This release adds a manual "Recalculate Word Counts" action in Settings, an agent-friendly Markdown
mirror for every documentation page, and a file-based "Create New Journal" flow that lets you pick
the exact file. It also fixes Day One JSON and TXT imports, and corrects two website issues around
screenshot layout and sitemap freshness.
Added
Agent-friendly docs mirrors (Copy page, llms-full.txt): Each documentation section page now has a
"Copy page" dropdown (Copy page, View as Markdown, Open in ChatGPT/Claude/Perplexity), a per-page
Markdown mirror at website/docs/.md, and a
discovery tag. website/llms-full.txt adds the full concatenated Markdown text of every docs
section as a sibling to the curated llms.txt. An nginx Link: rel="canonical" header canonicalizes
each Markdown mirror to its HTML page rather than noindex, so ranking signal consolidates onto the
HTML page.
Manual word count recalculation (TODO-0111, #275): Settings → Advanced now has a "Recalculate Word
Counts" button that rescans every entry in the journal and rewrites any stale word_count. The
recalculation skips locked entries and never touches an entry's last-modified date. It runs only
on demand, so there is no automatic or background recalculation.
Changed
File-based "Create New Journal" flow (TODO-0112): Creating a journal no longer forces a new folder
per journal. On Windows/macOS/non-Flatpak Linux, + Create New Journal now opens a native Save
dialog (pre-filled with the default location and diary.db) so you pick or rename the exact file,
the same way Open Existing already worked. On the Flathub build, Create still shows a dialog-free
form (a native save dialog there can return an unusable temporary path), now with an editable
Filename field so you can create a second journal in the same default folder. If a database file
already exists at the location you choose, the app refuses with a clear error instead of a raw
filesystem failure.
Fixed
Day One JSON import failed on entries with no text (#294): Day One omits the "text" key entirely
for blank entries, but the importer required it, so the whole file was rejected with a
missing-field error instead of importing anything. text now defaults to empty when absent, and
entries with no text content (after trimming) are skipped rather than imported as blank entries.
Day One TXT import didn't recognize the current export date format: the date parser only ever
accepted the legacy "DD MMMM YYYY" layout, but Day One's own current export guide documents "MMMM
D, YYYY at H:MM:SS AM/PM TZ" (e.g. "June 24, 2016 at 10:59:06 AM MDT"), and a real export in that
format would fail every entry's date parse. The importer now tries the modern format first and
keeps the legacy format as a fallback. Also fixed: tab-delimited Weather:/Location: metadata lines
that real exports place after the date line were previously leaking into the imported title/text
instead of being skipped, and a blank entry (title and body both empty, including one with an
unparseable date) is now silently skipped instead of importing an empty entry or, for the bad-date
case, failing the whole import.
Docs screenshot layout shift + sitemap freshness drift (SEO audit): Documentation pages now set
explicit image dimensions, read from the shipped image file at build time, so the page reserves
layout space and screenshots no longer shift as they load. The homepage and the fixed pages now
take their sitemap freshness date from a manually maintained date instead of the file's
modification time, so an unrelated CSS or JS rebuild no longer bumps those pages' freshness
signal.
Website docs: Added screenshots and diagrams to 10 of the 11 user guide pages (FAQ excluded as
reference-only). Screenshots are live-captured from the real dev app (journal creation, the editor
toolbar and multi-entry Timeline, calendar/search/import/export, three Preferences tabs,
Statistics, and two Backups panel states), plus two hand-built SVG diagrams (the plugin sandbox
boundary and the backup retention/rotation lifecycle) in the site's own dark/gold palette. Images
ship as WebP (q80) via a new image handler in the docs generator that wraps each in a captioned. The convention is documented in website/CLAUDE.md.
