2.0.144 is eighteen versions on from 2.0.126. Four things stand out: public pages are now cached for speed, a channel's public site can be password-protected, articles can be turned into AI voice content, and the headless API is complete and documented on every screen.
Faster public pages
Public channel, article, and list/search responses are now served through a read-through cache, so repeat visits don't re-query the database every time.
- Cached where it's safe. Only anonymous/public traffic and the read API are cached — the dashboard and editor always read live, so "I saved but it didn't update" can't happen.
- Edits show immediately. Saving or deleting content evicts the cache write-through; a short time-to-live is the backstop for anything else.
- Safe by default. The cache is never load-bearing — if it's unavailable the site just reads from the source, only a little slower.
Password-protect a channel's public site
- Hide a site behind a password. Turn it on in Channel → Edit and anonymous visitors must enter the password before any public page renders — ideal for letting a client review a site before it's truly public.
- You're never locked out. Signed-in channel members skip the prompt, and the API and MCP server are unaffected.
- Shareable. The password is stored encrypted but viewable in settings, so you can copy and re-share it. Changing it signs out everyone who had the old one.
AI voice content, with rcr.la
- Generate a reading script. One click on an article turns it into a narration script — the same content rewritten for the ear (tables, lists and headings spoken naturally, links read as their meaning) with ElevenLabs-style pause and emphasis tags. It's written in the article's own language, not translated to English.
- Send it to rcr.la for audio. Connect a per-channel rcr.la token, pick a voice, and send the script — the generated audio URL is attached to the article and exposed on the API.
- Editable. The script is yours to fine-tune in the article editor before it's voiced.
The headless API, completed and documented
OmniCon is a headless CMS, so every screen now shows exactly how to reproduce what it does over the API.
- API examples on every page. Each dashboard page has a reference panel with the real REST call — using that page's actual ids — and the equivalent MCP tool and natural-language prompt. Copy-paste runnable, and honest where a screen has no dedicated API.
- Two gaps closed. Galleries and channel folders now have full REST APIs (
/api/galleries,/api/folders), so every standalone resource is reachable over HTTP, not just from the dashboard. - Article permalink visible. The article detail page now shows the article's
permaName, so it's easy to copy straight into an API call.
New documentation
- Delivering content fast — how OmniCon serves content quickly and what your own app should do to cache it.
- API overview — the REST overview, now pointing at the performance guide.
Also in this release
- MCP file upload fixed. The
upload_filetool was returning the wrong endpoint; it now points at the correct API upload URL so the generated upload actually works. - Profile role badges. Your roles now show on the profile page — they were being fetched but never displayed.
All of this is live now, and nothing here needs action from you — caching is automatic, and password protection and voice are opt-in per channel. If you'd like a hand wiring up voice content or building on the API, drop us a note.