Skip to main content

Documentation Index

Fetch the complete documentation index at: https://docs.helpalive.com/llms.txt

Use this file to discover all available pages before exploring further.

The Agent → Test tab is your safety net before publishing. It lets you ask the Agent real questions using your draft settings — new knowledge sources, new tone, new custom responses — and see exactly what your users would see if you published right now. Open your dashboard and go to Agent → Test.

Why test before publishing

A draft change can mean: new docs you just crawled, a tone shift from formal to casual, a custom response you wrote for a tricky topic. Before that goes to your users, you want to see what it does. Test is the place to do it without risk. You’re chatting with the same Agent your users will get — but it’s pointing at your draft settings, not your live ones. A “Previewing draft” badge makes sure you never confuse the two.

Single test or batch

There are two ways to use Test: Single question. Type a question, press enter, watch the Agent reply. Quick way to spot-check one thing. Batch tests. Type a list of questions and run them all at once. The Agent answers each one and lays the results out side by side, with the current live answer next to the new draft answer so you can compare. Batch testing is what you reach for when you’ve made a bigger change — like adding a new knowledge source — and want to make sure you didn’t accidentally break answers to questions that were already working.

Simulating context

The Agent answers differently based on who’s asking. A free-tier user on the Billing page sees different responses than an admin on the Workflows page. To test those differences in Test:
  • Set the simulated page URL — pretend the user is on /billing, /workflows, or any other page in your product.
  • Set the simulated user — display name, role, plan. The Agent uses these the same way it would in production.
This lets you verify plan-aware welcome copy, role-specific guidance, and page-aware answers without setting up a real test account.

Saving test sets

If you have a regular set of questions you want to re-run after every meaningful change — your team’s “smoke tests” — save them as a test set. Saved sets show up in the test history and can be re-run with one click.

What test runs against

SourceWhat it uses
Knowledge sourcesYour draft sources (including ones you haven’t published).
Tone and rulesDraft tone, response length, custom rules from Personalization.
Custom responsesDraft custom responses.
Guidance templatesDraft templates.
AppearanceDraft appearance (so you can also visually preview the widget).
In short: everything you’ve changed since your last publish.

What it doesn’t do

  • Test traffic doesn’t appear in your real Topics inbox or your analytics. It’s a separate run.
  • Test isn’t a real user session. The Agent won’t remember conversations between test runs.
  • Test results aren’t shared with your end users. They only see the live, published Agent.

Rate limits

To prevent runaway test runs from running up your bill, Test has a per-user rate limit. If you hit the cap, wait a few minutes and try again. Most teams never come close.

Next steps

Publishing

When the test passes, ship it.

Personalization

Tone, custom responses, guidance templates.

Train

Add or update knowledge sources.