# SiteTrax.io Intelligence Documentation Migration Report

![SiteTrax.io Intelligence Documentation Migration Report](https://docs.sitetrax.io/uploads/images/gallery/2026-09/st-intel-migration-report-feature.png)

**Date:** 4 September 2026  
**Application version audited:** SiteTrax.io 3.0.0 at app.sitetrax.io  
**Scope:** Create the SiteTrax.io Intelligence documentation set and migrate references away from the legacy service.sitetrax.io experience.

## Summary

The SiteTrax.io Intelligence shelf existed but held no books. It now holds eight books and 35 pages. Thirty-two pages were written from scratch against the live application, two existing MCP pages moved onto the shelf with their content preserved, and the legacy service portal content was marked retired rather than deleted so that no existing URL breaks.

The live application was treated as the source of truth throughout. Where documentation and application disagreed, the application won, and each disagreement is recorded below.

## Pages created

<table id="bkmrk-bookpagesintroductio"><thead><tr><th>Book</th><th>Pages</th></tr></thead><tbody><tr><td>Introduction to SiteTrax.io Intelligence</td><td>5 — What is SiteTrax.io Intelligence?, Why Intelligence Matters, How Intelligence Uses SiteTrax.io Data, Core Intelligence Features, Intelligence Roadmap Overview</td></tr><tr><td>Getting Started</td><td>5 — Accessing SiteTrax.io Intelligence, User Requirements, First Login Experience, Navigating the Intelligence Interface, Common Workflows</td></tr><tr><td>AI Integrations</td><td>6 — Connecting ChatGPT, Connecting Claude, Understanding MCP, Using MCP with SiteTrax.io, Supported AI Platforms, Troubleshooting Connections</td></tr><tr><td>Operational Use Cases</td><td>8 — Container Visibility, Yard Operations, Gate Operations, Detention Analysis, Operational Investigations, Asset Tracking, Customer Service Workflows, Executive Reporting</td></tr><tr><td>Administration</td><td>5 — User Management, Permissions, Organization Configuration, Enabling Intelligence, Best Practices</td></tr><tr><td>FAQ and Troubleshooting</td><td>3 — Frequently Asked Questions, Known Issues, Troubleshooting Guide</td></tr><tr><td>Documentation Migration</td><td>1 — this report</td></tr></tbody></table>

Every page carries a branded feature image in the Intelligence Indigo palette.

## Pages moved and rewritten

Two empty placeholder pages, *Connect SiteTrax.io to ChatGPT from OpenAI* and *Connect SiteTrax.io to Claude from Anthropic*, existed in the MCP Server book with titles but no content. They were moved into AI Integrations and written as **Connecting ChatGPT** and **Connecting Claude**, matching their sibling pages in that book. Connection steps are sourced from OpenAI and Anthropic published documentation, and both pages cite their sources.

## Books moved

The **MCP Server** book moved from the SiteTrax.io Integrations shelf to the SiteTrax.io Intelligence shelf. Shelf membership was changed rather than the book itself, so `/books/mcp-server/...` and both page URLs are unchanged. This was deliberate: those URLs support the Microsoft Gallery for Connectors submission and must not break.

## Pages updated

<table id="bkmrk-pagebookchangesitetr"><thead><tr><th>Page</th><th>Book</th><th>Change</th></tr></thead><tbody><tr><td>SiteTrax.io API - Output (JSON)</td><td>SiteTrax.io API</td><td>Legacy hostname and portal terminology corrected</td></tr><tr><td>SiteTrax.io Capture - Android</td><td>SiteTrax.io Mobile</td><td>Legacy hostname and portal terminology corrected</td></tr><tr><td>Camera Installation Requirements and Guidelines</td><td>SiteTrax.io Gate</td><td>Legacy hostname and portal terminology corrected</td></tr><tr><td>Bring Your Own Camera (BYOC)</td><td>SiteTrax.io Gate</td><td>Legacy hostname and portal terminology corrected</td></tr><tr><td>Tutorial and Introduction to Snap</td><td>SiteTrax.io Snap</td><td>Legacy hostname and portal terminology corrected</td></tr><tr><td>Integrate Google Spreadsheet with Google Maps</td><td>Google Integrations</td><td>Legacy hostname and portal terminology corrected</td></tr></tbody></table>

All `service.sitetrax.io` links in these six pages now point at `app.sitetrax.io`, and references to the service portal now read SiteTrax.io Intelligence.

## Pages retired

All 22 pages of the **SiteTrax.io Service Portal** book now carry a retirement notice at the top directing readers to SiteTrax.io Intelligence. The book and shelf descriptions are prefixed **RETIRED**.

Nothing was deleted and nothing was renamed. Renaming the book or shelf would regenerate their slugs and break all 22 page URLs, so the names were left intact and the retirement signalled in the descriptions and on every page. Their original text is preserved as the historical record of the retired portal.

## Screenshots

Forty-six images were captured and uploaded to docs.sitetrax.io: 33 branded feature images and 13 screenshots of the live 3.0.0 application.

<table id="bkmrk-screenshotshowsintel"><thead><tr><th>Screenshot</th><th>Shows</th></tr></thead><tbody><tr><td>Intelligence hub</td><td>Profile status, coverage cards, Intelligence in the navigation</td></tr><tr><td>Assistant connections</td><td>ChatGPT, Claude and Microsoft Copilot cards</td></tr><tr><td>Dashboard</td><td>3.0.0 layout with period, project and totals</td></tr><tr><td>Assets gallery and asset detail</td><td>Readings, exceptions filter, record with image, payload and feedback controls</td></tr><tr><td>Videos list</td><td>Recordings with asset counts</td></tr><tr><td>Project users and integrations</td><td>Access management and outbound destinations</td></tr><tr><td>Interview steps 1, 2 and 7, and the published profile</td><td>Onboarding, adaptive detention inputs, published revision</td></tr></tbody></table>

The project users screenshot has real addresses replaced with example.com placeholders. No legacy 2.6.1 screenshots were reused in the new set, and the old screenshots remain only inside the retired book where they are correct as history.

## Conflicts found between documentation and application

<table id="bkmrk-findingevidenceresol"><thead><tr><th>Finding</th><th>Evidence</th><th>Resolution</th></tr></thead><tbody><tr><td>Microsoft 365 Copilot is documented as connectable, but the application says the connection is still being built</td><td>The Intelligence hub lists Copilot as coming soon</td><td>The Copilot page was kept intact because it supports the Microsoft certification submission. Supported AI Platforms and Known Issues now state the connection is not yet available, and the Copilot page is described as preparation.</td></tr><tr><td>A Reports section appears in legacy screenshots</td><td>Absent from 3.0.0 navigation; `/dashboard/reports` returns not found</td><td>Recorded in Known Issues so nobody hunts for it</td></tr><tr><td>Legacy screenshots show version 2.6.1</td><td>Current application is 3.0.0 with different navigation and dashboard</td><td>Fresh screenshots throughout the new set</td></tr><tr><td>One legacy screenshot was reused as the feature image on four different pages</td><td>Pages 31, 32, 33 and 35 of the retired book shared one image</td><td>Not corrected. Those pages are retired; the new set has a distinct image per page</td></tr></tbody></table>

## Product defects found while auditing

<table id="bkmrk-defecthow-to-reprodu"><thead><tr><th>Defect</th><th>How to reproduce</th><th>Why it matters</th></tr></thead><tbody><tr><td>Unhandled error on an unknown asset ID</td><td>Open `app.sitetrax.io/dashboard/assets/view/99999999`. The application shows *Unexpected Application Error! Cannot read properties of undefined (reading id)* instead of a not-found page. A valid ID loads correctly with or without a trailing slash.</td><td>The MCP link contract tells assistants to return links in exactly this form. Any assistant answer referencing a record the user cannot access lands on a crash screen, which reads as an outage rather than a permissions result.</td></tr><tr><td>Intelligence profile does not follow across accounts</td><td>A profile published in the browser session was not visible to the MCP connection, which was authenticated as a different account</td><td>Correct behaviour, since profiles are per user, but it produces generic answers with no explanation. Documented in Troubleshooting Connections and Known Issues. Consider surfacing the signed-in account in assistant responses.</td></tr></tbody></table>

## Quality assurance performed

- All 13 internal links across the shelf resolve.
- All 45 image references across the shelf resolve.
- All 32 new pages lead with a feature image. Both MCP Server pages carry the existing 440x250 branded images in their established position.
- No duplicate page titles were introduced.
- The only remaining `service.sitetrax.io` mentions on the shelf are five deliberate references, in code formatting, explaining that the hostname is retired. None are live links.
- Book and page ordering matches the specification, including ChatGPT before Claude in AI Integrations.

## Remaining gaps

<table id="bkmrk-gapwhywhat-would-clo"><thead><tr><th>Gap</th><th>Why</th><th>What would close it</th></tr></thead><tbody><tr><td>No screenshot of the first-run welcome panel</td><td>It only appears for an account with no published profile, and the audit account now has one</td><td>Capture on a fresh account before its interview is completed</td></tr><tr><td>Dashboard projects summary table not documented</td><td>The audit account has a single project, so a multi-project dashboard could not be observed</td><td>Re-audit the dashboard on an account with several projects</td></tr><tr><td>Organization-level administration is thin</td><td>No organization settings or roles console was reachable from the audit account. Administration is documented as project-scoped because that is what exists at this access level</td><td>Re-audit with an administrator account and extend the Permissions page</td></tr><tr><td>Roadmap page is deliberately minimal</td><td>Only Microsoft Copilot is stated as in progress anywhere in the product. Nothing else was invented</td><td>Product input on what may be published</td></tr><tr><td>Assistant-side screenshots</td><td>ChatGPT and Claude connector screens belong to those vendors and change frequently, and capturing them exposes an individual account</td><td>Optional. The written steps cite vendor documentation, which ages better than vendor screenshots</td></tr></tbody></table>

## Recommendations

1. **Fix the unknown-asset-ID error** before promoting assistant deep links widely. A not-found page stating the record does not exist or is not shared with you would remove the sharpest edge in the assistant experience.
2. **Decide the Microsoft Copilot position.** The page is valuable for the certification submission but describes a connection users cannot complete. Either keep it with the availability note now added, or move it behind a clearly marked preview section until the connection ships.
3. **Delete the retired Service Portal shelf and book** once you are satisfied nothing external links to it. Check inbound links first; the banners can carry it safely for a quarter in the meantime.
4. **Tidy three empty pages titled New Page** in the SiteTrax.io Gate book, and remove the file `test-upload.png` from the image gallery, which was created while verifying the upload path.
5. **Replace the demo Intelligence profile** on the audit account. It is published as revision 1 and labelled as a documentation demo in its custom instructions, and an edit draft is open against it. Republish with real answers or discard, whichever suits.
6. **Adopt one feature-image position.** The new pages lead with the image; the two MCP pages place it after the opening line. Either is fine, but pick one.
7. **Set a re-audit trigger on the application version.** This set was written against 3.0.0. When the version in the sidebar changes, re-check the navigation, the Intelligence hub and the assistant cards.

## Suggested future documentation

<table id="bkmrk-proposedwhyalerts-an"><thead><tr><th>Proposed</th><th>Why</th></tr></thead><tbody><tr><td>Alerts and Scheduled Digests</td><td>The only capability requiring a write scope, and currently undocumented for end users</td></tr><tr><td>Reference: status codes, asset types and capture types</td><td>Status codes such as A0 and capture types such as International Intermodal Vertical ID appear throughout the interface with no glossary. This is the single most requested kind of page in operational products.</td></tr><tr><td>Capture Coverage Planning</td><td>The coverage review tells operators to confirm both gate directions and map yard zones, but nothing explains how to do that well</td></tr><tr><td>Security and Data Handling</td><td>Consolidates OAuth, scopes, redaction, retention and what does and does not leave SiteTrax.io. Procurement asks for this.</td></tr><tr><td>Release Notes</td><td>Would have made the 2.6.1 to 3.0.0 documentation drift visible as it happened rather than at audit time</td></tr><tr><td>Interpreting SiteTrax.io Data</td><td>Record semantics are currently repeated as cautions across many pages. One canonical page would reduce duplication and drift.</td></tr></tbody></table>

<p class="callout info">This report is a working record, not end-user documentation. Archive or delete it once the recommendations have been actioned.</p>

## Phase 2 — site-wide consistency and answer-engine readiness

**Date:** 4 September 2026. A second pass covering the whole documentation site, not only the Intelligence shelf.

### Consistency fixes applied

<table id="bkmrk-findingactionthe-sit"><thead><tr><th>Finding</th><th>Action</th></tr></thead><tbody><tr><td>The SiteTrax.io Intelligence shelf had no description and no cover, against 650 to 870 characters and a cover on every other shelf</td><td>Branded cover created and a 799-character description written</td></tr><tr><td>The six new books carried 79 to 147 character descriptions against a house range of 279 to 1,273</td><td>All six rewritten to 442 to 615 characters, each opening with a definitional sentence</td></tr><tr><td>Two API pages contained stray in-body H1 headings competing with the page title, one of them three times</td><td>Demoted to H2. The site now has zero in-body H1 headings</td></tr><tr><td>Seven older pages had no feature image against 33 new pages that all did</td><td>Branded cards created and applied to the API test-environment page, Capture Android, both Snap pages, Drive, Google Spreadsheet and Zapier. Every page on the site that has content now carries a feature image</td></tr></tbody></table>

### Answer-engine and generative-engine readiness

The site was audited for how well answer engines and large language models can read and cite it. It is publicly readable and crawlable, which is the precondition for everything else. Four gaps were found and closed.

<table id="bkmrk-gapfixno-meta-descri"><thead><tr><th>Gap</th><th>Fix</th></tr></thead><tbody><tr><td>No meta description on any page. Only an Open Graph description was emitted, and on pages that open with a heading it concatenated words without spaces</td><td>A clean description is now built from the first substantial paragraphs and applied to both the meta description and the Open Graph description</td></tr><tr><td>No canonical URL</td><td>Added per page</td></tr><tr><td>No social or preview image, and no Twitter card</td><td>The page feature image is now used for both, which is why the feature-image backfill above mattered beyond appearance</td></tr><tr><td>No structured data anywhere on the site</td><td>Schema.org markup added: a site-level Organization, WebSite and SoftwareApplication graph, plus per-page BreadcrumbList and TechArticle, and FAQPage on qualifying pages. The FAQ page emits 19 question and answer pairs</td></tr></tbody></table>

All of this is delivered through Settings, Customization, Custom HTML Head. The site-level entity graph is static and therefore visible to every crawler. The per-page tags are built in the browser after the page loads.

<p class="callout warning">**Known limitation.** Search engines that render JavaScript, including Google and Bing, will see the per-page tags. Crawlers that do not render JavaScript, which includes some model training and retrieval crawlers, will see only the static site-level graph. The durable fix is to emit these tags server-side from the BookStack page template or through a reverse proxy. Treat the current implementation as a strong interim measure rather than the finished state.</p>

### Still outstanding

- **No sitemap.** `docs.sitetrax.io/sitemap.xml` returns not found, and BookStack does not generate one. This needs server-side work or a generator, and it is the highest-value remaining item for discovery.
- **Server-side rendering of the metadata** described in the limitation above.
- **Three empty pages titled New Page** remain in the SiteTrax.io Gate book. They are the only pages on the site without a feature image, and they should be filled or deleted.
- **A stray `test-upload.png`** remains in the image gallery from verifying the upload path.