Skip to content

Imports from Instapaper, Readwise Reader, Wallabag, Linkwarden and Karakeep, and a bookmarks file in the export - #169

Merged
NiallJoeMaher merged 8 commits into
mainfrom
feat/imports-and-bookmarks-export
Nov 10, 2026
Merged

NiallJoeMaher merged 8 commits into
mainfrom
feat/imports-and-bookmarks-export

Conversation

@NiallJoeMaher

Copy link
Copy Markdown
Contributor

Summary

People arriving from the apps Pluck sits beside could not bring their libraries, and the export could only be read back by a script.

 packages/extract/src/imports.ts          # each parser maps to the existing ImportEntry and nothing more
   pocket (CSV), raindrop (CSV), omnivore (zip), html (browser bookmarks)
+  instapaper (CSV)      URL, title, Selection as the note, Timestamp, Tags; a folder becomes a Tag
+  readwise (CSV)        URL, title, Document tags, Saved date; feed rows left out
+  wallabag (JSON)       url, title, tags, created_at, and the page Wallabag kept
+  linkwarden (JSON)     url, name, description as the note, tags; a collection's name becomes a Tag
+  karakeep (JSON)       link url, title, tags, note, createdAt; a manual list becomes a Tag
 export zip
   pluck.json, items/*.md, lists
+  bookmarks.html        # every Item with a link, in the format browsers and the apps above import, a folder per Status

Settings picks the source from the file's name, as before. A CSV, JSON or bookmarks file over 100 MB is refused at once (only Omnivore's zip is read a page at a time). Every place that names the import sources is updated, in all ten languages.

Evidence

  • imports.test.ts: one test per new source from a real-shaped fixture (quotes and commas in a CSV title or quotes in JSON, a missing title, a non-Latin title, a duplicate URL), and malformed JSON fails as unreadable.
  • import.test.ts: each of the five fixtures imports to the right Items with their Tags, notes and dates; a duplicate makes one Item; a second person is untouched; the wrong app's file fails as "could not be read".
  • export.test.ts: bookmarks.html is in the zip; read back with the repo's own bookmarks parser it gives the same URLs, titles, Tags and dates; &, < and quotes are escaped; a Note is absent.
  • Formats were checked against each project's own exporter source (Wallabag, Linkwarden, Karakeep), and against Wallabag's Instapaper importer and a 2023 example file for the two CSVs.
  • pnpm typecheck, pnpm lint and the unit suite pass, rebased on main. Every commit is signed off.

Merge Danger

Door: two-way

Blast Radius: import

  • Readwise Reader is the least certain: the shape of its tags cell and the choice to drop feed rows rest on third-party evidence, not a real export. Try one before launch.
  • Notes are written to bookmarks.html as <DD>, but Pluck's own bookmarks import does not read them back, and a Tag with a comma in it splits on re-import.
  • The new source names and two changed lines are in all ten languages; the translations were written here and have not had a native read. The Privacy page's source list changed; its "Last updated" date did not.

🤖 Generated with Claude Code

NiallJoeMaher and others added 8 commits November 10, 2026 22:20
…as the note

Settings, Import takes Instapaper's CSV (URL, Title, Selection, Folder,
Timestamp and, in exports since late 2025, Tags as a JSON array in one
cell). A folder's name becomes a Tag; Unread, Archive and Starred say where
a link sat, so they do not, and the star is dropped. An older export with
no Timestamp or Tags column reads too.

The source picker now lists its sources from one table of the file each
exports, which also decides whose file a dropped one is: the one its name
says, else the one picked if it is that kind of file, else the first source
that exports that kind. The import source type is the schema's, in the one
place.

Co-Authored-By: Claude <noreply@anthropic.com>
Signed-off-by: NiallJoeMaher <nialljoemaher@gmail.com>
…was kept

The CSV from Reader's "Export Library as CSV" (Title, URL, ID, Document
tags, Saved date, Reading progress, Location, Seen). Document tags is one
cell written as a Python list, read whichever quote it uses. Rows whose
Location is feed are what Reader fetched from a subscription, not what the
person kept, and stay behind. Reading progress, Seen and where a document
sat are dropped.

Co-Authored-By: Claude <noreply@anthropic.com>
Signed-off-by: NiallJoeMaher <nialljoemaher@gmail.com>
An array of entries: url, title, tags (labels, or objects with a label as
its API writes them), created_at and content. The content is the article
as Wallabag kept it, with no document round it, so one is put round it and
it is kept without fetching, as Omnivore's pages are. Where Wallabag's own
fetch failed the content is its notice, and that link is fetched afresh.
Archived, starred and annotations are dropped.

The fields read from a JSON export are a schema in shared-schemas: a field
that is missing or of another type reads as absent, a file of another
shape is refused with a short reason, and the import says it could not be
read.

Co-Authored-By: Claude <noreply@anthropic.com>
Signed-off-by: NiallJoeMaher <nialljoemaher@gmail.com>
Collections each hold their links: url, name, description (the note), tags
and dates. A collection's name becomes a Tag beside the link's own, but for
Unorganized, where links in no collection sit. importDate, the day a link
Linkwarden itself imported was first kept, wins over createdAt. A PDF or
image with no URL stays behind; pins, colours, icons and how collections
nest are dropped.

Co-Authored-By: Claude <noreply@anthropic.com>
Signed-off-by: NiallJoeMaher <nialljoemaher@gmail.com>
…lists

The bookmarks array of a Karakeep (once Hoarder) export: createdAt in Unix
seconds, title, tags, note and a content that is a link or a text. Only a
link has a URL, so a text and an uploaded file stay behind: an import makes
link Items and has no way to make a Note. A list a link sits in becomes a
Tag. Archived is dropped.

A file from another app than the one picked fails as unreadable and makes
nothing.

Co-Authored-By: Claude <noreply@anthropic.com>
Signed-off-by: NiallJoeMaher <nialljoemaher@gmail.com>
llms.txt, the home page's structured data and the Privacy page's "What we
store" (in all ten languages) list Instapaper, Readwise Reader, Wallabag,
Linkwarden and Karakeep beside Pocket, Omnivore and Raindrop. The line
under the drop box in Settings names the kinds of file (CSV, JSON, zip,
bookmarks HTML) instead of each source's, which the picker above it
already does: nine would not fit the box.

Co-Authored-By: Claude <noreply@anthropic.com>
Signed-off-by: NiallJoeMaher <nialljoemaher@gmail.com>
…and other apps import

One more file in the zip: the Netscape bookmark file every browser reads.
Each Item with a URL is a link with its title, the day it was plucked
(ADD_DATE) and its Tags (TAGS), with its note under it, in a folder for its
Status; a Status with no links gets no folder. Notes, Todos and Things have
no URL and are left out. Everything is escaped with the entities a
browser's own file uses.

Folders, not one flat list: every reader of the format takes them (browsers
keep them, other apps turn them into folders or tags, Pluck's own import
ignores them), and they are the only place a Status survives the trip. The
price is that a folder's lines wait in memory until every Item has gone
past, a few hundred bytes a link.

Settings, README.txt in the zip and llms.txt say the file is there.

Co-Authored-By: Claude <noreply@anthropic.com>
Signed-off-by: NiallJoeMaher <nialljoemaher@gmail.com>
…rker runs out of memory

Co-Authored-By: Claude <noreply@anthropic.com>
Signed-off-by: NiallJoeMaher <nialljoemaher@gmail.com>
@vercel

vercel Bot commented Nov 10, 2026 •

Copy link
Copy Markdown

The latest updates on your projects. Learn more about Vercel for GitHub.

Project Deployment Actions Updated
pluck Ready Ready Preview Nov 10, 2026 10:22pm UTC

Request Review

@NiallJoeMaher
NiallJoeMaher merged commit 9220c73 into main Nov 10, 2026
6 checks passed
@NiallJoeMaher
NiallJoeMaher deleted the feat/imports-and-bookmarks-export branch November 10, 2026 22:27
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant