Set Up Appian AI for Appian on Kubernetes

Appian AI features are available for self-managed environments that run on Appian on Kubernetes. This page explains which AI features you can run on Appian on Kubernetes, which deployment path each feature uses, and what you must set up before you can use each feature.

Appian on Kubernetes provides two deployment paths for AI features in a self-managed environment:

• Default deployments: Run a set of AI features on any supported Appian on Kubernetes platform that is not hosted on Amazon Web Services (AWS). Features route model calls to your own large language model (LLM) through a configured bring-your-own (BYO) AI Provider. This deployment supports offline environments. For requirements and setup, see Default deployments.
• AWS-based deployments: The AWS-based path runs AI features that require AWS infrastructure, including Amazon Elastic Kubernetes Service (EKS) and other AWS services. For requirements and setup, see AWS-based path.

Features and deployments

The following table lists the AI features available in a self-managed environment on Appian on Kubernetes, the deployment path each feature uses, and whether the feature requires Amazon Web Services (AWS):

AI feature	AWS required
Composer	No, but AWS-based deployments have access to full Composer functionality.
Generative AI skills that process text or visual elements and Execute Generative AI Skill smart service	No
a!genAiModels() function	No
DocCenter	No*
OCR-dependent document extraction and Advanced IDP Tools AI skills	Yes
Trainable AI skills: document classification, email classification, and structured document extraction	Yes
Smart Search	Yes
Documents Chat	Yes
Enterprise Copilot	Yes

^* DocCenter is supported with vision models. Features that depend on optical character recognition (OCR), such as bounding boxes and advanced layout, require Amazon Web Services (AWS).

Default deployments

Run AI features on any supported Appian on Kubernetes platform without Amazon Web Services (AWS) services. Route model calls to your own large language model (LLM) through a configured bring-your-own (BYO) AI Provider, and supports offline environments.

Supported platforms

AI features are supported on every platform that Appian on Kubernetes supports:

• Azure AKS
• Amazon EKS
• Red Hat OpenShift
• Google GKE
• Bare metal Kubernetes
• Offline environments with no internet connectivity

Prerequisites

To run the AI features, your environment must meet the following requirements:

• Appian: Appian on Kubernetes at Appian 26.9 or later.
• Appian Operator: Most recently released version.
• Kubernetes: v1.29 or higher.
• Bring-your-own AI Provider: A configured BYO AI Provider that routes model calls to your own large language model (LLM). Supported providers are Custom Endpoints, Azure OpenAI, and Google Cloud. For configuration details, see AI Services.
• Database: A PostgreSQL instance reachable from the cluster. The AI services store data in PostgreSQL and can share the same PostgreSQL instance as your Appian on Kubernetes site database, with each service using its own database, schema, and users. Appian has tested the AI services against a shared instance; a dedicated per-service instance has not been tested.

Set up

To get started using AI features, install Appian on Kubernetes if you have not already installed it, then complete the steps in the Set Up AI for Appian on Kubernetes setup guide.

The setup guide includes:

• Prerequisites, supported platforms, and the AI services that run without AWS.
• Guidance for mirroring the service images into your own registry and creating the required namespaces.
• Steps for creating the PostgreSQL databases and users and the Kubernetes secrets that the services reference.
• Steps for installing the AI services operator and deploying the AIStack custom resource (CR).
• Steps for exposing AI Skill Designer, configuring custom properties, connecting your own large language model (LLM) through a bring-your-own (BYO) AI Provider, and validating the AI features.

Visit Appian Community to access the full setup guide: Set Up AI for Appian on Kubernetes.

AWS-based deployments

The AWS-based path requires Amazon Web Services (AWS) infrastructure. The AWS-based path applies to the AWS-based features listed in Features and deployment paths.

Prerequisites for AWS-based deployments

To run the AWS-based AI features, your environment must meet the following requirements:

• Appian: Customers must use Appian on Kubernetes to access AI features in a self-managed Appian environment. The Appian platform version must be 25.3 or higher.
• Appian Operator: Most recently released version.
• Kubernetes: v1.29 or higher.
• AWS infrastructure: An Amazon Elastic Kubernetes Service (EKS) cluster and other AWS services.
• KEDA: Experience with KEDA.
• Knative: Experience with Knative infrastructure.
• Helm: Experience with Helm.

Set up the AWS-based path

To get started using the AWS-based AI features, install Appian on Kubernetes if you have not already installed it, then complete the steps in the Setup AI for Appian on Kubernetes Using AWS infrastructure setup guide.

The AWS setup guide includes:

• Comprehensive software, infrastructure, and tool prerequisites.
• Important disclaimers and best practices. The sample Terraform files are for demonstration, and customers are responsible for their environments, including data backups.
• Definitions of core concepts, such as IAM Roles for Service Accounts (IRSA) and chained role assumption.
• Guidance for configuring kubectl, setting up resources with example Terraform files, configuring IAM permissions for pods, and deploying Helm charts.
• A section for testing the connections to confirm that users can access multiple AI features.

Download the setup guide

To download the full installation guide, visit Appian Forum:

1. From the SUPPORT tab, click DOWNLOADS.
2. Click PLATFORM > LOW-CODE PLATFORM.
3. Select your Appian version.
4. Under Downloads, you'll find Setup AI for Appian on Kubernetes Using AWS infrastructure and a Terraform sample.

Disable Anthropic models

If your organization must stop using Anthropic models to meet regulatory or organizational requirements, then you can disable Anthropic models at the site level. Generative AI features then use another model provider instead.

To disable Anthropic models, you set a custom property in the Appian custom resource (CR) configuration file and prepare your environment so that generative AI features keep working. For the property and the full steps, see the Disable Anthropic models section of Post-Install Configurations.

Feedback