Integration documentation

Two calls: tell us what happened, and ask us before something risky happens.

Setup overview

Create your project.
Install the SDK.
Report events.
Protect sensitive actions.

Install

npm install @puppetwatch/sdk

Report events

puppetwatch.identify(currentUser.id);

puppetwatch.capture("analysis_requested", {
  target_type: "ai_inference",
  target_id: demo.id,
});

identify() associates activity with a pseudonymous account. capture() reports an event and its target metadata.

Protect a sensitive action

const decision = await puppetwatch.guard(currentUser.id, {
  targetType: "analysis_requested",
});

if (decision.action === "block") {
  // deny the request
}

capture() never blocks anything on its own. guard() is what lets you actually stop a bad actor. Call it right before the action happens, and act on what it returns (allow, block, require a CAPTCHA, and more).

Data sent in these examples

Value	Purpose
`currentUser.id`	Pseudonymous account identifier.
`analysis_requested`	Name of the event or sensitive action.
`target_type`	Type of resource involved in the captured event.
`target_id`	Identifier of the target involved in the captured event.
`targetType`	Action type submitted to guard for a decision.

Send pseudonymous IDs and event metadata. No emails or message content required.