Update app.py

app.py

@@ -12,7 +12,7 @@ logger = logging.getLogger(__name__)
 device = torch.device("cpu")
 logger.info(f"Using device: {device}")
 
-# Response cache with short prompts
+# Response cache with consolidated and new entries
 response_cache = {
     "hi": "Hello! I'm your financial advisor. How can I help with investing?",
     "hello": "Hello! I'm your financial advisor. How can I help with investing?",
@@ -47,6 +47,15 @@ response_cache = {
     "what is dollar-cost averaging?": (
         "Dollar-cost averaging is investing a fixed amount regularly (e.g., $100 monthly) in assets like ETFs, "
         "reducing risk by spreading purchases over time."
+    ),
+    "give me few investing idea": (
+        "Here are some investing ideas:\n"
+        "1. Open a brokerage account if you are 18 or older (e.g., Fidelity, Vanguard).\n"
+        "2. Deposit an initial amount you can afford (e.g., $100).\n"
+        "3. Buy a researched stock, ETF (e.g., VOO), or index fund.\n"
+        "4. Check your investments regularly and enable dividend reinvesting if desired.\n"
+        "5. Use dollar-cost averaging to buy the same investment regularly (e.g., monthly).\n"
+        "Consult a financial planner for personalized advice."
     )
 }
 
@@ -64,8 +73,15 @@ except Exception as e:
     logger.error(f"Error loading model/tokenizer: {e}")
     raise
 
-# Pre-tokenize minimal prompt prefix
-prompt_prefix = "You are a financial advisor. Provide concise advice. If no specific recommendations, suggest alternatives.\nQ: "
+# Pre-tokenize prompt prefix with few-shot example
+prompt_prefix = (
+    "You are a financial advisor. Provide concise, actionable advice in a numbered list when asked for ideas or steps. "
+    "If no specific recommendations, suggest alternatives. Avoid vague statements.\n\n"
+    "Example:\n"
+    "Q: Give me some investing tips\n"
+    "A: 1. Open a brokerage account (e.g., Fidelity).\n2. Start with $100 in an ETF like VOO.\n3. Research investments on Yahoo Finance.\n4. Invest regularly using dollar-cost averaging.\n\n"
+    "Q: "
+)
 prefix_tokens = tokenizer(prompt_prefix, return_tensors="pt", truncation=True, max_length=512).to(device)
 
 # Define chat function
@@ -91,8 +107,8 @@ def chat_with_model(message, history=None):
         with torch.no_grad():
             outputs = model.generate(
                 **inputs,
-                max_new_tokens=15,
-                do_sample=False,
+                max_new_tokens=30,  # Increased for detailed lists
+                do_sample=False,  # Greedy decoding for speed
                 pad_token_id=tokenizer.eos_token_id
             )
         response = tokenizer.decode(outputs[0], skip_special_tokens=True)
@@ -107,9 +123,10 @@ logger.info("Initializing Gradio interface")
 interface = gr.ChatInterface(
     fn=chat_with_model,
     title="Financial Advisor Chatbot (OPT-350m)",
-    description="Ask about investing! Powered by Meta AI's OPT-350m. Fast, direct answers.",
+    description="Ask about investing! Powered by Meta AI's OPT-350m. Fast, detailed answers.",
     examples=[
         "Hi",
+        "Give me few investing idea",
         "Hi, pretend you are a financial advisor. Now tell me how can I start investing in stock market?",
         "Do you have a list of companies you recommend?",
         "What's the difference between stocks and bonds?",