Introduction

TrustyAI

Alauda Build of TrustyAI is based on the TrustyAI Kubernetes operator. The operator simplifies the deployment and management of TrustyAI components on Kubernetes, including model explainability, fairness monitoring, LLM evaluation, and AI guardrails.

Main components and capabilities include:

TrustyAI Service: A service that deploys alongside KServe models and collects inference data. It enables model explainability, fairness monitoring, and drift tracking.
LM-Eval: A job-based architecture for deploying and managing LLM evaluations, based on EleutherAI's lm-evaluation-harness. The operator provides the LMEvalJob CRD for running and managing evaluation tasks. See Evaluate LLM for details.
AI Guardrails (FMS-Guardrails): A modular framework for guardrailing LLMs. The operator provides the GuardrailsOrchestrator CRD for orchestrating guardrail policies and components. See AI Guardrails for LLM safety for details.

For installation on the platform, see Install TrustyAI.

For usage of LMEvalJob and GuardrailsOrchestrator after installation, see the sections linked above.

Documentation

TrustyAI upstream documentation and the operator repository:

TrustyAI Operator (upstream): https://github.com/trustyai-explainability/trustyai-service-operator — Overview, installation, and usage of the Kubernetes operator.
TrustyAI Documentation: https://trustyai.org/ — TrustyAI project documentation.

#Introduction

#TOC

#TrustyAI

#Documentation

Introduction

TOC

TrustyAI

Documentation