Update app.py

app.py

@@ -14,7 +14,8 @@ logger = logging.getLogger(__name__)
 device = torch.device("cpu")
 logger.info(f"Using device: {device}")
 
-# Response cache
+# Load or initialize response cache
+cache_file = "cache.json"
 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?",
@@ -95,9 +96,27 @@ response_cache = {
         "4. Use dollar-cost averaging for regular investments.\n"
         "5. Monitor and diversify your portfolio.\n"
         "Consult a financial planner."
+    ),
+    "steps to invest": (
+        "Here are steps to invest:\n"
+        "1. Educate yourself using Investopedia.\n"
+        "2. Open a brokerage account (e.g., Fidelity).\n"
+        "3. Deposit an initial $100 after savings.\n"
+        "4. Buy an ETF like VOO after research.\n"
+        "5. Use dollar-cost averaging monthly.\n"
+        "Consult a financial planner."
     )
 }
 
+# Load persistent cache
+try:
+    if os.path.exists(cache_file):
+        with open(cache_file, 'r') as f:
+            response_cache.update(json.load(f))
+        logger.info("Loaded persistent cache from cache.json")
+except Exception as e:
+    logger.warning(f"Failed to load cache.json: {e}")
+
 # Load model and tokenizer
 model_name = "distilgpt2"
 try:
@@ -124,7 +143,7 @@ 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=0.8):
+def get_closest_cache_key(message, cache_keys, threshold=0.75):
     matches = difflib.get_close_matches(message, cache_keys, n=1, cutoff=threshold)
     return matches[0] if matches else None
 
@@ -138,12 +157,22 @@ def chat_with_model(user_input, history=None):
         closest_key = cache_key if cache_key in response_cache else get_closest_cache_key(cache_key, cache_keys)
         if closest_key:
             logger.info(f"Cache hit for: {closest_key}")
-            return response_cache[closest_key], history
+            response = response_cache[closest_key]
+            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
         
         # Skip model for short prompts
         if len(user_input.strip()) <= 5:
             logger.info("Short prompt, returning default response")
-            return "Hello! I'm your financial advisor. Ask about investing!", history
+            response = "Hello! I'm your financial advisor. Ask about investing!"
+            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
 
         # Construct prompt
         full_prompt = prompt_prefix + user_input + "\nA:"
@@ -153,7 +182,7 @@ def chat_with_model(user_input, history=None):
         with torch.cpu.amp.autocast(), torch.inference_mode():
             outputs = model.generate(
                 **inputs,
-                max_new_tokens=50,
+                max_new_tokens=40,
                 min_length=15,
                 do_sample=True,
                 temperature=0.7,
@@ -162,37 +191,55 @@ def chat_with_model(user_input, history=None):
                 pad_token_id=tokenizer.eos_token_id
             )
         response = tokenizer.decode(outputs[0], skip_special_tokens=True)
-        logger.info("Generated response")
-        torch.cuda.empty_cache()  # Clear memory
         response = response[len(full_prompt):].strip() if response.startswith(full_prompt) else response
+        logger.info(f"Chatbot response: {response}")
+        
+        # Update cache and save to file
+        response_cache[cache_key] = response
+        try:
+            with open(cache_file, 'w') as f:
+                json.dump(response_cache, f)
+            logger.info("Updated cache.json")
+        except Exception as e:
+            logger.warning(f"Failed to update cache.json: {e}")
         
         # Update history
         history = history or []
         history.append({"role": "user", "content": user_input})
         history.append({"role": "assistant", "content": response})
+        torch.cuda.empty_cache()  # Clear memory
         return response, history
     except Exception as e:
         logger.error(f"Error generating response: {e}")
-        return f"Error: {str(e)}", history
+        response = f"Error: {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
 
 # Create Gradio interface
 logger.info("Initializing Gradio interface")
-with gr.Blocks() as interface:
-    chatbot = gr.Chatbot(type="messages")
-    msg = gr.Textbox(label="Your message")
-    submit = gr.Button("Send")
-    clear = gr.Button("Clear")
+try:
+    with gr.Blocks() as interface:
+        chatbot = gr.Chatbot(type="messages")
+        msg = gr.Textbox(label="Your message")
+        submit = gr.Button("Send")
+        clear = gr.Button("Clear")
 
-    def submit_message(user_input, history):
-        response, updated_history = chat_with_model(user_input, history)
-        return response, updated_history
+        def submit_message(user_input, history):
+            response, updated_history = chat_with_model(user_input, history)
+            return response, updated_history
 
-    submit.click(
-        fn=submit_message,
-        inputs=[msg, chatbot],
-        outputs=[msg, chatbot]
-    )
-    clear.click(lambda: None, None, chatbot, queue=False)
+        submit.click(
+            fn=submit_message,
+            inputs=[msg, chatbot],
+            outputs=[msg, chatbot]
+        )
+        clear.click(lambda: None, None, chatbot)
+except Exception as e:
+    logger.error(f"Error initializing Gradio interface: {e}")
+    raise
 
 # Launch interface (conditional for Spaces)
 if __name__ == "__main__" and not os.getenv("HF_SPACE"):