Update app.py

app.py

@@ -14,12 +14,18 @@ logger = logging.getLogger(__name__)
 device = torch.device("cpu")
 logger.info(f"Using device: {device}")
 
-# Load or initialize response cache
-cache_file = "cache.json"
+# Response cache
 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?",
     "hey": "Hi there! Ready to discuss investment goals?",
+    "what is better individual stocks or etfs?": (
+        "Here’s a comparison of individual stocks vs. ETFs:\n"
+        "1. **Individual Stocks**: Offer high potential returns by investing in specific companies (e.g., Apple) but carry higher risk due to lack of diversification. Require active research.\n"
+        "2. **ETFs**: Provide diversification by tracking indices (e.g., VOO for S&P 500), reducing risk but with lower potential returns. Lower fees and less research needed.\n"
+        "3. **Recommendation**: Beginners should start with ETFs for stability; experienced investors may add stocks for growth.\n"
+        "Consult a financial planner for personalized advice."
+    ),
     "hi, give me step-by-step investing advice": (
         "Here’s a step-by-step guide to start investing:\n"
         "1. Open a brokerage account (e.g., Fidelity, Vanguard) if 18 or older.\n"
@@ -109,6 +115,7 @@ response_cache = {
 }
 
 # Load persistent cache
+cache_file = "cache.json"
 try:
     if os.path.exists(cache_file):
         with open(cache_file, 'r') as f:
@@ -143,14 +150,14 @@ prompt_prefix = (
 prefix_tokens = tokenizer(prompt_prefix, return_tensors="pt", truncation=True, max_length=512).to(device)
 
 # Fuzzy matching for cache
-def get_closest_cache_key(message, cache_keys, threshold=0.75):
+def get_closest_cache_key(message, cache_keys, threshold=0.7):
     matches = difflib.get_close_matches(message, cache_keys, n=1, cutoff=threshold)
     return matches[0] if matches else None
 
 # Define chat function
 def chat_with_model(user_input, history=None):
     try:
-        logger.info(f"Processing message: {user_input}")
+        logger.info(f"Processing user input: {user_input}")
         # Normalize and check cache
         cache_key = user_input.lower().strip()
         cache_keys = list(response_cache.keys())
@@ -176,14 +183,23 @@ def chat_with_model(user_input, history=None):
 
         # Construct prompt
         full_prompt = prompt_prefix + user_input + "\nA:"
-        inputs = tokenizer(full_prompt, return_tensors="pt", truncation=True, max_length=512).to(device)
+        try:
+            inputs = tokenizer(full_prompt, return_tensors="pt", truncation=True, max_length=512).to(device)
+        except Exception as e:
+            logger.error(f"Error tokenizing input: {e}")
+            response = f"Error: Failed to process input: {str(e)}"
+            logger.info(f"Chatbot response: {response}")
+            history = history or []
+            history.append({"role": "user", "content": user_input})
+            history.append({"role": "assistant", "content": response})
+            return response, history
         
         # Generate response
-        with torch.cpu.amp.autocast(), torch.inference_mode():
+        with torch.inference_mode():
             outputs = model.generate(
                 **inputs,
-                max_new_tokens=40,
-                min_length=15,
+                max_new_tokens=30,
+                min_length=10,
                 do_sample=True,
                 temperature=0.7,
                 top_p=0.9,
@@ -198,7 +214,7 @@ def chat_with_model(user_input, history=None):
         response_cache[cache_key] = response
         try:
             with open(cache_file, 'w') as f:
-                json.dump(response_cache, f)
+                json.dump(response_cache, f, indent=2)
             logger.info("Updated cache.json")
         except Exception as e:
             logger.warning(f"Failed to update cache.json: {e}")