SciFlow · Better3

Why it exists

Keeping up with a field means a dozen journal alerts, two preprint servers and a search you re-run every Monday. SciFlow is the trading-floor version of that: one dense, chronological flow of everything new in microbiology, polled from Europe PMC and bioRxiv every hour, with the reading pane next to it so you never leave the list. The question it answers is the only one that reliably brings a researcher back: have I missed anything in my area?

SciFlow in dark mode: the filter bar with date range, source and search, the flow of the latest bioRxiv preprints on the left, and the paper pane on the right listing popular subjects with counts — **fig. 01** — the flow. Date range, source and a single search box that understands journals, authors, subjects and free text. The status line reads like a prompt because the whole interface is built to be driven from the keyboard.

What it does


ingest	Europe PMC and bioRxiv, scoped to microbiology at query time by a vocabulary over title and abstract plus a journal whitelist, so off-topic records are never fetched
filter	date window, articles or preprints, open access, full text, journal, subject, author, affiliation, free text; semantic search over abstracts
read	title, abstract, authors, affiliations and MeSH subjects in the pane; j / k to move, Enter to open, ? for keys
save	sign in to keep named filter sets; each set shows what is new since your last visit, counted as inventory, never as a score
keep	bookmarks export to BibTeX and RIS, so the workflow ends where yours does
share	every paper has a permalink with its own preview card for Bluesky, X, LinkedIn, Facebook and email
subscribe	any filter as an Atom feed; an opt-in weekly digest every Tuesday morning

SciFlow with a preprint open in the paper pane: preprint, bioRxiv and open-access labels, the title in large monospace, and the abstract set in a justified serif column — **fig. 02** — a paper open. The list stays put and the pane fills; reading marks are written on a short dwell, never on scroll-past, so the unread count stays something you can trust.

How it is built

A FastAPI backend over PostgreSQL with pgvector, a scheduler running seven jobs at cadences from fifteen minutes to weekly, and a React front end. Every job is idempotent and re-reads a trailing window, so a machine that was off overnight catches up on the next tick. One filter object serves as the URL state, the API request and the saved set, so the private dashboard needed no second pipeline. The whole thing runs on one small VPS behind Caddy, with 409 backend tests and a tagged deploy for every release.

The same SciFlow flow in light mode: white background, hairline rules, monospace throughout — **fig. 03** — light theme. Square corners, hairline rules, IBM Plex Sans and JetBrains Mono; the same design language the Manzat tool borrows.

Try it

Open sciflow.eu. No account is needed to read; sign in to save sets, bookmark and subscribe to the digest.