Update app.py

app.py

@@ -6,14 +6,14 @@ from sentence_transformers import SentenceTransformer
 from huggingface_hub import InferenceClient
 
 # Load embedding model
-model = SentenceTransformer('all-MiniLM-L6-v2')
+model = SentenceTransformer('all-MiniLM-L6-v2")
 
 # File paths
 TEXT_FILE = "combined_text_documents.txt"
 EMBEDDINGS_FILE = "policy_embeddings.npy"
 INDEX_FILE = "faiss_index.bin"
 
-# Load policy text from the file
+# Load policy text from file
 if os.path.exists(TEXT_FILE):
     with open(TEXT_FILE, "r", encoding="utf-8") as f:
         POLICY_TEXT = f.read()
@@ -22,7 +22,7 @@ else:
     print("❌ ERROR: combined_text_documents.txt not found! Ensure it's uploaded.")
     POLICY_TEXT = ""
 
-# Split text into chunks
+# Split text into chunks for FAISS indexing
 chunk_size = 500
 chunks = [POLICY_TEXT[i:i+chunk_size] for i in range(0, len(POLICY_TEXT), chunk_size)] if POLICY_TEXT else []
 
@@ -79,22 +79,26 @@ def respond(
         if val[1]:
             messages.append({"role": "assistant", "content": val[1]})
 
-    # 🔹 Retrieve policy info related to the query
+    # 🔹 Retrieve relevant policy info from FAISS
     policy_context = search_policy(message)
 
     if policy_context:
-        # 🔹 Force the LLM to use retrieved policy text
-        system_context = f"""
-        You are an expert in Colorado public assistance policies. 
-        Below is relevant information retrieved from official policy documents:
+        # 🔹 Display retrieved context in chat
+        messages.append({"role": "assistant", "content": f"📄 **Relevant Policy Context:**\n\n{policy_context}"})
+
+        # 🔹 Force the LLM to use the retrieved policy text
+        user_query_with_context = f"""
+        The following is the most relevant policy information retrieved from the official Colorado public assistance policies:
 
         {policy_context}
 
-        Based on this information, answer the user's question accurately and concisely.
+        Based on this information, answer the following question:
+        {message}
         """
-        messages.append({"role": "system", "content": system_context})
-
-    messages.append({"role": "user", "content": message})
+        messages.append({"role": "user", "content": user_query_with_context})
+    else:
+        # If no relevant policy info is found, use the original message
+        messages.append({"role": "user", "content": message})
 
     response = ""
     for message in client.chat_completion(
@@ -108,7 +112,6 @@ def respond(
         response += token
         yield response
 
-
 # 🔹 Gradio Chat Interface
 demo = gr.ChatInterface(
     respond,