High-volume sites get a faster, fuzzy article search on its own database capacity and at-a-glance check summaries — alongside a batch of Reader rendering fixes.

Curvenote SCMS¶

Faster, fuzzier works search. The public works listing and search is now backed by a dedicated search index with full-text and fuzzy matching over accent-normalized titles, authors, DOIs, and affiliations. It runs on its own database connection pool, so search on a busy site no longer competes for capacity with the rest of the platform, and unfiltered browsing counts come straight from the index.

Fixes & polish. A GET to the login endpoint now returns a proper “method not allowed” instead of a server error; a cancelled works search releases its database connection immediately instead of running to completion; version tags on the work timeline stay stable through re-orders and re-extracts; and a terminally failed check can now surface a specific recovery hint.

Curvenote Reader¶

The shared rendering core shipped a batch of quality fixes:

• Cleaner figure lightbox. Equation images are excluded from the figure gallery.

• Reliable hash links. Anchored links now scroll correctly even from inside menus and popovers.

• DOI-preferring citations. Citation links prefer a DOI when one is available.

• Tidier document outline. Implicit front- and back-matter (abstract, data availability, acknowledgements) is no longer listed as headings in the on-page outline, and figure and equation numbering starts from the article body, not the abstract.

• Smarter abbreviations. Abbreviations match on word boundaries (so ABBR no longer matches inside ABBReviated), and when several casings share one expansion the panel shows a single canonical entry.

openRxiv Labs — Curvenote Reader¶

A smoother listing and search. The filter, sort, subject, and date controls at the site root stay live and stable while results load: filter chips and the search box reflect your pending choice immediately, and a loading indicator sits under the toolbar.

Cleaner articles. openRxiv inherits all of the shared-core wins above — including the tidier outline (no more stray “Data Availability” heading), the smarter abbreviation matching, DOI-preferring citations, and the lightbox fixes.

The headline this week is for editors and operators running busy SCMS sites: the job pipeline behind publish, checks, and converters is now fully decoupled and asynchronous. Reader also starts serving supplementary files directly, alongside another JATS release.

Curvenote SCMS¶

A fully decoupled job pipeline. Publish, checks, and converters now run asynchronously end-to-end on a dedicated queue — with retries, dependency chaining, and clean terminal-failure cascades. No more submissions stuck waiting on a dispatch round-trip.

Check-services maintenance mode. Platform admins can now put individual check services into maintenance from the extensions admin. Dependent controls show clear disabled-state tooltips, and the upload flow skips a check that’s in maintenance and continues — with an informational note — rather than blocking the submission.

Version history, now visible. The submission version popover now is shown for both works and submissions listings: status-coloured site chips, version tags on the work timeline, popovers trimmed to a “see all” view, and version labels on work details.

Faster ETL endpoints. The published-article API route is materially faster; ETL re-registration now unpublishes the stale copy first, so a re-extracted article never leaves an old version live; a new “exports exist” check ships; and word-count check messages now include the filename.

Curvenote Reader¶

Reader’s shared rendering core picked up SCMS client-identity plumbing: every server-side fetch to the SCMS now identifies itself, and the article-listing transform sends the same headers — which, as a bonus, fixes article listings that weren’t appearing on preview submission pages.

openRxiv Labs — Curvenote Reader¶

Supplementary files served directly. PDFs, datasets, and Office documents attached to preprints are now served straight through the openRxiv theme at a dedicated route — so download links for tables, figures, and appendices resolve to the source file instead of 404-ing or bouncing through the image optimizer.

JATS processing. Five fixes, all of which surface in Reader output:

• Supplementary “media” files now render as download links instead of silently dropping.

• JATS <statement> blocks become MyST proofs — theorems, lemmas, propositions carry through with their meaning intact.

• Abstracts with richer title structures parse cleanly.

• Nonstandard cross-reference types are accepted, so more in-article links resolve.

• BibTeX export covers more citation fields.

Cleaner request logs. Every SCMS request from the openRxiv theme now carries an identifying Curvenote user-agent, so partners running their own SCMS can attribute traffic to the Reader app.

Curvenote Theme¶

Richer links. Citations can now surface figures inline, and links to Hugging Face, GEO, GitHub, and Figshare render as rich cards and previews.

Theme-server requests to the SCMS now send a descriptive user-agent and client headers, so it is now possible to attribute SCMS traffic to a specific theme app and runtime.

Curvenote Sites¶

Editors using the Sites extension get the same finished version popover on submission detail and listing views, and the activity feed now reflects the decoupled job pipeline — surfacing converter-task activities. The submissions listing date filter was refreshed too.

This was one of the biggest Curvenote Reader polish windows of the cycle: a rebuilt landing page, clearer submission listings, faster cold loads, richer links, and a JATS release that makes papers render more faithfully end-to-end.

Curvenote SCMS¶

General platform improvements this week: a refreshed Submissions UI, faster works-listing search, quicker DOI resolution, a new content check on upload, and direct publishing from the ETL pipeline.

Curvenote Reader¶

Faster at scale. A new runtime cache sits in front of content lookups (with a tested memory fallback), so cold paths stay snappy even under heavy load.

Under the hood. Improved math rendering, a sharper image optimizer, more consistent analytics, and a Content Security Policy at the edge.

openRxiv Labs — Curvenote Reader¶

A rebuilt landing page. A new front page and navigation with refreshed messaging and tuned layout widths, and an openRxiv Labs banner.

Clearer recent-submissions listings. Publication dates now use a calendar-day difference instead of a rolling 24-hour window, so “how recent” matches what readers expect. The ambiguous “Today” filter is gone, and “Published” is now labelled “Posted” to match preprint terminology.

Better manuscript reading. A friendly rate-limit page instead of a hard error; a clearer path for brand-new preprints that haven’t been processed yet; an article-version notice with link prefetch so moving between versions is instant; and a refreshed mobile article bar with more visible feedback links.

JATS processing. The pipeline that turns publisher XML improved:

• Math renders for images as children.

• Figure and table numbering is preserved from the source paper, so “Figure 3” stays “Figure 3”.

• Bibliographies serialize cleanly — references that used to break a build now work.

• Figure and table titles render as styled text rather than naked inlines.

• Abbreviations expand cleanly, with no stray parentheses in the glossary.