Update app.py

app.py

@@ -1,3 +1,4 @@
+import os
 import gradio as gr
 from sentence_transformers import SentenceTransformer
 from transformers import AutoTokenizer, AutoModelForCausalLM, pipeline
@@ -5,20 +6,29 @@ import faiss
 import numpy as np
 from langchain.text_splitter import RecursiveCharacterTextSplitter
 import fitz  # PyMuPDF
+from huggingface_hub import login
 
-# Load models
+# Authenticate with Hugging Face to access gated models
+hf_token = os.environ.get("HUGGINGFACE_TOKEN")
+if hf_token is None:
+    raise ValueError("Please set the HUGGINGFACE_TOKEN environment variable")
+login(token=hf_token)
+
+# Load embedding model
 embed_model = SentenceTransformer("BAAI/bge-base-en-v1.5")
 
+# Load LLM model and tokenizer with 4bit quantization
 model_id = "mistralai/Mistral-7B-Instruct-v0.1"
-tokenizer = AutoTokenizer.from_pretrained(model_id)
+tokenizer = AutoTokenizer.from_pretrained(model_id, use_auth_token=hf_token)
 model = AutoModelForCausalLM.from_pretrained(
     model_id,
     device_map="auto",
-    load_in_4bit=True
+    load_in_4bit=True,
+    use_auth_token=hf_token
 )
 llm = pipeline("text-generation", model=model, tokenizer=tokenizer)
 
-# Globals
+# Globals for FAISS index and document texts
 index = None
 doc_texts = []
 
@@ -35,7 +45,7 @@ def extract_text(file):
     else:
         return "❌ Invalid file type."
 
-# File processing
+# File processing: chunk text, create embeddings, build FAISS index
 def process_file(file):
     global index, doc_texts
     text = extract_text(file)
@@ -52,9 +62,10 @@ def process_file(file):
 
     return "✅ File processed successfully. You can now ask questions!"
 
-# Context + LLM response
+# Generate answer using retrieved context and LLM
 def generate_answer(question):
-    if index is None:
+    global index, doc_texts
+    if index is None or len(doc_texts) == 0:
         return "⚠️ Please upload and process a file first."
     
     question_embedding = embed_model.encode([question])