- Release notes
- Overview
- Getting started
Healing Agent user guide
Recovery strategies
Healing Agent employs several recovery strategies tailored to various scenarios. These recovery strategies are both AI-driven and deterministic.
Healing Agent employs AI-based and deterministic recovery strategies to generate additional possible matches. The final target element is identified by comparing the outcomes of all recovery strategies to ensure high accuracy and minimize false positives.
If you disable the Healing Agent AI recovery methods, Healing Agent uses Levenshtein distance to perform text matching, as a fallback.
When one or more attributes of the target or anchor selector change, Healing Agent generates a new selector and compares it to the original. If the new selector matches the same element, Healing Agent executes the activity against it.
tag and aaname attributes changed:
|
Initial selector |
Selector generated by Healing Agent |
|---|---|
<webctrl id='Country' tag='H2' type='' class='' aaname='Country' /> | <webctrl id='Country' tag='LABEL' type='' class='' /> |
aaname, class, and text selector attributes changed:
|
Initial selector |
Selector generated by Healing Agent |
|---|---|
<webctrl visibleinnertext='Sign in with another account' tag='P' type='' class='MuiTypography-root MuiTypography-body1 jss32
css-9grhbv' check:text='Sign in with another account' /> | <webctrl aaname=' Log in with another account' tag='AP-TYPOGRAPHY' type='' class='jss32 light m block inherit hydrated'
check:text='Log in with another account' /> |
Activities sometimes need time before an action can be executed. In such situations, Healing Agent knows how much to wait, and provides fix suggestions when needed.
Healing Agent can leverage AI-based recovery strategies.
If a pop-up, overlay or modal window obstructs a target element, the Healing Agent takes appropriate action:
-
If the pop-up belongs to the automated application, the Healing Agent closes the pop-up.
-
If the obstruction is caused by an external window unrelated to the application, the Healing Agent minimizes the window and re-tries the activity.
Labels associated with input fields, checkboxes, and dropdowns may be rephrased with an equivalent semantic meaning.
For example, an automated application had two input fields labeled as Name and Surname. In a subsequent application update, these two labels changed to First name and Last name. Healing Agent applies semantic targeting recovery strategies to repair the automation.
If the original target configuration did not use Computer Vision as the primary or fallback method, Healing Agent attempts to recover the element using the Computer Vision service.
This recovery leverages an element screenshot captured during the automation creation. An API key is not required for this step.
