Skip to content
Anslation

Self-healing automation that asks before it changes a bot.

When a page changes and a step's target stops matching, AnsBot tries the fallback targets it has learned and, if one works, continues the run and proposes that target as a repair. A repair never applies itself: it is tested on a copy of the bot in a shadow run with the same inputs, and only a person with publish rights can approve and publish it.

From a broken step to a fixed bot. Every repair is tested first.

  1. A target stops matching

    The page changed and a step's element is no longer found.

  2. The run tries its fallbacks

    Other targets saved for that step, and selectors that worked in earlier runs, are tried in turn.

  3. A repair is proposed

    A fallback that worked becomes a repair suggestion; if none worked, the failed page is used to suggest new targets.

  4. A shadow test runs on a copy

    The repair goes into a copy of the version and runs in test mode with the same inputs.

  5. A person approves and publishes

    Only after the shadow run succeeds can someone with publish rights approve the repair and publish that version.

Where repairs come from. And what happens next.

Sources of AnsBot repairs
SourceHow it is foundWhat happens next
Fallback that workedThe run found the element with a fallback target.Can go straight to a shadow test.
Suggestion from the failed pageA model where the Anslation model gateway is enabled, otherwise rules that read the page's structure.Scored below the level that starts a shadow test: a person reviews it and changes the step if it is right.
Selector memorySelectors that worked for each step in successful runs.Tried as fallbacks in later runs.

What self-healing does not do. On purpose.

It does

  • Leaves the published version untouched until a person publishes the repaired copy.
  • Asks for a new shadow test if the repaired step changes after its test.

It does not

  • Apply a repair on its own: the repair autopilot only starts shadow tests.
  • Send pages behind a sign-in to a model, unless the bot allows it.
  • Change a bot's logic: a repair replaces how a step finds its element.

Why repairs wait for a person. Three reasons.

  • A new target can match the wrong elementTwo buttons with similar text look alike to a selector; a shadow run and a person catch it.
  • The live bot keeps working meanwhileRuns keep using the published version until the repaired copy is published.
  • Every decision is on the recordThe approval is kept in the version's review history.

Self-healing: questions.

Still have a question? Ask us

What does self-healing mean in AnsBot?

Automation that notices when a page has changed and finds another way to the same element instead of failing. In AnsBot that happens through fallback targets and repair suggestions, and a person approves each repair before it goes live.

Does AnsBot fix bots automatically?

Not on its own: it finds and tests fixes, and a person approves them. A repair runs on a copy of the bot in a test-mode shadow run with the same inputs, and publishing it needs that run to succeed and someone with publish rights to approve.

Is AI used to repair bots?

Only to suggest targets after a step fails, and only where the Anslation model gateway is enabled; otherwise rules that read the page suggest them. AI suggestions are scored below the level that can start a shadow test, and pages behind a sign-in are not sent to the model unless the bot allows it.

What if the page changes again before I approve?

If the repaired step changes after its shadow run, approval is refused and the repair needs a new shadow test.

Pages change. Your bots can keep up, with you in charge.

Tell us the web task and the site it runs on. We read every request and reply by email.

Office hours: Monday to Saturday, 10:00–18:00 IST+91 74158 85838