Newsroom OS – Discover & AI-Visibility OS for WordPress Publishers

تبدیلی لاگ

1.8.0

• RELEASE: Major public release since 1.7.4 — bundles Discover Score v2, Breaking News Mode + live coverage tools, Author Performance Discover layer, Classic Editor stability fixes, duplicate tag guard fix, and readme/positioning refresh (see 1.7.5–1.7.12 entries below for detail).
• DOCS: WordPress.org readme — Stable tag 1.8.0, dedicated Classic Editor & Gutenberg section, FAQ, installation step, and plugin tags for block/classic editor discoverability.
• SAFE: No database migrations. Existing posts, tags, schema, tasks, Focus/Discover scores, and settings stay intact. Safe direct upgrade from 1.7.4+ on small and large sites.

1.7.12

• FIX: Classic tag duplicate guard — AJAX check used invalid assign_terms capability (always denied for Authors/Editors), so the UI silently added tags without the “keep or use existing?” prompt. Fixed taxonomy capability check; classic script no longer bypasses on 403. Improved typo detection (e.g. NEWSOS vs NEWSSS). Classic script also loads when the classic Tags metabox is shown alongside the block editor.

1.7.11

• NEW (Pro): Author Performance — Discover readiness layer — per-writer avg Discover Score, up to 3 coaching weak spots (16:9 hero, headline, topic, OG, author profile), Discover column on cards, and Focus + Discover per article in drill-down. Uses saved _newsai_discover_score / breakdown meta from the editor (Phase 2 GSC traffic stays separate).

1.7.10

• CHANGE: Discover Score v2 aligned with Google Discover best practices: hero 1200px+, 16:9, 300k+ pixels, descriptive alt (35 pts); headline 55–70 chars with anti-clickbait penalties (25 pts); softer timeliness/trends (10 pts); saved OG image from Yoast/Rank Math when post is saved (5 pts). Not scored in-editor: max-image-preview:large, Core Web Vitals.

1.7.9

• DOCS: Positioning & readme refactor — hero “Discover & AI-Visibility OS”, five-pillar narrative (Discover / Semantic / Workflow / AI / Knowledge), clearer free vs Pro story; updated screenshot list and plugin short description. No feature code changes.

1.7.8

• FIX: Classic Editor crash — Maximum call stack size exceeded in blocks.min.js when Breaking/live-update code called Gutenberg APIs (getEditedPostContent, createBlock, wp.data.subscribe) on Classic Editor screens. Block APIs now run only when block-editor-page + core/block-editor store are present; Classic uses TinyMCE/textarea only.

1.7.7

• NEW: Live coverage editor tools — collapsible “How to write live coverage” guide, Insert live update (H3) button (Gutenberg heading block or Classic insert), and a warning when Breaking is active but the post has no H2/H3 headings (live liveBlogUpdate needs structured headings).
• FIX: LiveBlog schema — no more fallback liveBlogUpdate built from the first 24 words when Breaking/LiveBlog is on or when content looks like a pasted plugin readme; avoids bogus updates on announcement posts.
• i18n: Greek, English, German, and Spanish catalogs updated for Discover Score, Breaking News Mode, and new editor strings (PHP .po + JS newsai-script.json).

1.7.6

• NEW: Breaking News Mode — replaces the editor “LiveBlogPosting” checkbox with a unified workflow: enable breaking coverage (LiveBlog schema + liveBlogUpdate), optional “developing story” badge on the article, IndexNow / Google ping on enable, Resolve now, and auto-expire (24/48/72h, hourly cron). AJAX toggle works in Gutenberg; classic editor uses the same metabox save path.
• NEW: Settings → Schema: breaking auto-expire hours and frontend badge toggle.

1.7.5

• NEW: Discover Score — separate from Editorial (Focus) score: hero image width/aspect/alt, cached regional trend token fit (daily refresh + on-demand when empty), entity/tag/category/geo cues, author E-E-A-T fields, internal links / storytelling; headline subscore (50–65 char sweet spot, numbers, power chars, recency cues, specificity) merged in the editor. Publication Readiness meta box: twin gauges, Editorial / Discover tabs under Score Details, combined Next Best Action across both tracks. Postmeta: _newsai_discover_score, _newsai_discover_score_breakdown. Filter: nros_discover_score_weights.
• NEW: includes/discover-score.php — AJAX newsai_discover_score_partial, newsai_save_discover_score, cron nros_refresh_trend_keyword_cache (cleared on plugin deactivation).

1.7.4

• FIX: JSON-LD dates use site local time — datePublished, dateModified, dateCreated, LiveBlog coverage times, and video uploadDate now output ISO 8601 with the WordPress timezone offset (e.g. 2026-05-16T09:45:00+03:00) instead of UTC (+00:00), aligning with wp-admin display and Rank Math Open Graph.
• FIX: Site Health REST false alarm — removed a UTF-8 BOM (3 bytes) at the start of includes/admin-ui.php that WordPress reported as charsout=3 on activation and that could prefix REST loopback responses so Site Health showed “The REST API did not process the context query parameter correctly” even when the endpoint returned HTTP 200.
• FIX: REST-safe i18n — on /wp-json/* requests the plugin no longer reloads its text domain or overrides plugin_locale for newsroom-ai-assistant, so core REST and Site Health checks stay stable while the per-site language picker (1.7.1+) continues to work in wp-admin.
• FIX: Per-site UI language picker — restored the 1.7.3 gettext loading path (load_textdomain with determine_locale() lookup) after an experimental init-only loader caused the dashboard to stay in English regardless of Settings → Language.
• FIX: Task assignment dropdown uses capability => edit_posts instead of deprecated who => authors (WordPress 5.9+).
• IMPROVEMENT: the_content / timeline shortcode and speakable snippet bail out during REST/JSON requests (no mutation of API payloads).

1.7.3

• IMPROVEMENT: Author E-E-A-T split (posts vs profile) — single posts (Balanced & Full presets) now embed only a pointer-style Person in @graph (@type, @id, name, url, avatar image) with NewsArticle.author as { "@id": "…/author/name/#author" }. Full Person markup (description, jobTitle, sameAs, alumniOf, knowsAbout) is emitted only on the author archive (ProfilePage + full entity + mainEntityOfPage), matching the same #author @id so crawlers resolve one canonical author entity per writer.
• IMPROVEMENT: thumbnailUrl uses lead-sized images — prefers full, 2048×2048, 1536×1536, large, medium_large (widest available among those), so the URL is not the small thumbnail crop. Falls back to the attachment file URL, then to the previous widest-candidate logic. Image candidates also accept URLs when WordPress returns 0×0 dimensions in the src tuple, filling width/height from attachment metadata when possible.
• IMPROVEMENT: NewsArticle.image avoids 150×150 (and similar micro-crops) winning aspect slots: only variants whose larger side is ≥ 600px compete for 1:1 / 4:3 / 16:9 picks, with a stronger width penalty so wide lead images beat perfect-ratio thumbnails.
• ADD: nros_attachment_image_src_dims() — fills missing featured-image width/height from attachment metadata when wp_get_attachment_image_src returns zeros; nros_schema_parse_speakable_selectors() — normalizes bare class tokens and space-separated classes on one element (e.g. entry-content clear → .entry-content.clear) for valid Speakable cssSelector values.
• IMPROVEMENT: SpeakableSpecification — speakable is now emitted on single posts for Balanced and Full presets when CSS selectors are configured (still off for Minimal). Bare tokens like entry-title are normalized to .entry-title in JSON-LD. Settings preview mirrors this; a warning appears if selectors are set but the preset is Minimal. Saving other options no longer clears speakable when the field is absent from a partial POST.
• IMPROVEMENT: Speakable on WebPage too — the same SpeakableSpecification is attached to both NewsArticle and WebPage in @graph (Google’s primary example uses WebPage). Developers can adjust selectors with nros_schema_speakable_css_selectors (receives array $selectors, int $post_id, WP_Post $post).
• IMPROVEMENT: Speakable concise lede snippet — optional visible paragraph (class .nros-speakable-snippet, default 400 characters from excerpt or trimmed lede) prepended on single posts; the selector is auto-prepended to SpeakableSpecification.cssSelector when enabled (Settings → Schema & SEO). Filters nros_speakable_snippet_text and existing nros_schema_speakable_css_selectors allow overrides.
• FIX: Speakable max characters vs. full article in validators — the character cap applies only to .nros-speakable-snippet; selectors like .entry-content still matched the whole article in the DOM, so Rich Results Test could show huge repeated text. With the snippet enabled, the engine now omits common wide-body selectors from cssSelector by default (filter nros_speakable_body_selector_substrings). New checkbox omit wide body selectors (default on). nros_speakable_finalize_schema_selectors() centralizes this for JSON-LD and the settings preview. Snippet-only mode: snippet on + empty Speakable CSS field still emits .nros-speakable-snippet and prepends the lede (Balanced / Full).
• IMPROVEMENT: Schema JSON-LD transients bumped to v154 (busts cached @graph after Speakable selector changes; supersedes v153 keys).

1.7.2

• NEW: Schema richness presets (Minimal / Balanced / Full) under Settings → Schema & SEO.
  • Minimal — no separate tag entity graph (about / mentions / keywords), no video object, no speakable; author is a lightweight inline name + URL only; still uses the slim publisher on posts (below).
  • Balanced — tag entities run with caps (about ≤ 1, mentions ≤ 3, keywords ≤ 8); slim Person in @graph (drops alumniOf / knowsAbout and short bio); video allowed; no speakable.
  • Full — same richness as before (full Person, all tag entities, video, speakable when configured). Single posts still use the slim publisher node so VAT/legal is not repeated on every URL.
• NEW: Article body in JSON-LD — choose omit, full post text, or first N words only (N clamped 30–2000, default 200). Cuts duplicate payload on long articles while keeping headline, dates, images, and other core NewsArticle fields.
• IMPROVEMENT: Organization (VAT, legal ID, full address, contact, etc.) is output as the full Organization node on the homepage (and blog index / custom feed cases handled by the engine). Category archives, tag archives, author archives, and single posts add a slim publisher in @graph: same @id as the home organization (#organization on the site URL) with @type, name, url, and logo (with dimensions when known). NewsArticle.publisher keeps pointing at that @id so knowledge-graph linking stays valid site-wide.
• IMPROVEMENT: NewsArticle.image is a list of Schema.org ImageObject entries (url + width + height when WordPress provides them). The engine samples registered sizes (full, large, medium_large, medium, thumbnail, 1536x1536, 2048x2048), de-duplicates by URL, then outputs up to three images tuned for 1:1, 4:3, and 16:9 aspect ratios (Discover-oriented), not every intermediate file size.
• IMPROVEMENT: Schema JSON-LD transients version bumped to v146 so cached markup refreshes after upgrade without manual cache flush.

1.7.1

• NEW: Full multilingual UI (Greek, English, German, Spanish) — the entire plugin (settings tabs, dashboards, sidebar, editor strings, email subjects, error messages) now flows through native WordPress gettext under the newsroom-ai-assistant text domain. Includes compiled .mo catalogs and wp_set_script_translations JSON for the editor JS, so writers see consistent translations from the dashboard through to the Gutenberg sidebar.
• NEW: Per-site UI language picker — Settings → Site Details → Language. Switch between EL / EN / DE / ES independently of the global WordPress locale, perfect for international newsrooms where the chief and the writers prefer different UI languages.
• IMPROVEMENT: WordPress 6.7+ Just-In-Time textdomain loader compatibility — the plugin now loads its translation catalog directly with the resolved request locale (determine_locale()), so __() / _e() calls hit the right bucket on the new WP_Translation_Controller. Previously, the picker could silently fall back to English on admin screens served under a different user-locale than the site locale.
• IMPROVEMENT: Predictable script translation file routing — JS catalogs now resolve to readable filenames (newsroom-ai-assistant-{locale}-newsai-script.json) instead of the default md5-of-relative-path naming, so packaging translations no longer breaks if the plugin folder is renamed (e.g. newsroom-ai-assistant vs newsroom-ai-assistant-pro).
• FIX: Instant Indexing “Test now” false negative — the IndexNow self-test no longer reports “Key file unreachable (HTTP 200)” when the key file is actually served correctly. Strict body equality was tripping on UTF-8 BOMs and minor whitespace added by some CDNs / caching layers. The verifier now strips BOM and accepts any 200 response whose body contains the configured key. The real Bing/Yandex submission was always working — only the in-dashboard verification was lying about it.
• FIX: Focus Score never drops on republish — when a post transitions to publish (or is re-published after an edit), the pinned analytics score is now max(previous_pinned, current_live). Earlier behavior could overwrite a higher historical score with a lower mid-edit value, deflating author averages on the Performance dashboard. Live, in-editor saves remain unchanged.
• FIX: Editor sidebar rendering bug — a stray ?> sequence inside a comment block on certain PHP parsers was causing the Newsroom OS meta box to print raw PHP into the Gutenberg sidebar. Sidebar now renders cleanly on all PHP 7.4 – 8.3 builds.
• FIX: Decision Engine score persistence honors the 10-second debounce on the client and an idempotent server write — repeated saves of the same value no longer bust the Author Performance transient cache (was causing unnecessary re-aggregation queries on long edit sessions).
• IMPROVEMENT: Pro license activation now surfaces the real reason when the NewsOS API rejects a key — the previous flow swallowed the response and printed a generic “License validation failed.” The error now includes the HTTP status and a snippet of the server response (e.g. License validation failed (HTTP 404). Server response: rest_no_route), so license issues can be diagnosed without enabling WP_DEBUG.
• IMPROVEMENT: License response parser tolerates UTF-8 BOM and leading whitespace before the JSON payload — previously, a license server whose PHP files were saved with a BOM (a very common WordPress packaging mistake) would silently fail every activation because json_decode() returned null on the BOM-prefixed body.

1.7.0

• NEW (Pro): Author Performance dashboard with per-writer cards, inline expandable details, top recent articles + categories, 7d/30d/90d live AJAX toggle, personal “Your own performance” card for chiefs.
• NEW (Pro): Focus Score persistence — the client-side Publish Confidence is now saved to postmeta (_newsai_focus_score) on a 10-second debounce and pinned at publish time (_newsai_focus_score_at_publish) so analytics aren’t skewed by later edits.
• NEW (Pro): Groq AI provider with free-tier support (no credit card required) — drop-in next to OpenAI and Gemini, using llama-3.3-70b-versatile by default.
• NEW (Pro): Discover-optimized Title Suggestions — five headlines per run, each in a different editorial angle (facts / impact / aftermath / authority / Discover hook), strictly 55–70 characters, with live char counter + angle badges in the editor sidebar.
• NEW (Pro): Geo-aware AI language — AI prompts now answer in the correct language for the configured geo target (50+ locales: GR, BR, MX, ES, JP, IL, AR, DE, FR, IT, PL, RO, TR, RU, +more), even on aggressive English-leaning models. New filter newsai_ai_geo_to_locale for custom multilingual setups.
• NEW: Chiefs can now also access “My Inbox” via a dedicated submenu, so they can operate in writer-mode for their own assignments without losing the chief dashboard.
• NEW: Personal performance card inside “My Inbox” for every writer (Pro) — posts, tasks, avg Focus Score, trend, top categories.
• IMPROVEMENT: Trend Radar fallback uses stable Google News topic RSS first, then localized search query — fixes empty feeds on EL/BR/MX installs.
• IMPROVEMENT: Capability-based access tightened. “Chief” now strictly means edit_others_posts everywhere (menu, dashboard widget, AJAX endpoints, email notifications) — peer writers can no longer leak into each other’s threads.
• IMPROVEMENT: Gemini default model upgraded to gemini-2.5-flash with transparent auto-migration of deprecated 1.5-* model names.
• IMPROVEMENT: AI Settings page now links directly to OpenAI / Gemini / Groq API-key dashboards with one-line guidance — including the free Groq + free Gemini paths so writers can start with zero spend.
• IMPROVEMENT: Pro purchase CTAs across the plugin (banner, lock cards) point to newsos.io and carry UTM tracking.
• FIX: Author Performance expanded view now spans the full grid row with a clean article list (no more clipped tables on narrow screens).

1.6.0

• Editorial Intelligence: Publish Confidence + Next Best Action gauge, refined Auto-Timeline rules, soft profile presets.
• Wizard copy updated to reflect new settings model.

1.5.0

• Schema Engine v3 (Knowledge Graph build-out) and full E-E-A-T author profile support.
• Smart Tag Suggester rebuilt with confidence levels and phrase-aware matching.