Skip to content

docs(changelog): migrate full history into Mintlify, rewrite 2026#623

Open
jordanc-relevanceai wants to merge 11 commits into
mainfrom
changelog-migration
Open

docs(changelog): migrate full history into Mintlify, rewrite 2026#623
jordanc-relevanceai wants to merge 11 commits into
mainfrom
changelog-migration

Conversation

@jordanc-relevanceai
Copy link
Copy Markdown
Collaborator

Summary

Migrates the full Relevance AI changelog from relevanceai.com/changelog into the docs site as Mintlify <Update> blocks, with a Relevance-AI-themed visual treatment.

  • 158 entries pulled from the sitemap (March 2023 → May 2026)
  • Split by year: /changelog (2026, 18 consolidated entries), /changelog/2025 (107), /changelog/2024 (17), /changelog/2023 (8), with cross-links between years
  • 2026 entries rewritten for clarity (less marketing, more facts), fact-checked against the relevance-api-node repo (3 substantive corrections — Confluence is 5-min polling not real-time; Claude Opus 4.6 is 1M context for both variants; AWS Bedrock Guardrails forwards the ARN, AWS handles the filters)
  • Same-date entries consolidated into a single Update with ## sub-headings per feature
  • Product nouns (Agent, Tool, Workforce, Knowledge, Chat, Phone Agent, Meeting Agent) capitalized
  • Visual: indigo vertical rail down the left of the entries with fade-in / fade-out at the ends, indigo dot per entry, soft shadow on hero images
  • Added "Changelog" anchor in the sidebar (icon: newspaper) next to Community and Contact Support, pointing at /changelog
  • One small style.css fix: the existing 20rem right padding on #content-area is now scoped via :has(div#content-side-layout) so it only kicks in when the right TOC is actually rendered (otherwise wide-mode pages got a dead zone on the right)

Test plan

  • /changelog loads with 2026 entries + year cards at the bottom
  • /changelog/2025, /2024, /2023 all load
  • Year cards link to the right place; "Earlier changes" section appears at the bottom of every year page
  • Sidebar Changelog anchor (newspaper icon) navigates to /changelog in-place, not as an external link
  • Timeline rail is continuous between entries (no overlap brightness, no gaps); fades at the top of the first and bottom of the last
  • Hovering the date column reveals the Mintlify anchor-link icon without it being covered by the dot
  • No banned-words / lint regressions
  • Visual check in both light and dark mode

🤖 Generated with Claude Code

@jordanc-relevanceai jordanc-relevanceai self-assigned this May 13, 2026
jordanc-relevanceai and others added 7 commits May 20, 2026 16:07
Adds 158 entries (March 2023 – May 2026) to a Mintlify <Update>-based
changelog page, wires it as a sidebar anchor next to Community and
Contact Support, and scopes the existing right-padding rule so it only
applies when the right TOC sidebar is actually rendered (fixes wide-mode
pages getting a dead-zone on the right).

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Main /changelog now holds only 2026 entries (26) plus a CardGroup
linking to /changelog/2025, /changelog/2024, /changelog/2023. Each year
page mirrors the same format and cross-links to the others. Cuts the
initial page from ~5700 lines to ~750 to reduce time-to-first-paint and
image-load on the main page.

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Trim marketing scaffolding ("you can now…", "transform", closing CTAs)
and rewrite every 2026 entry around what changed, why it matters, and
how to get to it. Cross-checked technical claims against the node API
repo and corrected three:

- Confluence Knowledge Sync: 5-minute polling, not real-time
- Claude Opus 4.6: both variants are 1M context; the "1M" entry exists
  for pricing-tier separation
- AWS Bedrock Guardrails: clarify Relevance forwards a guardrail ARN;
  the filters are configured AWS-side

Same-date entries are merged into a single <Update> with ## sub-headings
per feature, dropping 26 entries to 18. Enterprise-gated features use
<Note> callouts. Removed every "Other improvements / General fixes and
UI improvements" filler block.

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Adds a Relevance-AI-themed visual treatment to the changelog pages:
indigo vertical rail in the left margin (continuous across all entries,
fading in at the top of the first entry and out at the bottom of the
last), an indigo dot per entry aligned with the date label, and a
soft shadow on hero images. Scoped to pages whose data-page-title
starts with "Changelog" so it applies to the main page and the three
year archives.

Also a casing sweep on the 2026 entries to capitalize the product
nouns Agent, Tool, Workforce, Knowledge, and Chat (plus compound
Phone Agent / Meeting Agent), per the docs style guide.

Two small style.css fixes:
- Scope the existing 20rem right padding on #content-area to only
  apply when the right TOC is actually rendered (otherwise wide-mode
  pages get a dead zone on the right).
- Push Mintlify's anchor-link button further left so it doesn't sit
  underneath the timeline dot.

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Mirrors all 199 changelog images from cdn.relevanceai.com into
images/changelog/ in the repo (~11MB) and rewrites the 206 image
references across the four changelog files to local paths, so the
docs no longer depend on the external CDN or the relevanceai.com
changelog page staying up.

Also converts the 10 video links (4 YouTube, 6 Wistia previously
wrapped via cdn.embedly.com) into responsive 16:9 iframes using the
same paddingTop: 56.25% wrapper pattern the repo uses for Supademo
embeds, so they play inline instead of being blue text links.

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Was 4px too high. Updated top: 2.25rem → 2.5rem so the dot's center
sits at the date pill's vertical midpoint.

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Documents the expansion of Relevance MCP access to users with the Member
role in RBAC-enabled organizations, previously gated to Editor and above.

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Derives alt text from each entry's heading so screenshots are accessible
and the markdown-lint accessibility rule passes.

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
@jordanc-relevanceai jordanc-relevanceai marked this pull request as ready for review May 20, 2026 06:19
Mintlify skipped the initial check while the PR was in draft. Empty
commit to retrigger the external check now that the PR is open.

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
@github-actions
Copy link
Copy Markdown
Contributor

🎯 Vibe check

Reviewed: 4 MDX files + docs.json (4 MDX files with issues, 0 clean)

Scores

Dimension Score What's holding it back
🔴 Consistency 3/10 Systematic banned-word violations across 2024.mdx and 2025.mdx — "powerful", "supercharge", "leverage", "unlock", "effortlessly", "robust", "seamlessly" — plus British spellings ("customise", "specialised", "optimisations", "categorisation") and heading-case failures in 2024.mdx.
🟡 Technical clarity 7/10 Broken demo link in 2024.mdx:1261 (routes to /changelog, not a demo page). Raw URLs in 2025.mdx instead of markdown links. Otherwise UI references are specific in 2026 content.
🟡 Non-technical clarity 7/10 The 2026 content is terse and clear. Archive pages (especially 2025.mdx) use verbose "You can now…→ bullets→ start doing X today!" boilerplate throughout, which is repetitive but not confusing.
🟡 Structure 7/10 Heading capitalization inconsistencies in 2024.mdx (multiple Title Case headings, one emoji in a heading). Overall changelog architecture — year archive pages with CardGroup cross-links — is sound.

Score key: 🟢 9–10, 🟡 6–8, 🔴 1–5. Scores are a single overall judgment about the whole PR.

Overall vibe: The new 2026 entries in changelog.mdx are genuinely the strongest writing in the PR — factual, terse, well-cased, and free of banned words. The archive pages (2024.mdx and 2025.mdx especially) carry significant legacy debt: they were migrated from an older format without a quality pass, so they're packed with the exact words the style guide bans and with British spellings throughout. The structural reorganization (splitting changelog by year, cross-linking via CardGroup) is a good call; the content standards now need a clean-up pass on the archive files to bring them in line.

🔧 Issues (32)

Banned words — changelog/2024.mdx

  • changelog/2024.mdx:38 — "powerful prompt features" → remove "powerful"; describe what the features do
  • changelog/2024.mdx:388 — "extremely powerful feature for debugging" → drop "powerful"
  • changelog/2024.mdx:978 — "robust filtering capabilities" → "robust" is banned; use "comprehensive filtering"
  • changelog/2024.mdx:1015 — "You can effortlessly incorporate these resources" → "effortlessly" is banned; cut the word
  • changelog/2024.mdx:1063 — "leverage the powerful tools they've built" → both "leverage" (use "use") and "powerful" are banned
  • changelog/2024.mdx:1261 — "To unlock this feature" → "unlock" is banned; just say "To use this feature" or "Access this on the business plan"

Banned words — changelog/2025.mdx (systematic; flagging worst offenders)

  • changelog/2025.mdx:14 — "significantly outperforms…powerful" → remove
  • changelog/2025.mdx:239 — "This powerful notification system" → remove "powerful"
  • changelog/2025.mdx:750 (Update description) — "Supercharge AI workflows with Gemini 2.5 flash image generation" → "supercharge" is banned; rewrite to "Gemini 2.5 Flash image generation"
  • changelog/2025.mdx:1317 — "Supercharge your image generation capabilities today!" → remove the CTA line or rewrite without "supercharge"
  • changelog/2025.mdx:1641 — "Supercharge your AI workflows with these powerful new models today!" → both "supercharge" and "powerful" banned
  • changelog/2025.mdx:1427 — "MRP Workforce: Unlock Powerful Multi-Integration Workflows Without Custom Tools" (bold heading) → "unlock" and "powerful" are both banned
  • changelog/2025.mdx:2265 (Update description) — "Unlock new ways to build, share, and trigger your AI Workforce" → "unlock" is banned; "New ways to build, share, and trigger your Workforce"
  • changelog/2025.mdx:2512 — "unlock the true potential of collaborative artificial intelligence" → remove "unlock"
  • changelog/2025.mdx:24 — "Leverage Google's latest AI advancements in your applications" → "leverage" is banned; use "Use" or "Access"
  • changelog/2025.mdx:362 — "integrate seamlessly with your team's communication" → "seamlessly" is banned; cut it
  • Note: "powerful", "leverage", "seamlessly" each appear 10+ times across 2025.mdx; a global search-and-replace pass is needed

British spellings — changelog/2024.mdx

  • changelog/2024.mdx:555 — "customise" → "customize"
  • changelog/2024.mdx:628 — "customising" → "customizing"
  • changelog/2024.mdx:648 — "customise" → "customize"
  • changelog/2024.mdx:1255 — "specialised roles" → "specialized"
  • changelog/2024.mdx:1269 — "customise more capabilities" → "customize"
  • changelog/2024.mdx:1344 — "optimisations" → "optimizations"

British spellings — changelog/2023.mdx

  • changelog/2023.mdx:101 — "Added categorisation to templates" → "categorization"

Heading capitalization — changelog/2024.mdx

  • changelog/2024.mdx:162### Key Features → "Key features"
  • changelog/2024.mdx:208### Example Use Cases → "Example use cases"
  • changelog/2024.mdx:231### Agent Building Improvements → "Agent building improvements"
  • changelog/2024.mdx:233### Direct Tool Integration → "Direct tool integration"
  • changelog/2024.mdx:244### In-Line Tool Creation → "In-line tool creation"
  • changelog/2024.mdx:285### New Agents List View → "New Agents list view"
  • changelog/2024.mdx:826### Enhanced Import and Export for Tools → "Enhanced import and export for tools"
  • changelog/2024.mdx:841### Expanded Step View in Tool Builder → "Expanded step view in tool builder"
  • changelog/2024.mdx:867### Kill Switch for Running Tools → "Kill switch for running tools"
  • changelog/2024.mdx:769### **New LLM Models Added to Tool Builder** → drop the bold inside a heading, and fix case: "New LLM models added to tool builder"

Emojis in content — changelog/2024.mdx

  • changelog/2024.mdx:330### 🔌 New Messaging Integrations (Teams Plan+) → CLAUDE.md prohibits emojis in documentation content; remove 🔌 and fix case: "New messaging integrations (Teams Plan+)"
  • changelog/2024.mdx:291–298 — Inline emojis in bulleted list items (📊 🔍 🔄 📈 💡) → remove per no-emoji rule
  • changelog/2024.mdx:332–352 — More emojis in sub-bullets (🔔 💬 🤝 📱) → remove

Broken link

  • changelog/2024.mdx:1261[book a demo](https://relevanceai.com/changelog) routes to the changelog page, not a demo booking URL. Should be https://relevanceai.com/book-a-demo or the appropriate booking link.

Raw URLs (not using markdown link format) — changelog/2025.mdx

  • changelog/2025.mdx:1605 — bare URL https://relevanceai.com/docs/integrations/popular-integrations/slack dropped inline in text; wrap in a descriptive markdown link
  • changelog/2025.mdx:2359 — bare URL https://relevanceai.com/docs/integrations/llm-integrations/anthropic in paragraph; same fix
🏗️ Page structure (1)
  • changelog/2024.mdx — "Other improvements" sub-headings appear twice inside the August 11 2024 Update block (lines ~455 and ~469), creating a duplicate section header within the same entry. The second should be removed or merged with the first.
🔋 Credit usage
Item Count
Files reviewed 4 MDX + docs.json
Context pages read 0 additional
Total lines processed ~3,300

Files read: changelog.mdx (311 lines), changelog/2023.mdx (214 lines), changelog/2024.mdx (1,366 lines), changelog/2025.mdx (~2,800 lines read in chunks), docs.json (grep only)

@github-actions
Copy link
Copy Markdown
Contributor

🎯 Vibe check

Reviewed: 5 files (3 with issues, 2 clean)

Scores

Dimension Score What's holding it back
🔴 Consistency 3/10 2025.mdx has 77 hits on banned words; 2024.mdx has 12. Emojis used as bullet markers throughout 2025.mdx (CLAUDE.md: "No emojis in documentation content"). British spellings in both archive files. Title-case headings throughout 2024.mdx.
🟡 Technical clarity 7/10 Most instructions are specific and accurate. A few vague UI references ("go to a Workforce image"), bare URLs instead of formatted links, one content typo, and stale 2023-era external doc links.
🟡 Non-technical clarity 6/10 changelog.mdx (2026 content) is clean and direct. The migrated 2024–2025 content is verbose, with a full paragraph of marketing copy per entry and repeated "Start [doing X] today!" CTAs that add noise without value.
🟢 Structure 8/10 Navigation between archive pages via <CardGroup> is well done. Three archive pages (changelog/2023, changelog/2024, changelog/2025) are not registered in docs.json — confirm Mintlify serves them via links alone, or add them to nav.

Score key: 🟢 9–10, 🟡 6–8, 🔴 1–5.

Overall vibe: The 2026 entries in changelog.mdx are exactly what this should look like — factual, tight, product terms capitalized correctly, no banned words. The problem is the migrated 2024 and 2025 content: both files read like marketing emails, with banned words on nearly every entry, pervasive emoji bullet markers, and British spellings scattered throughout. The new structure (archive pages + CardGroup navigation) is a clear improvement, but the content quality debt in the historical entries is significant.

🔧 Issues (40)

Banned words — changelog/2025.mdx (77 total hits; representative instances below)

  • changelog/2025.mdx:25 — "Leverage Google's latest AI advancements" — leverage is banned
  • changelog/2025.mdx:337 — "This powerful upgrade" — powerful is banned
  • changelog/2025.mdx:384 — "Powerful Chain-of-Reasoning Tool Calling" — Powerful is banned
  • changelog/2025.mdx:753 — "Supercharge Your Agents with Meeting Intelligence" — supercharge is banned
  • changelog/2025.mdx:1293 — Update description "Supercharge AI workflows with Gemini 2.5 flash image generation" — supercharge is banned
  • changelog/2025.mdx:1317 — "Supercharge your image generation capabilities today!" — supercharge is banned
  • changelog/2025.mdx:1415 — "Start supercharging your research capabilities today!" — supercharge is banned
  • changelog/2025.mdx:1427 — "Unlock Powerful Multi-Integration Workflows" — unlock + powerful are both banned
  • changelog/2025.mdx:1641 — "Supercharge your AI workflows with these powerful new models today!" — two banned words in one sentence
  • changelog/2025.mdx:2265 — Update description "Unlock new ways to build, share, and trigger your AI Workforce" — unlock is banned
  • changelog/2025.mdx:2512 — "unlock the true potential of collaborative artificial intelligence!" — unlock is banned
  • changelog/2025.mdx:2549 — "You can now build powerful, custom AI agents" — powerful is banned

Banned words — changelog/2024.mdx (12 total hits)

  • changelog/2024.mdx:37 — "surface all of our powerful prompt features" — powerful is banned
  • changelog/2024.mdx:49 — "having a seamless chat experience" — seamless is banned
  • changelog/2024.mdx:112 — "you often need seamless design customization" — seamless is banned
  • changelog/2024.mdx:557 — "seamlessly linking everything into one unified process" — seamlessly is banned
  • changelog/2024.mdx:977 — "robust filtering capabilities" — robust is banned
  • changelog/2024.mdx:1011 — "you can effortlessly incorporate these resources" — effortlessly is banned
  • changelog/2024.mdx:1063 — "leverage the powerful tools they've built" — leverage + powerful are both banned
  • changelog/2024.mdx:1262 — "start leveraging the full capabilities" — leveraging is banned

British spellings

  • changelog/2024.mdx:556 — "customise it" → "customize it"
  • changelog/2024.mdx:973 — "new Activity Centre" → "new Activity Center" (also in the heading)
  • changelog/2024.mdx:977 — "command centre" → "command center"
  • changelog/2024.mdx:1253 — "specialised roles" → "specialized roles"
  • changelog/2024.mdx:1269 — "customise more capabilities" → "customize more capabilities"
  • changelog/2024.mdx:1347 — "app speed optimisations" → "app speed optimizations"
  • changelog/2025.mdx:1065 — heading "Default Organisation & Project" → "Default Organization & Project"
  • changelog/2025.mdx:1071 — "in the correct organisation and project" → "organization"

Heading case violations — changelog/2024.mdx

  • changelog/2024.mdx:161### Key Features### Key features
  • changelog/2024.mdx:188### Example Use Cases### Example use cases
  • changelog/2024.mdx:232### Agent Building Improvements### Agent building improvements
  • changelog/2024.mdx:234### Direct Tool Integration### Direct Tool integration
  • changelog/2024.mdx:243### In-Line Tool Creation### In-line Tool creation
  • changelog/2024.mdx:285### New Agents List View### New Agents list view
  • changelog/2024.mdx:330### 🔌 New Messaging Integrations (Teams Plan+) → remove emoji + sentence case: ### New messaging integrations (Teams plan+)
  • changelog/2024.mdx:333### LinkedIn Integration### LinkedIn integration
  • changelog/2024.mdx:343### WhatsApp Personal Integration### WhatsApp Personal integration
  • changelog/2024.mdx:769### **New LLM Models Added to Tool Builder**### New LLM models added to Tool Builder
  • changelog/2024.mdx:843### Enhanced Import and Export for Tools### Enhanced import and export for Tools
  • changelog/2024.mdx:851### Expanded Step View in Tool Builder### Expanded step view in Tool Builder
  • changelog/2024.mdx:867### Kill Switch for Running Tools### Kill switch for running Tools
  • changelog/2024.mdx:1225### **Better Integration Setup**### Better integration setup

Heading case — changelog.mdx

  • changelog.mdx:45## Agent Performance Observability## Agent performance observability

Emojis in content

  • changelog/2025.mdx➡️ is used as a bullet marker in every single entry across ~107 updates. CLAUDE.md says "No emojis in documentation content." Replace with plain markdown list items (-).
  • changelog/2024.mdx:291📊, 🔍, 🔄, 📈 used as bullet decorators. Remove.

Typos and artifacts

  • changelog/2025.mdx:301 — "with a single click n Fine-Grained Access" — "n" should be "in"
  • changelog/2024.mdx:327 — lone . character on its own line, a content artifact
  • changelog/2025.mdx:327 — Update description attribute reads "Anthropics's Claude Opus 4.5 now available" — double apostrophe (Anthropics'sAnthropic's) and double space before "now"

Inconsistent product name

  • changelog/2025.mdx:199description="Open AI&#x27;s GPT 5.2..." — "Open AI" should be "OpenAI" (no space); same pattern at line 473

Bare/poor link text

  • changelog/2025.mdx:1605 — Raw URL https://relevanceai.com/docs/integrations/popular-integrations/slack pasted inline instead of formatted as a markdown link
  • changelog/2025.mdx:2514[https://relevanceai.com/workforce](https://relevanceai.com/workforce) — link text is the URL itself, use descriptive text: [Workforce overview](https://relevanceai.com/workforce)
  • changelog/2025.mdx:222[**https://openai.com/index/introducing-gpt-5-2/**](...) — link text is the URL; use "OpenAI's announcement" or similar

Stale external links — changelog/2023.mdx

  • changelog/2023.mdx:153 — links to https://documentation.relevanceai.com (old SDK docs URL, likely redirects or 404s)
  • changelog/2023.mdx:169 — links to https://chain.relevanceai.com/templates (old chain builder URL, likely dead)
🧩 Component suggestions (1)
  • changelog/2025.mdx — The ➡️ emoji bullet pattern is used as a list style across all 107 entries. Beyond violating the no-emoji rule, these lists would render better as plain markdown lists (-). Each sub-feature point is a short standalone bullet, which is exactly what markdown lists are for. The emoji adds no semantic value here.
🏗️ Page structure (2)
  • changelog/2023.mdx, changelog/2024.mdx, changelog/2025.mdx — None of the three archive pages are registered in docs.json. The only changelog entry in docs.json (line 637) is the anchor pointing to /changelog. In Mintlify, pages not in docs.json may not appear in the sidebar or may not render at all. Either add them as nested pages under the changelog anchor, or confirm that Mintlify serves them as "unlisted" pages reached via CardGroup links.

  • changelog/2024.mdx (general) — The marketing CTA pattern "Start [doing X] today!" or "If you're interested in trying out Relevance AI, you can [sign up] or [book a demo]" appears after nearly every entry. On a documentation page, these are noise — readers are already users. The 2026 entries in changelog.mdx correctly omit these CTAs. Worth cleaning on a follow-up pass when the 2024/2025 content is refactored to match the current tone.

✅ Clean files (2)

changelog.mdx (2026 entries — well-written, no banned words, correct capitalization), changelog/2023.mdx (no banned words, heading case is clean; stale external links are a minor legacy issue)

🔋 Credit usage
Item Count
Files reviewed 5
Context pages read 0
Total lines processed ~5,340

Files read: changelog.mdx (311 lines), changelog/2023.mdx (214 lines), changelog/2024.mdx (1,365 lines), changelog/2025.mdx (~3,350 lines), docs.json (100 lines partial)

The top fade-in relied on `.update-container:first-child::before`, which
matches on local dev but drops out on Mintlify production where the first
child of the content area can differ. Move the gradient to a `::before`
on `.mdx-content[data-page-title^="Changelog"]` so it renders regardless
of child ordering.

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
@github-actions
Copy link
Copy Markdown
Contributor

🎯 Vibe check

Reviewed: 4 files (3 with issues, 1 clean)

Scores

Dimension Score What's holding it back
🔴 Consistency 3/10 changelog/2025.mdx has 108 grep hits for banned words (seamlessly, supercharge, leverage, powerful, unlock); changelog/2024.mdx has 10 more plus 7+ British spellings (customise, organised, specialised, reorganising, optimisations, Activity Centre); both archive files use emoji pervasively — all explicitly prohibited by CLAUDE.md.
🟡 Technical clarity 6/10 2026 entries are precise. Two dead external links in changelog/2023.mdx; a bare unparsed URL in changelog/2025.mdx; most 2025/2024 instructions say "go to X and look for Y" without naming exact UI labels.
🟡 Non-technical clarity 6/10 The 2026 entries explain features plainly. The archive entries (2025, 2024) bury the actual change inside marketing copy — "unprecedented", "game-changing", and "Start building X today!" sign-offs add noise without adding meaning.
🟡 Structure 7/10 Cross-year <CardGroup> navigation is consistent and appropriate. changelog/2024.mdx mixes ### **Bold** (heading + bold, redundant) and proper ### headings; the style shift from archive pages to the 2026 entries is jarring for readers who follow year-links from the main page.

Score key: 🟢 9–10, 🟡 6–8, 🔴 1–5.

Overall vibe: changelog.mdx (the 2026 entries) is in great shape — concise, factual, zero emojis, clean UI references. The three archive files are clearly imported from a previous CMS (Webflow based on image URL patterns) and carry significant style debt: banned words, British spellings, and emoji markers saturate changelog/2025.mdx and changelog/2024.mdx, making them non-compliant with CLAUDE.md. The structural bones (CardGroup navigation, ## Earlier changes anchor) are sound, but the content in the archive files needs a cleanup pass before this ships.


🔧 Issues (25)

Banned words — changelog/2025.mdx (108 total occurrences across "seamlessly", "supercharge", "leverage", "powerful", "unlock")

  • changelog/2025.mdx:751**Fireflies Integration: Supercharge Your Agents with Meeting Intelligence** — "Supercharge" is banned. Describe what the integration does instead.
  • changelog/2025.mdx:1293description="Supercharge AI workflows with Gemini 2.5 flash image generation" — banner description with banned word; renders as subtitle.
  • changelog/2025.mdx:1317 — "Supercharge your image generation capabilities today!" — banned word + marketing CTA.
  • changelog/2025.mdx:1641 — "Supercharge your AI workflows with these powerful new models today!" — both "Supercharge" and "powerful" are banned.
  • changelog/2025.mdx:2728**New OpenAI Models: Supercharge Your Agents with GPT-4.1 Series** — banned word in bold heading.
  • changelog/2025.mdx:363 — "workflows that integrate seamlessly with your team's communication" — "seamlessly" is banned.
  • changelog/2025.mdx:1103 — "You can now seamlessly reference connected agents" — "seamlessly" is banned.
  • changelog/2025.mdx:1365 — "You can now seamlessly integrate your Notion pages" — "seamlessly" is banned.
  • changelog/2025.mdx:1805 — "seamlessly integrate your organization's custom models" — "seamlessly" is banned.
  • changelog/2025.mdx:2494 — "sophisticated multi-agent systems that collaborate seamlessly" — "seamlessly" is banned.
  • changelog/2025.mdx (throughout) — "Leverage X", "➡️ Leverage Google's latest AI" pattern appears in dozens of ➡️ bullet lists. "leverage" is banned — replace with "use", "connect to", "apply", etc.
  • changelog/2025.mdx (throughout) — "powerful" used in nearly every entry (e.g. "powerful new capabilities", "powerful research capability"). "powerful" is banned — describe the actual capability instead.

Banned words — changelog/2024.mdx

  • changelog/2024.mdx:387 — "an extremely powerful feature for debugging multi-agent systems" — "powerful" is banned.
  • changelog/2024.mdx:977 — "real-time dashboards and robust filtering capabilities" + "This powerful command centre" — both "robust" and "powerful" are banned, and "centre" is British spelling.
  • changelog/2024.mdx:993 — "Approval Mode allows you to seamlessly rotate through all pending approval tasks" — "seamlessly" is banned.
  • changelog/2024.mdx:1063 — "Users who want to leverage the powerful tools they've built" — both banned.
  • changelog/2024.mdx:1261 — "To unlock this feature… start leveraging the full capabilities" — "unlock" and "leverage" are both banned.
  • changelog/2024.mdx:1285 — "seamlessly linking everything into one unified process" — "seamlessly" is banned.

British spelling — changelog/2024.mdx

  • changelog/2024.mdx:555 — "understand how to customise it" → "customize"
  • changelog/2024.mdx:557 — "organised within a new sidebar" → "organized"
  • changelog/2024.mdx:629 — "you can customise it" → "customize"
  • changelog/2024.mdx:1255 — "agents with specialised roles" → "specialized"
  • changelog/2024.mdx:1269 — "reorganising the panels" → "reorganizing"; "customise more capabilities" → "customize"
  • changelog/2024.mdx:1347 — "app speed optimisations" → "optimizations"

British spelling — changelog/2025.mdx

  • changelog/2025.mdx:1067**Default Organisation & Project** → "Organization"; same entry line 1071 uses "organisation" (lowercase). Inconsistent within a single entry.

Emoji usage — changelog/2024.mdx

  • changelog/2024.mdx:53–55 — ✅ emoji bullets throughout December 12 Agent Chat entry. CLAUDE.md: no emojis in documentation content.
  • changelog/2024.mdx:289–299 — 📊🔍🔄📈💡 emojis used as list bullets in the November 3 list view entry.
  • changelog/2024.mdx:331–342 — 🔔💬🤝📱 emojis in LinkedIn/WhatsApp integration section.

Emoji usage — changelog/2025.mdx

  • changelog/2025.mdx (throughout) — ➡️ used as a bullet marker in every single entry. This violates the no-emoji rule and is unnecessary — standard - markdown bullets convey the same structure.

Dead / outdated links — changelog/2023.mdx

  • changelog/2023.mdx:152 — Link text "documentation" points to https://documentation.relevanceai.com — old domain, likely dead. Replace with the current docs URL or a root-relative path.
  • changelog/2023.mdx:168https://chain.relevanceai.com/templates — old domain, likely dead. Remove or replace.

Bare URL — changelog/2025.mdx

Marketing CTAs — changelog/2025.mdx and changelog/2024.mdx

  • changelog/2025.mdx (throughout) — Every entry ends with "Start building X today!", "Try X today!", "Experience Y now!" These are marketing prompts, not documentation. The 2026 entries in changelog.mdx don't do this — drop the sign-offs.
🧩 Component suggestions (3)
  • changelog/2024.mdx:769### **New LLM Models Added to Tool Builder** — redundant bold inside heading; heading already provides weight. Either ### New LLM models added to Tool Builder or a plain **...** line, not both. Same pattern appears at lines 937, 953, 973, 989, 1011, 1133, 1139 throughout the file.

  • changelog/2025.mdx and changelog/2024.mdx (throughout) — "To access this feature, go to Settings > X > click Y" appears in nearly every 2025 entry. These are multi-step procedures — wrapping in <Steps> would add visual progress indicators and make the path scannable. Example for the Spend Controls entry: Step 1: Go to Settings → Spend Controls. Step 2: Click Start on "credits" or "actions".

  • changelog/2025.mdx (throughout) — The ➡️ bullet format with 5–6 items per entry reads as a feature benefits list. Where items aren't navigable links or parallel choices, a prose sentence is cleaner. For the entries that already follow the 2026 style (short paragraph + optional sub-list), no change needed — but the ➡️ pattern should be dropped in the archive cleanup.

🏗️ Page structure (1)
  • changelog/2025.mdx and changelog/2024.mdx — The writing style is fundamentally different from the 2026 entries: verbose marketing intros ("You can now…" + bold recap headline + ➡️ list + CTA sign-off vs. the 2026 style of one tight paragraph and a UI path). A reader following the "2025" card from the main changelog will land in a noticeably different editorial voice. Consider a brief cleanup pass to strip CTAs, replace emoji bullets, and cut the marketing recap sentences — the core feature descriptions are solid, they just need the wrapper removed.
✅ Clean files (1)

changelog.mdx (2026 entries — sentence case headings, no banned words, no emojis, precise UI paths, correct <Note> usage)

🔋 Credit usage
Item Count
Files reviewed 4 MDX + docs.json (partial)
Context pages read 0 additional
Total lines processed ~3,900

Files read: changelog.mdx (311 lines), changelog/2024.mdx (~1,366 lines), changelog/2023.mdx (215 lines), changelog/2025.mdx (~1,900 lines of ~2,200+ total, read in 4 chunks), docs.json (100 lines, partial)

Non-doc files: docs.json — changelog registered at line 637 (/changelog href), structure consistent with existing nav. Images and style.css — binary/CSS, no documentation content reviewed.

…d tags

Each accordion bullet is prefixed with a colored chip naming the surface
the change relates to (MCP, Triggers, Tools, etc.) so the same surface
gets the same colour across the whole changelog. Accordion body and
summary fonts are shrunk so the supporting info doesn't overshadow the
main entry body.

Also widens the `main > div` cap to 100% on changelog pages so the page
doesn't centre at 1400px and leave a dead zone on wide monitors.

Includes placeholder example fixes / improvements / extras under both
the May 13 and May 1 entries — replace with real items before publishing.

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
@github-actions
Copy link
Copy Markdown
Contributor

🎯 Vibe check

Reviewed: 4 MDX files (3 with issues, 1 clean) + docs.json + style.css

Scores

Dimension Score What's holding it back
🔴 Consistency 2/10 77 banned-word uses in 2025.mdx alone (powerful 30+, seamlessly 10+, leverage 10+, supercharge 4, unlock 4); 12 more in 2024.mdx. British spellings throughout 2024.mdx (customise, organised, specialised, Activity Centre, optimisations). Emojis used as bullets (➡️ in every 2025 entry) and in headings (🔌, 📊 etc. in 2024) violate the no-emoji rule.
🟡 Technical clarity 6/10 Two broken or wrong links: 2023.mdx:152 links to the old documentation.relevanceai.com domain; 2024.mdx:1261 [book a demo] points to /changelog instead of a demo page. UI references in newer entries are specific; older entries are more vague ("click the button" style).
🟡 Non-technical clarity 7/10 Feature descriptions in 2025.mdx generally explain what changed and why, which is good. The marketing CTA closings ("Start [X] today!") add noise without content. The 2026 entries in changelog.mdx are crisp and usable.
🟡 Structure 7/10 Year-split structure is sensible and the changelog.mdx (2026) sets a clean pattern with proper ## Heading hierarchy and Fixes/Improvements/Extras accordions. The 2025/2024/2023 files use a completely different format — bold pseudo-headers in Title Case, ➡️ bullets, and CTA closings — creating a jarring shift in tone and structure when navigating between years.

Score key: 🟢 9–10, 🟡 6–8, 🔴 1–5.

Overall vibe: The 2026 entries in changelog.mdx are well-written and follow the content standards — clean headings, no banned words, good technical detail. The problem is the three archived year files (2025.mdx, 2024.mdx, 2023.mdx), which appear to have been migrated from a marketing-heavy CMS format without cleanup: virtually every entry violates the banned-word list, emojis are used as decorative bullets throughout, and British spellings appear across 2024. These files need a style pass before shipping.

🔧 Issues (47)

Banned words — changelog/2025.mdx (77 total hits, representative sample)

  • changelog/2025.mdx:137powerful feature → describe what it does instead
  • changelog/2025.mdx:239This powerful notification system → drop "powerful"
  • changelog/2025.mdx:271makes building powerful tools easier → "makes building tools easier"
  • changelog/2025.mdx:335This powerful upgrade → "This upgrade"
  • changelog/2025.mdx:363integrate seamlessly with your team's communicationseamlessly is banned
  • changelog/2025.mdx:511Seamless conversation flow → banned
  • changelog/2025.mdx:563allowing seamless continuation → banned
  • changelog/2025.mdx:751Supercharge Your Agents with Meeting Intelligencesupercharge is banned (also in the Update description attribute)
  • changelog/2025.mdx:755powerful meeting data capabilities → drop "powerful"
  • changelog/2025.mdx:1317Supercharge your image generation capabilities today! → banned
  • changelog/2025.mdx:1427Unlock Powerful Multi-Integration Workflows → both unlock and powerful are banned
  • changelog/2025.mdx:1641Supercharge your AI workflows with these powerful new models today! → two banned words
  • changelog/2025.mdx:2265 — Update description Unlock new ways to build...unlock is banned
  • changelog/2025.mdx:2512unlock the true potentialunlock is banned

Banned words — changelog/2024.mdx (12 total hits)

  • changelog/2024.mdx:37our powerful prompt features → drop "powerful"
  • changelog/2024.mdx:387extremely powerful feature for debugging → drop "powerful"
  • changelog/2024.mdx:977robust filtering capabilities and powerful command centrerobust and powerful are both banned
  • changelog/2024.mdx:993Approval Mode allows you to seamlessly rotateseamlessly is banned
  • changelog/2024.mdx:1011you can effortlessly incorporate these resourceseffortlessly is banned
  • changelog/2024.mdx:1015effortlessly integrate the necessary resources and ensuring seamless communication → both banned
  • changelog/2024.mdx:1063leverage the powerful tools they've built → both leverage and powerful are banned
  • changelog/2024.mdx:1261start leveraging the full capabilitiesleveraging is banned
  • changelog/2024.mdx:1295Here's a powerful new capability → drop "powerful"

British spellings — changelog/2024.mdx

  • changelog/2024.mdx:555customisecustomize (also lines 629, 647)
  • changelog/2024.mdx:557organisedorganized
  • changelog/2024.mdx:977Activity CentreActivity Center; command centrecommand center
  • changelog/2024.mdx:1255specialised rolesspecialized (also line 1259)
  • changelog/2024.mdx:1269reorganisingreorganizing; customisecustomize
  • changelog/2024.mdx:1285seamlessly linking (banned word) + check surrounding context
  • changelog/2024.mdx:1347optimisationsoptimizations

British spellings — changelog/2025.mdx

  • changelog/2025.mdx:1067**Default Organisation & Project:**Organization
  • changelog/2025.mdx:1071correct organisation and projectorganization

Emojis in documentation content

  • changelog/2025.mdx — ➡️ used as bullet points in every single Update entry throughout the 3,619-line file. Replace with standard markdown bullets (-).
  • changelog/2024.mdx:291–299📊, 🔍, 🔄, 📈 inside a bullet list; 💡 **Pro Tip:** at line 299; 🔌 in a ### heading at line 331; 👇 at line 24. Remove all.

Heading capitalization — changelog/2024.mdx

  • changelog/2024.mdx:156### Custom Task Views### Custom task views
  • changelog/2024.mdx:160### Key Features### Key features
  • changelog/2024.mdx:186### Example Use Cases### Example use cases
  • changelog/2024.mdx:232### Agent Building Improvements### Agent building improvements
  • changelog/2024.mdx:234### Direct Tool Integration### Direct tool integration
  • changelog/2024.mdx:242### In-Line Tool Creation### In-line tool creation
  • changelog/2024.mdx:285### New Agents List View### New agents list view
  • changelog/2024.mdx:333### LinkedIn Integration### LinkedIn integration
  • changelog/2024.mdx:345### WhatsApp Personal Integration### WhatsApp personal integration
  • changelog/2024.mdx:769### **New LLM Models Added to Tool Builder**### New LLM models added to Tool Builder
  • changelog/2024.mdx:829### Enhanced Import and Export for Tools### Enhanced import and export for Tools
  • changelog/2024.mdx:843### Expanded Step View in Tool Builder### Expanded step view in Tool Builder
  • changelog/2024.mdx:867### Kill Switch for Running Tools### Kill switch for running Tools

Broken / wrong links

  • changelog/2023.mdx:152https://documentation.relevanceai.com is the old domain. Link to the current docs page instead (e.g. a root-relative / path to the relevant SDK page if it exists).
  • changelog/2024.mdx:1261[book a demo](https://relevanceai.com/changelog) sends users to the changelog, not a demo page. Fix the href.
🏗️ Page structure (2)
  • changelog/2025.mdx vs changelog.mdx — The 2026 entries use ## Heading tags, clean ### Other improvements sections, and no filler CTAs. The 2025 entries (and 2024/2023) use **Bold Title: Marketing Subtitle** pseudo-headers in Title Case, then a ➡️ bullet list, then a marketing CTA closer ("Start [X] today!"). The format mismatch means navigating from a 2026 entry to a 2025 entry feels like landing on a different site. Consider applying the 2026 entry format (proper headings, no CTA closers, no ➡️ bullets) to the older files, or at minimum to any new entries added to those files.

  • changelog/2025.mdx — The bolded pseudo-headers throughout (e.g. **Gemini 3 Flash Preview: Faster, Cheaper AI with Pro-Level Performance**) use Title Case, which violates the sentence-case rule for all text acting as section headers. These would read: **Gemini 3 Flash Preview: faster, cheaper AI with pro-level performance** or better, replace with actual ## Heading tags to match the 2026 format.

✅ Clean files (1)

changelog.mdx (2026 entries) — No banned words, no British spellings, no emojis, headings in sentence case, accordions used correctly for Fixes/Improvements/Extras. This is the gold standard for the rest of the files.

🔋 Credit usage
Item Count
Files reviewed 4 MDX + docs.json + style.css
Context pages read 0 (changelog is self-contained)
Total lines processed ~3,619 (2025.mdx, partial) + 1,365 (2024.mdx) + 214 (2023.mdx) + 355 (changelog.mdx) + 260 (style.css) ≈ 5,813

Files read: changelog.mdx (355 lines), changelog/2025.mdx (~1,200 lines sampled across 4 reads), changelog/2024.mdx (1,365 lines), changelog/2023.mdx (214 lines), style.css (260 lines)

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant