Product · Changelog

Changelog

New features, improvements, and fixes for Winnow Probiotics.

v0.3.0 — The strain directory opens to everyone

Added

  • The strain directory is open to everyone at probiotics.winnowlabs.com. It holds each strain's taxonomy, its deposit numbers and the research that names it. Reading needs no account. Sign in when you want to comment or save a strain or paper.
  • Share a link to any page and it unfurls with a preview card.

Improved

  • Each strain page's title names the strain itself, not just its species. Browser tabs and link previews can tell two strains of one species apart.
  • The site is built to be used on a phone. Links, filters and buttons are big enough to tap, and methodology tips open inside the screen. Escape closes a tip.
  • Captions, dates and counts have enough contrast to read comfortably in light mode and in dark mode.
  • Body system pages list the 24 strains with the most papers, so a big hub like the gut stays quick to load on a phone.

v0.2.67 — A lighter journal on a phone

Improved

  • Strain and paper pages load less JavaScript up front. The graph library is now fetched only on the pages that actually draw a nodemap or a tree.

v0.2.65 — A directory that fits on a phone

Added

  • A new research digest covers the batch that came in on September 13. It runs to 303 papers across 37 strains, and it looks at where paper counts and evidence strength pull apart. It joins the earlier editions under Digests once it clears editorial review.

Improved

  • The strain directory used to render all 508 strains into one page. On a phone that was a document about 157 screens tall. It loads 24 at a time now, so the page opens quickly and the first strain card sits near the top.
  • Six filters stacked on a phone put two full screens of controls above the first result. They fold into one control now that tells you how many you have set. The search box stays out in the open, since that is the one most people reach for first.

Fixed

  • Tapping a filter on an iPhone used to zoom the whole page in, and it did not zoom back out when you were done. That is fixed. The menu button, the shop bag and the methodology tips are all big enough to hit with a thumb now.
  • The page behind an open menu used to scroll along with it. It stays put now, and Escape closes the menu. The strain map page is shorter on a phone too, so a vertical swipe carries on past it instead of getting caught inside it.
  • Some research pages scrolled sideways on a phone. A handful of paper titles are a single unbroken 200-character string, and there was nowhere for the line to break. Long titles wrap now, and on a narrow screen the paper rows give the title its own full-width line.
  • The home page headline was sized for a monitor and ran to six lines on a phone, which pushed the first card off the opening screen. It scales with the screen now, so you land on the page and can see where to go next.

v0.2.63 — A new home page and tidier research pages

Added

  • Research paper pages now carry the metadata search engines and Google Scholar read. Every page gets a canonical URL, share previews and structured data. The citation tags Scholar asks for are all-or-nothing per paper, so the 4,293 papers where we hold authorship emit them and the rest emit none.

Improved

  • Paging through research results, flipping the sort or toggling a filter no longer makes the search work out what your query meant all over again. That step is cached for a few hours, so only the first search for a phrase pays for it. Research search now shares the same per-visitor limit Atlas uses, and says so when a run of searches has hit it.

Fixed

  • The changelog pager no longer breaks on a bad page number. Asking for page zero quietly served the oldest releases under the wrong label, and a non-numeric page took the whole page down. Pages now clamp to the range that exists.
  • Forty-two paper links pointed at a URL that returns a 404. One publisher registers a second DOI for the PDF version of an article and we were storing it as given. Thirty of those rows turned out to be duplicates of papers already in the corpus, so they have been merged and the per-strain paper counts they inflated came back down.
  • Strain pages no longer carry sections that exist only to say they are empty. Ninety strains have no linked paper and 277 have no curated citation, and those pages now note the absence once instead of walking through it. The amber sequenced dot is gone as well, since it flagged a gap in our records rather than anything about the organism.
  • The retired Microplastic / Toxin Interaction category no longer renders on strain pages. Three tags replaced it at the end of August but it stayed in the list, so every strain showed a fourth empty panel carrying its old description.
  • The markup cleanup now covers plain-language summaries and the translated title and abstract, not just the title and abstract it started with. 285 summaries on production were carrying leftover publisher markup that the earlier pass could not see, so it never cleaned them. They get picked up the next time the cleanup runs.
  • Asking for a page past the end of the research results now lands on the last page. It used to sort the whole corpus and hand back an empty listing, and a page number with enough digits took the page down outright.
  • Renamed strains redirect to their current page. A taxonomic rename had left both the old and the new name serving a full page and both showing in the directory. Eighteen of the 36 retired entries resolve to exactly one successor and now point there.

v0.2.62 — Cleaner titles and a closed door

Improved

  • Every evidence category on a strain page now shows its paper count in the header. Some panels used to show a genome rating there instead, with nothing saying which number you were looking at. Where a live query reproduces the count, it links to those papers. The heading above them also dropped a line that described the layout and talked up the company.

Fixed

  • Paper titles are cleaned of leftover publisher markup as they come in. Species names that publisher XML had run together are split back apart, and accented letters and symbols decode properly instead of showing as raw entity codes. Titles already in the corpus get the same pass in a separate cleanup run.

v0.2.61 — Clearer strain pages

Added

  • Paper counts on a strain page are now links. Click the number beside a body system, a symptom or an evidence category and you land on those papers for that strain, past the twenty row limit the page used to stop at.

Improved

  • Classifications explain themselves before they link. Anything that scores or categorises now carries a small info icon with a one line explanation, and the full write up sits behind a link inside it. Clicking a heading no longer drops you on the methodology page.

Fixed

  • The Identity signal is gone from strain pages. It turned green whenever we had an NCBI taxon ID and called that a taxonomic match, but every ID we held named the species rather than the strain. The ID now sits in the page header beside the deposit numbers instead of being dressed up as a check.
  • Strain search finds deposit numbers. Searching 17938 now turns up DSM 17938, which used to return nothing even when the card in front of you showed that number. Renamed genus and species are searchable too.

v0.2.59 — Faster strain pages

Improved

  • Strain pages and browsing load faster.

v0.2.58 — Shareable strain directory filters

Added

  • Strain directory filters now live in the URL, so you can share or bookmark a filtered view.

Fixed

  • The strain directory search feels right now. Pressing Enter no longer reloads the page, the back button works as you would expect, and clicking a strain in the tree keeps it selected.