Update app.py

app.py

@@ -9,19 +9,10 @@ import re  # Import the regular expressions module
 model_name = "Qwen/Qwen2.5-72B-Instruct"
 client = InferenceClient(model_name)
 
-def llm_inference(user_sample):
+def llm_inference(messages):
     eos_token = "<|endoftext|>"
     output = client.chat.completions.create(
-        messages=[
-            {
-                "role": "system",
-                "content": "You are a Python language guide. Write code on the user topic. If the input is code, correct it for mistakes."
-            },
-            {
-                "role": "user",
-                "content": f"Write only python code without any explanation: {user_sample}"
-            },
-        ],
+        messages=messages,
         stream=False,
         temperature=0.7,
         top_p=0.1,
@@ -50,7 +41,12 @@ def is_math_task(user_input):
     Simple heuristic to determine if the user input is a math task.
     This can be enhanced with more sophisticated methods or NLP techniques.
     """
-    math_keywords = ['calculate', 'compute', 'solve', 'integrate', 'differentiate', 'derivative', 'integral', 'factorial', 'sum', 'product']
+    math_keywords = [
+        'calculate', 'compute', 'solve', 'integrate', 'differentiate',
+        'derivative', 'integral', 'factorial', 'sum', 'product',
+        'average', 'mean', 'median', 'mode', 'variance', 'standard deviation',
+        'limit', 'matrix', 'determinant', 'equation', 'expression'
+    ]
     operators = ['+', '-', '*', '/', '^', '**', 'sqrt', 'sin', 'cos', 'tan', 'log', 'exp']
     user_input_lower = user_input.lower()
     return any(keyword in user_input_lower for keyword in math_keywords) or any(op in user_input for op in operators)
@@ -58,12 +54,38 @@ def is_math_task(user_input):
 def chat(user_input, history):
     """
     Handles the chat interaction. If the user input is detected as a math task,
-    it generates Python code to solve it, strips any code tags, executes the code,
-    and returns the result.
+    it performs a two-step process:
+    1. Generates an explanation of how to solve the task.
+    2. Generates Python code based on the explanation and executes it.
     """
     if is_math_task(user_input):
-        # Generate Python code for the math task
-        generated_code = llm_inference(f"Create a Python program to solve the following math problem:\n{user_input}")
+        # Step 1: Generate Explanation
+        explanation_messages = [
+            {
+                "role": "system",
+                "content": "You are a Python language guide. Provide a concise explanation on how to approach the following mathematical task without calculating the answer."
+            },
+            {
+                "role": "user",
+                "content": f"Provide a short explanation on how to solve the following mathematical problem: {user_input}"
+            },
+        ]
+        explanation = llm_inference(explanation_messages)
+        
+        # Step 2: Generate Python Code using Explanation and User Prompt
+        code_prompt = f"Based on the following explanation, write a Python program to solve the mathematical task. Ensure that the program includes a print statement to output the answer.\n\nExplanation: {explanation}\n\nTask: {user_input}"
+        
+        code_messages = [
+            {
+                "role": "system",
+                "content": "You are a Python developer. Write Python code based on the provided explanation and task."
+            },
+            {
+                "role": "user",
+                "content": f"{code_prompt}"
+            },
+        ]
+        generated_code = llm_inference(code_messages)
         
         # Strip code tags using regex
         # This regex removes ```python and ``` or any other markdown code fences
@@ -73,11 +95,25 @@ def chat(user_input, history):
         # Execute the cleaned code
         execution_result = execute_code(cleaned_code)
         
-        # Prepare the responses
-        assistant_response = f"**Generated Python Code:**\n```python\n{cleaned_code}\n```\n\n**Execution Result:**\n```\n{execution_result}\n```"
+        # Prepare the assistant response
+        assistant_response = (
+            f"**Explanation:**\n{explanation}\n\n"
+            f"**Generated Python Code:**\n```python\n{cleaned_code}\n```\n\n"
+            f"**Execution Result:**\n```\n{execution_result}\n```"
+        )
     else:
         # For regular chat messages, use the AI's response
-        assistant_response = llm_inference(user_input)
+        messages = [
+            {
+                "role": "system",
+                "content": "You are a helpful assistant."
+            },
+            {
+                "role": "user",
+                "content": user_input
+            },
+        ]
+        assistant_response = llm_inference(messages)
     
     # Append to chat history
     history.append((user_input, assistant_response))
@@ -87,14 +123,14 @@ with gr.Blocks() as demo:
     gr.Markdown("# 🐍 Python Helper Chatbot")
     with gr.Tab("Chat"):
         chatbot = gr.Chatbot()
-        msg = gr.Textbox(placeholder="Type your message here...")
+        msg = gr.Textbox(placeholder="Type your message here...", label="Your Message")
         msg.submit(chat, inputs=[msg, chatbot], outputs=[chatbot, chatbot])
     
     with gr.Tab("Interpreter"):
         gr.Markdown("### 🖥️ Test Your Code")
-        code_input = gr.Code(language="python")
+        code_input = gr.Code(language="python", label="Python Code Input")
         run_button = gr.Button("Run Code")
-        code_output = gr.Textbox(label="Output")
+        code_output = gr.Textbox(label="Output", lines=10)
         run_button.click(execute_code, inputs=code_input, outputs=code_output)
     
     with gr.Tab("Logs"):