Fix: with “Card insertion point: section end”, the suggestion preview now shows exactly where the link will land, and applying uses the very same position — previews and results can no longer disagree.
Fix: “section end” placement now respects lists, CTAs, summary sections and existing link cards instead of occasionally dropping a card mid-flow or at the very end of the article.
Fix: classic-editor content without explicit paragraph tags (plain text separated by blank lines, or old-style text with double ) is now split into real paragraphs for placement, so cards land between paragraphs instead of after whole sections.
Fix: headings inside Group/Columns blocks are now recognized as section boundaries.
Fix: the highlighted anchor phrase in suggestion previews now matches reliably on Japanese and mixed Japanese-Latin text (Unicode-normalized, case-insensitive matching).
Fix: the “Leave a review” button no longer renders as low-contrast accent-on-dark (CSS specificity hardening).
New: a “How to use” guide is available from the header on every screen, plus a one-line legend above the suggestions list.
New: suggestion previews for link cards show the sentence before and after the exact insertion point.
Improved: link-format settings only show the fields relevant to the selected insertion method.
Improved: on new non-Japanese installs the default insertion method is now an inline text link (Japanese installs keep link cards).
(Pro) Fix: saving unrelated settings can no longer silently switch AI suggestions off when the two AI flags had drifted apart.
(Pro) New: the Gemini API key is validated on save (format check plus a live call), the key field is protected from browser password auto-fill, and the UI now says clearly when a stored key can no longer be decrypted (for example after a site migration).
(Pro) Improved: the LLM model is a two-option choice (accuracy: gemini-3.5-flash / speed and cost: gemini-3.1-flash-lite); the embedding configuration is pinned to the supported setup and no longer user-editable.
(Pro) Improved: connection-test errors are shown in Japanese on Japanese-language sites.
0.9.6
Fix: the “Leave a review” button could render as terracotta underlined text instead of a button; it now renders consistently.
Improve: the review prompt now appears after you apply a change (the Suggestions tab) rather than on the dashboard.
Tidy: the keyword-cluster view maps by article title and notes where the optional Pro query inference would help.
0.9.5
New: link suggestions now also work on SQLite-based WordPress (for example, the in-browser WordPress Playground) via a portable fallback used only when MySQL FULLTEXT search isn’t available. Behaviour on standard MySQL/MariaDB is unchanged.
0.9.4
Fix: changing which post types are indexed now takes effect properly. Previously, de-selecting a post type and re-indexing left the old type’s posts in the analysis, graph and orphan views; they are now removed when you save the setting or run a full re-index.
New: the first-run screen now lets you choose which post types to index before the first index, so you start with the right content instead of having to fix the scope afterwards.
Fix: the getting-started checklist no longer tells you to “approve” or “review” suggestions before any exist — its third step now guides you to generate your first suggestions.
0.9.3
Clearer listing name in the plugin directory (“Internal Linking & Orphan Page Finder”) so the right people can find it. No functional changes.
0.9.2
Clearer plugin-directory listing (tags and description) so the right people can find it. No functional changes.
0.9.1
Onboarding: the quick-start steps are now clickable — each jumps to (or runs) its action — and show your progress.
New: hover tooltips that explain the dashboard metrics, the link-health factors, and suggestion details (direction, relevance) in plain language.
The review request now appears after a few applied links, so it lands at a more natural moment.
Fixed the in-product “Learn about Pro” link and added a contextual pointer in the suggestions queue.
Small UX polish.
0.9.0
Onboarding: a dismissible 3-step quick-start checklist on the dashboard.
A gentle, one-time review request after your first applied change (dismissible).
Clearer in-product pointers and small UX polish.
0.8.0
Renamed to Relinka for Internal Linking.
Now a fully free, fully functional internal-linking toolkit: link analysis & visualization, keyword cluster graph, internal-link health score, broken-link finder, and relevant link suggestions — all running locally with no account, API key, or external service.
Suggestions are generated on your own server from your content; nothing is sent off-site.
Card styles are now enqueued via the standard WordPress style API.
Hardened REST permission checks on link-data, revert, and clear-suggestions endpoints.
0.7.19
Fix: link-card placement is now drift-proof — if a card was added above a pending suggestion, the suggestion re-anchors to the right spot instead of landing a paragraph too high. The preview shows exactly where the card will go.
Fix: cards are no longer inserted in the middle of a bulleted/numbered list or between a lead-in and the list it introduces; a list at the end of a section is still a valid spot.
Internal: placement geometry is now shared by suggestion generation, preview, and apply, so all three always agree.
Pricing: the free monthly AI cap is now 15 (one AI run per article/direction). Non-AI suggestions remain unlimited; Pro lifts the cap.
0.7.18
AI: more suggestions per article — the inbound shortlist no longer over-filters, so several genuinely relevant articles can be proposed to link to a page (up to your per-post limit), not just one.
Cards: when “use your theme’s card” is selected, a bare internal link no longer falls back to WordPress’s default post-embed box — the theme renders its own card (or a clean link).
Fix: titles containing apostrophes/quotes now display correctly in the admin (no more “’”).
Pricing: the free monthly AI limit is enforced consistently (1–30); the unlimited tier is the Pro upgrade.
0.7.17
AI: smarter anchors — suggestions now use natural, descriptive multi-word phrases taken from your text, instead of single generic words.
AI: more relevant keyless (no-API-key) suggestions, with a clearer shared-topic reason for each.
Improved English-language placement guards and noindex handling for All in One SEO.
Pricing: the free monthly AI-suggestion allowance is now 30 per month.
Foundation for an optional Pro tier (unlimited AI). The WordPress.org build remains fully free; Pro is opt-in and never required.
0.7.14
New: theme-agnostic link cards — three designs plus an option to use your theme’s own card, with a live preview.
New: internal-link health score (0–100) with one-click fixing of orphaned pages.
New: keyword cluster graph, now with a smooth, non-blocking layout on large sites.
Safety: noindex pages are excluded, and page-builder posts (Elementor, Divi, WPBakery, Beaver, etc.) are protected from card insertion.
Safety: rel / nofollow and open-in-new-tab controls for inserted links.
AI: opt-in and off by default; clear data-flow disclosure; monthly free-AI usage counter; built-in guide to get a free Google Gemini key.
AI: automatic retry on rate limits and clearer quota/error messages.
Performance: faster, lighter indexing via batched embeddings.
UX: major admin refresh — clearer dashboard, accessible tabs and dialogs, full keyboard support, plain-language labels, and a modern flat/rounded design.
0.6.0
Section-aware semantic placement and link distribution.
Batched, lower-dimension embeddings for faster, lighter indexing.
