Troubleshooting Ask Inductive Automation AI Widget

Ask Inductive Automation AI is a pilot widget in the lower-right corner of the User Manual. When it fails to open or displays an unspecified error, first separate a local browser problem from a service-side or embedded-content problem. The same failure has occurred in Chrome and Firefox on macOS and Debian XFCE, and in Brave and Chrome, including a case with the ad blocker disabled. That pattern makes a single browser, operating system, or ad-blocking extension an unlikely sole cause.

Manual Version and Widget Availability

1. Open the intended version of the Inductive Automation User Manual directly, rather than an old tab or cached navigation result.
2. Confirm that the Ask AI control appears in the lower-right corner. The pilot is part of the manual interface, so an absent control and a visible control that returns an error are different fault states.
3. Click the control once and wait for the Ask Inductive Automation AI popup. Do not begin changing browser settings until this basic state is recorded.

Do not move on until the failure is classified as absent widget, failed popup, or failed answer.

Clean Browser Reproduction

1. Open the same manual page in a private or incognito window. This reduces interference from cached site data and extensions, although it does not bypass organization-wide filtering.
2. Reload the page and activate the widget without attaching a file. A text-only question is the smallest useful test because it removes file selection, file parsing, and upload handling from the path.
3. If the error remains, repeat the test in a second browser. Keep the manual version, page, question, network, and test time unchanged.
4. Record whether the popup fails before a question is entered or only after the question is submitted.

Disabling an ad blocker alone does not prove that all browser filtering is inactive. Privacy protections, tracking prevention, script controls, DNS filtering, endpoint security, and network gateways can independently block embedded services or their requests. Brave can also apply browser-level protections independently of an installed blocker.

A repeatable failure across Chrome and Firefox on macOS and Debian XFCE, plus Brave and Chrome, shifts the next check toward a shared service or network dependency. The confirming result is either success in one clean browser or the same failure stage in two browsers.

Console and Network Capture

1. Open the browser developer tools before reproducing the problem.
2. Clear existing console and network entries so unrelated page messages do not obscure the event.
3. Reload the manual page, click the Ask AI widget, enter a short text question, and submit it once.
4. Inspect the console for blocked scripts, cross-origin failures, content-security messages, uncaught exceptions, or failed resource loads.
5. Inspect the network panel for the request initiated by the widget. Record the request URL as displayed by the browser, request method, HTTP status, elapsed time, and response body or browser-generated failure reason.
6. Capture a screenshot with the visible error and relevant developer-tool entry. Remove cookies, authorization values, file contents, and other sensitive data before sharing it.

Do not move on until the capture identifies whether the failure occurs before transmission, during transport, in the response, or while rendering the response.

Filtering and Network Isolation

1. Temporarily disable extensions for the test profile, not only the extension labeled as an ad blocker.
2. Review built-in browser privacy controls for the manual page and repeat the same text-only question.
3. If the request still fails, run the test from an independent network that does not use the same DNS filter, proxy, firewall, or TLS inspection path. Use an approved network and follow site security policy.
4. Compare the network entry from both paths. A request that succeeds on the independent network isolates the problem to the original network path or its security controls.
5. If both networks fail at the same stage and time, retain the captures rather than adding broad allow rules without a confirmed blocked destination.

The widget is powered by Kapa.ai and draws from solved technical content, Help Center knowledge-base articles, the SDK Guide, and the selected User Manual version. The browser must therefore load the manual interface and complete the widget's supporting requests. A visible manual page does not prove that every dependency needed by the popup and answer request is reachable.

The proving check is a controlled A/B result: the identical question succeeds on one network and fails on the other, or it fails with the same request evidence on both.

Site Data Reset and Minimal Query

1. Close other tabs for the same manual site so they do not retain an older application state.
2. Clear site-specific cached content and storage for the manual page using the browser's site-data controls.
3. Reopen the selected User Manual version and confirm that the lower-right widget loads.
4. Submit a short text question without using Attach files.
5. Only after a text response succeeds, test Attach files with a non-sensitive sample permitted by company policy.

The attachment option adds context to a question, but it also adds a separate input and transfer path. Testing attachments before basic question submission works can hide the original failure behind file-related behavior. Likewise, changing several privacy or network controls simultaneously destroys the comparison needed to identify the responsible layer.

Do not move on until a minimal text question either produces an answer or generates a fresh console and network capture after the site-data reset.

End-to-End Acceptance Test

1. Open the required manual version in the target browser and confirm the widget appears in the lower-right corner.
2. Open the popup and submit a short question that can be answered from the selected manual version.
3. Confirm that the reply renders without the prior error.
4. Open at least one direct source link supplied with the response and confirm that the source list appears at the bottom of the answer.
5. Test Attach files separately if that function is required, using approved non-sensitive content.
6. Use the thumbs-up or thumbs-down control to rate the result. For an unresolved fault, send the manual version, page, timestamp with time zone, browser and operating system, failure stage, sanitized screenshot, console message, and network status through an official Inductive Automation support channel.

The commissioning check passes only when the widget opens, accepts a question, renders an answer, displays its source list, and opens a cited source. File attachment passes as a separate function after the basic path is operational.

FAQ