastAPI App ---
 app = FastAPI(title="Legal Aid Chatbot API")