Onboard your agent ecosystem
This section guides you through each stage of agent management, from onboarding your cloud environment to securing, monitoring, and managing deployed agents.
Before you begin
Make sure you have the required cloud account information, permissions, and authentication details available. Cloud credentials allow AI Trust Manager to securely connect to your cloud infrastructure for agent discovery and management.
Identify the cloud provider: Determine whether you are connecting an AWS or GCP environment.
Prepare authentication details: Have the required credential information or federation configuration ready for your selected authentication method.
Verify permissions: Ensure the credentials have the permissions required to discover the cloud resources and agent infrastructure you want to manage.
Prepare the target environment: Identify the cloud accounts, projects, subscriptions, or resources that should be connected.
Onboard agents
You can complete the agent onboarding steps in sequence as a single workflow, or perform each step individually. This flexibility lets you complete onboarding at your own pace and return to any step as needed.
Use this workflow to complete the entire agent onboarding process:
Opmerking
All stored credentials are encrypted at rest. Only ORG_ADMIN users can create or delete credential profiles.
In the AI Trust Manager menu, select Overview > Dashboard > Onboard clour environment.
Select your cloud environment AWS or GCP, and select Next.
Select a Cloud environment or add an environment.
After completing your cloud environment details, select Next.
Select a cluster from the list and select Use this cluster, or create a cluster.
Let op
Plan for approximately 15–20 minutes to complete the deployment. Actual deployment time may vary depending on your environment and configuration.
After configuring your cluster settings, select Next.
Select a registry from the list or select Add registry to create agent registry settings.
When you select a registry, you can view the list of agents present in the registry.
Select the agents you want to add to AI Trust Manager and select Configure selected agents.
On the Selected agents page, configure the following settings for each agent:
Opmerking
Agent configuration varies based on your cloud provider (AWS or GCP) and exposure type (Cluster or VPC).
Specify the No. of replicas, number of instances of the agent workload to run concurrently.
Multiple replicas can improve availability and provide capacity for concurrent requests.
Select the agent image Version or image version to deploy. This determines the specific agent configuration that will run.
Provide a custom Agent name (optional) to identify the deployed agent. If left blank, a system-generated name may be assigned.
Define the Exposure: setting to fine how an agent can be accessed:
Select the Cluster to make the agent accessible within the cluster or connected internal environment.
(OR)
VPC to expose the agent to resources within the configured virtual private cloud.
Specify the primary network Port on which the agent listens for incoming requests.
Specify Additional ports the agent requires for communication.
Define Allowed sources to access the agent.
Opmerking
Enter comma-separated IPv4 CIDRs or AWS security group IDs. Examples: 10.0.0.0/16, sg-12345678.
Specify the Managed policy ARNs to attach predefined permissions to the agent.
Define an Inline policy (JSON) specify the permissions, rules, or restrictions that apply to the agent. The policy is stored and managed as part of the agent configuration.
Specify the Project roles, the permissions granted to access and manage resources in the selected Google Cloud project.
After configuring the deployment settings for the agents images, select Done and Next.
Review and apply recommended security policies for the selected agents and deploy them in your environment.
Select a Passport profile from the list or create a passport profile.
Select Deploy.
You can review passport attributes, trust verification status, capabilities, and other security information to understand and verify each agent's identity.
When the agents are imported and deployed in your environment, select Finish to complete the onboarding process.
Opmerking
To view the progress of agent deployment:
Select View logs to view detailed logs for the deployment to monitor its progress, identify errors, and troubleshoot deployment issues.
Or select Active runs in the upper right to review the status of your clusters and agent deployment.