Update app.py

app.py

@@ -1,7 +1,7 @@
 import yaml
 import re
-from smolagents import CodeAgent, HfApiModel
-from tools.final_answer import FinalAnswerTool
+from smolagents import HfApiModel  # Removed CodeAgent import.
+from tools.final_answer import FinalAnswerTool  # Retained in case you need it.
 from Gradio_UI import GradioUI
 
 # Updated system prompt: Only output the final, direct advice in plain text.
@@ -28,14 +28,11 @@ def remove_code_snippets(text):
     text = re.sub(r"\[Step \d+: Duration .*", "", text)
     # Remove lines that mention code snippet instructions.
     text = re.sub(r"Make sure to include code with the correct pattern.*", "", text)
-    # Finally, remove any remaining lines that seem to be debug logs.
+    # Remove any remaining lines that seem to be debug logs.
     lines = text.splitlines()
     cleaned_lines = [line for line in lines if not re.search(r"Step \d+|Duration", line)]
     return "\n".join(cleaned_lines).strip()
 
-# Use only the final_answer tool.
-final_answer = FinalAnswerTool()
-
 # Set up the model with a reduced token limit.
 model = HfApiModel(
     max_tokens=1024,
@@ -44,45 +41,24 @@ model = HfApiModel(
     custom_role_conversions=None,
 )
 
-# Load prompt templates from YAML.
+# Load prompt templates if needed (unused here but kept for compatibility).
 with open("prompts.yaml", 'r') as stream:
     prompt_templates = yaml.safe_load(stream)
 
-# Ensure the final_answer key exists in prompt_templates to prevent KeyError.
-if "final_answer" not in prompt_templates:
-    prompt_templates["final_answer"] = {"pre_messages": "", "post_messages": ""}
-
-# Initialize CodeAgent with a low verbosity level to reduce extra debug output.
-agent = CodeAgent(
-    model=model,
-    tools=[final_answer],
-    max_steps=1,
-    verbosity_level=0,
-    grammar=None,
-    planning_interval=None,
-    name="Hypertension Prevention Advisor",
-    description=system_prompt,
-    prompt_templates=prompt_templates
-)
-
 def run_agent(user_input):
     """
-    Runs the agent, then removes any internal chain-of-thought, step logs, and code snippets
-    before returning the final plain-text answer.
+    Constructs a prompt from the system prompt and the user input,
+    then directly calls the model to generate an answer.
+    After generation, it removes any chain-of-thought, debug logs, or code snippets.
     """
-    raw_response = agent.run(user_input)
-    print("Raw Agent Response:", raw_response)  # Debugging output (optional)
+    prompt = f"{system_prompt}\nUser Input: {user_input}\nFinal Answer:"
+    try:
+        # Try using the model's run method if available.
+        raw_response = model.run(prompt)
+    except AttributeError:
+        # Otherwise, assume the model is callable.
+        raw_response = model(prompt)
     
-    if not raw_response.strip():
-        return "I'm sorry, but I couldn't generate a response. Please try again."
-    if "final_answer" not in raw_response.lower():
-        return "Error: The response did not use the `final_answer` tool. Please try again."
+    print("Raw Model Response:", raw_response)  # Debug output (optional)
     
-    clean_response = remove_code_snippets(raw_response)
-    words = clean_response.split()
-    if len(set(words)) < 5:
-        return "I'm unable to generate a meaningful response. Please refine your query."
-    return clean_response
-
-# Launch the Gradio UI.
-GradioUI(agent).launch()
+    if not raw_respon