Agent Compute environment
The Agent Compute environment (ACE) runs your agents in your virtual private cloud. Each ACE supports multiple workspaces, offering flexibility to physically separate workspaces on different compute resources or logically separate them using Kubernetes namespaces. Each ACE can be configured on a different AWS account.
Example Agent Compute environments
- Production: An environment managing your production workspaces (e.g., Finance and HR).
- Pre-production: A separate environment for staging workspaces (e.g., Sandbox, Finance Staging, HR Staging).
We recommend starting with at least two Agent Compute environments:
- One for production agents.
- One for pre-production, covering everything from development to acceptance testing.
Workspace
Agents are deployed to workspaces, which are also used to manage user permissions. Each workspace is tailored to specific purposes or departments. For instance, a Sandbox workspace is used for development and testing, while a Finance workspace supports day-to-day operations of the Finance team. All users with access to the Finance workspace can use or manage all agents in this workspace.
Example workspaces
- Sandbox: A shared development workspace for testing and demonstrating agents.
- Finance: A production workspace for the Finance team operations.
- Finance Staging: A workspace used for agent testing before deploying them to the production Finance workspace.
- Marketing: The production workspace for Marketing agents.
- Marketing Staging: The staging environment for Marketing development agents.
Set up your workspaces according to your organization structures or the purposes of the agents you intend to use. Assign role permissions based on workspace access, ensuring clarity on who can access which agents and resources.
Deployed agents
Deployed agents are accessible to business users through Work Room. Once deployed, an agent is assigned to a workspace with essential configurations—LLM, secrets, OAuth clients, and feedback channels—automatically applied. Agents remain continuously operational and ready to provide service at all times.
Document sources
Document sources refer to the repositories where agents look for files to process, especially in scenarios involving Document Intelligence.
Getting started with document sources
Start using document sources by following these steps:
- Navigate to the document intelligence tab in the Agent Compute view
- Look for the workspace for which you to enable it and click the "Configure" button
- In Work Room, a Documents table will now be displayed
