Description

Rediscover your notes: open a random note, resurface notes from this day in past years, and see writing statistics for your whole library (counts CJK correctly).

Additional Information

Links:	Repository Homepage View source
Maintainers:	smiletalker
Version:	1.0.1
Minimum app version:	3.0
Downloads: This version:	6
Last updated:	2026-09-26T19:11:10Z

Recall — a Joplin plugin

Rediscover your own notes, and see how much you've actually written.

• Open a random note — one click (or toolbar button) jumps you to a random note. Great for serendipitous review.
• On This Day — surfaces notes you created on this calendar day in previous years, the way photo apps show "memories". Joplin has no built-in equivalent.
• Daily Recall digest — generates a single note combining On This Day plus a handful of random notes, with clickable links. Optionally built automatically on startup.
• Writing statistics — the word count of your whole library: totals, a 12-month heatmap, and breakdowns by year and by notebook.

Writing statistics

Screenshot generated from a synthetic library (node tools/demoData.js), not from anyone's real notes.

Every other Joplin word-count plugin counts the current note, and all of them tokenise with /\b\w+\b/, where \w is ASCII-only. A 500-character Chinese note reports 0 words. Recall counts the whole library, and counts CJK correctly:

Text	/\b\w+\b/	Recall
今天下午去看了房子，中介说学区还不错。	0	17
读完这本书花了三天，里面讲的 survivorship bias 很有意思。	2	19
これはテストです	0	8
Today I went to see the apartment.	7	7

How it counts. Han characters, hiragana and katakana count per character — the convention in Chinese and Japanese publishing. Every other script, including Hangul and Cyrillic, counts per word. Before counting, markdown syntax, URLs, attachment links, HTML tags and inline code are stripped, and fenced code blocks are excluded entirely and reported separately. Full-width punctuation is reported on its own line rather than folded into the total. Recall's own generated notes are never counted.

Performance. Counts are cached per note and only recomputed when a note's updated_time changes. On a 3,400-note library the first run reads every body in ~70 paged API calls; later runs re-read nothing and only refresh the metadata. The cache lives in a plugin setting (~220 KB at that size) and can be rebuilt from the panel.

Where it lives. Open it from Tools → Recall → Writing statistics. It is a dialog, not a panel, because this is a look-at-it-occasionally feature and a panel would hold a column of the layout open for it permanently. From the dialog you can copy the figures as Markdown or rebuild the cache.

Usage

After installing, you get:

• A die icon in the note toolbar → open a random note.
• A Tools → Recall menu with every action.

Settings

Settings → Recall:

Setting	Default	What it does
Random notes in Daily Recall	3	How many random notes the digest includes
Exclude completed to-dos	on	Skip finished to-dos when picking notes
Target notebook ID	(empty)	Where generated notes are stored; empty = default
Generate Daily Recall on startup	off	Auto-build the digest when Joplin starts

The "On This Day" and "Daily Recall" notes are upserted per day — re-running updates the same note instead of creating duplicates.

Building from source

npm install
npm test            # counting assertions, including the CJK cases above
npm run dist        # produces publish/com.github.smieltalker.recall.jpl

Install the resulting .jpl via Joplin → Settings → Plugins → gear icon → Install from file.

Checking the counter against your own library

tools/ holds two harnesses that run the real code against a real library with the plugin API stubbed out (tools/apiStub.ts), so you can check both the numbers and the layout without launching Joplin:

npm run harness

# numbers: totals, per-year and per-notebook breakdowns, cache behaviour
<dump> | node .harness-build/tools/statsHarness.js

# layout: renders the dialog to a standalone HTML file
# (a 5th argument of "light" or "dark" picks one theme; omit it for both)
<dump> | node .harness-build/tools/previewHarness.js out.html \
          src/webview/stats.css src/webview/stats.js

<dump> can be node tools/demoData.js, which emits a deterministic synthetic library — that is how the screenshot above is produced, so no real notes are involved:

node tools/demoData.js | node .harness-build/tools/previewHarness.js \
  docs/preview.html src/webview/stats.css src/webview/stats.js dark

To check it against your own library instead, pipe NDJSON on stdin out of a read-only query against database.sqlite — one {"t":"n","v":{...note fields...}} line per note and one {"t":"f","v":{...folder fields...}} per notebook. Piping rather than dumping to a file keeps your notes off disk.

License

MIT