Skip to main content
This page explains the core concepts of Actions and Triggers in the platform, and how you can use them to build powerful, responsive AI systems.

​
Actions

Actions enable your models to dynamically call external functions or APIs, making them more useful for real-world tasks. By augmenting their capabilities with up-to-date information, automation, and integration with other systems, Actions allow AI agents to interact with the world beyond static data.

​
Why Use Actions?

  • Access real-time data: Fetch the latest information from external sources (e.g., weather, stock prices).
  • Automate workflows: Perform tasks like sending emails, updating databases, or triggering cloud functions.
  • Integrate with other systems: Connect your AI with tools like Slack, GitHub, or custom APIs.

​
Example: Simple Calculator Action

Below is an example of a simple calculator action that allows an LLM to perform mathematical calculations: 
def evaluate(self, expression: str) -> int:
    """Evaluate a mathematical expression and return the result as an integer.

    Args:
        expression (str): A string containing a valid mathematical expression.
                          Supports basic arithmetic operations (+, -, *, /, **).
                          Example: "2 + 2", "3 * 4", "10 / 2"

    Returns:
        int: The result of evaluating the expression.
    """
    return eval(expression)
Note: This action can be called by the model to compute results on demand, enabling dynamic and context-aware responses.

​
Triggers

Triggers act as a notification system for your AI applications, enabling your agents to respond dynamically to external events occurring within your integrations.

​
How Triggers Work

  • Event monitoring: Triggers watch for specific events in applications like GitHub (e.g., new pull request), Gmail (e.g., new email), or any other integrated service.
  • Agent notification: When the event occurs, the trigger notifies your agent, which can then take appropriate action (e.g., respond, process, escalate).

​
Example Use Cases

  • GitHub: Notify an agent when a new issue is created, so it can triage or label the issue automatically.
  • Gmail: Alert an agent when an important email arrives, prompting it to draft a reply or update a CRM.
  • Custom integrations: React to webhook events from any service, enabling endless automation possibilities.

​
Best Practices

  • Use actions to extend your agent’s capabilities with real-time data and automation.
  • Design triggers to respond to meaningful events and avoid unnecessary notifications.
  • Document each action and trigger clearly, including expected input and output.
  • Test your actions and triggers to ensure reliability and security.
  • Regularly review and update your actions and triggers to support new workflows and integrations.
By combining Actions and Triggers, you can build powerful, responsive AI systems that both act on demand and react to the world around them.