Before you integrate the Guardrails API, you can use the online testing feature to validate detection performance. On the console, you can submit content for testing, select a detection model, and view the detection results without writing any code.
The online testing feature has been upgraded and migrated to the Security Operations Agent. For the new entry point and complete operating instructions, see the documentation.
Prerequisites
Before you begin, ensure that:
You have activated Guardrails on the Guardrails service activation page.
Online testing incurs charges. For more information, see billing overview.
Procedure
Choose the procedure that matches the region in which you use AI Guardrails.
Shanghai and Hangzhou
Log in to the Guardrails console.
In the left-side navigation pane, click Security Operations Agent.
On the Security Operations Agent page, click the Online Test card. Alternatively, start the conversation by requesting online testing.
Complete the online testing interaction in the conversation. The migration notice at the beginning of this topic links to the complete procedure.
Other regions
Log in to the Guardrails console.
In the left-side navigation pane, click Online Testing.
Select a template based on the type of content that you want to test.
In the input box, enter up to 2,000 characters of content. You can also select a preset text template to populate the input field.
Below the input box, select a detection model.
Input Detection Model: Select AI Input Content Moderation_Pro (
query_guard_pro_ec) to detect input sent to a large language model.Output Detection Model: Select AI Output Content Moderation_Pro (
response_guard_pro_ec) to detect content generated by a large language model.Click Run test.
View the Detection Results.
Enable detection features
To enable or adjust detection features, follow these steps in Protection Config:
In the left-side navigation pane, choose .
In the detection item configuration list, find the desired service and click Management.
In the protection dimension area, use the toggles to enable or disable detection features such as Sensitive content detection and Prompt injection detection. Some features are billed separately. For more information, see billing overview.
On a protection dimension card, click Configuration Management to manage individual detection rules.