Skip to content
Topicary

Changelog

Product updates and release notes.

0.28.0betaSep 2, 2026

Structure suggestions you can inspect, edit, revert — and an honest content-debt score

  • Every structure suggestion now shows a concrete preview of exactly what accepting will create — you approve specific changes, not a sentence describing them
  • Edit before applying: uncheck individual items or rename things inside a suggestion; the record keeps what you actually applied, not what was originally proposed
  • Accepted suggestions can be reverted — the confirmation lists exactly what will be removed, and nothing else is touched. A reverted suggestion may come back if the evidence still supports it; a dismissed one never does, and dismissed suggestions can now be restored
  • A history of resolved suggestions (accepted, reverted, dismissed) lives alongside the pending list
  • Content debt rescored: quality findings now carry 50% of the weight (staleness 30%, verification 20%) and are severity-weighted — one warning counts 6× an informational note — so a fresh-but-broken topic outranks an old-but-fine one; new topics start at zero, and a verified topic can never score worse than a never-verified one of the same age. Factor bars are labeled with their weight share, and every project was rescored
  • Permission-refused writes no longer fail silently: we audited all 34 write paths in project actions and fixed the 8 that could report success on a write permissions had blocked — worst case, a reviewer-role member's editor saves were discarded while the UI showed saved. A regression test now holds every write path to account
0.27.0betaSep 1, 2026

Demand worklist redesigned: calm, one-glance triage

  • Rows are quiet by default: actions appear on hover or keyboard focus (always visible on touch), removing roughly a hundred persistently rendered controls from a full worklist
  • One view strip — All / To create / To improve, with Later, Snoozed, and Done beside them — and every count is live and honest: the numbers on the pills always sum to what the list shows
  • The toolbar is one row: search, an effort selector (quick wins vs. deep work), and a labeled sort — down from three rows of controls
  • Snoozing is one click and condition-first: the default is "until the evidence changes", so an item wakes when its underlying signal actually moves, with fixed durations available behind the same button
  • Rows lead with a one-line plain-text summary of the issue; raw internal IDs are gone from titles (a broken link now names the deleted topic it pointed to)
  • The worklist now shares the same header, toolbar, and row composition as the rest of the app instead of its own dialect
0.26.1betaAug 18, 2026

Git sync you can trust, plus a reliability sweep

  • Git sync hardening: an editor left open on stale content can no longer clobber newer synced changes; topic locks are respected during conflict resolution; and changes that only moved history no longer show up as false "modified in both places" conflicts
  • Content writes are pinned to the branch you are working on — concurrent work on different branches can no longer bleed into each other
  • Fixed: the AI answer widget was dead on Topicary-hosted documentation sites — a routing bug swallowed its requests; it now works on every published site
  • Validation errors now reach you with their actual message in production instead of a generic failure screen
  • Every delete and other destructive confirmation now reports its failure inline in the dialog with the real reason, instead of closing as if it succeeded
0.26.0betaAug 5, 2026

Demand worklist: provenance, closed loops, and AI that learns from your edits

  • Priority is now a single integer out of 10, and the priority badge is the only color carrier on the list — red and amber mean urgency, nothing else competes for attention
  • Runs of three or more similar items collapse into an expandable group with dismiss-all and undo
  • Every locatable finding links to the exact spot in the topic: click through and the editor scrolls to and highlights the offending passage
  • Drafts created from the worklist carry their provenance, and rejecting every suggestion in a batch reverts it cleanly — with undo
  • Loops actually close: mark items verified or fixed, archive topics (removed from maps and republished) with unarchive, and expand the evidence behind any gap before acting on it
  • New demand signals: reader feedback, review verdicts, and site searches that returned nothing now feed the worklist alongside existing sources
  • Inline AI learns from your decisions: content you accept becomes an approved example that guides future drafts, and suggested actions are ranked by what you have actually accepted — with cold-start guards so sparse data changes nothing
  • Automated content scans now show their run status on the dashboard, so a scan that did not run can never pass for one that found nothing
0.25.0betaAug 5, 2026

Style rules in the editor, doc-test failures in plain language, open signup

  • Style linting runs live in the editor: 6 Vale-compatible rule types (existence, substitution, occurrence, repetition, consistency, capitalization) flag violations as you type — no server round-trip — with a style panel listing every hit
  • Bring your existing Vale configuration: import a .vale.ini and rule packs directly; rule types the in-editor engine cannot run yet are listed as skipped, never silently dropped
  • Doc-test failures now read like writing feedback, not CI logs: the editor shows what the test expected, what the docs said, and where — end to end in writer language
  • Signup is open: create an account directly, no waitlist
0.24.1betaJul 29, 2026

Content health scan: honest, grounded, and no longer failing silently

  • The nightly health scan now records a per-run status — a scan where some checks could not complete is reported as partial, not mistaken for a clean bill of health
  • Aging-topic review now reads the actual content of each topic instead of judging from its title, so its suggestions point at real issues in the text
  • Docs-as-tests now checks internal cross-references (links to deleted or unpublished topics), validates multi-document YAML correctly, and verifies concrete documented API URLs — with a Run checks now button, separated clearly from external CI test ingestion
  • Content debt now scores each topic only on signals about that topic (staleness, findings, verification); project-wide open-gap count is shown as its own context, not folded into every topic's score
  • Structure suggestions now run on a weekly schedule for active projects, not only when triggered by hand
0.24.0betaJul 28, 2026

Honest AI search readiness: evidence-density factor, clearer scoring language

  • AI search readiness gains a seventh factor, evidence density: concrete specifics (exact values, limits, defaults, versions) score above vague quantifiers — the one content edit with experimental support for improving AI citation (Aggarwal et al., GEO, KDD 2024)
  • Scoring language now says exactly what the audit is: a check of the page itself — structure, fetch size, parseability, evidence — not a prediction of whether AI engines will cite you. No black-box visibility numbers
  • The pages-to-improve list is now ranked by observed AI demand: we log AI agent fetches of your .md pages and AI-crawler page views (humans are never logged), so the weak pages AI is actually pulling rise to the top, each with a 30-day demand count
0.23.0betaJun 17, 2026

AI-agent tooling: docs audit tool, paginated llms.txt, GEO fetch-size, webhook + Zapier

  • Free public tool "Are your docs AI-ready?" at topicary.com/tools: a no-signup audit that scores page structure and cross-page consistency (near-duplicate passages, numeric contradictions, terminology drift, broken cross-references)
  • llms.txt and llms-full.txt now split into navigable per-section files on large sites, so AI agents that silently truncate big files still get complete content
  • GEO scoring adds a "fetch size" factor: flags pages too large for an AI agent to read in a single fetch, with a split-the-page fix
  • New topic.published webhook event so external automations can react the moment a topic goes live (now 10 event types)
  • Zapier integration (private beta): Topic Published and Topic Updated triggers, plus Create Topic and Find Topic actions over the REST API
0.22.0betaJun 16, 2026

Content Demand worklist + AI authoring improvements

  • New Demand page: a prioritized "what to write next" worklist combining content gaps, lifecycle recommendations, and content-debt scoring into one ranked list of create / improve actions
  • AI edits are now obvious and reversible: every AI change highlights the affected region, flashes it on apply, and offers an inline Undo (inline AI, AI chat, and AI map organization)
  • AI map organization shows a clear before/after: moved and new topics are marked and highlighted in the tree, with a one-click revert after applying
  • Clearer AI errors: configuration problems are now distinguished from temporary failures instead of a generic "try again"
  • Project member sharing: invite teammates to a project by email with admin, author, or reviewer roles (Settings → Team)
  • Faster published sites: pages load individually with hover prefetch instead of shipping the whole site up front; larger sites benefit most
0.21.0betaJun 16, 2026

SME review, end to end

  • Must-verify checklist: flag the exact topics a reviewer has to sign off on; required topics show a "Verification required" badge, an "I verify this is accurate" confirmation, and live verified/total progress
  • Approval can now gate verification: when enabled, an SME approving a topic marks it human-verified, feeding content-debt and lifecycle scoring
  • Content freeze: optionally snapshot topics when a review starts, so reviewers always see exactly what they were asked to review even while authors keep editing
  • Login-free reviewers can suggest edits, not just comment: propose a specific text replacement that authors apply or dismiss in one click
  • Automatic reminders for due-soon and overdue reviewers, plus a one-click manual nudge per reviewer
  • Reviewer engagement tracking: per-reviewer tracking links show who has opened their review and who has finished
  • One-click sign-off export: download a CSV (or full JSON audit) of every approve / request-changes decision, with pending items shown explicitly
  • Configurable review-link expiry (7 / 30 / 90 days) with one-click extend, and an approved / changes-requested rollup per session
0.20.0betaJun 15, 2026

Lossless DITA round-trip

  • DITA task steps import in full: step choices, nested notes and images, and profiling conditions on step content are now preserved
  • DITA images referenced by key (keyref via an image keymap) resolve and upload, no more broken images
  • DITA export preserves everything nested inside lists and steps: images, callouts, and conditional content keep their profiling attributes
  • Inline images round-trip in both directions
  • Verified import → export → re-import is lossless on a 57-topic DITA project (conrefs, conditions, images, reltables, glossary all 1:1)
0.19.0betaJun 10, 2026

Lossless imports and conditional inline content

  • Flare imports now preserve condition colors, conkeyref resolution, and image references
  • DITA imports handle conkeyref attributes and nested image paths
  • Inline conditional content: apply conditions to text spans within a paragraph, not just full blocks
  • Large import batching to prevent timeouts on 200+ topic projects
  • Zip uploads via Supabase Storage to bypass Vercel's 4.5 MB payload limit
0.18.0betaJun 3, 2026

Translation management and TMS integration

  • XLIFF 2.0 export and import with 3 modes: merge into existing, create new topics, or overlay as translations
  • Phrase TMS integration: send content for translation, auto-poll for completion, download translated XLIFF on finish
  • Phrase webhook handler for real-time completion callbacks
  • Translation panel in editor: per-locale status grid with stale indicators, coverage summary
  • Multi-locale publishing: assign a locale to each publication target, translated content renders automatically
  • Stale translation detection: editing a source topic flags all translations as needing update
0.17.0betaMay 28, 2026

Content branching and release management

  • Topic-level branching: create named branches, switch between them, rename or delete
  • Three-way merge with conflict resolution dialog (source vs target, accept per block)
  • Release lifecycle: draft → published → archived, with comparison dashboard showing added/removed/changed topics across releases
  • Publication versioning with reader-facing version switcher on published sites
  • Track changes (suggest mode): insertions in green, deletions in red strikethrough, accept/reject per suggestion or in bulk
  • Suggestions panel with multi-author grouping, initials avatars, and jump-to-position
  • Email notification to topic owner when a different author submits a suggestion
0.16.0betaMay 22, 2026

Real-time collaboration and approval workflows

  • Real-time co-editing via Yjs CRDT with live cursors, presence indicators, and colored name labels
  • Offline change queue: edits buffer locally and sync when connection resumes
  • Configurable approval workflows with named stages, role-based transitions (admin, author, reviewer), and workflow history
  • Presence avatars on project and topic pages: see who's online and what they're editing
  • Boolean condition expressions: combine dimensions with AND/OR/NOT instead of flat include/exclude
  • Condition expression builder in publish manager with visual tree editing
0.15.0betaMay 16, 2026

LLM-ready output and AI content intelligence

  • Every published site automatically generates llms.txt, llms-full.txt, per-page .md URLs, sitemap.md, and a smart 404
  • ?ask= query endpoint on every published page: AI agents can ask questions and get grounded answers
  • MCP server with 4 read-only tools (search_docs, get_topic, list_topics, get_site_structure)
  • AI content graph: community detection, duplicate passage detection, reuse candidates, cross-reference suggestions, terminology variant detection
  • GEO scoring: 5-factor analysis (structural clarity, semantic completeness, schema potential, answer density, internal linking) per published page
  • Content gap detection: clusters zero-result and low-confidence AI queries into suggested new topics
  • Content debt scoring: per-topic 0 to 10 score combining staleness, findings, verification, and gaps
  • Nightly lifecycle agent: scans for stale content, expired verifications, near-duplicates, low GEO scores, orphans, and ambiguous topics, then creates recommendations and never mutates content
0.14.0public betaMay 8, 2026

Public beta: enterprise integrations and API

  • Topicary is now open for signups: free during beta with all features unlocked
  • REST API with bearer token authentication, rate limiting (1000 req/min per key), and plan gating
  • API endpoints for topics, components, maps, targets, and webhooks, with full CRUD
  • Webhook event system: 10 event types, HMAC-SHA256 signed payloads, automatic retry with exponential backoff
  • SAML 2.0 SSO with XSW protection, SHA-1 rejection, and role-capped provisioning (author/reviewer only)
  • Bidirectional GitHub sync with push/pull, content-hash diffing, and conflict detection with resolution dialog
  • Zendesk Help Center integration: push published content as articles with sync state tracking and divergence detection
  • Salesforce Knowledge integration: same pattern as Zendesk with OAuth2 auth and UrlName dedup
  • Configurable search engine adapters: built-in pgvector, Algolia, and Elasticsearch
0.13.0beta

AI authoring assistant and inline AI

  • Multi-turn AI chat panel with conversation persistence, history sidebar, auto-resume, and @/# mentions for topic and component references
  • Inline AI (Cmd+Shift+I): Cursor-style prompt overlay on selected text with streaming diff preview
  • Partial accept/reject on AI suggestions: accept per-block instead of all-or-nothing
  • Proactive component extraction: editor highlights paragraphs that match 2+ other topics at ≥85% similarity and suggests extracting them as reusable components
  • Contextual prompt suggestions: dynamic quick-action chips based on active findings (duplicate → 'Check duplicates', readability → 'Simplify')
  • AI-powered map organization: suggests hierarchical tree structure based on community detection and cross-references
  • Content validation rules engine: 10 rules (broken refs, undefined variables, heading hierarchy, missing alt text, and more) with on-demand scan and severity-grouped results
0.12.0beta

Editor overhaul and transactional email

  • Editor layout rebuilt: prose stays centered regardless of which panels are open
  • Three-zone workspace with overlay panel drawers (left panels for navigation, right panels for tools and metadata)
  • Dashboard management: project deletion, inline rename, multi-select with batch operations
  • Dedicated publishing overview page with per-target status
  • Branded transactional email for all auth flows and a welcome email on signup
  • Navigation performance improvements across the heaviest pages
  • Published site fixes: configurable start page, inline links, heading permalinks
  • Import flow now shows a completion summary and links to the imported map
  • Dark mode and error handling audit across the full app
0.11.0beta

Advanced authoring and multi-format export

  • DITA XML export (topics and maps): standards-compliant exit path for teams that need XML portability
  • Word (.docx) export via the publishing pipeline
  • Inline component reuse (text fragments inserted within paragraphs, changes propagate to all instances)
  • Where-used panel for components: see every topic referencing a component before editing
  • Orphan detection for unused components
  • Map versioning with block-level diff and restore
  • Publication versioning with named versions and reader-facing version switcher
  • Embeddable SDK packages: @topicary/embed-react, @topicary/embed-web (custom elements), @topicary/embed-node (Node.js client)
0.10.0beta

Design system and published site polish

  • Landing page redesign with warm editorial aesthetic
  • Brand motifs (herringbone stripes, chevrons, zigzag accents)
  • Console-depth button system for marketing pages
  • Instrument Serif display typography with IBM Plex Sans body and IBM Plex Mono code
  • Published site improvements: JSON-LD schema markup (HowTo, FAQPage, TechArticle), OG image generation, XML sitemap, robots.txt
  • Suggested 'See Also' links computed from page embedding similarity at publish time
  • Password protection for published sites with scrypt hashing and HMAC session tokens
0.9.0beta

Published site enhancements

  • AI-powered search widget on published sites (Gemini-backed, plan-gated, two-tier rate limit)
  • Code copy buttons and tabbed code groups
  • GitHub sync (push content to a repo on publish)
  • OpenAPI 'Try It' interactive API explorer with CORS proxy
  • Dark mode on published sites
  • Reader feedback widget (per-page 'Was this helpful?' votes)
  • Custom CSS injection on published sites
0.8.0beta

Security hardening

  • Full security audit: RLS policy review, auth checks on all RPCs, service-role policy fixes
  • SSRF protection with numeric IPv4 range checks and full IPv6 normalization (no regex bypass)
  • Rate limiting via Upstash Redis (shared across serverless isolates)
  • Timing-safe comparisons on all secret verification paths
  • WCAG accessibility audit: 30+ findings fixed
  • Plan limit enforcement at the database level
  • Server action error sanitization: no internal details leak to client
  • 12 missing database indexes added, 5 redundant indexes removed
0.7.0alpha

Content organization and health

  • Topic cross-references (TopicLink atoms) with link picker: survive renames, auto-resolve at publish time
  • Taxonomy system with color-coded tag groups and tag picker in editors
  • Content status workflow (Draft, In Review, Published) with configurable transitions
  • Content health dashboard: orphan detection, stale content flags (30/90 day), broken reference tracking
  • Search analytics and reader feedback analytics on published sites
  • Custom topic type definitions with structured field schemas
0.6.0alpha

Editor experience

  • Slash command menu (16+ block types)
  • Floating bubble toolbar on text selection
  • Table editor with column/row operations toolbar
  • Code block syntax highlighting (100+ languages)
  • Callout blocks (info, warning, tip, danger)
  • Drag-and-drop reordering in maps with projected drop indicators
  • Image upload (paste, drag-and-drop, file picker) via Supabase Storage
  • Global search (Cmd+K / Ctrl+K) across topics, components, maps, and variables
  • Keyboard shortcuts across the editor with platform-aware display (Cmd on Mac, Ctrl on Windows)
0.5.0alpha

SME review and collaboration

  • Token-based SME review: generate a unique link, reviewer opens it with no account and no login
  • Inline comments on topics with threaded replies
  • Per-topic approve/reject workflow
  • Email invitations for reviewers with branded template
  • Review session management in map editor: assign topics, track status
0.4.0alpha

Billing and onboarding

  • Free, Pro, and Team pricing tiers with transparent published pricing
  • Stripe checkout and customer portal
  • Plan-gated features (AI, advanced analytics, PDF, API access)
  • First-run onboarding with sample project
  • Referral program with $50 Stripe balance credit on conversion
0.3.0alpha

AI assistant and 7-format import

  • AI writing assistant powered by Claude: draft, rewrite, expand, summarize, improve
  • 7-format import: Markdown, HTML, DITA (topics, maps, conrefs), Confluence, MadCap Flare (full project), Word, OpenAPI 3.x
  • Flare import preserves topics, snippets, TOC hierarchy, variables, and conditions
  • DITA import maps conrefs to reusable components and preserves map hierarchy
  • Drag-and-drop file import with format auto-detection and preview cards
0.2.0alpha

Multi-channel publishing

  • 4-stage publishing pipeline: resolve components → filter conditions → substitute variables → render output
  • Hosted documentation sites with search, sidebar navigation, and dark mode
  • PDF generation with branded cover page, table of contents with dot leaders, running headers, and custom footers
  • Markdown export (single topic or full map as zip)
  • Condition filtering and variable substitution at publish time: same source, different outputs
  • Custom domains on published sites
0.1.0alpha

Foundation

  • Project creation with subdomain routing
  • Rich text editor with 1-second debounce auto-save
  • Reusable components: save any selection, insert anywhere, changes propagate
  • Conditional blocks: define dimensions and values, tag content with include/exclude rules
  • Variables: key-value pairs with default sets and per-target overrides
  • Content maps with drag-and-drop hierarchy and ordering
  • Version history with block-level visual diff and one-click restore