Repository navigation
Releases
paperus v0.1.5
paperus now carries an OpenAlex key, waits politely when the day's budget runs dry, and finds your preprints' citations even after the journal renamed them.
- A rescued group says so: under Groups, an account that is
deactivated is marked as such, and when a site administrator names a
new admin for a group left without one, everyone in it is told ("Root
made Ben an admin of Lab, which had none left"). - OpenAlex key: OpenAlex now gives each address a small daily budget
for its lists ($0.10, about 1,000) and ten times that with a free key.
An optional key under Administration → Settings → Server (or
OPENALEX_API_KEYin.env) is sent with every request to OpenAlex
(authors, published versions, your papers, times cited); never shown
again once set. When the day's budget is used up anyway, OpenAlex's
lookups wait until it renews (midnight UTC) and go on from there, with
one line in the log instead of a failure an hour. - Times cited for preprints too: a preprint OpenAlex merged into its
journal version (often renamed on publication) was not found by its
arXiv DOI in a list; it is now asked for alone (free at OpenAlex).
paperus v0.1.4
paperus now fetches your new papers by itself, throws a small party when they get published or cited, and keeps the site administrator's hands off your group.
- Automations, and your papers by themselves: a new section under
Settings, "Automations", shows what paperus does by itself, and lets
each owner choose. "Add my papers" (yours alone, on): each morning, the
papers OpenAlex lists under your ORCID iD that are not in a library you
can see yet are added (one you remove stays out; new ones celebrated),
into where your new papers go or an owner you choose, such as your
group. "Published versions" and "Free PDFs" (on) are each owner's: your
library's here, a group's on its page under Groups, switched by its
admins. What the whole server does (errata, reference lists, authors,
OCR) is listed there too. - Your paper is published: congratulations! When one of your own
papers (you are among its authors by your ORCID iD, set under Settings)
is published, the bell says so with a few sparkles: when its published
version is found for your preprint, or when the paper comes into a
library less than a week after it came out, however it comes and
whoever adds it. Once a paper. - Cited! When a paper new in the library cites one of yours (you are
among its authors by your ORCID iD), the bell says so, with a star and
a few sparkles: "“…” cites your paper “…”". Once a citing paper; only
where you can read both; not for your own papers citing each other. - Times cited: each paper's citations as OpenAlex counts them, shown
beside its DOI ("Cited 123 times") and a new way to sort the library
("Times cited", most first). Looked up when a paper is new, then each
week, 50 papers to a request (only DOIs are sent; a switch under
Administration → Settings). When one of your own papers passes 10, 50,
100, 250, 500, 1,000… citations, the bell congratulates you, with
sparkles (the first count sets where you start). - More under the bell: a paper shared with you ("Ann shared “…” with
you"), being added to a group by someone ("Ann added you to Lab"; to a
group of people, what is shared with you there), and a correction
(erratum) found for one of your papers. A notification can be
dismissed (×), or all at once ("Clear all"). - One notice a group while unread: papers added to a group, and
people added to it, come together in one notice for as long as it is
unread, whoever adds them and however much later ("Ann, Ben and Clara
added 50 papers to Lab"): back from a week away, one line a group. - A group's admins hear who joins: "Ann added Ben and Clara to Lab",
under the bell of the group's other admins; members are not told of
each new one. - Groups are run by their own admins: a site administrator creates
top-level groups and sees every group (its admins, how many people and
papers), but no longer runs other people's groups: no adding or
removing people (themselves neither), renaming, deleting or switching.
For a group whose admins are all gone, they may name a new admin. A
group left by its last member, with nothing in it, is deleted. - People to add, found as you type: under Groups, from two letters
on, people are suggested by name or username (a group's admin who is
no site administrator saw no suggestions before). - Saving an edit no longer fails now and then: saving a paper's
details (Edit) while paperus was still working on the same paper (just
after its upload: its reference list, its authors) could end in
"Something went wrong." (the two waited for each other); now one waits
for the other. The same for saving a note.
paperus v0.1.3
paperus now reads your papers the way you do: across the columns, over the page break, and politely past the running header — and you can finally share one with just Anna.
- Share with single people: a paper's Share… now offers people as
well as groups. Choose Anna (found by name, anyone on the server), and
the paper goes to a small group of the two of you, "Anna & Ben", made
the first time and found again after that, with more people as one
wishes. Everyone in it is equal (adds people, renames it); these groups
show under Groups in a section of their own, "Shared with people", and
as owners everywhere else. - On a phone, a long press picks a paper: it starts picking (taps
then pick more), instead of the Select button above the list, which is
gone; the bar at the bottom stays. - Search reads PDFs as fluent mode does: a PDF's text now comes from
its page layouts, in reading order across columns and pages: a
paragraph over a page break is one passage, and running headers, page
numbers, arXiv's margin stamp and the labels inside plots no longer
land in mid-sentence. Figure captions are passages of their own (the
text runs on past them), and each section starts a new passage. After
the update, the layouts and passages of all PDFs are made again in the
background, a few at a time (Administration → Jobs shows how far);
search keeps working meanwhile, and passages whose text stayed keep
their vectors. - Fluent mode reads tables across the page whole: a table over both
columns (between full-width rules, as REVTeX and booktabs draw them)
was read in halves, its left half joining the left column's text; now
row by row. A contents page's page numbers, standing apart at the right,
are no longer taken for a margin note and left out. - A new upload's page shows its title: now and then a paper's page
kept the file name as its title after the upload, although the title
had been found (the page asked at the very moment it was saved, and saw
the file read but the old title); now it reads both at once.
paperus v0.1.2
paperus now skips the publisher's cover page, files the referees' grumbles under "Peer review", and quietly checks every few months whether anyone had to say "erratum".
- Errata are found by themselves: a correction a journal published to
a paper becomes a version of it of its own kind ("erratum", with its
DOI and details), never the version cited or shown, and the paper's
BibTeX cites it with it (addendum = {Erratum: Phys. Rev. A 109, 059901 (2024), doi:…}), as journals ask. Looked for rarely at Crossref (a few
hundred papers a week, each again every half year while under three
years old); APS's errata, which Crossref does not link, by their titles.
A setting under Server turns it off. - Peer review files: a new kind of attachment, "Peer review", for
referees' reports and the authors' replies. The browser extension
recognises them on the page and saves them as such (Nature's "Peer
Review File", reviewer reports), and links to a paper's published
reviews (PLOS's, eLife's); before, Nature's came in as an anonymous
supplement. Supplementary files saved by the extension now say what they
are (supplement, code, data, peer review) and are named as the page's
link reads when the site's own name says nothing ("Peer Review
File.pdf", not "…_MOESM2_ESM.pdf"). - Browser extension: supplementary files at your choice. A new option,
"Save a paper's supplementary files with it" (on, as before): when off,
the popup lists them after saving the paper, each with what it is, to
tick those you want. Links to code and data are kept either way. - A more compact library filter: one line per rule, its name beside
it (above it on a phone). The tags are one line: "has", "has any of" or
"has not" and "Add a tag…", the tags chosen as chips ("or", "not"); the
types the same way instead of a dozen boxes. "Owned by" is in the
filter, as buttons; the menu beside Sort is gone. The panel stays open
when a rule leaves no papers (before, it closed). - Filter by journal: "Journal" in the filter and in smart collections,
with suggestions from one's papers (the most papers first). - Cover pages are passed over: the page a site puts in front of a
paper (IOP's "To cite this article", Taylor & Francis', JSTOR's,
ResearchGate's, a repository's licence sheet) is recognised when the
file is read. The PDF opens after it, fluent mode shows a bar instead
("Show"), and its words stay out of search and passages; the file
itself is unchanged. Files read before are looked at once on the update. - An eighth page twin, Nurture, from a Nature paper as saved: four
other files on another site (Supplementary Information, Reporting
Summary, Peer Review File, Source Data), each checked for what the
extension saves. Nature's link texts end in "(download PDF )", so the
Peer Review File would have been saved as "Peer Review File (download
PDF ).pdf": that ending is left out now. Springer Nature's file names
(…_MOESM3_ESM.pdf) mark a supplement wherever they are served. Making
twins keeps link texts that say what a file is (peer review, reporting
summary, source data) and replaces ORCID iDs with placeholders; two
twins had the real authors' iDs and were made again. - The extension no longer keeps a database's front page as a paper's data
("deposited in the EMDB …"): at EBI's and NCBI's sites only entries
count. - Error messages say what the server or the connection answered, instead
of only "Something went wrong".
paperus v0.1.1
"Select all" now actually selects all, and reading files no longer waits while arXiv does its three-second breathing exercises.
- Reading files no longer waits behind lookups: reference lists,
open-access copies, journal versions and authors are looked up in
queues of their own, one per service (arXiv, Crossref, OpenAlex; under
Jobs as one card, "Looking things up"), so files are read at full speed
instead of between the pauses arXiv asks for (after a large import,
about 10 files a minute), and no service's lookups wait behind
another's. Lookups waiting at the update move there too, and "Reading
files" counts only files. - Reference lists of journals' papers from Crossref first: on our
papers as complete as arXiv's LaTeX sources, more often with DOIs, and
without arXiv's 3 s pause and download; preprints still from arXiv. - "Last active" instead of "last login" among the users: any use of
paperus (website, app, extension, assistant), not only logging in,
which happens every few weeks. - The jobs charts' times no longer overlap: full minutes only.
- Files in the trash (or of a paper in it) get no OCR and no vectors until
restored, so trashing a large import of scans right away frees the ML
service; a file deleted while its scans are read ends that quietly
instead of with an error. - Imports say where papers were already: besides "already here",
how many are in other libraries one sees ("1 already in Lab"). A paper
in a group one may only read, with files the group's copy lacks, is
added as one's own with them (before, its files were dropped without a
word). The bell tells when an import is done (it may take long), with
its counts. - Checkboxes on a computer: every row's box is always there, and the
list's own box above them selects all of its papers (a dash when some
are picked; a click then clears). On a phone, Select as before. - Select all in the selection's bar ("Select all 1,429"): every paper
the list's filters match, also those not loaded yet, changed by the
server in one go (up to 50,000). Before, only the rows loaded so far
could be picked, and changing more than 1,000 picked papers failed; now
they are sent in parts. - Whose papers the library lists, at hand: a menu beside Sort and
Filter chooses All, Only mine or one of one's groups (for people in a
group). - arXiv's versions of a preprint are versions of the paper: a PDF
named with its arXiv version ("2401.12345v3.pdf") goes to the paper's
preprint labelled "v3" (made if there is none), instead of becoming a
second main file; the version cited is the published one, else arXiv's
newest. Uploads, the browser extension, open-access copies and imports
all count; existing preprints whose PDF names its version get the label
on the update. - A paper's attachments in one row: code, data, supplementary files
and slides, each chip saying what it is (before, links showed their long
names and supplementary files hid under "Details and files"); "+
Attachment" adds a link or uploads a file in that role. - The search models in a table, with what they take in memory: the
ML service measures each model once loaded (RAM, or the GPU's memory),
and the settings show it with the scores (English and multilingual),
size and release date, the speed here, and whether the models in use fit
together, with a warning when they do not. A model that does not fit is
loaded after the idle ones are unloaded; on the GPU a reranker or OCR
model then runs on the CPU. - Lists (library, collections, Recent) no longer stop loading at times
while scrolled to their end, with only "Show more". - A long group name wraps on a phone instead of pushing the group's
buttons off the screen. - Recommended models, for a CPU and for a GPU: choosing a model, the
administration marks the ones to use where the ML service runs (setup
too): arctic-embed-m-v2 for search and PP-OCRv6 for scans everywhere,
the bge-reranker-v2-m3 reranker on a GPU. On 200 test questions
arctic-embed-m-v2 was the best search model, and without a reranker as
good as any with one, so a CPU, which does not rerank by default, loses
little.
paperus v0.1.0
Version 0.1: officially a beta, which in academic terms means it has passed peer review by one very patient reviewer.
The first beta. What paperus does:
- A research library on your own server, for one person or a group:
papers, books, lecture notes and meeting notes, with their versions
(preprint, journal version), files, notes, highlights and comments;
shared with groups and subgroups, private things kept private. - Papers come in easily: by DOI or arXiv ID, by dropping PDFs (their
details read from the file), from Zotero, Mendeley, JabRef or EndNote,
and with the browser extension (Firefox; Chrome as a zip), which saves a
publisher's page with its PDF, supplementary files and code links, and
gets missing PDFs with your own access. - Open-access copies, published versions and authors are found for
you: PDFs from arXiv and Unpaywall, a preprint's journal version when it
appears, authors linked to people by ORCID. - Search by words and by meaning across titles, abstracts, full texts
(scans read with OCR) and notes; similar papers and reference lists. - Reading: the PDF in the browser or a reflowing reading mode, with
highlights; the app for Linux, Windows and macOS, also offline. - Tidy up: duplicates, papers without details or PDFs, authors to
check, tags, collections and smart collections, archive and trash. - Your AI assistant (Claude, ChatGPT and others, over MCP) searches
and reads the library for you, citing the passages, and may add papers or
notes when you let it. - Yours to keep: backups and exports (BibTeX, the whole library); open
source (Apache-2.0); database changes come as migrations, each release
tried as an update of the one before.
What changed since 0.0.29:
- Split a version off its paper: a version merged in by mistake (a
paper's code from Zenodo became a "published" version of the paper, cited
over its own, with its title) or another work's DOI becomes an item of its
own with its files: "Split off" on each version of a paper that has
several. - Code and data attached to their paper, not merged: software and
datasets offer "Attach to a paper…" instead of "Merge into…": the paper
links to them (their DOI) and takes their files as code or data; the item
goes to the trash. The paper is suggested by its title. - Papers whose titles share words: suggested in the merge and attach
dialogs (stems, without stop words: "Code of Spin qubits" finds "Spin
qubits"), also when no DOI is shared. - Search lists titles first: above the passages, up to five papers
whose titles have all the searched words ("In titles"); the ranking below
is as it was. - The browser extension, on a paper one has: a click no longer saves it
all again; the popup says it is in paperus and offers the page's
supplementary files that are not ("Add to this paper"), and "Save again".
Supplementary files are listed under the paper, never as papers of their
own. - The PDF to the version the page is about: saving from a preprint's
page put its PDF on the journal version; it goes to the preprint now. The
popup says which version it saved to (Journal version, Preprint, …). - Captchas on Manage → Missing: a paper stopped by a check for robots
is tried again at the end in a tab in front, where such checks often pass
by themselves (or one answers them); up to three minutes, then the PDF is
saved and the tab closes. - The extension's tabs out of sight: the tabs it opens for its work are
hidden in Firefox (new permission: hide tabs) and open in a minimised
window in Chrome. - "Get the PDF with the extension" on a paper's page, next to "Upload
files", for a version without one. - The Manage counts in the menu follow at once: a merge on a Manage
page left its number standing until the next page or minute. - "Probably the same person" without collaborations' namesakes: two
names that fit ("A. Huber" and "Anna Huber") on one paper's author list
were suggested as one person, with all the paper's other authors as
co-authors in common (2919 on a big collaboration's paper). Two names on
one author list are two people: no longer suggested. - arXiv papers' authors by their journal version: arXiv's records
give names only, no ORCID iDs, and OpenAlex mostly files a preprint
under its journal version's DOI, so asking it by the arXiv DOI often
found nothing and the authors were linked by name alone. A preprint
whose journal version is found later is now asked about again, by the
journal's DOI (papers that have one already, once, after the update);
a journal's DOI is asked before arXiv's. - Preprints' journal versions within minutes: new preprints were
checked a few dozen an hour, so a large import of preprints took half a
day. Now they are checked soon after an import or a file is read
(otherwise within five minutes, e.g. from the browser extension), up to
500 a run, 50 to a request to arXiv with its pause between, OpenAlex a
moment apart; the hourly job is left for the monthly rechecks. A journal
version found has its authors looked up at OpenAlex by its DOI at once
(ORCID iDs), not at the next hour.
paperus v0.0.29
Your papers can now line up by author, year, journal or title, which is more than can be said for your desk.
- Sort the library and collections: by date added, title, first
author, year, publication date or journal, either way (a button beside
the choice turns the order round); dates start newest first, words A to
Z. Papers without the value come last either way; an author's papers go
by year, a year's by first author. Kept in the address and remembered
on the device; in the API assortandorder. Papers keep their
first author and journal as cited for this (migration 0012 fills them
in for the papers there are). - Edit a paper's details: Edit had only the title and owner; now also
its authors (rows to reorder, with suggestions from the library's
authors as one types; a picked one keeps their ORCID iD), year, type,
journal and abstract. Each change is kept over later lookups;
"Restore original" takes it back. Saving no longer fails when paperus
itself changed the paper meanwhile (its authors linked, an abstract
found); only someone else's edits are not overwritten. - "Details and files" in sight: a row under "Similar papers" and
"References" on a paper's page (with how many versions and files), not
only in its menu. - The corner trays close by themselves: uploads and tasks (getting
PDFs with the extension, adding papers) after 10 seconds once done, 30
when some PDFs were not got; a failed upload stays until closed; never
while the pointer or the focus is on them. - The extension's button shows whether you have the paper: when a
paper's page (or its PDF) opens, a green check if it is in paperus with
the PDF of the version the page shows, an orange "!" if that PDF is
missing (a click saves it; the tooltip names another version's PDF if
there is one). Only the page's DOI or arXiv ID is looked up in your
paperus; the extension's settings turn it off. - The extension's popup waits while the pointer is on it: it closed
under the mouse in Firefox (its wait listened for events Firefox's popup
does not send). - A page's list saves to the default owner set in paperus: with "My
default owner" in the extension's settings, the list's owner menu
started at one's own library and saved there. - Getting PDFs for Missing always ends: a paper that failed
unexpectedly stopped the run's report, and its task hung at "x of y". - "Add to a collection" waits for one's collections: clicked while
the page still loaded them, it asked for a new collection's name. - ScienceDirect's PDFs through its check for scripts: its PDF link
now shows a check that needs scripts (pdfft?crasolve…) before the PDF,
in a tab too; the extension looked at its tab out of sight while that
check still ran, and got nothing. In Firefox it now follows the tab from
the start, wherever the site sends it (redirects, checks, the PDF's
signed address), and keeps the first PDF that arrives; it gives up when
the tab rests on a web page (a login) for 15 seconds, or after a minute.
The same for a PDF open in Firefox's viewer. Tried in Firefox. - AIP's and ACS's PDFs past Cloudflare's "Just a moment…": when the
extension's own fetch of a PDF met a check for robots, it gave up
without trying a tab, as a tab could not get past one. In Firefox it can
now (a check that needs only scripts lets the tab through, as it lets
the reader), so the PDF link is opened in a tab out of sight there too;
a check that wants a person still says so, after 15 seconds. Tried in
Firefox on AIP and ACS. - A seventh page twin, The Journal of Comical Physics (AIP Perishing):
its PDF link sends the page's fetch on to another site, which the
extension then fetches itself. Making twins no longer takes a site's own
name in an address for a person's (AIP's/aip/became "someone"). - Old Wiley DOIs, whole: DOIs in the old SICI form carry angle
brackets (10.1002/1521-3978(200009)48:9/11<965::AID-PROP965>3.0.CO;2-V);
the extension cut them before the<, so the DOI was not found and
the paper saved with the page's details and the cut DOI. Whole now, from
the page's tags, its address and its links. - Copy buttons over plain HTTP: a paperus on the local network
(http://…:8000) has no clipboard API in the browser, so "Copy" next to
a new access token did nothing (it selected the token, unseen). It
copies there now (as the browser still lets a click), or says to copy
by hand; the same for the OIDC redirect address and a note's text that
could not be saved. - A paper's DOI under its authors: with its PDF open, a paper's page
showed its DOI only under "Details and files". It is under the authors
now, always, a link to the paper at its publisher (via doi.org). - A new collection from a paper's page: the collection picker's last
choice is "New collection…": name it, Enter, and the paper is in it (a
group's collection for a group's paper one may change, else one's own).
With none to choose from, the + asks for the name right away.
paperus v0.0.28
Six publishers walk into a test suite: EEEK, Seance, APES, ScienceIndirect, Nurture and the American Comical Society. Only their PDFs come out.
- A PDF open in Firefox's viewer, where the site turns the extension
away: Firefox lets no extension read its viewer, so the extension
fetched the PDF's address itself, and some sites answered that with a
web page (a check for robots, a login). The address is now loaded again
in a tab out of sight and a copy of the PDF kept as Firefox receives it
(new permissions in Firefox: webRequest, webRequestBlocking,
webRequestFilterResponse). An expired signed address says so. - The extension saves IEEE Xplore's papers: their pages have no
citation_*tags, so the popup took one for a page listing papers (its
references) and offered those. Their details (DOI, authors, venue,
pages, PDF link) are read from the page's own data now, and the PDF from
the frame that IEEE's PDF link (stamp.jsp) shows it in. - Science's DOIs and ACS's titles, as the pages give them: of a
page'sdc.Identifiertags the extension read only the first, on
Science's pages its own ID, not the DOI (saving worked by the DOI in
the address); ACS's titles kept the line break inside their tag. - Page twins for the extension: publishers' pages as the extension
sees them, without their content, with unrelated titles, invented
authors and parody names, checked inextension/twins/pages/;npm testthere (and in CI) reads each as the extension does. The first six:
EEEK Xplore, Seance Advances, Phys. Rev. Lol., ScienceIndirect, npj
Quantum Misinformation and ACS Fauxtonics. They catch this release's
page bugs (IEEE, Science's DOI, ACS's title). Where a PDF link forwards
(ScienceIndirect's page in between and check for robots, EEEK's frame),
a twin's chain plays the network and the PDF is fetched as in the page. - ScienceDirect's PDFs in Firefox, saved from the article's page: its
PDF link, opened in a tab out of sight, reaches the PDF; the extension's
own fetch of that address gets a check for robots ("Security
verification") instead. The PDF is now loaded again there and a copy
kept as Firefox receives it, as for a PDF open in Firefox's viewer. - The paper's PDF, not its supplement: on a page without a
citation_pdf_url(Science Advances) the extension took the first PDF
link naming the paper's DOI, which could be the supplementary material
(/doi/suppl/…/…_sm.pdf), and saved it as the paper. Supplements are
passed over now, and the article's own PDF (/doi/pdf/<DOI>, or its
reader's/doi/epdf/<DOI>) comes first.
paperus v0.0.27
A paper rarely travels alone: now its supplements, its code, its data and the rest of its author's oeuvre come along for the ride.
- Links to a paper's code and data: on its page, next to its notes
("+ Code or data"; GitHub is guessed as code, Zenodo or Figshare as
data), kept through merges and exports, shown to AI assistants. Files
get the roles "code" and "data"; the same file uploaded again to a
paper in the same role is not stored twice. - The extension brings a paper's supplementary material: the
supplementary files its page links (Nature, Elsevier, Wiley, ACS,
Science, PNAS, PLOS, IOP, bioRxiv, arXiv's ancillary files; up to 10,
100 MB each) are saved with it, and links to its code and data where the
article says they are its own (its abstract, a "Code availability"
section, "the data are available at …"), never the reference list's. - Pages that list papers or files: on a course page, a publication
list or a journal's contents, the extension's popup lists the files and
papers it links, marks those you have, and saves the ticked ones, each
as a paper of its own. Files over 100 MB only when you ask ("Save
anyway"). - Someone's Google Scholar profile: the popup lists all their papers
(at most 1,000) and adds the ticked ones, by arXiv ID or by the DOI
paperus finds for the title at Crossref (same title, first author and
year), else as listed. The library lookup and adding a paper by its
title are in the API too (title:…,find_doi). - The extension's log keeps nothing of one's own: web addresses keep
their site, path and parameter names, not the values (signed PDF
addresses, sessions, proxy tickets, search words); libraries appear by
id. Its settings page keeps the last 20 steps (to copy into an issue),
not 400.
paperus v0.0.26
Same extension, shorter cover letter: Firefox's reviewers get the short version, and slow pages get twenty seconds.
- Extension 0.0.25's changes reach Firefox: its upload to
addons.mozilla.org was refused, as the notes for Mozilla's reviewers had
grown too long. They are short again, and the build refuses long ones. - Saving never waits for ever at "Reading the page…": a page busy with
its own scripts (Science Advances) could keep the browser from running
the extension's. The page's details are read after ten seconds at most;
with no answer after twenty, the save goes on from the page's address. - CI: the browser tests give the worker's reading of an upload its own
minute (WebKit ran out of time).