TrustyAI service

The TrustyAI service is a REST service that provides the integration between XAI and metrics algorithms provided by the TrustyAI core and external models.

Although it can be used as a standalone service, it is designed to be used as part of Open Data Hub, deployed in OpenShift/Kubernetes and managed by the TrustyAI operator.

The main functionality of the service is to provide a REST API to calculate metrics and explainability. The service can consume model data (inputs and outputs) for metrics calculation either via a consumer endpoint, or cloud events.

To provide explanations, the service performs inferences to a ModelMesh/KServe-deployed model using gRPC.

External endpoints

TrustyAI service external endpoints are only accessible by authenticated users. The authentication is done oauth-proxy service.

This generally requires getting a user token (or a token from a service account with sufficient privileges) and passing it to the service.

To get the token on OpenShift, you can use the following command:

TOKEN=$(oc whoami -t)

Then, you can pass the token to the service using the Authorization header:

curl -H "Authorization: Bearer $TOKEN" $TRUSTYAI_ROUTE