AI Agent Engineering > Deploy the agent > Deployment properties
  

Deployment properties

When you deploy an agent flow, you need to configure the deployment properties like the memory connection, runtime environment, and REST API endpoint properties.
The following table describes the deployment properties:
Property
Description
Memory
Optional. Connection to the memory provider you want to use when the AI agent runs. You can choose a PostgreSQL or SQLite connection, or you can omit this property.
PostgreSQL is a full-featured, relational database. Choose a PostgreSQL connection when you need a robust, scalable solution, when concurrent access might be required, when you need advanced database features like JSON support or full text search, or when you need strong data integrity and backups. For example, you might choose a PostgreSQL connection for a complex, multi-task AI agent that runs in a production environment.
SQLite stores memory in a file on the Secure Agent machine. Choose a SQLite connection when you want a lightweight, simple solution, your AI agent runs locally, the memory needs are small to moderate, or you don't need high concurrency. For example, you might choose a SQLite connection for an AI agent that uses a small data set and doesn't perform multiple, simultaneous database write operations.
If you don't choose a memory connection, the Secure Agent uses SQLite, and the memory is stored in a file on the Secure Agent machine. You might omit the memory connection if you're deploying your AI Agent for testing.
Runtime environment
The Secure Agent that runs your AI agent. You can choose one Secure Agent as your runtime environment. If a Secure Agent group contains multiple agents, choose one agent within the group.
The Secure Agent runs each time you invoke your AI agent, either by performing a test run or when you send requests to the AI agent's REST endpoint using a tool like Postman or cURL.
Note: There's no metering in the preview release of AI Agent Engineering, so you don't consume IPUs when your AI agent is invoked.
Expose as REST API
Exposes the AI agent as a REST endpoint. This option is always selected.
URL Suffix
Last part of the REST endpoint URL after the protocol, domain, and base path. The suffix can contain alphanumeric characters, hyphens (-), underscores (_), and periods (.).
Enter a suffix that's clear, meaningful, and follows RESTful conventions. For example: order-fulfillment
URL Preview
REST endpoint URL that's generated when you deploy the agent flow. The URL preview includes the suffix you entered. For example:
https://localhost:19445/a2e/lP9rKXiMOW6gc7Example/order-fulfillment
The endpoint URL contains the following parts:
  • - Protocol. The protocol is https://.
  • - Domain. The domain is localhost:<port number>. In the domain, "localhost" refers to the Secure Agent machine. The port number is the port number used by the AI Agent Runtime service of the Secure Agent. By default, the AI Agent Runtime service port number is 19445.
  • - Base path. The base path is /a2e/<agent flow ID>/, where the agent flow ID is the FRS ID of the agent flow.
  • - Suffix. The suffix you entered in the URL Suffix field.
Authentication Type
Authentication type you use to access the AI agent.
Choose one of the following authentication types:
  • - Anonymous. Allows clients to access the AI agent without providing credentials.
  • You might choose anonymous authentication when you want your AI agent to be freely accessible to the public, during early testing, or when your AI agent runs behind a firewall and you don't need additional authentication.
    Note: Anonymous authentication is not secure. It can be unsafe in development environments and is not recommended for production environments.
  • - Basic. Allows clients to access the AI agent using their Informatica Intelligent Cloud Services user name and password.
  • You might choose basic authentication for internal deployments or during development when you need to limit access.
  • - Session ID. Allows clients to access the AI agent using a session ID that they get when they log in to Informatica Intelligent Cloud Services using the REST API. This is the most secure method.
  • You might choose session ID authentication when your AI agent provides personalized responses based on a user's previous interactions or preferences that are tracked in their session.
You can select both Basic and Session ID authentication simultaneously. When you do this, clients can use either type to access the AI agent.