Overview
Page titles, meta descriptions, canonical URLs, robots, Open Graph / X cards, JSON-LD and XML sitemaps for ExpressionEngine 7.5+ (PHP 8.2+). Built for sites that use Structure for their page tree.
MetaCraft outputs nothing until a template uses {exp:metacraft:head}. It never edits templates, Structure data, SEO Lite data or channel entries.
The same guide, with screenshots and search, is under Documentation in MetaCraft's sidebar in the control panel. It's also a website: open docs/site/index.html in a browser, or upload docs/site/ anywhere that serves static files.
- GETTING-STARTED.md — install, upgrade and first setup, step by step
- TEMPLATES.md — the head and sitemap tags, every parameter, examples
- SITEMAPS.md — sitemap sources, setup, static mode, cron, hosting notes
- SCHEMA-MARKUP.md — marking structured data values in your pages
- TROUBLESHOOTING.md
- CHANGELOG.md · FUTURE.md
Install
- Copy
metacraft/tosystem/user/addons/metacraft/(withouttests/, see Deploying). - Control Panel → Add-Ons → MetaCraft SEO → Install.
- Open Add-Ons → MetaCraft SEO. The Overview walks you through the rest (GETTING-STARTED.md has every step).
To upgrade, replace the folder, then click Update on MetaCraft in Add-Ons (or run php system/ee/eecli.php addons:update --addon=metacraft) to run the database updates.
Uninstalling removes MetaCraft's tables and its publish tab from every layout. Static sitemap files are left in place. If you used static mode, click Remove static files on the Sitemap screen before uninstalling, or delete sitemap.xml / sitemap-N.xml by hand afterwards.
How values are chosen
Every value is resolved from the most specific source that has one:
| Order | Source | Badge |
|---|---|---|
| 1 | Template tag parameter ({exp:metacraft:head title="…"}) |
Template |
| 2 | The entry's SEO tab override | Custom |
| 3 | A mapped channel field (e.g. a "Summary" field used as the description) | Mapped |
| 4 | Channel defaults | Channel |
| 5 | Global defaults | Global |
| 6 | Automatic (entry title, entry URL, current page URL) | Auto |
Channel-level sources (2–4) only apply in channels that are enabled in MetaCraft. Entries in other channels — and pages that aren't entries at all — use template parameters, global defaults and automatic values.
The same resolver drives the head tag, the publish-tab preview, the sitemap and the audit, so they always agree.
The control panel
Add-Ons → MetaCraft SEO has four screens: Overview, Content SEO, Sitemap and Settings, plus Documentation: this guide, with search, for everyone who can open MetaCraft. Settings and Sitemap changes need a MetaCraft admin (Permissions). Other roles don't see Settings in the sidebar and get a read-only Sitemap.
CP links carry EE's session in the URL (&S=…). Open screens from the sidebar or from links inside MetaCraft. A bookmarked or pasted CP URL without a valid session asks you to log in again.
Setup
Settings → General
- Site name: used in titles and as the default name everywhere. Blank = EE's site name.
- Base URL: every canonical and sitemap URL is built from it. Blank = EE's base URL. Never include
index.php; the field warns if you do. A Canonical example line shows the result as you type. - Title format: tokens
{title},{site_name},{channel_name},{separator}(click a token to insert it). Default{title} {separator} {site_name}. The Example box renders an entry title and a page without one. Unknown tokens and a format without{title}get a warning. - Separator:
|,-,–or·. - Default description: used when nothing more specific exists. Usually best left blank (no description tag), so search engines write their own snippet.
- Search indexing / Link following: the site-wide robots default. Choosing Noindex shows a red warning. Only use it on staging. With Index and Follow there's no robots tag in the page source; that's the default search engines assume.
Settings → Social
og:site_name (blank = site name), the default share image (File Manager), X card type (Large image or Summary) and X @handle.
Settings → Structured data
Turn JSON-LD on or off. Turning it off keeps the values below.
- Who publishes this site: the entity type, plus name, URL, logo and profile links (
sameAs, one full URL per line; a bad line is named). Entity types:- General:
Organization(no public location),LocalBusiness,ProfessionalService; - Medical:
MedicalOrganization,MedicalBusiness,MedicalClinic,Physician,Dentist.
- General:
- Location and contact (LocalBusiness types, including MedicalClinic, Physician and Dentist): address, telephone, email, latitude/longitude, price range, area served, and opening hours. Hours are rows of days with opening and closing times; tick Closed to say a day is closed, and days you leave out say nothing. If a day is in two rows, only the first counts (the screen warns you).
- Medical details (medical types): specialties (schema.org's list), accepting new patients (Not stated / Yes / No) and accepted insurance (one per line). schema.org only defines these for MedicalOrganization, MedicalClinic, Physician and Dentist, so they aren't output for plain MedicalBusiness.
- WebSite: the site's name and URL.
- Example output: the site-wide JSON-LD built from the saved settings.
Sites with several locations (a clinic with ten offices) keep the brand here, usually as Organization or MedicalOrganization, and give each location its own entry in a channel set to LocalBusiness (location). See Structured data per channel.
Settings → Channels
A table of every channel on the site: an on/off switch, plus where its title, description and image come from. A broken field mapping shows in amber. Edit opens the channel's drawer:
- Title format: overrides the site format for this channel (blank = site format), with an example.
- Field mappings: SEO title, Open Graph title and Open Graph description from a channel field. Only compatible fields are listed:
| Target | Fieldtypes |
|---|---|
| SEO title, description, OG title, OG description | Text input, Textarea, Rich Text (HTML stripped) |
| Social image | File |
| Canonical URL | Text input, URL. Anything that isn't a full http(s) URL or a path starting with / is skipped. |
- Description comes from: a field (the fixed text below it is the fallback when the field is empty), the same fixed text for every entry, or nothing.
- Social image comes from: a File field, one image for every entry, or nothing.
- Canonical URL comes from: the page's own URL (recommended) or a field.
- Channel defaults: Search indexing and Link following (Site default / on / off).
- Structured data: the channel's type and where each of its properties comes from. See below.
- Show the MetaCraft SEO tab: adds the tab to the publish form, including custom publish layouts, and removes it again when you turn it off.
A mapping to a field that was deleted, moved out of the channel's field group, or changed type keeps showing as Missing field (#N) with a warning until you choose another. At runtime it's ignored, never fatal.
All drawers save together with Save settings. URL patterns and sitemap inclusion are on the Sitemap screen (SITEMAPS.md).
Structured data per channel
Pick the channel's type, then map each property to a channel field. Only fields that fit are listed (a Date field for a start date, a URL field for a ticket link). Properties are marked:
- Required: without a value the entry outputs a plain
WebPageinstead of a half-filled type (Google treats a broken Event as worse than none). The summary line says how many required properties aren't mapped yet. - Optional: added when the entry has a value.
- Automatic: filled from MetaCraft's own values (the SEO title, the entry title, dates, the social image, the site's organization).
| Type | Required | Optional |
|---|---|---|
| WebPage | — | — |
| MedicalWebPage | — | Last reviewed (Date), Reviewed by (Text) |
| FAQPage | Questions and answers (Grid) | — |
| Article, BlogPosting | — | Author (entry author or Text), Image (File; else the social image) |
| Event | Start date (Date), Venue name, Venue address (Text) | End date, Tickets link (URL), Image |
| Product | Price (Text with a currency, e.g. $25, 25.00 USD; a price without one counts as empty) |
Brand, Image (File; else the social image) |
| Service | — | Service type, Area served |
| Person | — | Job title, Photo |
| Physician | — | Specialty, Telephone, Photo |
| LocalBusiness (location) | Street, City (Text) | Region, Postcode, Country (else the country in Settings › Structured data), Telephone, Opening hours (Grid), Latitude, Longitude, Photo |
The Required marks follow what Google needs before it will show a rich result (an event's venue address, a product's price), so a page that outputs a type is eligible for it.
Person pages output a ProfilePage whose main entity is the Person. Event, Product, Service, Physician and location pages output a WebPage whose main entity is the thing.
A location uses the site's own type when that's a kind of local business (a MedicalClinic site gets MedicalClinic locations), MedicalClinic under a MedicalOrganization, and LocalBusiness otherwise. Each location links to the site's organization as its parent.
Grid fields feed two properties:
| Property | The Grid needs columns named (or labelled) |
|---|---|
| Opening hours | day, opens, closes |
| Questions and answers | question, answer |
A Grid without the right columns is listed but can't be chosen, with the missing column named ("needs a Closes column"). For the day column, a Select or Checkboxes column with Monday…Sunday is the most reliable; MetaCraft also reads text like Mon–Fri, Monday, Wednesday or Weekdays. Times can be 8:00 AM or 08:00; Closed in the Opens column marks the days closed. Rows MetaCraft can't read are skipped and listed in the editor's Structured data preview.
Each entry can choose a different type in its editor's Advanced section. The channel's mappings are used for whichever type the entry outputs.
Settings → AI
AI suggestions are off until an admin turns them on here. While they're off, MetaCraft never contacts an AI provider and no AI controls appear anywhere. Page capture, page marks and suggestions from page HTML still work.
- Enable AI suggestions: the switch. The disclosure below it says what is sent and when (Privacy).
- Provider: Anthropic (the only one in 1.3).
- API key: your own Anthropic key, from the Anthropic Console. It's stored encrypted and never shown again; the screen shows only its last four characters. Replace or Remove it, or Test key (a free call that checks the key, nothing is generated or charged). To keep the key out of the database, set it in
config.phpinstead:$config['metacraft_anthropic_key'] = 'sk-ant-…';. The screen then shows it as locked. - Models: one for titles and descriptions, one for structured data. Claude Haiku 4.5 is the default for both; Claude Sonnet 5.5 costs twice as much and is better with long or messy pages.
- Usage: requests, tokens and estimated cost for this month and last month, counted by MetaCraft. Your Anthropic invoice is the final word.
- Page capture (advanced): see Reading pages.
Settings → Advanced
- Trailing slashes in URLs: Keep as is, Remove or Add, applied to every URL MetaCraft outputs. On Structure sites choose Add. Structure adds a trailing slash to page URLs when an entry is saved, but drops it for every page when you drag-reorder the tree, until each page is saved again. With Add, canonical and sitemap URLs stay the same either way. When the choice doesn't match Structure's own "add trailing slash" setting, an amber box offers a one-click Use Add / Use Remove. MetaCraft never changes Structure's setting.
- Who can change settings: see Permissions. Every role on the site is listed; roles that can't use the control panel are marked no CP access.
- Import from SEO Lite…: shown while SEO Lite's data table exists (Switching from SEO Lite Pro).
Every Settings tab has one sticky save bar: unsaved changes are flagged, Discard reloads the saved values, and ⌘S / Ctrl+S saves. Errors are listed in the bar, marked on the field and shown as a red dot on the tab. Nothing is saved until they're fixed.
Editing an entry's SEO
There are three ways in, and all three edit the same values:
- The MetaCraft SEO publish tab on the entry form. It saves with the entry.
- The full editor (Edit SEO from Content SEO or the Overview). It's its own page with its own Save SEO and doesn't touch the rest of the entry.
- Quick edit in Content SEO: the SEO title and description in a drawer, with a search preview. Open full editor links to the rest.
The editor has three sections:
- Search appearance: title (plus "use exactly as written"), description, canonical URL, index and follow.
- Social: OG title, description and image, the image's alt text, plus optional X-specific overrides (Share image alt text).
- Advanced: sitemap inclusion, structured data type, extra robots directives (
noarchive,max-image-preview:large…).
Each field shows the inherited value as a placeholder and a badge with its source. The search and social previews update as you type. Reset (or leaving a field blank) inherits again.
The preview's Structured data tab shows the JSON-LD the page outputs, as last saved: the type, a Test in Google link (Google's Rich Results Test for the page's URL), a note when a required property is empty, and any Grid rows that were skipped.
Above the preview, a line says when the entry's page was last read, with Re-read. In 1.3 the editor can also suggest values from that page (Page content and suggestions). From 1.4, an Images panel under that line checks the page's images for alt text (Image alt text).
The publish tab only saves when the tab itself was on the submitted form, so bulk edits, Channel Form and API saves never clear SEO data. "Save as new entry" copies everything except the canonical URL.
The full editor and quick edit save only the fields you changed. If someone else saved the entry's SEO after you opened it, you're asked whether to overwrite their changes.
Permissions
- Overview, Content SEO and the read-only Sitemap: anyone whose role has access to the MetaCraft add-on (EE's normal add-on access).
- Settings, Sitemap changes and the SEO Lite import: Super Admins, plus the roles chosen under Settings → Advanced → Who can change settings. This is checked on every request, including form posts. Other roles see a "Settings are limited to MetaCraft admins" page.
- Editing an entry's SEO (publish tab, full editor, quick edit): the same rules as editing the entry in EE: the channel must be assigned to the role, and editing other people's entries needs that permission. MetaCraft must be on for the entry's channel. Otherwise the editor explains why and nothing can be saved.
- Suggestions for an entry (Re-read, Suggest, Use, Dismiss): anyone who can edit that entry. Bulk suggestion jobs and Settings › AI: MetaCraft admins only.
Overview and Content SEO
These screens audit what MetaCraft would output for every published entry. These are candidate entries: public status, entry date in the past, not expired, and either in an enabled channel or in the Structure tree.
Overview shows the totals (indexable, noindex, needs attention), what needs attention most, the sitemap's status, and any templates that would output SEO tags twice (TEMPLATES.md).
Needs attention means a missing description or social image (with nothing to inherit), a duplicate title (case- and whitespace-insensitive) or a duplicate canonical URL. Inherited channel or global values don't count as missing.
Content SEO lists every candidate entry with coverage chips for each value: custom, inherited, missing or noindex. Hover or focus a chip to see the value and where it comes from. Filter by status (Needs attention, Duplicate title, Using entry title, Schema incomplete…), by channel or by search. Using entry title, Schema incomplete and Images missing alt text are informational and don't count toward Needs attention. Rows with images missing alt text show a chip with the count ("2 images") that opens the entry's editor. Each row has Quick edit and Edit SEO.
Results are cached for about an hour. They refresh whenever an entry, MetaCraft setting or channel setting changes, or when you click Re-check now. This is a check of your configuration, not data from search engines.
Page content and suggestions
MetaCraft 1.3 can read each entry's rendered page and suggest values for its SEO fields and structured data. Every suggestion is reviewed by a person; nothing is applied on its own, including in bulk.
Reading pages
MetaCraft reads an entry's public page with an ordinary GET request to your own site (user agent MetaCraft-Capture/1.3). It keeps the page's main text, any page marks and what it recognised in the HTML. It never reads pages during a page view or while an entry is being saved: saving only notes that the page has changed.
Pages are read:
-
when you click Re-read (or Read the page) in the editor, or ask for a suggestion and the page has changed;
-
from the command line, for entries changed since they were last read:
php system/ee/eecli.php metacraft:capture # changed entries php system/ee/eecli.php metacraft:capture --all # every entry in enabled channels php system/ee/eecli.php metacraft:capture --entry=12 # one entry (also --channel=N, --site=N)Run it from cron (for example every 15 minutes) so pages are ready before anyone asks.
Only public entries are read: closed, future-dated and expired entries are skipped, and so are entries without a URL. Redirects, error responses and pages over 5 MB aren't read, and the editor says why (TROUBLESHOOTING.md).
If the server can't fetch its own pages by their public URL (basic auth on staging, a CDN that blocks the server, a different internal host), set Settings › AI › Page capture: a Capture URL that replaces the base URL for reading only, and a basic auth username and password (stored encrypted). Test capture reads the last page read successfully with those settings. The capture URL is never used in canonical or sitemap URLs.
Where suggestions come from
| Source | What | Applied |
|---|---|---|
| Page marks | data-schema-prop / data-schema attributes (SCHEMA-MARKUP.md) |
Used directly, shown with a Marked badge |
| Page HTML | phone and email links, <address>, <time datetime>, FAQ-style questions |
Suggested for review |
| AI (when on) | titles, descriptions and structured data written from the page text | Suggested for review |
Structured data values resolve in this order: a value used on the entry → a page mark → the channel's mapped field → MetaCraft's automatic value.
In the editor
- Suggest next to the SEO title, description and Open Graph title and description (with AI on), or Suggest N empty fields above the sections. A suggestion card shows the text and what it replaces, with Use, Edit and Dismiss. Use fills the field; it's saved with the rest of the form.
- If the page hasn't changed since the last suggestion, the earlier one is shown and nothing is sent. Suggest again (sends a request) asks anew.
- The Structured data tab lists Suggested from the page: marked values, and suggestions from the page HTML and AI with Use / Dismiss (and Undo after using). Suggest with AI asks for structured data. Used values are saved straight away and the JSON-LD updates.
Bulk jobs
With AI on, MetaCraft admins get a bulk job on Content SEO's Missing description and Schema incomplete views ("Suggest descriptions for 48 entries…"), and from the Overview's Needs attention card.
- Confirm: how many entries will be sent, how many are skipped because their page hasn't changed, how many pages haven't been read yet, and an estimated cost range. Nothing is sent until you click Start.
- Running: entries go 5 at a time. Stop, or closing the drawer or tab, pauses the job; Content SEO then offers Resume. Entries already done aren't sent again.
- Review: current value and suggestion side by side, with Use, Edit, Dismiss and Undo, or Use all shown. Failed entries say why.
- Done: a summary of what was applied, dismissed and failed.
Suggestions you don't review stay on each entry, for review in its editor.
Privacy
- What is sent, and where: when someone asks for a suggestion or starts a bulk job, MetaCraft sends that entry's title, URL and the main text of its page (up to 24,000 characters; longer pages are trimmed and the editor says so), plus the list of fields or properties wanted, to Anthropic's API (
api.anthropic.com), using your API key. Nothing else from the entry, other entries, members or settings is sent. - When: only on Suggest, Suggest N empty fields, Suggest again, Suggest with AI, or a bulk job's Start / Resume. Never on page views, entry saves or captures. Test key sends no content.
- With AI off: nothing is sent to any AI provider. The only requests MetaCraft makes are page captures to your own site.
- Stored on your server: each entry's captured text, marks and findings (
metacraft_page_content), suggestions and their status (metacraft_suggestions), and monthly totals of requests and tokens (metacraft_ai_usage). An entry's page content and suggestions are deleted with the entry. - The key is stored encrypted with EE's encryption key (or kept in
config.php), is never written to logs, and is never sent to the browser. - Anthropic's handling of API data is covered by your agreement with Anthropic. MetaCraft doesn't ask the provider to keep or train on anything.
Costs
Requests are billed to your Anthropic account at its list prices (October 2026: Claude Haiku 4.5 $1 in / $5 out, Claude Sonnet 5.5 $2 / $10 per million tokens). A typical page is 2,000–7,000 input tokens and a description about 60 output tokens, so a description costs well under a cent with Haiku. Bulk jobs show an estimate before anything is sent, Settings › AI shows this month's usage, and unchanged pages are never sent twice.
Image alt text
MetaCraft 1.4 checks the images on each entry's page for alt text, and outputs alt text for the share images it puts in the head tag. No AI is involved, and nothing is sent anywhere new.
What's checked
When MetaCraft reads an entry's page (Reading pages), it also lists the images in the page's main content: the same part it reads text from (<main>, or the <body> without its header, navigation and footer). Every <img> and <input type="image"> counts, including lazy-loaded images (data-src, srcset, <picture>). Image files are never downloaded.
Each image gets one of four states:
| State | Means | |
|---|---|---|
| Missing alt text | The image has no alt attribute. Screen readers usually read out the file name instead. |
An issue |
| Possibly unhelpful | The alt text is the file name (IMG_2041.jpg), a generic word (image, photo, banner…) or under 3 characters. |
A warning |
| Decorative | alt="", role="presentation" or role="none", or aria-hidden="true" on the image or around it. Screen readers skip it, which is right for images that add nothing. |
Fine |
| OK | Anything else. MetaCraft can't judge whether the description is accurate. | Fine |
Not checked: images in the site header, navigation and footer (logos and icons are the same on every page, so check them once in the templates), inline <svg>, CSS background images, <video> posters, tracking pixels (1×1), and the <noscript> copies lazy-loading scripts add.
Where you see it
- In the editor, an Images panel under the capture line: "5 images: 1 missing alt text, 1 possibly unhelpful, 1 decorative", or "All 12 images have alt text". Show which lists them, issues first, with a thumbnail, the file name, the state and the current alt text. The panel only appears once the page has been read by MetaCraft 1.4 or later. Before that it says Re-read the page to check its images.
- In Content SEO, the Images missing alt text status filter, and a chip on each such row. Like Schema incomplete it's informational: it doesn't count toward Needs attention, and entries whose pages haven't been read don't appear.
After upgrading to 1.4, run php system/ee/eecli.php metacraft:capture once: the upgrade marks every page read before 1.4 as changed, so this reads them again for their images. Re-reading never sends anything to AI and doesn't count as changed content.
Fixing alt text
Alt text is part of the entry's content, so fix it where the image was added. MetaCraft only reports it; it never changes your content.
- PageCraft: select the image, open its settings and fill in the alt text (or leave it empty and mark the image decorative if it adds nothing).
- A file field: the alt text comes from your template. Many templates output the file's description from the file manager as the alt text.
- A template image: edit the
altattribute in the template.
Then save the entry and click Re-read in the editor.
Share image alt text
The editor's Social section has an Alt text field under the social image (and under the X image when the entry has its own). MetaCraft outputs og:image:alt and twitter:image:alt from:
- what's typed in that field, for this entry;
- otherwise the image's description in the file manager;
- otherwise nothing. MetaCraft never makes alt text up from the title or the file name.
The X image uses the Open Graph alt text only when it's the same image. Images given by the image= template parameter only use the file description route, because the entry's alt text describes the entry's own image.
Switching from SEO Lite Pro
MetaCraft copies SEO Lite's per-entry data once, then SEO Lite can be removed. MetaCraft never edits templates or SEO Lite data, so every step can be undone until you uninstall SEO Lite.
- Work on a staging copy of the site and database.
- Configure MetaCraft first: Settings → General (title format, separator, base URL), Social, Structured data, Advanced (trailing slashes), and turn on the channels SEO Lite was used on.
- Dry run the import: Settings → Advanced → Import from SEO Lite…, or
php system/ee/eecli.php metacraft:import-seolite(add--site=2on other sites). This changes nothing. It lists what would be copied:- Conflicts: the entry already has a different MetaCraft value. MetaCraft's value is kept.
- Skips: e.g. a canonical that isn't a valid URL.
- Adjustments: SEO Lite titles that already end in your site name have that suffix removed. Titles that already contain your branding are imported as "exact", so the site name isn't added twice.
- Run the import (the CP button, or the same command with
--apply). Only empty MetaCraft fields are written. Channels with imported data are enabled and get the SEO tab. Running it again is safe. - Compare output side by side. Make a scratch template group with a copy of your page template, using
{exp:metacraft:head}in place of{exp:seo_lite …}(this project'smetacraft_testgroup is an example). View a few pages of each type in both versions and compare the<head>source. - Swap the tag in the shared head (layout or partial): add
{exp:metacraft:head}and remove{exp:seo_lite …}and any hard-coded<title>, description, canonical orog:tags. - Check the Overview: it should not list any templates that may output duplicate tags.
- Deploy, repeat steps 3–4 on production, then (after a quiet period) uninstall SEO Lite.
Deploying
Ship the add-on without tests/. docs/ ships. Dev tooling (PHPUnit, Composer, test scripts) lives outside the add-on in metacraft-dev/ and is never deployed.
rsync -a --delete \
--exclude 'tests/' \
--exclude '.DS_Store' \
system/user/addons/metacraft/ user@server:/path/to/site/system/user/addons/metacraft/
The same in a deploy tool's exclude list:
system/user/addons/metacraft/tests/
metacraft-dev/
To keep page content fresh for suggestions, add the capture command to cron (Reading pages).
On the real site, create its own sitemap.xml template and route (SITEMAPS.md).
License
MetaCraft is commercial software, licensed per production website: see LICENSE.md. It includes Lucide icons under the ISC License (LICENSE-lucide.txt).
Development
metacraft-dev/
├── composer.json, phpunit.xml PHPUnit 11 (dev only; vendor/ is git-ignored)
├── ee-run.php boots EE to run a script: ddev exec php metacraft-dev/ee-run.php <script> [args]
├── docs-site/ builds docs/site/ from these Markdown files (build.php) and its screenshots (shots.mjs)
├── js/ JS/PHP title-counter parity check
└── scripts/ fixtures, integration checks, query counts, perf
- Unit tests:
ddev exec "cd metacraft-dev && vendor/bin/phpunit" - Integration checks:
metacraft-dev/scripts/integration-*.php,access-check.php,bulk-stale-check.php - Query count for one head tag:
metacraft-dev/scripts/head-tag.php /some/uri/ --sql - Audit / template scan from the CLI:
scripts/audit.php,scripts/template-scan.php - Suggestions end to end with a fake AI provider:
scripts/suggestions-check.php; the AI gate:scripts/ai-gate-check.php - User guide site: edit the Markdown, then
php metacraft-dev/docs-site/build.php(pages, links and figures are set indocs-site/pages.php; the build fails on a broken link). Screenshots: seedocs-site/shots.mjs. - UI checks:
scripts/suggest-ui-fixture.php setup|teardown(stored page for entry 52) andscripts/ai-ui-toggle.php on|off(AI on with a fake key; stub requests in the browser)
The project repository is whitelist-only. It tracks docs/metacraft/, metacraft-dev/ and this add-on.
Before the first deploy from this project, remove the local test content:
- template group
metacraft_test(and with it thesitemap.xml,sitemap-N.xmlandmc-test/…routes):ddev exec php metacraft-dev/ee-run.php metacraft-dev/scripts/templates-phase1.php --remove - channels
mc_test,mc_test_off,mc_struct_pages,mc_struct_listingand their entries (CP → Channels). Structure drops its rows for those channels. - fields
mc_test_*and the field groupmc_test_group.