📘

Custom Signals Are Not Supported Yet!

Cove doesn't yet support custom signals. This documentation isn't quite finalized, so please don't rely on it.

Do you already have custom Signals that you use - or would like to start using - in your Trust & Safety workflows? Maybe you have custom machine learning models, user attributes or inferences, or item-level metrics that you're not already sending to Cove with each Item?

If you'd like to use any custom Signals to help you detect harmful Items, you can easily do so. All you have to do is:

1. Expose that Signal to Cove through a public-facing REST API. Cove should be able to send GET requests to that API with relevant data about a piece of content, and receive the Signal's computed result as a response.
2. Add your new Signal in the Signals tab in your Cove dashboard.

Creating a Custom Signal REST API

Let's consider an example: you have a custom machine learning model that computes a user's trustworthiness score. You'd like to use that score in your Cove Rules to help determine whether any given Item is harmful, but you don't want to include the user's trustworthiness score by default every time you send us an Item. Instead, you only want to check that score after you've determined that a few other conditions are true. In order to integrate that custom machine learning model into your Cove rules, you need to expose it to Cove through a Signal REST API.

You must create a REST API endpoint that can accept incoming GET requests from the public internet (i.e. requests from outside your VPC). If you'd like to gate this endpoint with some form of key-based authentication (which is highly recommended), you can easily provide Cove with the means to authenticate.