Update app.py

app.py

@@ -1,11 +1,15 @@
 import json
+import os
 import gradio as gr
 from huggingface_hub import InferenceClient
 
 # Function to call the Llama 3.1 8B model through Hugging Face API
 def call_llama_model(user_query):
-    # Initialize the inference client
-    client = InferenceClient("meta-llama/Meta-Llama-3.1-8B-Instruct")
+    # Initialize the inference client with access token
+    client = InferenceClient(
+        "meta-llama/Meta-Llama-3.1-8B-Instruct",
+        token=os.environ.get("HF_TOKEN")
+    )
     
     # Define the addition function schema
     function_schema = {
@@ -29,14 +33,18 @@ def call_llama_model(user_query):
     
     # Create the system prompt with function definition
     system_prompt = f"""You have access to the following function:
-    {json.dumps(function_schema, indent=2)}
-    
-    Your task is to extract two numbers from the user's query and call the add_numbers function.
-    Format your response as JSON with the function name and parameters.
-    Only respond with valid JSON containing the function call, nothing else.
-    """
+{json.dumps(function_schema, indent=2)}
+
+When given a query about adding numbers, you must:
+1. Extract the two numbers from the query
+2. Call the add_numbers function with these numbers
+3. Respond ONLY with a valid JSON object in this exact format:
+{{"function_call": {{"name": "add_numbers", "parameters": {{"num1": [first number], "num2": [second number]}}}}}}
+
+DO NOT include any explanation, just the JSON.
+"""
     
-    # Call the model
+    # Call the model with the appropriate format for Llama 3.1
     response = client.text_generation(
         prompt=f"<|system|>\n{system_prompt}\n<|user|>\n{user_query}\n<|assistant|>",
         max_new_tokens=256,
@@ -52,6 +60,9 @@ def process_addition(query):
         # Get model response
         model_response = call_llama_model(query)
         
+        # Create a debug output with the raw response
+        debug_info = f"Raw model response:\n{model_response}\n\n"
+        
         # Try to parse the JSON response
         try:
             # Find the JSON part in the response (it might have additional text)
@@ -61,8 +72,9 @@ def process_addition(query):
             if json_start >= 0 and json_end > json_start:
                 json_str = model_response[json_start:json_end]
                 response_data = json.loads(json_str)
+                debug_info += f"Parsed JSON:\n{json.dumps(response_data, indent=2)}\n\n"
             else:
-                return f"Error: No valid JSON found in response: {model_response}"
+                return f"Error: No valid JSON found in response.\n\n{debug_info}"
             
             # Check if it has a function call
             if "function_call" in response_data:
@@ -70,68 +82,21 @@ def process_addition(query):
                 params = response_data["function_call"]["parameters"]
                 
                 if function_name == "add_numbers":
-                    num1 = params["num1"]
-                    num2 = params["num2"]
+                    # Extract the numbers
+                    num1 = float(params["num1"])
+                    num2 = float(params["num2"])
                     result = num1 + num2
                     
                     # Return a formatted response
                     return f"""
-                    Model parsed your query as:
-                    - First number: {num1}
-                    - Second number: {num2}
-                    
-                    Function called: {function_name}
-                    Result: {result}
-                    """
-                else:
-                    return f"Unknown function: {function_name}"
-            else:
-                return f"No function call found in response: {response_data}"
-                
-        except json.JSONDecodeError as e:
-            return f"Error parsing JSON: {str(e)}\nRaw response: {model_response}"
-            
-    except Exception as e:
-        return f"Error: {str(e)}"
+### Input Processed Successfully
 
-# Create Gradio interface
-def create_demo():
-    with gr.Blocks() as demo:
-        gr.Markdown("# Llama 3.1 Function Calling Demo: Addition")
-        gr.Markdown("Enter a query asking to add two numbers (e.g., 'Add 25 and 17' or 'What's 42 plus 58?')")
-        
-        with gr.Row():
-            with gr.Column():
-                query_input = gr.Textbox(
-                    label="Your Query",
-                    placeholder="Add 25 and 17"
-                )
-                submit_btn = gr.Button("Calculate")
-            
-            with gr.Column():
-                output = gr.Textbox(label="Result")
-        
-        submit_btn.click(
-            fn=process_addition,
-            inputs=query_input,
-            outputs=output
-        )
-        
-        gr.Examples(
-            examples=[
-                "Add 25 and 17",
-                "What is 42 plus 58?",
-                "Can you sum 123 and 456?",
-                "I need to add 7.5 and 2.25",
-                "What's the total of 1000 and 2000?"
-            ],
-            inputs=query_input
-        )
-    
-    return demo
+**Numbers extracted:** {num1} and {num2}
+
+**Function called:** `{function_name}`
 
-# Create and launch the demo
-demo = create_demo()
+**Result:** {result}
 
-if __name__ == "__main__":
-    demo.launch()
+**JSON Function Call:**
+```json
+{json.dumps(response_data, indent=2)}