There are three ways to use Averment, depending on how much control you need over what happens after a decision is made.

Use this when you just want to see what Averment returns for a given action. You call decide() and inspect the signal yourself.

This is useful for testing and exploration. It is not recommended for production because you are responsible for reading the signal and acting on it. If you skip that step, nothing is enforced.

Use this when you want Averment to control whether your code runs. You pass your action as a function and Averment decides if it executes.

Averment evaluates the decision and returns one of four outcomes: allow, caution, review, or block. Only allow runs your function. For everything else, you choose how to respond using a strategy.

The default behavior. If the action is not allowed, Averment throws an error. Use this when you want a hard stop on anything that is not explicitly safe.

Use this when you want to ask the user to verify before continuing. If Averment returns review, your confirmation function runs. If it returns true, the action proceeds.

Use this when you want to route actions to a review queue instead of throwing an error. If Averment returns caution, your onCaution function runs so you can handle it gracefully.

Use this when you need full control over what happens for each possible outcome. You provide a function for each action and decide exactly what to do. There are no restrictions on what each handler does.

For example, you could allow the action on allow, queue it for review on caution, trigger an AI feedback loop on review, and route to an operator on block. Each outcome is yours to define.

The block handler is required.

A clear, natural-language description of the action.

Structured data about the action. Only send what's relevant.

Category of the action.