On-Premises: Setting up Automation.AI for AI-Augmented Workflow Creation

This page describes the specific configuration use case for setting up Automation.AI to enable the AI-Augmented Workflow Creation feature. For more information, see AI-Augmented Workflow Creation with the Automation Assistant.

Important! The configuration on this page is currently supported only for manual, on-premises installations (using standalone Docker containers or a VM). If you are deploying the Automic Automation Kubernetes Edition, see AAKE: Setting up Automation.AI for AI-Augmented Workflow Creation.

Use Case Framework

The AI-Augmented Workflow Creation feature exclusively supports Google Vertex AI (Gemini) as its Large Language Model (LLM) provider. Other LLM providers are not supported for this specific use case.

Implementing this capability requires modifying the application.properties file of your Automation.AI component to enable a dual-database architecture and activate the specialized agentic Spring profile. For more information, see Profiles - Spring Boot.

Architecture Specifics

While the general architecture supports multiple LLM endpoints and standard database layouts (as illustrated in the architecture overview, see Architecture of the Generative AI Capabilities), this specific use case is constrained to Google Vertex AI (Gemini) and requires the following infrastructure modifications:

  • Dual-Database Isolation: The application requires two distinct logical database destinations—a main database for standard application data and a dedicated vector database for embeddings.

  • Vertex AI Target: The system must connect specifically to the Google Vertex AI API to handle chat and embedding tasks.

Prerequisites

Before configuring the application properties file, ensure your environment meets these specific infrastructure criteria:

  • Java: A supported version of Java matching the Automation.AI requirements established in the compatibility matrix, see compatibility matrix.

  • Network Access: The host running Automation.AI must have outbound HTTPS access to Google APIs and your AE REST API endpoint.

  • Google Cloud Project: A configured Google Cloud Platform (GCP) project with the Vertex AI API enabled and a service account JSON key file saved securely on the Automation.AI host.

Database Requirements

To separate operational workflows from vector searches, you must provide connection parameters for two separate logical databases.

Main Database

The main database is the standard Automation.AI database. It stores your application data, such as chat memory and MCP configurations, and uses Liquibase to manage its schema.

  • Connection string format: jdbc:postgresql://<MAIN_HOST>:<MAIN_PORT>/<MAIN_DATABASE>

For more information on how to prepare the Automation.AI database, see Preparing the Automation.AI Database.

Vector Database

The vector database isolates the vector_store table and vector-specific schemas.

  • Connection string format: jdbc:postgresql://<VECTOR_HOST>:<VECTOR_PORT>/<VECTOR_DATABASE>

  • Extension Requirement: The target database instance must have the pgvector extension installed and active.

  • Deployment state: This database can reside on a managed PostgreSQL instance, a separate VM, or an isolated Docker container running an official image (such as pgvector/pgvector:pg16). If using Docker on the same host, expose the container on a distinct port to prevent conflicts with the main database.

Note: When the Automation.AI service starts with the agentic profile active, it automatically executes vector-specific Liquibase changelogs to initialize the necessary database tables.

Configuring the application.properties File

To apply this use case, update the standard application.properties file located in your Automation.AI component installation directory. Replace all <placeholders> with your environment-specific values.

Active Spring Profile

Activate the agentic-object-creation profile alongside your required logging profile. This profile instructs the Automation.AI component to bind the vector database to Vertex AI for agentic creation routines.

spring.profiles.active=console,agentic-object-creation

Vertex AI (Gemini) Properties

Configure the connection details, service account location, and model IDs for both chat and embedding tasks.

Property Value/Description
automation.ai.model.name vertex.ai.gemini
spring.ai.vertex.ai.gemini.project-id Your GCP project ID
spring.ai.vertex.ai.gemini.location Your GCP region (for example, us-central1)
spring.ai.vertex.ai.gemini.credentials-uri file:<absolute-path-to-service-account.json>
spring.ai.vertex.ai.gemini.chat.options.model The Gemini model ID (for example, gemini-2.5-pro)
spring.ai.vertex.ai.embedding.project-id Your GCP project ID
spring.ai.vertex.ai.embedding.location Your GCP region
spring.ai.vertex.ai.embedding.text.options.dimensions Must match automation.ai.vector-db.vector-dimension and the Liquibase vector changelog for your release.

Vector Database Properties

Define the credentials and parameters for the dedicated pgvector target.

Property Value/Description
spring.datasource.vector.url JDBC URL for the pgvector database
spring.datasource.vector.username Vector database username
spring.datasource.vector.password Vector database password
spring.datasource.vector.driver-class-name org.postgresql.Driver
spring.ai.vectorstore.type pgvector
spring.ai.model.embedding.text vertexai
spring.ai.model.embedding.multimodal vertexai
automation.ai.vector-db.change-log classpath:db/changelog/db-vector.changelog-master.yaml
automation.ai.vector-db.schema-name public

Main Database Properties

Define the credentials and parameters for the standard PostgreSQL target.

Property Value / Description
spring.datasource.url JDBC URL for the main database
spring.datasource.username Main database username
spring.datasource.password Main database password
spring.jpa.hibernate.ddl-auto none
spring.liquibase.enabled true
spring.liquibase.change-log classpath:db/changelog/db.changelog-master.yaml
spring.ai.chat.memory.repository.jdbc.initialize-schema never

Automation Engine OpenAPI (MCP Tools)

Configure the connection to your JCP REST host using HTTPS:

automation.ai.mcp.openapi.external.providers.<NAME>.definition-location-url=https://<AE_HOST>:<PORT>/ae/api/v1/openapi2/swagger.json

automation.ai.mcp.openapi.external.providers.<NAME>.base-url=https://<AE_HOST>:<PORT>/ae/api/v1

automation.ai.mcp.openapi.external.providers.<NAME>.included-method-types=GET,POST

automation.ai.mcp.openapi.external.providers.<NAME>.included-parameters-in-context=client_id,Authorization

Note: If your environment employs private Certificate Authorities for TLS, add the appropriate PEM or JKS trust configurations. For more information, see Enabling and Defining TLS/SSL for the MCP Server.

Starting the Service

Navigate to your Automation.AI component execution directory and run the launcher package:

cd <AUTOMATION.AI_DIR>

java -Xms512m -Xmx2g -jar automation-ai-launcher.jar

Verifying the Setup

To verify that the configuration applied successfully, access the status endpoint via a browser or curl command at http://<HOST>:8080/api/v1/status.

The response payload must confirm that:

  • The main database status is connected.

  • The model provider is identified as vertex.ai.gemini and successfully connected.

  • The active profiles array includes agentic-object-creation.

See also: