smolagents is a lightweight Python framework developed by Hugging Face for building AI agents. An AI agent can use a Large Language Model (LLM) to understand a task, choose actions, use tools, and work toward a result.

For example, a plant shop chatbot could use an AI agent to look up plant prices and recommend suitable plants to customers.

1. Why use smolagents?

smolagents offers several advantages:

• Simple: Easy to learn and get started.
• Flexible: Supports different LLM providers.
• Code-first: Agents can write Python code to perform actions.
• Tool integration: Agents can use Python functions and external tools.

It is a good choice when you want to experiment with AI agents without building a complex system from scratch.

2. CodeAgent vs. ToolCallingAgent

smolagents supports two main agent types.

CodeAgent generates Python code to perform actions. For example, it can write a loop to retrieve prices from multiple countries and find the cheapest option.

ToolCallingAgent generates structured tool calls, typically using JSON, to request specific actions. For example, it can call a get_plant_price tool to retrieve a plant’s price.

The main difference is how they represent actions: CodeAgent uses Python code, while ToolCallingAgent uses structured tool calls. Both can use tools — they differ in how actions are expressed.

3. Connecting an LLM

smolagents supports different ways to connect language models, including:

• TransformersModel for models running through Transformers.
• InferenceClientModel for Hugging Face inference services.
• LiteLLMModel for connecting to multiple model providers.
• OpenAIServerModel for services with an OpenAI-compatible API.

This flexibility allows developers to choose a model based on their requirements and budget.

4. Creating Tools

Tools allow an agent to interact with external systems or perform specific tasks. In smolagents, you can define a tool using the @tool decorator. For example:

from smolagents import tool

@tool
def get_plant_price(plant_name: str) -> str:
    """Get the price of a plant."""
    return f"The price of {plant_name} is 500000 VND."

An agent can use this function when it needs to retrieve a plant’s price. In a real application, the function could call an API or query a database instead.

5. When Should You Use smolagents?

smolagents is useful when you want to:

• Build a simple AI agent quickly.
• Experiment with tool calling and code execution.
• Connect an LLM to Python functions.
• Automate tasks involving multiple steps.

However, applications with complex workflows may require additional architecture and orchestration.