Implementation
Implementation maintains callable models, Agents under test, and service credentials for reuse by evaluators and evaluation tasks.
For users and administrators who configure services. Verification baseline: the project's current implementation as of 2026-10-30. Entry points in deployed environments may vary by version and permissions.
Search keywords: integration, connection, models, Agent, credentials, connectivity.
1. Choose by goal
- Prepare a scoring model or model under test: Models.
- Connect an Agent under test: Agent.
- Maintain service keys for services other than AE Agent: Credentials.
2. Recommended configuration order
- First, confirm which type of service to connect and your account's permissions. With TE SSO, AE Agent automatically uses login credentials. For other services, an administrator must prepare credentials first.
- Create a scoring model under Model. For online Agent evaluation, also add the evaluation target under Agent.
- Test connectivity for the configurations you created. Confirm service availability before creating evaluators and evaluation tasks.
3. Available entry points and limitations
Regular users can see Model and Agent; administrators can see Credentials. A model can serve as the Judge or as an evaluation target. The model under test and the scoring model have different roles and must be selected separately.
This version does not expose the Data reporting or AI Monitor configuration tabs, and the evaluation dataset list has no default AI Monitor import entry point. Do not infer that a page is usable from backend APIs or older instructions. If you need these capabilities, ask an administrator to check the deployed version and feature availability.
Configuration visibility is constrained by the current project, owner, and sharing settings. A successful service connection means only that this connectivity check passed. You still need a small-scale evaluation to validate business inputs, actual outputs, and scoring performance.
4. Next steps and related pages
- Create a model configuration: Models.
- Create an Agent configuration: Agent.
- Prepare the calling identity: Credentials.
- Check access scope: Permissions and members.
Return to the manual home page:

