Use a local transformers model instead of InfererenceReferenceModel

app.py

@@ -1,14 +1,15 @@
 import os
 import gradio as gr
 from huggingface_hub import login
-from smolagents import DuckDuckGoSearchTool, ToolCallingAgent, InferenceClientModel
+from smolagents import DuckDuckGoSearchTool, ToolCallingAgent, InferenceClientModel, TransformersModel
 
 login(os.environ.get('HF_TOKEN'))
 
 def create_agent():
+    model_id = "meta-llama/Llama-3.2-3B-Instruct"
     tools = [DuckDuckGoSearchTool()]
-    model = InferenceClientModel(tool_choice="auto")
-    return ToolCallingAgent(tools=tools, model=model)
+    model = TransformersModel(model_id=model_id)
+    return ToolCallingAgent(tools=tools,  model=model, add_base_tools=True)