Creating an Agent With Tools

AI Agents with Hugging Face smolagents

Adel Nehme

VP of AI Curriculum, DataCamp

Creating a Code Agent (No Tools)

from smolagents import CodeAgent, InferenceClientModel

agent = CodeAgent(
    tools=[],
    model=InferenceClientModel()
)

agent.run("Calculate the average of the list [23, 45, 67, 89]")

Executing parsed code:
  numbers = [23, 45, 67, 89]                                                                               
  average = sum(numbers) / len(numbers)                                                                            
  final_answer(average)                                                                                             
Final answer: 56.0
[Step 1: Duration 4.14 seconds| Input tokens: 1,900 | Output tokens: 109]
56.0

Why Use Tools with Code Agents?

The agent we defined can already solve many tasks using:

An LLM model
Python code

But it may also need access to external information:

Example: live web data

That's where tools come in!

Adding a Web Search Tool

from smolagents import CodeAgent, InferenceClientModel, WebSearchTool

agent = CodeAgent(
    model=InferenceClientModel(),
    tools=[WebSearchTool()]
)

Code Agent With Web Search Tool Output

agent.run("What's the tallest building in the world right now?")

Executing parsed code:
  tallest_building_info = web_search("tallest building in the world 2023")                                
  print(tallest_building_info)                                                                                

# Search results omitted for brevity...

Executing parsed code: 
  final_answer("Burj Khalifa, Dubai, 828 meters")                                                              
Final answer: Burj Khalifa, Dubai, 828 meters
[Step 2: Duration 2.97 seconds| Input tokens: 5,078 | Output tokens: 153]
Burj Khalifa, Dubai, 828 meters

Built-in Tools (by Category)

Category	Tools
Information Retrieval	`ApiWebSearchTool`, `DuckDuckGoSearchTool`, `GoogleSearchTool`, `WebSearchTool`, `WikipediaSearchTool`
Web Interaction	`VisitWebpageTool`
Code Execution	`PythonInterpreterTool`
User Interaction	`UserInputTool`
Speech Processing	`SpeechToTextTool`
Workflow Control	`FinalAnswerTool`

¹ https://huggingface.co/docs/smolagents/main/en/reference/default_tools

Tools From the Hugging Face Hub

Using Community Tools from Hugging Face

from smolagents import load_tool

# Load remote tool from Hugging Face
model_downloads_tool = load_tool(
    repo_id="example-repo/hf-model-downloads",
    trust_remote_code=True
)

# Create agent with remote + built-in tools
agent = CodeAgent(
    tools=[model_downloads_tool, WebSearchTool()],
    model=InferenceClientModel()
)

agent.run("Find the most downloaded image classification model on Hugging Face")

google/vit-base-patch16-224-in21k

Let's practice!

AI Agents with Hugging Face smolagents