The self-protection guard refused every change to the admin's own account.
It now only refuses blocking and deleting yourself; a quota change is fine.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
- The "online reputation checks" rspamd profile used groups_enabled,
which means ONLY those groups: with the switch on, a message got the
RBL lookups alone (1 rule instead of 19, score 0.00) and Bayes never
ran, so every spam lesson was ignored. It now disables nothing.
- A reply or reply-all teaches the answered message as ham (not
forwards or new mail, not one's own Sent/Drafts, never over an
explicit spam mark; once per message). message.sent now carries the
draft's kind and reply_to_message_id.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
The first start tried the default languages once; a DNS failure at that
moment left them missing for good. Languages asked for are now kept as
pending in state.json and retried in the background (30 s, doubling to
15 min) until they are in, across restarts; an install from the admin
panel joins the same queue. A state from before this, with nothing
installed, counts as an unfinished first start. The admin panel shows
what is waiting.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
- Email translation moves from LibreTranslate to our own translator
service (translator/): Mozilla's Firefox Translations models on the
Bergamot engine, fastText language detection (176 languages). Polish,
English, German, French, Spanish and Italian on first start (~350 MB
instead of ~10 GB); Admin panel > Email translation installs and
removes languages. Built and pushed as the `translator` image.
- Every message is translated paragraph by paragraph as plain text;
names travel with their sentence and are kept only where the model
changed them; a sentence start the model copied ("Enclosed") is tried
again in lower case; list markers are not sent
- The Translate button shows only for installed languages
- New logo: favicon, app/maskable/iPhone icons, notification badge,
email logo and calendar copy regenerated; BIMI logo rebuilt in SVG
Tiny PS with its gradients; masks render in browsers
- Launch screen background is the app's dark grey
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
- Snooze (with a Snoozed folder), keyboard shortcuts with a Settings
section, mail templates (Compose > More > Templates), a Subscriptions
folder with bulk unsubscribe/delete, and tracker blocking with a
Privacy toggle (on by default)
- Own dialogs instead of browser popups; emoji picker in compose; compose
windows survive a visit to Settings
- Sentence completion learned from the user's sent mail (no AI)
- Interface in English and Polish (one file per language, English as the
fallback), picked from the browser on first run, then kept
- Email translation on the server via LibreTranslate: Translate button
and "Translate to...", messages in several languages translated part
by part, names/addresses/emoji/file names/phone numbers kept, an
animated line while translating
- Antispam "Mark as spam" loads with the message, not when the menu opens
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
- Invitations: each user can invite two friends by email (sign-up works
through the link even while registration is closed); admin-editable
template, subject and sender.
- Invitation sender defaults to the system-mail From address, which the
SMTP account is allowed to send as (was postmaster@APP_DOMAIN, refused
with 553 when the two differ).
- Sign-in, sign-up and forgot-password fields show the fixed
@domain after the typed name; the registration status reports it.
- System mail gets a Date header; the invitation's plain-text part keeps
its paragraphs and line breaks.
- SMTP EHLO uses SMTP_HELO_NAME / APP_DOMAIN, not the container name.
- DNS recommendations for the mail domains.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Security (ultimo-security-report.md):
- Security headers on every response, the app shell included: they lived at
server level and nginx dropped them in any location with its own header.
CSP, framing, nosniff, referrer, permissions (docker/nginx-security.conf).
- Plain HTTP forwarded by the reverse proxy is redirected to HTTPS; HSTS on
HTTPS responses. server_tokens off.
- Cross-site state-changing /api requests refused (login CSRF), WebSocket too.
- Validation errors in the usual error shape, without the request model.
- New-password policy: max 128, common/own-address/repeated refused,
breached-password check (HIBP k-anonymity, fails open).
- Message HTML: links keep only http(s)/mailto (sanitizer v6); the HTML
endpoint carries its own no-script CSP.
- Outbound fetches pinned to the checked IP (DNS rebinding); external
IMAP/SMTP servers re-checked before each connection; CGNAT and mapped
addresses counted as non-public.
Fixes from testing:
- IDLE loop catches up before idling: mail that came while the worker was
away no longer waits up to 10 minutes.
- Restore from Spam/Trash/Archive moves only the removed messages, not your
reply in Sent.
- Narrow desktop list stacks sender/subject/snippet (subject was squeezed
to nothing at the default width), with hover actions.
- Phone: wide emails scale to fit the screen.
- Singular forms ("1 message"); no zip button for a single attachment;
app passwords no longer claim "never used" for mail apps.
- ruff format across the backend.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Hosted mailboxes (SPEC App. D), behind HOSTED_MAILBOXES (off: unchanged):
- Stalwart mail server, provisioned by `python -m app.hosted.provision`;
directory read from a Postgres view, app passwords only for mail apps
- native account per user, the worker signs in as master for the user
- raw messages of hosted mailboxes only cached, fetched back on demand
- migrations 0025 (accounts.kind, directory view), 0026 (IMAP ids bigint)
Mail:
- dark message mode switch in the reading pane (Thunderbird's algorithm)
- moves and folder creation quote the folder name ("Deleted Items"):
swipe-to-delete / Trash no longer bounce back
- thread unread counters: lock before recount, sorted lock order,
UIDVALIDITY re-import recounts, nightly repair_threads
- sidebar accounts sorted alphabetically, more space above headings
Settings: Security headings renamed; antispam accounts are explicit
(none ticked = none filtered), input widths, Activity hidden.
Plugin bundles served with Cache-Control: no-cache.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
A reminder notification links to its day, event and occurrence
(/p/calendar?date=…&event=…&start=…). The calendar opens on that day with the
event's details showing — the full calendar on a computer; on a phone any
calendar link leads to Mail with the Calendar panel slid in (the big-screen
page never opens there), scrolled to the event, details open.
The panel's agenda kept no scroll position across re-renders: opening an
event's details lower down jumped the list back to the top. It stays put now.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
App passwords (SPEC App. D §D.5): one generated password per app, shown once,
revocable on its own, revoked on a password reset; never accepted by the web
sign-in. Settings → Security → App passwords (migration 0024).
CalDAV / CardDAV server (SPEC Appendix E), built in: /dav/ with
/.well-known/caldav|carddav, HTTP Basic with app passwords only (cached checks,
throttled failures). The calendar and contacts plugins serve their own tables as
collections (Plugin(dav=...)):
- calendars as iCalendar: series with overrides, time zones, all-day, attendees
and RSVP, alarms <-> reminders; MKCALENDAR / PROPPATCH / DELETE of calendars;
subscriptions read-only; writes to upstream-synced calendars flow on upstream;
- address books as vCard (3.0 served, 3.0/4.0 read), groups as CATEGORIES;
- what a client stores is served back verbatim until the web app edits it;
- sync tokens / CTags from change logs filled by triggers (calendar0008,
contacts0002), so every write path counts; trimmed nightly after 90 days.
REPORTs: calendar-query (time range), calendar-/addressbook-multiget,
addressbook-query, sync-collection. nginx and the Vite proxy route /dav/.
Settings shows the server and user name to enter; install guide §5.4.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
A job every minute finds the event reminders that fell due — repeating events
and moved occurrences expanded as the calendar view does — and pushes each
once to every device with notifications on (sdk.push, new: plugins can notify
through the new-mail push delivery). A reminder missed while the worker was down
is still sent within 10 minutes; cancelled events and declined invitations stay
quiet. Title, then "In 15 minutes, 14:00 · Room 2" in the user's zone and clock
format; it stays until dismissed and shows even with the app open; clicking it
opens that day (/p/calendar?date=…, new deep link).
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Search showed at most the first 50 matching messages ("1–50 of 50"). The
first page now carries the number of matching conversations, further pages
load as the list is paged through (a conversation spanning two appears once;
oldest-first loads them all, up to 40 pages), and the Unread/Pinned tabs filter
on the server, so their count covers every result.
Bulk Delete / Archive / Restore over selected results acts at once: the rows
leave the list (greyed instead when archived from a search, which still finds
them), the search results are refreshed, and the reading pane returns to
"Select a conversation". Each move landing on the server refreshes them again,
and a hidden row only comes back when no move has landed for 30 seconds — a
long batch no longer brings deleted results back.
Settings → Accounts → Notifications: shorter description.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Attachments are no longer stored apart from their message (they were kept
twice: base64 inside the raw message and decoded on their own). The indexer
records the MIME part (attachments.part_index, migration 0023); downloads,
zip, inline images, thumbnails and forwarding take it out of the message;
videos stream from a short-lived copy (kept a day, 1 GiB per user). A
background pass on worker start removes the copies stored before, each only
after checking it can be taken out again (spec App. D §D.6 item 1, criterion
corrected to <= 60%).
A nightly job (03:47, after the Trash/Spam purge) removes files nothing
refers to any more — deleted or purged messages, previews of attachments
gone, uploads no draft has used for a week, expired streaming copies. Files
younger than an hour are left alone; a sent message waiting for its Sent copy
is held until its job is done. Account removal keeps files a draft still uses.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
The message frame is set up as soon as its document is parsed instead of at
its load event, which waits for every image (remote ones included): it no
longer sits at its first height until the slowest image arrives.
A reply quoted an original's inline (cid:) images as this server's own URLs,
and they were sent like that — broken for the recipient, and mangled copies
("http:///api/…") made browsers look up a host named "api". Outgoing mail
(and the IMAP Drafts copy) now carries them as inline parts with new
Content-IDs; one that can no longer be had is left out. Shown messages load
such URLs from this server (sanitizer v5).
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
The webmail is now a PWA: web app manifest, icons (any + maskable, a status-bar
badge), theme-coloured title bar following the chosen theme, and an app-icon
badge with the unread Inbox count. The service worker is always installed;
offline access is on only when it is registered as /sw.js?offline=1, so
turning it on or off is an update that applies at once.
Notifications (Settings → Accounts → Notifications, per device) use Web Push.
Encryption (RFC 8291) and VAPID signing (RFC 8292) are implemented in
app/push/webpush.py — pywebpush needs a newer cryptography than we pin — and
checked against the RFC's worked example. The server makes its key pair on
first use (private half encrypted with MASTER_KEY). New mail found by the live
sync in an Inbox enqueues a job 5 s later that tells the user's devices about
messages still unread there, under 2 hours old and not from the user; more
than three become one "N new messages". A subscription belongs to the session
it was made under: signed out or revoked, nothing more is sent there. Endpoints
are limited to the known push services (no SSRF). Nothing shows while the app
is in front (except on Apple, which requires it). Tested with real pushes
through Google's service in Chromium and Opera.
The mobile drawer gets "Install app": Chrome's own install dialog where the
browser offers one, else short steps for iPhone/iPad, Opera and Firefox, with
"Don't show this again". docs/nas-install.md §5.3 covers installing,
notifications, Public URL and a later Android app (TWA).
Migration 0022: push_subscriptions, app_settings.vapid_*.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Deleting an event the user organizes, with other guests, asks "Would you like
to inform the participants that this event is canceled?" (Yes, inform them and
delete / No, just delete / Cancel; for a recurring one after which occurrences).
Yes sends each guest an iTIP CANCEL from the organizer's mailbox — of one
occurrence via RECURRENCE-ID — or, for "this and following", the shortened
series. Invitations now carry their RRULE, so recurring events repeat for
guests too.
An organizer's cancellation is applied as the email arrives (message.new hook)
or, for older mail, when the card is opened — always, whatever the settings:
only from that event's organizer and not older than the calendar's copy; one
occurrence becomes an EXDATE. The card says "Event was removed from calendar".
Settings → Calendars → Invitations now work, defaulting to "Do not
automatically update my calendar": "Only update events that I have already
added" applies the organizer's changes (keeping the user's answer), "Add new
events…" also adds new invitations, unanswered — from contacts or anyone — to
the chosen calendar, and "mark as read and file" archives such messages. With
no automatic updates, a changed invitation says so on its card.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Invitations received by email (iMIP): above the message body, a card with
when, what, where (or "Join video call"), the organizer, what else is in the
calendar then (a conflict, else the previous event that day), and Yes / No /
Maybe. Yes and Maybe put the event in the calendar (or update it), No takes
it out — and is still shown as the answer given; each sends the organizer
the iTIP REPLY, then a "Reply sent" toast, and an open calendar refreshes.
A cancellation offers "Remove from calendar". It is the calendar plugin's
message.banner slot (ui/invite.js, imip.py, /invitations endpoints); the SDK
gains the user's own accounts and settings. An Outlook-style invitation sent
only as text/calendar is now kept as invite.ics too.
Calendar: clicking an empty day in the side panel's agenda opens a new event
on that date; a video-call link can be removed (✕ beside Copy); the page
guests see after answering, and our invitations' footer, say Ultimo Calendar
(with the logo).
Attachments: videos (mp4, mov, webm, …) play in the full-screen preview,
paused until played; attachments are now served from disk with HTTP Range,
so a video starts at once and can be seeked.
People tab: everyone in the thread, the user included, A–Z, under the name
any message gives them. Initials circles share one colour rule in the list,
the reading pane and the People tab. The contact view drops its duplicate
"Message" button. A general corner toast (webmail:toast) joins Undo's.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Attachments in the reading pane are tiles with a preview: images and a PDF's
first page, rendered on the server (Pillow, pypdfium2) as WebP of at most
480 px, cached beside the blob store and in the browser. Tiles are all one
size — at most 230 px wide, narrower so three still fit in a row — and frame
each preview by its shape: close to the tile's it fills it, very tall / wide /
square ones show whole over a blurred copy, small ones are not enlarged, PDFs
show the top of the page. The name bar (checkbox, type, name, size, download)
slides in on hover and adapts to the tile's width; files without a preview
get a large type icon. Previews of opened conversations are kept offline.
Forward now carries the original's attachments: new compose attachments
sharing the stored blobs (POST /messages/{id}/forward-attachments).
Mailbox display / Settings → Appearance: "Show a dividing line between
rows" (off by default, per device), with a new switch. Settings drop the
"On a phone…" note, and "Language" reads "Spellchecking language".
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
v1.3.26's IMAP4_SSL subclass put its create_client ahead of aioimaplib's,
which is where the TLS context is made — so port 993 got plain TCP, the
server waited for a handshake, and every SSL account timed out ("Can't
reach the mail server"). The context is now made in our create_client too,
and a test checks an SSL client really gets one (it fails without the fix).
With IMAP down, a send could still be delivered by SMTP while filing its copy
in Sent failed — the send then counted as done and the copy was lost. The
copy is now kept and filed by file_sent_copy_job, retried (arq Retry) for
about 24 hours, then shown in Sent as usual.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Sending: a send the server could not take (unreachable, timeout, 4xx) is no
longer failed at once. It waits in Scheduled ("retrying") and is tried again
after 1, 2, 5, 10, 30 min, then hourly, for 24 hours — on the server, so it
works with the browser closed. A permanent refusal (5xx, wrong password) is
"failed" at once, with the server's reason. GET /outbox lists both;
POST /outbox/{id}/retry (Retry now) and /cancel (Edit, Delete) act on them,
and they are left out of Drafts. Mail delivered but not filed in Sent now
counts as sent, and a conditional claim keeps two jobs from sending one
message twice. The app says when a send failed and when it finally went.
Actions: arq retries a job only on Retry, never on a plain exception, so the
"retry x5, then revert" for flags and moves never happened. Flag, move,
archive, restore, delete and empty-folder jobs now raise Retry while the
server is unreachable (~24 hours of growing gaps), skip a change a later one
replaced, stop at the first connection failure instead of 20 s per message,
and delete/empty keep the rows until the server has the messages gone.
Account health: 3 failed connections in a row mark an account unreachable
(amber in the sidebar and Settings, cleared on the next success); 3 refused
sign-ins stop its syncing as auth_failed until the password is entered
again, so the provider does not lock the account. Sync loops back off 30 s
doubling to 10 min and log one line per expected failure; a refused
connection fails at once with its reason (the connect task is kept), and a
silently dead IDLE connection is noticed within ~11 min instead of ~29.
Also: the empty reading pane is clipped to its width when dragged very
narrow; tests no longer leave live accounts and outbox rows in the database.
Migration 0021: outbox attempts / next_attempt_at / retry_until, accounts
sync_error / sync_error_since.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Compose, Reply and Forward work offline. Drafts are kept on the device
(IndexedDB, lib/scheduled.ts), with files attached offline; Send becomes
"Schedule" and puts the message in a new Scheduled folder. When the server
is reachable again, scheduled messages are sent through the normal outbox
(Undo offered for a single one) and offline drafts become server drafts —
files uploaded first, each step recorded so nothing is created or sent
twice, under a Web Lock so two tabs never send the same message. Drafts
written offline show in Drafts as "On this device". Sign-out sends what
is scheduled, then wipes the device's copy.
Also: offline failures from keyboard shortcuts show "Not available
offline"; no recipient suggestions are asked for offline; a reply no
longer quotes an error page when the original is not saved.
The undo countdown counted to the server's deadline on the browser's
clock, so a few seconds of skew showed (18 s instead of 10). /send now
also returns undo_seconds, and the composer counts down from that.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
A self-hosted LanguageTool (dev: docker/docker-compose.yml, started by
`make up`) checks what the user writes in the compose window, through
POST /api/spellcheck — the browser never talks to the engine directly.
- Marks are CSS highlights over the text, never written into the message: red
wavy for spelling, blue for grammar/style. Clicking one shows the message,
up to five suggestions (applied as typed, so Ctrl+Z undoes them) and Ignore.
The popup, like the link popup, opens above its anchor when there is no
room below.
- Checked after a pause in typing; the signature and quoted original are
skipped; marks on words just edited drop at once; a check overtaken by a
newer one, by turning checking off, or by closing the window is discarded.
- With "auto", text is checked in chunks of lines, each in its own detected
language — LanguageTool settles on one language per request, so a mixed
Polish/English message had every English word flagged as a Polish typo.
- Settings > Appearance: on/off and language (Auto-detect or any of the
engine's languages); the compose window's More menu overrides it for one
message. The browser's own spellcheck steps aside while ours runs, and comes
back when it is off or the server has no engine (503 from the endpoint).
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Found on the NAS, where nothing had been learned in weeks while the UI said
"Marked as spam": the source tree predated v1.3.2, so api/worker-* had no
controller password and every lesson got 401 — and the trail of state that
left behind kept it broken after the password was fixed.
- /learn records a lesson only when rspamd accepted it. Recording refused ones
made the menu claim "Marked as spam" and its dedup blocked every retry.
- Bootstrap: done-markers move to a new key. The old one may have been written
by the "-0" bug (a total outage read as a clean run of zero), which left an
account marked done with nothing taught and never retried; every older
install re-runs the bootstrap once, harmlessly (rspamd skips what it knows).
- "Learn from my Spam folder now" in Settings > Antispam filtering queues the
bootstrap for the user's accounts at once, instead of the hourly sweep.
- webmail.sh update now recreates rspamd: it reads the controller password only
when created, so a changed .env reached the app but never the engine.
- doctor's advice for the 401 names both real causes (a pre-v1.3.2 compose
file, or an rspamd started with an old password) and the fix for each; the
old "update v1.3.2" advice could not help.
- docs/nas-install.md: how to refresh the source tree on a tarball install.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
System mail: the password-reset email (and the admin panel's test email) now
opens with the Ultimo logo, centered. It is a PNG rendered from
icons/ultimo-logo.svg and embedded in the message (multipart/related, cid:),
so it shows where remote images are blocked and in clients that do not
render SVG. The PNG is listed in package-data so the Docker image carries it.
Trusted devices: a 2FA sign-in offers "Don't ask for the code on this device
in future". Ticked with a correct code, the browser gets its own HttpOnly
trusted-device cookie (token in Redis, 90 days, renewed on use); later
sign-ins to the same account from that browser need only the password. Other
accounts and browsers are unaffected. The trust is revoked on a password
reset, when 2FA is turned off, and when an admin blocks or deletes the user;
signing out keeps it.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Branding: the page titles and app name are now "Ultimo Email".
Admin panel (now two columns with a divider):
- Users: search by login or by a mail account's address or name (results
only, nothing listed until something is typed). A user opens a window to
change their quota, block/unblock them, or delete them; block and delete ask
for confirmation, and an admin cannot do either to their own account.
- Blocking (users.blocked_at, migration 0019) refuses sign-in with a clear
message and revokes every session; a new user.signed_out event sends the
user's open tabs to the sign-in screen at once. Any "not signed in" answer
now does the same, not only /auth/me. Deleting removes mail accounts (sync
stops), plugin rows (cleared inside the user's RLS context, since plugin
tables have no FK to users), blobs, the search index, sessions and the user.
- Outgoing system mail (migration 0020): SMTP server, port, security, login,
password (vault-encrypted, never sent back), From address and the public
address reset links point to, plus a "Send test" that reports the server's
error.
Password reset:
- "Forgot password?" on the sign-in page asks for the user name and always
answers "If this account exists you will receive an email to your recovery
email address"; the mail goes out after the response, so neither the reply
nor its timing reveals an account, and it is capped at 3 per hour.
- The emailed link (token in the URL fragment, single use, 30 minutes) opens a
page to set a new password. With 2-step verification on, the code is checked
before anything changes; five wrong codes burn the link.
- Settings > Profile: the recovery email can be added, changed or removed,
confirmed with the current password, which must be typed — autofill and
paste are refused (TypedPasswordInput).
Also adds frontend/icons/ultimo-logo-bimi.svg, an SVG Tiny PS version of the
logo for a BIMI record.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
A missing avatar was a 404, which the browser reports as an error for every
such sender on every page. It is now an empty 204: the <img> still fails to
decode it and the initials show through, but nothing is logged, and the miss
carries the same one-day browser cache as a hit so it stops being re-fetched.
On a phone the mail list's rows sit 20px apart instead of 7px.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
The shared Redis client decodes replies as UTF-8 text, but the avatar cache
holds raw image bytes and a b"none" miss marker: an image could not be read
back at all, and a miss came back as str and crashed on .partition(b"\n").
Only a sender's first request worked; after that, a 500. The cache now goes
through a binary-safe client, get_redis_bytes().
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
<title> is not an allowed tag, so nh3 unwrapped it and kept its text —
"mailer", "Email template" or the subject rendered before the content.
It is now dropped with its content, both in the sanitized HTML and in the
text derived from HTML-only mail (snippet and search). SANITIZER_VERSION
goes to 4 so stored bodies re-sanitize from the raw blob on next render.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
The sender column showed participants[0], and participants are rebuilt
newest message first — so replying to a thread put the user in that slot
and the list named them instead of the person who wrote in.
Thread summaries (list and search) now carry `senders`: the distinct From
addresses of the thread's visible messages, oldest first, with the same
Trash/Spam visibility as the message count. The list joins them and shows
the user's own account addresses as "me" ("me, Alice" for a thread the
user started), and the avatar prefers the other side of the conversation.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
`_teach` signalled failure by negating the count taught so far. That
works for every case but the one that matters: when the FIRST message is
refused nothing has been taught yet, and `-0` is `0` — not negative. The
caller's `< 0` test then read a total engine outage as a clean run of
zero, recorded `_bootstrapped` for the account, and the hourly sweep
skipped it from then on. The account most in need of a warm start was
the one guaranteed never to get one.
Which is precisely what happened alongside the missing controller
password: every learn was refused 401, and each account was quietly
retired on its first sweep.
Return None for "the engine stopped accepting these" so failure cannot
be confused with a count, and decline to mark an account whose samples
were all unreadable.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_018SqJYt7jD7oUEkoY68sBZo
Rspamd answers on two ports and only one of them asks for a password:
the scan worker takes /checkv2 anonymously, the controller demands
`Password:` for /stat and /learnspam. `webmail.sh init` generates
RSPAMD_CONTROLLER_PASSWORD and docker-compose hands it to the rspamd
container — but the shared x-backend-env block that feeds api and
worker-* never carried it. The plugin reads it from the environment,
found "", and sent no header.
So classification worked and learning never did. Every learn returned
401 and the plugin swallowed it, as it is built to: an engine problem
must not lose mail. Bootstrap logged "engine unavailable", passive
learning from Inbox->Spam moves taught nothing, and the single visible
trace was `"bayes": null` on GET /api/plugins/antispam/stats, on an
install whose verdict counts were ticking up normally.
Verified against a live engine: with no password /stat and /learnspam
both 401 and rspamd.learn() returns False; with it, stats returns and
learn returns True.
- pass RSPAMD_CONTROLLER_PASSWORD to the app containers, in both the
production and dev compose files
- doctor now makes the real call from inside the api container, with
the password that container actually holds, and says plainly that
nothing will be learned if it fails
- a test reads both compose files, because this is a missing line that
review already let through once
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_018SqJYt7jD7oUEkoY68sBZo
Reported from the running app: a message showed as unread in the list — bold,
blue dot — while the reading pane showed it as read, and only a manual "mark as
read" reconciled them.
The two views read different things. The pane reads the message's own flags; the
list reads `threads.unread_count`, denormalized from them. `apply_delta` updated
message flags with a bare UPDATE and never recomputed the thread, so a flag set
anywhere but in this app — read on a phone, filed by a server-side rule — moved
one and not the other. Auto-mark-on-read could not fix it either: it sees a
message that is already \Seen and correctly does nothing, leaving the counter
wrong forever.
Expunges had the same hole, which also left `message_count` overstated.
`apply_delta` now collects the affected threads and re-denormalizes them after
the deletes, so a thread emptied by an expunge is removed rather than recounted.
Migration 0018 repairs rows that already drifted — recomputed from the messages,
so it is a no-op on a healthy database. One thread was stale here; installs that
have been syncing longer will have more.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_018SqJYt7jD7oUEkoY68sBZo
Warming Bayes hung on a single event — `account.state_changed` reaching "live".
That covers an account added from now on and nothing else. An account that went
live before this plugin existed never gets one; neither does one whose Junk
folder was empty at the time and filled up later, or one whose owner enabled
spam-folder sync afterwards. A worker restarting at the wrong moment loses it
too. The result was a feature that quietly required someone to run it by hand
per account, which is not a feature.
An hourly sweep now walks every live account and offers each a bootstrap. That
works because the per-account job was already cheap to repeat: it returns at once
for an account already done, and declines — without marking it done — when the
corpus is still too small to learn from. So the sweep simply keeps asking until
there is something worth learning, then does it once.
`s.accounts.live()` is the new SDK surface it needs: cross-user by design,
because a sweep runs for the instance rather than a request. Ids and state only,
so reading anything still goes through the user-scoped accessors.
One failing account cannot stop the sweep — the mailbox that raises is logged and
the rest still get their turn.
docs/plugins.md now says this in general terms next to `JobSpec`: a state-change
hook is one-shot, so pair it with a sweep when the work must eventually happen
for everything.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_018SqJYt7jD7oUEkoY68sBZo
docs/plugins.md described an SDK that no longer exists. It now covers
messages.raw/sample, actions.move_to_special, the envelope's account_id /
mailbox_special_use / account_state, the message.changed transition payload, the
message.actions.menu slot, Slot.title and .wm-menu-item — with the reasons, not
just the signatures. M3's acceptance is that a plugin can be written from this
file alone, and it had drifted.
Acceptance 4 (§C.8): a test that teaches 20 spam + 20 ham to a real engine and
checks the classifier generalises to held-out mail, keeps ham under the banner
threshold, and does not leak between users. It skips when no engine is reachable.
It does NOT assert the literal move_threshold — a synthetic fixture cannot reach
12.0 once the SMTP-time symbols are zeroed, which is the whole point of the
tuning. Recorded as OQ-8 rather than quietly asserting something weaker.
Bayes ignores training until min_learns, stock 200. For a personal mailbox that
means every correction appears to be ignored for months, so it is now 20.
The bootstrap would have made this filter worse. Accounts do not sync their
provider's spam folder by default, so the real corpus here is 3 junk beside 1,294
aged inbox — teaching both as found trains "everything is ham". It now refuses
below 10 spam samples (leaving the account unmarked so a later run can do it
properly) and otherwise caps ham at 3x the spam count.
Run against the live account, it correctly declined:
"only 3 spam samples (need 10); enable spam-folder sync to build a corpus"
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_018SqJYt7jD7oUEkoY68sBZo
"Mark as Spam" moved the message but left it on screen. Two causes, both fixed
generally rather than for this one action.
`move_job` published message.changed without a thread_id, so the WS handler
invalidated the message list but never the thread the user was reading — the
panel had no reason to refetch. It now names the thread.
And a thread whose last visible message has gone should not linger. MailLayout
closes the panel when the open thread comes back empty, whoever emptied it —
core action or plugin.
`move_to_special` now also publishes the move immediately rather than waiting
for the IMAP job, since a plugin acting on the user's behalf has no client-side
handle to update optimistically the way the core endpoint's caller does.
That exposed an ordering bug: the optimistic event reaches the passive-learning
hook, which would re-teach the lesson /learn had just taught. /learn now records
`learned` before moving, so the hook's dedup stands down — and records it with
an upsert, because provider-filed spam has no verdict row of ours to update.
One real mistake on the way: the event first reported the destination folder as
the source, which would have made every move look like a no-op to anything
reading the pair.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_018SqJYt7jD7oUEkoY68sBZo
A button pinned to every message header is a persistent control; marking spam is
a verb you reach for occasionally, like Reply or Delete. It now lives in the
Actions menu, between Labels and Delete.
That needed a slot inside the dropdown, which the catalog had no name for:
`message.actions.menu`. The host wraps it so any click closes the menu, leaving
a bundle to do nothing but act.
The UI kit gains `.wm-menu-item`, @applying the exact utilities ThreadView's own
menu items use — the same trick the sidebar shell already relies on, so a plugin
verb is indistinguishable from a core one rather than approximately similar.
Labels follow the folder as before: "Mark as Spam" in the inbox or archive,
"Mark as not spam" in the spam folder, nothing in Sent/Drafts/Trash. A message
already taught shows that, disabled, instead of offering the lesson again.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_018SqJYt7jD7oUEkoY68sBZo
Spam / Not spam appeared on every message in every thread. "Not spam" on an
inbox message is nonsense — it was never filed as spam — and a pair of buttons
everywhere is noise that trains nobody.
The slot now shows a single action chosen by where the message actually is:
"Spam" in the inbox or archive, "Not spam" in the spam folder, nothing at all in
Sent, Drafts or Trash, where no move is a verdict about spam. A message already
taught says so instead of offering the same lesson again.
That needs the message's folder, which the slot was never told. Rather than a
core change to the slot props, /verdict/{id} becomes /context/{id} and returns
the folder alongside the verdict — one round-trip, answered for any message
including ones never classified.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_018SqJYt7jD7oUEkoY68sBZo
Explicit teaching shipped with C1 (the Spam / Not spam buttons). This adds the
two halves that need no buttons at all.
Passive learning (§C.5). Moving mail into Spam says "this is spam" more honestly
than any button, because the user is doing it for their own reasons rather than
to train a filter; moving it back out says the opposite. Only those two
transitions teach — Spam to Trash is someone tidying up, Inbox to Archive says
nothing about spam. Teaching the same message the same thing twice is skipped,
since users move things back and forth, but reversing a lesson is allowed. If
the engine is down nothing is recorded, so a later move can teach it again.
This needed the transition, not just the destination: `message.changed` carried
only where a message landed, so `move_job` now also publishes
previous_mailbox_id and both special_use values.
Provider bootstrap (§C.5). An account going live enqueues a job that teaches
Bayes from mail the user already owns — up to 500 messages from the provider's
Junk folder as spam, and up to 500 Inbox messages older than 30 days as ham.
This is the constructive use of the history that classification deliberately
refuses to act on: the same backfill, read instead of judged. It sleeps between
messages so live classification never queues behind training, stops rather than
grinding if the engine disappears, and marks the account done in KV so it runs
once.
That needed `PluginMessages.sample`: the SDK could fetch a message by id but not
enumerate a folder. Ids only — bodies are pulled one at a time, so a 500-message
sample never materialises 500 bodies.
Verified against the running engine: teaching moved its `learned` counter 0 -> 2.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_018SqJYt7jD7oUEkoY68sBZo
Dev compose gains rspamd + redis-rspamd, matching production. The dev API and
worker run on the host rather than in the compose network, so the ports are
published on loopback and RSPAMD_URL points at 127.0.0.1 — documented in
.env.example.
Contacts now cap a verdict at "banner": mail from someone in the user's address
book is never moved to Junk. Capped rather than exempted, because the way this
fails in practice is a contact's account being compromised and spamming its own
address book — the warning still has value even when the move would be wrong.
An explicit deny rule still wins; that is the user overruling us deliberately.
The lookup reads the contacts plugin's table directly and forgives its absence:
no contacts plugin means no contacts to check, not an error. Addresses are
stored as the user typed them, so both sides are lower-cased rather than relying
on the jsonb containment index, and secondary addresses in the emails array
count as well as the primary. It runs only when the verdict would otherwise be a
move, so the common path pays nothing.
Also fixes a test that asserted stats degrade "without an engine" by relying on
no engine being present — true until this commit. It now simulates the absence.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_018SqJYt7jD7oUEkoY68sBZo
message.new fires for imported history as well as new arrivals, and EventBus
awaits hooks inline. As written, adding an account would have scored every
message in its backfill — filing years-old inbox mail into Junk on the server,
and adding ~70 ms of serialized HTTP per message to the backfill loop (twelve
minutes on a 10k-message account). It is also when the filter is least able to
judge, because Bayes has learned nothing yet.
The envelope now carries account_state, and the hook returns unless it is
"live". The backfill window belongs to the C2 learning bootstrap (§C.5), which
reads that history as training data instead of acting on it.
Tests pin the gating: backfilling accounts are not even read from the blob
store, non-inbox folders are ignored, a live inbox arrival still classifies, and
an unreachable engine still delivers.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_018SqJYt7jD7oUEkoY68sBZo
Completes phase C1 — until now the plugin classified mail that nobody could see
or steer.
Router (§C.6): settings with merge semantics, allow/deny rules, a per-message
verdict lookup, learn, and stats. Two behaviours worth naming. A banner
threshold above the move threshold is silently capped rather than refused,
because a banner past the move score can never be seen — everything that far up
is already in Junk. And /learn moves the message even when the engine is
unreachable: teaching is best-effort, but filing is what the user actually
asked for.
UI: settings.page (thresholds in plain language, per-account toggles, the two
lists, the privacy switch, activity); message.banner for the uncertain middle,
offering Move to Junk / Not spam / Always trust sender; message.actions with
Spam / Not spam. The banner mounts on every message but renders only for a
`banner` verdict — a spam warning on ordinary mail is worse than no warning.
Bundle serving is allow-listed by filename, so /ui/../hook.py is a 404 rather
than a file read.
Tests cover the defaults, merge semantics, the threshold cap, rule
normalisation and idempotency, stats degrading to empty when Rspamd is absent,
and RLS isolation: one user cannot see another's rules or settings, nor delete a
rule whose id they know.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_018SqJYt7jD7oUEkoY68sBZo
Second-pass spam filtering per SPEC Appendix C. The engine is Rspamd in its own
container; this plugin is glue and policy, never a spam engine of its own.
Classification (§C.4) hooks message.new, acts only on inbox arrivals of enabled
accounts, and runs deny list -> allow list -> engine -> thresholds. Every
failure path ends in "deliver normally, unclassified": an unreachable engine, a
missing blob, an account with no Junk folder, a failed verdict write. Nothing
here may delay or lose mail, so no service declares depends_on: rspamd.
Two containers, started with everything else rather than behind a profile:
rspamd, and a redis of its own. The separate redis is not incidental — the app's
redis holds sync leases and arq queues under noeviction, and an ever-growing
Bayes corpus must never be able to fail a job enqueue. This one takes RDB
snapshots and evicts, because its data is derived and retrainable.
Network checks (RBL/SURBL/fuzzy) are off: they leak message-derived hashes and
URLs to third parties, which a self-hosted server's users did not ask for. The
engine makes no outbound connections in the default configuration.
The controller password is generated into .env like MASTER_KEY. It is internal;
humans will reach Rspamd through the admin panel, authenticated as the ordinary
administrator, rather than typing it.
local.d/groups.conf zeroes six SMTP-time symbols — DATE_IN_PAST,
PREVIOUSLY_DELIVERED, HFILTER_HOSTNAME_UNKNOWN, RCVD_NO_TLS_LAST, ARC_NA,
DMARC_NA — and greylisting is disabled. Measured against 398 real delivered
messages, those checks were firing on 44-71% of ordinary mail and pushing 11.3%
of it over the banner threshold; scoring mail at rest for facts true of every
delivered message is simply the wrong question. After: median 3.09 -> 0.70,
p95 7.00 -> 3.40, banner rate 11.3% -> 1.0%. The spec's 12.0/6.0 thresholds were
never the problem and are unchanged. Symbols are zeroed rather than removed so
verdicts stay explainable, and they matter again at the M6 edge (§C.7).
Still to come in C1: the §C.6 router and UI slots. C2 adds the learning loop —
until Bayes is trained, genuine scams and legitimate bulk mail overlap in the
5-8 band and no threshold separates them.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_018SqJYt7jD7oUEkoY68sBZo
Three strands that only make sense together, so they land together.
Registration gate (migration 0016). POST /auth/register was open to anyone
who could reach the port. Now the first account on an empty instance is
always allowed, becomes the administrator, and closes registration behind
itself; after that it takes an admin reopening it. The decision is made
under a row lock on app_settings so two simultaneous first signups cannot
both see an empty table and both claim administrator. Existing installs
keep their oldest account as admin and start closed.
Instance default quota (migration 0017). users.quota_bytes is snapshotted
at creation from app_settings.default_quota_bytes, so changing the default
affects later accounts only and never resizes a mailbox that exists.
The panel is a separate app, not a page in the mail client. admin.html is
its own Vite entry (5.8 KB next to the client's 1.7 MB; they share only the
vendor chunk), with its own administrator sign-in. One nginx now serves two
listeners: :80 is the mail client and 404s both /admin.html and
/api/admin/*, :81 is the panel. docker-compose publishes them as WEB_PORT
and ADMIN_PORT. The boundary is nginx refusing paths on one listener, not
process isolation — CurrentAdmin still guards every admin route — so
ADMIN_PORT must be restricted to the LAN, which the runbook now covers for
both DSM and QNAP.
Also QNAP: deploy/webmail.sh finds Container Station's docker under
/share/*/.qpkg/container-station/bin, which is off an SSH session's PATH,
before anything shells out to it. docs/nas-install.md gains a QuTS hero
runbook (§2.6), splits the DSM firewall section into container-to-container
(§8.1) and the egress case that makes account-add time out while
autodiscovery appears to work (§8.2), and scopes both to Synology.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_018SqJYt7jD7oUEkoY68sBZo
CalDAV account connect (Settings → Calendars) with RFC 6764 discovery,
per-calendar vaulted credentials, and full two-way sync verified live
against NextCloud:
- caldav.py: discovery, calendar-query REPORT (pull), PUT/DELETE with
quoted ETags (If-Match/If-None-Match), iCloud partition-redirect
handling, ICS build/parse; clear error codes (auth/private-host/
unreachable) surfaced to the connect UI.
- sync.py: ctag-gated pull + push of local changes; dirty rows protected
from being clobbered; 15-min cron + immediate enqueue on every
create/edit/delete so changes reach the server in ~1s.
- Dirty tracking via a BEFORE INSERT/UPDATE trigger (migration 0007),
suppressed during sync with a txn-local GUC; deletes tombstone then
DELETE server-side. Writability now driven by server privileges.
- migration 0006: event uniqueness NULLS NOT DISTINCT — fixes duplicate
events on every full re-pull (NULL recurrence_id never matched ON CONFLICT).
- 5 CalDAV parsing unit tests.
Calendar UI (page.js): all-day events now render in Day/3-days/Week views
(were month-only). They share the timed-event column layout so overlapping
events sit side-by-side, fill the day column vertically with a sticky label,
and are matched by start day (fixes UTC-midnight events spilling into the
next local day).
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Sender avatars (privacy-safe, server-fetched, cached):
- New GET /api/avatar resolves Gravatar (d=404) then the sender domain's favicon
under the SSRF guard, Redis-cached (hit+miss); gated by SENDER_AVATARS_ENABLED.
GuardedResponse gains raw bytes. Shown in the mail list, reading pane, People
panel, and Contacts (overlay that falls back to initials on a miss).
PDF preview:
- Password-protected PDFs now prompt in-place and retry (pdf.js PasswordException),
with a clear error for other load failures instead of a blank overlay. The
password is remembered per attachment until the thread is left; clicking the
field no longer dismisses the overlay.
Contacts:
- Autosave reply recipients: message.sent now carries to/cc (name+email); the
contacts plugin subscribes and adds any non-contact recipients to the Autosaved
group. Contact avatars show the gravatar/favicon via /api/avatar.
Mail views / retention:
- Clicking an account shows all its mail except Trash/Spam (was inbox-only).
- Daily cron autopurge_trash_spam permanently deletes Trash/Spam older than
TRASH_RETENTION_DAYS (default 30; 0 disables) — server EXPUNGE + local rows.
UI polish:
- Ultimo signet logo in the Mail/Calendar/Contacts sidebar header and collapsed
rail; consistent 11px top margin and vertically-aligned header row across states
(fixed-size flex icon buttons); SVG favicon on the browser tab.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01U72FZiF1W1pb8XxyWBTREA
Gmail sync (SPEC: Gmail sync):
- Provider fetches X-GM-LABELS when X-GM-EXT-1 is advertised; parses custom labels
(system labels dropped, mUTF-7 decoded) onto RemoteMessage.
- Engine syncs only system folders + All Mail for Gmail, skips every label folder,
and dedupes the All Mail pass by Message-ID — kills the INBOX/All-Mail duplication.
- Custom labels ride the message.new event; the labels plugin subscribes and maps
them to chips as "<name> [Gmail]" (per-message + thread aggregate + known set),
so labels fill in gradually with the sync window. Non-ASCII labels decode right.
Labels self-healing (message-driven reconcile):
- Per-message keys are the source of truth; t: aggregates and _labels are derived.
Reconcile drops labels of deleted messages, regroups survivors by their current
thread (fixing JWZ thread merges), rewrites changed t: keys, and rebuilds _labels
from live messages — keeping user labels, dropping orphaned Gmail ones.
- New SDK reads: sdk.threads.existing / sdk.messages.threads_of. Sidebar refreshes
live via a plugin.labels.changed WS event on new labels.
Account deletion: prune threads left empty after the cascade (they were user-scoped,
so only messages cascaded) + de-list the account from surviving cross-account threads.
Sync SPAM: opt-in per-account sync_spam (migration 0015, default off) — the engine
skips the Junk folder in backfill and poll unless enabled; wizard checkbox added.
Robustness / UX:
- WebSocket auto-reconnects with backoff and resyncs on (re)connect, and account
add/delete invalidates threads/labels — no more hard refresh for live updates.
- Backfill banner + sidebar ring also require done < total, so a completed account
can't leave a stuck "X of X".
- Scrub lone UTF-16 surrogates from parsed headers/bodies so bad messages no longer
crash the backfill on Postgres text/JSONB.
- Ultimo signet logo in the Mail/Calendar/Contacts sidebar header (and at the top of
the collapsed rail), via the shared plugin-ui.css + MailLayout.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01U72FZiF1W1pb8XxyWBTREA
Calendar plugin (webmail_plugin_calendar, non-removable core plugin):
- Local calendars/events CRUD with RLS isolation; recurrence expanded at
query time (DST-correct), EXDATE + detached-override merge.
- Calendar app bundle: month/week/day/3-day views, mini-month, event form
(color, show-as, timezone combobox, repeat, reminders, guests w/ recipient
autocomplete, Jitsi video call), event popover with RSVP badges, resizable/
collapsible left panel + theme toggle, overlap layout.
- iMIP invites: sdk.mail sends REQUEST from the organizer's mailbox (HTML +
text/calendar + invite.ics); public /rsvp endpoint (vault-sealed token)
records accepted/declined/tentative per attendee.
- "This and following" edit + scoped delete (this / this-and-following), with
a delete confirmation dialog for recurring events.
Core:
- Settings: time_format (12/24h, default 24h) + jitsi_server; new Integrations
section; date order follows time format (en-GB vs en-US) across Mail.
- Plugin SDK: mail (send from a user's account), vault (envelope encrypt),
multi-method SSRF-guarded http; AppSwitcher product switcher; full-bleed
PluginPage.
Mail HTML rendering:
- Sanitizer allows <style> blocks + class (CSS sanitized: drop script-like,
strip @import/url() when remote images blocked); SANITIZER_VERSION 3 so
already-indexed mail re-sanitizes on view.
- Attachments: a Content-Disposition: attachment part is a real attachment
even when it also carries a Content-ID (fixes hidden invite.ics).
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Server-side deletions now propagate to the DB. Previously poll() detected new
messages and CONDSTORE flag deltas but never expunges (the old TODO), so mail
deleted in another client lingered as phantoms outside the small Drafts reconcile.
- provider: enable QRESYNC when advertised (_maybe_enable_qresync + qresync_enabled
property). poll() harvests VANISHED (expunged UIDs) and changed FLAGS in one pass
via a raw `SELECT <mbox> (QRESYNC (uidvalidity modseq))` — issued as a low-level
SELECT because aioimaplib only surfaces untagged VANISHED lines on a sync command,
not on an async UID FETCH. CONDSTORE-only servers (e.g. Gmail) keep the CHANGEDSINCE
flag delta and fall back to the engine reconcile.
- parsing: parse_vanished() for `VANISHED [(EARLIER)] <uid-set>` with range expansion.
- engine: apply_delta now drops expunged rows from the search index, un-links any
local draft mirroring a vanished UID, and publishes message.deleted for the UI.
- tests: parse_vanished unit cases + apply_delta expunge (removal, draft un-link,
modseq advance).
Verified live: the worker issues `SELECT "INBOX" (QRESYNC (uidvalidity modseq))`
against QRESYNC-capable accounts. ruff + mypy --strict clean; 296 tests pass.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01Rcj3TpydMTzTSmU7kunzf8
lukasandClaude Opus 5.5