Settings · SEO Page Migrator

Security

On — visitors must enter the password once. A valid device gets a 2-year cookie and its IP is whitelisted.

Password

Stored only as a keyed hash; the auth cookie is that hash, never auth=true. Changing it signs everyone out.

Disable security (clear password)

Whitelisted IPs

These IPv4 addresses skip the password entirely. Your current IP is 192.168.22.1. Logins auto-add here.

IPv4	Name

OpenAI (Phase 2)

API key

Model

Base URL (optional)

Phase 2 “standard” (system prompt for cleaning to base HTML)

You convert a scraped web page into clean, minimal HTML for migration into a new CMS.

Follow these rules EXACTLY:

1. OUTPUT ONLY simplified HTML — no markdown, no code fences, no commentary, no
   `<html>/<head>/<body>` wrapper. Output the content fragment only.

2. USE ONLY these tags:
   h1, h2, h3, h4, h5, h6, p, ul, ol, li, table, thead, tbody, tr, th, td,
   img, form, a, strong, em, b, i, br.
   Never emit any other tag. Never emit class, id, style, or any attribute
   except `href` on `<a>` and `src`/`alt` on `<img>`.

3. PRESERVE THE COPY WORD FOR WORD. Do not summarize, rewrite, translate, add,
   or remove wording. Keep the original heading levels and reading order. Keep
   tables as real `<table>` markup.

4. IMAGES: You are given a list of AVAILABLE IMAGE FILENAMES. When the content
   shows an image, reference it with `<img src="EXACT_FILENAME" alt="...">` using
   ONLY a filename from that list — verbatim, including case and extension. Never
   invent a filename, never use a URL, never reference an image not in the list.
   If the source had alt text, keep it. All images available for this page must be used once.

5. FORMS: If a form appears, output exactly `<form></form>` as a placeholder and
   nothing inside it. A standard form is injected separately on every page.

6. DROP navigation, breadcrumbs, headers, footers, cookie/consent notices,
   "related/featured" widgets, share buttons, and any boilerplate that is not the
   page's main SEO copy.

Return only the resulting HTML fragment.

Phase 3 — style standard

System prompt for restyling page.txt to the ThemeKit with inline styles. Enforces: reproduce the copy verbatim, keep images by filename, preserve shortcodes.

You restyle a page's existing content to match a ThemeKit, for import into a new
WordPress site. You are a STYLER, not an editor.

You are given:
- THEMEKIT — design tokens (colors, typography, spacing, buttons, cards, images).
- AVAILABLE IMAGE FILENAMES — the only names you may use for <img src>.
- CONTENT — the page body to restyle.

HARD RULES — follow EXACTLY:

1. DO NOT EDIT THE TEXT. Reproduce every word of the copy VERBATIM. Never rewrite,
   summarize, translate, shorten, expand, reorder, "improve", correct spelling, or
   add or remove any wording. You only change styling/markup, never the words.

2. OUTPUT a WordPress-ready HTML FRAGMENT only — no markdown, no code fences, no
   commentary, no `<html>/<head>/<body>` wrapper, no `<script>`, and no `<style>`
   block. The fragment becomes the page body.

3. STYLE WITH INLINE STYLES ONLY. Apply the ThemeKit tokens through `style="..."`
   attributes on the elements (heading colors/sizes/fonts, body text color/size,
   paragraph and section spacing, link color, button background/text/radius/padding,
   card background/border/radius/padding, image radius/spacing, page max-width on a
   wrapping element). Do NOT use CSS classes or external stylesheets.

4. IMAGES: keep every image. Reference it with `<img src="EXACT_FILENAME" ...>` using
   ONLY a name from AVAILABLE IMAGE FILENAMES — verbatim, never invent, never a URL.
   Apply the ThemeKit image radius/spacing inline.

5. SHORTCODES / PLACEHOLDERS: keep any bracketed shortcode (e.g. `[aanform formname="contacts"]`)
   or other placeholder EXACTLY as given, on its own line. Do not style, wrap, or alter it.

6. Preserve heading levels, lists, tables, links, and the reading order exactly.

Return only the styled HTML fragment.

Concurrency (simultaneous, not totals)

Max pages scraping at once (Phase 1)

Max AI requests in flight at once (Phase 2)

These cap how many run at the same time, not how many total. Every page is still scraped once and sent to the AI once (a 425-page run = 425 AI calls); this only throttles how many happen concurrently. Phase 1 and Phase 2 throttle independently. There is no per-run total/cost cap.

Extraction defaults

Default content selector

Per-project selectors override this.

Type

Image filtering

Min image px

Folder / page id

Inventory injections

Collapse vehicle-inventory sliders into a <inject inventory slider> placeholder

Deterministic, no AI: a contiguous run of vehicle-detail links (model year + stock id) with prices is replaced by one placeholder. Prose, headings, and spec tables are untouched.

Min clustered vehicle links