BoostMedia AI Content Generator

Changelog

2.1.11

• Fix: LiteSpeed Cache deep compatibility — uses official LiteSpeed PHP API (litespeed_control_set_nocache) to prevent private cache from serving stale REST responses

2.1.10

• Fix: REST API responses now include no-cache headers to prevent LiteSpeed Cache and other caching plugins from serving stale data

2.1.9

• Fix: PHP timeout on shared hosting during AI link analysis and image generation (set_time_limit extended for long-running backend calls)
• Fix: Link analysis batch size reduced from 10 to 3 to stay within web server connection timeouts (LiteSpeed, Nginx, etc.)

2.1.8

• Fix: Content analysis table migration — added missing created_at column for existing installs
• Fix: Field pattern analyzer handles nested array meta values without warnings
• Fix: Content Types analysis display renders field objects correctly (was showing React error #31)
• Fix: Gemini output token limit increased for Technical QAs (8K → 32K) — sites with 46+ meta fields no longer hit truncation
• Fix: Smart retry on token limit — automatically doubles token budget on retry instead of repeating with same limit
• Fix: Specific error message when AI response is too large (instead of generic “format issue”)

2.1.6

• New: External links bulk select checkboxes (matching internal links UX)
• New: Expandable summary component — click to expand/collapse long summaries
• New: Bulk delete endpoint for external links
• Fix: External links table shows title, summary, keywords, AI status columns
• Fix: Deploy path consistency across sites

2.1.5

• New: External links AI enrichment — paste a URL, AI scrapes and generates title, summary, keywords
• New: External links table redesigned to match internal links (title, summary, keywords, AI status)
• New: “Complete with AI” button for external links with progress bar
• New: Polylang/WPML language auto-assigned on generated posts (no manual assignment needed)
• New: Internal plugin fields (Polylang, WPML, Wordfence, etc.) excluded from Technical QA scanning
• New: Links table per-page selector (50/100/200/500) with Select All checkbox
• New: Premature ready guardrail — AI must show confirmation card before completing Technical QA
• Fix: Reporter gender enforced in article generation (masculine/feminine Hebrew grammar)
• Fix: Question card answers use gendered singular (not plural) matching reporter gender
• Fix: Multilingual prompt pipeline — 9 fixes: grammar instruction scoped to Hebrew only, binary language normalization replaced with proper map, single language source of truth, topic examples scoped, sprint/rewrite language aligned
• Fix: Shared language utility (bc-backend) — normalizeLanguageCode, getLanguageLabel, isHebrewLanguage
• Fix: Link collection rules displayed correctly (dirty ref bug from Wave 3)
• Fix: Empty link rules blocked from collecting (was silently collecting everything)
• Fix: Content Language setting saves correctly (was resetting to Auto)
• Fix: term_scope column expanded VARCHAR(200) → TEXT (Hebrew category slugs overflow)
• Fix: App icon loads via Vite import (no more 404)
• Fix: Production build uses type=”module” for ES module syntax
• Fix: bcSettings injected via wp_add_inline_script (no more & encoding issues)
• Fix: Copyscape rewrite language uses reporter writing language (not site language)
• Fix: generation.worker copyscape language uses shared helper
• Fix: PHP normalize_data_language expanded to support ar/fr/ru codes

2.1.4

• Fix: Reporters page no longer refetches list on every row click
• Fix: “All” reporter filter now shows archived reporters with badge
• Fix: AI reporter builder — save button always visible, close warns about unsaved progress
• Fix: Reporter builder session null safety after failed create
• Fix: Removed unused set_default method (cross-language risk)
• Fix: Cancel while creating reporter always exits create mode
• Fix: Specialization tag keys no longer collide
• Fix: Logs route hidden when debug mode is off (redirect to dashboard)
• Fix: Logs query limit capped at 500 rows
• Fix: Log status ‘submitted’ mapped to valid ENUM ‘success’
• Fix: Clear logs shows error on failure (was silent)
• Fix: Dashboard recent activity handles unknown action types gracefully
• Fix: API error messages shown correctly across all pages (was “Unknown error”)
• Fix: Content plan delete checks database result
• Fix: Content plan post_length guarded against null
• Fix: Update install requires explicit target version
• Fix: Download URL allowlist tightened to /downloads/ path prefix
• Fix: Settings validation whitelist for language fields
• Fix: Purchase overlay timer no longer resets on re-render
• Fix: Confetti animation positions stable (memoized)
• Fix: Content Types date display guards against invalid dates
• Fix: Chat poll aborts on page navigation (no stale requests)
• Fix: Save toast timer cleaned up on unmount
• Fix: Preview URL works on subdirectory WordPress installs
• Fix: API nonce refreshed on each request (prevents stale nonce after long sessions)
• Fix: Dashboard help text no longer mentions credit balance
• Verified: Copyscape cost ~2 BoostCoins is accurate for typical articles

2.1.3

• Fix: Onboarding banner now checks correct API field (was showing on sites with generated content)
• Fix: Onboarding banner no longer appears on API errors
• Fix: Dashboard stat cards show “—” on error instead of misleading zeros
• Fix: Generated Content — status counts updated after delete
• Fix: Jobs — cursor-based pagination replaces broken offset system (no more duplicates/gaps)
• Fix: Jobs — sprint child articles always loaded for visible sprints
• Fix: Jobs — “completed today” count includes all terminal statuses
• Fix: Content Plans — “Run” marks plan only after sprint succeeds (was marking before)
• Fix: Create Content — technical save blocked when structure data missing (was silently skipping)
• Fix: Create Content — settings completion protected against double-click
• Fix: Create Content — post type from Content Types page prefills generator
• Fix: Updates — sidebar badge clears after install without page reload
• Fix: Updates — duplicate check requests deduplicated
• Fix: Settings — TypeScript defaults aligned with PHP (sample_count, post_count, copyscape)

2.1.2

• Fix: Generated Content page now has pagination with “Load More” (was limited to first 20 rows)
• Fix: Deleting a generated post also removes the WordPress post (no more orphans)
• Fix: Generated Content stats (total, published, draft) now come from server, not partial page
• Fix: Delete/publish errors shown to user (was silent)
• Fix: Jobs — completed jobs no longer appear as “running” in history (ghost jobs fixed)
• Fix: Jobs — backend returns real total count for correct pagination
• Fix: Jobs — consistent page size prevents overlap and duplicate rows
• Fix: Jobs — refresh resets pagination properly
• Fix: Jobs — failed job title matches job type (generation/sprint/other)
• Fix: Links — taxonomy terms now refresh when switching data language
• Fix: Links — unsaved rule edits preserved on refresh
• Fix: Links — fast language switching no longer causes stale data
• Fix: Links — bulk actions protected against double-click

2.1.1

• Fix: Dashboard no longer shows fake stats on API failure — proper error states
• Fix: Pending publication count now queries real data (was hardcoded 0)
• Fix: Removed fake +12% trend from Content Created card
• Fix: Dashboard footer shows real plugin version (was hardcoded v1.0.0)
• Fix: Content Types — sample posts viewer now reads correct API response shape (was always empty)
• Fix: Content Types — analysis display reads correct response shape
• Fix: Content Types — removed mock data from error catches, errors properly surfaced
• Fix: Usage — “Purchased BoostCoin” card now shows purchased-only credits (was showing combined total)
• Fix: Usage — error state shown on credits fetch failure (was showing zeros)
• Fix: Usage — no more full-page spinner on credit refetch (subtle refresh indicator instead)
• Fix: Create Content — planId without intent now defaults to review (was blank)
• Fix: Create Content — removed references to unrendered ‘generate’ step
• Fix: Create Content — save failure no longer advances wizard (was silently losing data)
• Fix: Create Content — settings error no longer advances to technical
• Fix: Create Content — double-click guard on Intent submit
• Fix: Settings — save blocked when settings fail to load (prevents overwriting with defaults)
• Fix: Settings — double-submit guard on save and reset
• Fix: Delete logs now checks TRUNCATE result
• Fix: “Show all” content types button now navigates to Content Types page

2.1.0

• Security: API key no longer exposed in settings REST response (masked with last 4 chars)
• Security: Changelog HTML sanitized server-side (wp_kses_post) and client-side before rendering
• Security: Download URL restricted to HTTPS only
• Security: Admin warning when WordPress security keys are not configured (weak encryption)
• Security: Debug log warning when OpenSSL is unavailable for encryption
• Fix: Internal links $wpdb->update() format array corrected — keywords now stored as JSON (was silently cast to 0)
• Fix: Content Plan Runner handles missing technical rules gracefully (no more PHP fatal on cron)
• Fix: Settings save guards against overwriting real API key with masked display value

2.0.9

• New: Guided onboarding — 4-step setup flow with pulsing indicators on sidebar tabs
• New: Per-page setup banners guide new users through Usage, Content Types, Links, and Reporters
• New: Dashboard setup progress card shows completed/remaining steps
• New: Settings page pings on Auto-publish and Smart Linking toggles (dismiss on hover)
• New: Help button pings until first click
• New: Per-user onboarding state (tracked via user meta, persists across devices)
• New: Updated help content (“מדריך לעמוד”) for all 10 pages
• Change: Initial BoostCoins increased from 50 to 60 for new registrations
• 35 new Hebrew translations for onboarding and help content

2.0.8

• Fix: Sprint generation no longer returns 400 — intent_state properly mapped to content_plan
• Fix: Play button on Content Plans submits sprint directly → navigates to Jobs page
• Fix: Incomplete plans show edit icon instead of play
• New: Prerequisites guard on Create Content — blocks when no scanned post types or reporters
• New: Sidebar tabs reordered — Dashboard → Setup → Agency → System
• New: Logs tab hidden when debug mode is off
• New: Auto-detect site language on activation (sets working language)
• Change: Default sample count → 10, default post count → 10
• Change: “Working Language” in Settings renamed to “Default working language”

2.0.7

• New: Registration Monitor — daily health check with exponential backoff (self-healing connection)
• New: Management key — secure per-site key enables automatic API key recovery
• New: Deferred registration — 15-second cron after activation (more reliable than sync)
• New: Publishing gates — Copyscape/image checks before auto-publishing
• Fix: Backend registration no longer returns 409 for existing domains
• Fix: Site + credit grant wrapped in single transaction (no orphan sites)
• Fix: Credit API routes now require site authentication
• Security: Management key endpoint for authenticated key rotation

2.0.6

• Fix: Save button stays on current step — no more jumping to Settings (generation counter approach)
• Fix: Editing saved plans shows accordion checkmarks correctly
• Fix: Intent/sprint state restored when reloading plans
• Fix: Sprint exit requires HIGH confidence articles (3+ findings) — no more medium-only batches
• Fix: Server-side confidence validation — model can’t inflate confidence beyond evidence
• Fix: Hebrew question card options use neutral form (not plural “you”)
• Fix: Publishing gates — articles with Copyscape issues or missing images saved as draft
• Fix: Drip scheduling works even without saved plan ID
• Fix: Singular Hebrew grammar for single article (“נוצר מאמר אחד”)
• Fix: “Findings” translated to Hebrew (“ממצאים”)
• Fix: Sprint always saves plan first to ensure valid plan_id
• Improvement: Sprint Think prompt enforces honest confidence assessment
• Improvement: Link analysis debug logging for keyword investigation

2.0.4

• New: Publishing settings per plan — draft, publish immediately, drip daily, or drip weekly
• New: Fully autonomous recurring plans — research, generate, and publish without human intervention
• New: Scheduled plans section in Jobs tab — shows upcoming runs greyed out with next run time
• New: Cron precision — recurring plans checked every 15 minutes (was daily)
• New: Consecutive failure protection — plans auto-deactivate after 3 failed runs, not on first failure
• Improvement: All Hebrew UI buttons use proper שם פעולה (action noun) form — 102 strings updated
• Improvement: Sprint research angles now produced in the reporter’s language (Hebrew for Hebrew plans)
• Improvement: Archived reporters hidden by default — new “Archived” filter to show them
• Improvement: Reporter name shown in progress bar info strip
• Fix: Settings accordion sections can be reopened after completion
• Fix: Reporter section requires explicit confirmation (no auto-advance on mount)
• Fix: Internal links keywords “0” — normalized to valid JSON arrays
• Fix: Technical QA JSON parse errors — multi-strategy recovery + retry

2.0.3

• New: Filter Generated Content by plan — click “View articles” from Jobs tab to see only that plan’s articles
• New: Plan origin badges on job cards — shows one-time vs scheduled
• New: Plan ID shown on generated article cards
• Fix: Settings accordion sections can be reopened after completion
• Fix: Reporter section no longer auto-marks complete without user interaction
• Fix: JSON parser handles missing values pattern from Gemini (“key”:, → “key”:null)
• Fix: “Research Scope” and job status badges translated to Hebrew
• Fix: No more stray “0” rendered in job cards (React conditional rendering fix)
• Fix: Button icons positioned inline with text on job cards

2.0.2

• New: Jobs tab — track all active research sprints and article generation in one dashboard
• New: Active job badges on sidebar navigation
• New: Job history with costs, duration, and per-article breakdown
• New: Run Again / View Articles actions from job history
• Improvement: First 3 wizard steps merged into compact “Settings” accordion
• Improvement: Plan save bar visible on all steps — save new plans directly from the wizard
• Improvement: Step progress bar fits all steps comfortably (7 in sprint mode, 5 in legacy)
• Fix: Old wizard URLs (?step=select, ?step=configure) redirect to Settings

2.0.1

• Fix: Research sprint now runs in the background (non-blocking) — user can navigate away and return
• Fix: Sprint progress shows real-time data (actual round count and findings)
• Fix: Question cards during intent collection show contextual options (not just “Other”)
• Fix: Sprint prompt instructs reporter to always include clickable options
• New: Sprint resume — page refresh doesn’t lose sprint progress
• New: Sprint polling every 5 seconds with live status updates

2.0.0

• NEW: Research-first content generation — articles emerge from iterative research, not pre-planned topics
• NEW: Research Sprint engine — AI reporter explores your field through multiple research rounds, building knowledge iteratively
• NEW: Research Map — persistent tree structure that stores all research knowledge across plan runs
• NEW: Article Formation with Gravitational Model — articles backed by 2+ research findings, validated against user intent
• NEW: Intent Collection step — conversational chat to understand audience, goals, and research preferences
• NEW: Sprint Progress step — real-time research sprint status display
• NEW: Article Review step — review proposed articles with confidence indicators and backing findings before generation
• NEW: Reporter Memory — research maps persist between runs, preventing content repetition
• NEW: Time-aware research branches — branches track freshness and revisit schedules
• NEW: Automatic sprint mode — system adapts research depth based on article count (no user configuration needed)
• NEW: Recurring plans use research sprints with persistent memory across runs
• Improvement: Credit refund on permanent generation failure
• Improvement: Batch generation continues after single article failure with retry option
• Improvement: Recurring plans correctly track generated titles for dedup
• All existing v1.x plans continue to work unchanged

1.10.2

• Fix: Credits are now refunded when article generation permanently fails
• Fix: Batch generation continues after a single article failure (remaining articles still generated)
• Fix: Partial generation success shows per-article breakdown with retry option
• Fix: Recurring plans correctly track previously generated titles to prevent repetition
• Fix: Plan ID stored on generated posts for accurate plan-to-article tracking
• Fix: Recovery cards display in Hebrew (not English) during Hebrew conversations
• Fix: Technical QA auto-advances at 60%+ understanding
• Fix: Recovery cards not forced when AI asks a natural follow-up question
• Fix: Empty question cards auto-populate with Yes/No options
• Fix: New content flows isolated from previous flow context
• Fix: Additional Hebrew translations for captured technical answers
• Fix: “Start over” button styled consistently with plugin design
• Fix: Bulk link action buttons positioned in top toolbar
• Fix: AI instructed to synchronize topic tree prose with JSON data

1.10.1

• Fix: Technical QAs phase no longer gets stuck — auto-ready promotion + confirmation card for Round 4
• Fix: Article angles in topic tree display correctly (title→angle field name mismatch resolved)
• Fix: Pillar and cluster titles normalize from alternative field names
• Fix: Category prompt is conditional — single category skips question, multiple asks about distribution
• Fix: Removed contradicting “no accumulation” subscription text (coins now roll over)
• Fix: Navigation chevrons positioned correctly relative to button text in RTL

1.10.0

• New: Image regeneration with AI-interpreted user notes (Gemini rewrites prompts for Imagen)
• New: Image lightbox with full-size preview and regenerate button
• New: Link collection rules displayed as consolidated tag clouds
• New: Bulk “Remove from all posts” action for internal links
• New: Article count changes auto-notify the AI reporter
• New: Recurring plans generate broader topic angles for content variety
• New: Credit balance correctly displayed in content creation wizard
• New: Technical answers displayed with Hebrew labels
• New: Review step matches Generated Content controls (Edit in WP, Publish, Copyscape badges)
• New: Session persistence — page refresh no longer resets content creation progress
• New: Dynamic year injection — AI reporter always knows the current year
• New: Copyscape enabled by default for new installations
• Fix: Pricing overhaul — correct ILS-based conversion (1 BoostCoin = ₪1) with 3x markup
• Fix: Usage page shows both per-article and per-100-words rate
• Fix: “Play” on existing content plan no longer redirects to empty review page
• Fix: Inline CSS spans stripped from generated content
• Fix: Article slugs limited to 3-5 words for SEO
• Fix: Topics interpreted naturally (no literal subject repetition)
• Fix: Generated images no longer contain text/watermarks
• Fix: Copyscape uses API-reported costs for maximum accuracy
• Fix: JetEngine fields from other post types no longer leak into scans
• Fix: Malformed AI responses trigger recovery cards instead of stuck conversations
• Fix: Category assignment and post_author handled automatically (not asked by AI)
• Fix: RTL chevron direction on navigation buttons
• Fix: Reduced padding in question cards for better readability
• Fix: In-app edit removed from review step (WordPress edit only)

1.9.9

• New: Regenerate article images with custom AI-interpreted notes
• New: Full-size image preview lightbox
• New: Link collection rules displayed as compact tag clouds
• New: Article count changes automatically update the content plan
• New: Broader topic angles for recurring content plans
• New: Hebrew labels for technical settings answers
• New: Full controls in content review step (Edit, Publish, Status badges)
• Fix: Correct credit pricing (1 BoostCoin = ₪1)
• Fix: Content plan “Play” button works correctly for re-generation
• Fix: Clean HTML output (no inline CSS spans)
• Fix: SEO-friendly article slugs (3-5 words)
• Fix: Natural article titles (no literal topic repetition)
• Fix: Images generated without text/watermarks

1.9.8

• New: Switched image generation to Google Imagen 4
• New: Subscription credit rollover (unused credits preserved)
• Fix: Inline CSS stripped from generated content
• Fix: Short SEO-friendly slugs
• Fix: Natural topic interpretation

1.9.7

• New: Copyscape plagiarism checking (web + private index)
• New: Deterministic post-generation internal linking
• New: Per-100-words transparent pricing
• New: Research grounding with retry mechanism
• Fix: All WordPress Plugin Check (PCP) errors and warnings resolved

1.8.0

• Fix: Content planning conversation state improvements
• Fix: Stale question cards properly cleared between rounds

1.0.0

• Initial release
• JetEngine scanning
• Content style analysis
• AI content generation
• Post scheduling