minor update the code

app.py

@@ -1,6 +1,37 @@
 import gradio as gr
 from infer import run_search, question_list
 
+import subprocess
+import time
+import atexit
+from urllib.parse import urlparse
+
+# ... (keep all your other imports like transformers, torch, requests, re, gr) ...
+
+# --- NEW: Server Launch Block ---
+# Insert this block *before* you load the model
+# -----------------------------------------------------------------
+print("Attempting to start retrieval server...")
+
+# Start the server as a background process
+# subprocess.Popen does not block, unlike os.system
+try:
+    server_process = subprocess.Popen(["bash", "retrieval_launch.sh"])
+    print(f"Server process started with PID: {server_process.pid}")
+
+    # Register a function to kill the server when app.py exits
+    def cleanup():
+        print("Shutting down retrieval server...")
+        server_process.terminate()
+        server_process.wait()
+        print("Server process terminated.")
+    
+    atexit.register(cleanup)
+
+except Exception as e:
+    print(f"Failed to start retrieval_launch.sh: {e}")
+    print("WARNING: The retrieval server may not be running.")
+
 def gradio_answer(question: str) -> str:
     print(f"\nReceived question for Gradio: {question}")
     try:

setup.sh

@@ -2,7 +2,7 @@
 
 ```bash
 bash install_cuda.sh
-conda create -n faiss_env python=3.10 && conda activate faiss_env
+# conda create -n faiss_env python=3.10 && conda activate faiss_env
 bash install_env.sh
 
 bash download_corpus.sh
@@ -11,5 +11,5 @@ python download_model.py
 bash retrieval_launch.sh
 # sanity check
 python infer.py
-python demo.py
+GRADIO_SERVER_PORT=7890 python demo.py
 ```