Mobile GPS location, chat-driven device presets, and PR page pagination

When a test deletes, updates, or otherwise acts on a specific row in a list or table, the agent later has to confirm the outcome. Agents now anchor that verification to an item's distinguishing text rather than its position, and correctly recognize that a different row shifting into where the acted-on one was is evidence of success. This makes list and table operations noticeably more robust on longer web tests where earlier screenshots are pruned to stay within context limits.

The chat assistant could already create, rename, and modify test plans, but had no way to remove them – asking it to "archive the Smoke Tests plan" would fail. You can now ask it to archive one or several plans at once. Scheduled runs attached to the archived plans are cancelled automatically so they stop firing. The test cases inside remain intact.

If your project has "Auto-review on pull requests" disabled and you kick off reviews on demand via the GitHub Action or the API, the GitHub check previously stayed stuck on "Auto-review is off" for the duration. It now correctly transitions to "QA.tech is reviewing this PR" as soon as the review starts, and reaches a pass or fail verdict when it completes. The merge gate works as expected regardless of how the review was triggered.

On the test result page, selecting the final "Assess" step used to hide the Console, Network, Page, Test & Agent, and Graph inspector panels – you had to click a specific intermediate step to see them. The inspector now remains available on the assessment step, surfacing the entire run's network and console output, including any network errors, right below the verdict text.

The Custom Fields type – key-value pairs that the agent receives as key: value lines during each test run – is now documented at docs.qa.tech. The page also explains that configs are QA.tech's approach to variables: there's no {{variable}} placeholder syntax; the agent resolves config values from their name, description, and stored content at run time.

When the device preset settings UI gained egress region support, the chat agent lagged behind – you could configure a region in settings but not through chat. The agent now accepts an egress region on preset creation and edits, validates it against your plan's enabled regions (listing valid choices if you pick an unavailable one), and auto-fills locale and timezone from the region's defaults unless you override them. Listing your presets now shows each one's current egress region too.

Previously, the iOS and Android sections were hidden entirely if mobile testing wasn't active for your account. They're now always visible: default presets are displayed alongside a "Talk to support" prompt so you can see what's available before committing. The "Create preset" button stays disabled and custom preset edit and delete actions are blocked until mobile testing is active on your account.

GitHub auto-merge requires every required check to end in a recognized terminal state. When a QA.tech review concluded inconclusively – a "comment" verdict, or a review manually unblocked via "Unblock merge" – the check was left in a state GitHub did not count as complete, causing auto-merge to stall. The check now concludes with "skipped", the non-blocking terminal state GitHub recognizes for required checks, so you can keep QA.tech PR reviews as a required status check without blocking auto-merge.