Internal Linking Engine

Turn on Internal Linking Engine for one site already known to Sites Registry. Creates the site's settings row in review_first mode by default -- no content is touched until a plan is previewed and explicitly applied.

Update an existing site's Internal Linking Engine settings (mode, limits, CTA targets, exclusions, languages). Only given fields change.

List every site with Internal Linking Engine settings (enabled or not), across every connected site.

Read one site's Internal Linking Engine settings in full.

Cache metadata for a batch of posts already fetched via wordpress-hub, so the Relevance Engine can score them without re-fetching. Call this once per post before get_relevant_targets/preview_internal_links can use it as a candidate or source.

One-glance index health for a site: how many posts indexed, by language, and when last indexed.

Find the top-N internal-link targets for one already-indexed source post, using deterministic metadata + term-overlap scoring (categories/tags/product_type/title). Language isolation is a hard filter -- a candidate whose lang differs from the source post's lang is never returned, even if it scores highest (the climtec.md RU/RO lesson). Excluded URLs from the site's settings are also dropped before scoring.

Build a preview linking+CTA plan for a site -- NEVER writes anything. Takes raw content Webbee already fetched via wordpress-hub.get_post_content, and (optionally) anchor/CTA insertions Webbee already drafted via an LLM pass, and validates every proposed insertion's find_exact_substring occurs EXACTLY ONCE in that post's current raw content -- any insertion that doesn't is dropped, never guessed. Returns a plan_id; nothing is applied until apply_internal_links is called after Webbee has actually written each surviving diff via wordpress-hub.replace_post_content_text.

Read one linking plan in full -- every post's proposed diff, for review before applying.

List linking plans, optionally filtered by site and/or status.

Record that a pending_review plan's diffs were actually written to the site -- call this ONLY AFTER Webbee has already called wordpress-hub.replace_post_content_text for each surviving insertion in the plan (using expected_state_token per post for safety). This function does not write to any site itself; it only transitions the plan's own status and increments the site's confirmed_applies_count toward its full_auto threshold.

Reject a pending_review plan without applying it. Nothing was ever written, so this is purely a status change.

Roll back an already-applied plan: returns the original find/replace pairs (the inserted anchor/CTA text as 'find', the pre-insertion original snippet as 'replace') for Webbee to write back via wordpress-hub.replace_post_content_text, restoring each post's content. This function does not write to any site itself -- it only returns the rollback instructions and transitions the plan's own status once Webbee confirms.

List the linking runs dashboard -- one row per plan lifecycle event, newest first.