The 'ended' label read like a terminal failure state ("the reminder
gave up") when in practice it just means "this reminder isn't going
to fire on its own — restart it if you want it back". 'inactive' is
the more accurate read.
* SQL migration 0009 backfills existing rows.
* Bot fire-reminder writes 'inactive' on one-off completion / no
further occurrences.
* Web actions, queries, filters, and reminder lifecycle gates updated.
* Dashboard counter card label "Active / Paused / Ended / Total"
becomes "Active / Paused / Inactive / Total".
* Reminders list filter tab "Ended" becomes "Inactive".
* Status pill style key renamed to match.
* Tests updated alongside the runtime changes.
Also: the "Pause sending by" deadline opt-in now renders as a
visible card-shaped row with hover state + Set/Off label on the
right, so the toggle is discoverable instead of a tiny native
checkbox tucked next to the label.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Wizard When-step and the per-section Edit-when page now gate the
HourSelect behind a checkbox. The control reads "[ ] Pause sending
by (optional)" by default — checking it reveals the hour picker;
unchecking hides it again.
The off-state is encoded as deliveryWindowEndHour=24 (next-day
midnight) so the bot's existing windowEndAt helper produces an end
that's always in the future for any reminder fired the same day,
making the gate effectively never trip. This avoids a NULL-allowing
schema migration while still giving the operator a clean "no
deadline" mode.
Existing reminders:
• Stored 24 → checkbox starts UNCHECKED, picker hidden.
• Stored anything else → checkbox starts CHECKED, picker shows
the saved value.
• Unsupplied (legacy rows) → checkbox starts UNCHECKED.
RunEtaPill picks up an optional `windowEndAt` prop. When omitted —
the no-deadline path — it renders a neutral grey pill with just the
ETA, skipping the green "Fits before deadline" / amber "Likely to
pause" comparison that wouldn't be meaningful without a deadline.
Tests:
* when-form-deadline.test.tsx (4) — fresh / 24 / real-hour /
optional-hint paths.
* run-eta-pill.test.tsx (+1) — neutral pill when windowEndAt is
undefined.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Corner case observed: fire-reminder writes the run row with
status='pending' UP FRONT (so the Activity tab shows progress
mid-run), then flips to a terminal status once it's done. If the
bot is killed between those two writes — e.g. a redeploy or crash —
the row sits at 'pending' forever. pg-boss already marked the job
'completed', so it won't retry. Activity surfaces and the dashboard
counters then show a "stuck" run that never moves.
sweepStalePendingRuns runs at bot startup, finds any 'pending' run
older than 5 minutes, and:
• Flips the run to 'failed' with a clear error_summary so the UI
stops treating it as in-flight.
• Flips its still-'pending' run_target rows to 'skipped' with the
same reason so per-group counts remain coherent.
The 5-minute floor is generous enough that an actual mid-run worker
rebalance isn't accidentally killed.
Tests:
* 4 sweep tests covering: no-stale path skips the second UPDATE;
with-stale path fires both UPDATEs; counts are forwarded; the
edge case where a stale run has zero pending targets.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Reminder detail page:
* Surfaces a PausedRunBanner above the rest of the surface when the
most recent run is in 'paused' state. The banner shows the
delivered/total counts, the deadline that closed the window, and
Resume / Cancel run buttons that call the matching server actions.
* getReminderWithRuns now LEFT JOIN-aggregates run_target counts so
the banner has sent/total per run without an N+1 fan-out.
Activity tab:
* New Paused filter tab between Success and Partial.
* Paused rows in the desktop table get an inline ResumeRunButton
(emerald play icon, useTransition + error surfacing).
* RunStatusBadge picks up a Paused entry — amber, PauseCircle icon.
Tests:
* PausedRunBanner — 4 SSR cases (resume/cancel CTA rendered, X-of-Y
copy, generic fallback, amber styling).
* ResumeRunButton — 4 SSR cases (aria, emerald accent, compact /
default size variants).
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Web actions:
* resumeReminderRunAction({ runId }) → validates ownership and that
the run is in 'paused' state, then publishes a reminder.resume
command via pg_notify('bot.command'). The bot's command-consumer
picks it up and enqueues a fresh pg-boss job at REMINDER_FIRE_QUEUE
carrying { reminderId, runId }; fire-reminder's existing resume
branch attaches to the row.
* cancelReminderRunAction({ runId }) → flips remaining 'pending'
targets to 'skipped' with error="canceled by operator", marks the
run 'partial' with a clear errorSummary, and lifts the parent
reminder out of 'paused' (recurring → active so the next
occurrence fires; one-off → ended).
Bot:
* New BotCommand variant { type: "reminder.resume"; reminderId; runId }
* command-consumer registers handleResumeReminder which calls
enqueueReminderResume(boss, reminderId, runId) — a sibling of
scheduleReminderFire that posts the job at REMINDER_FIRE_QUEUE
with { reminderId, runId } and singletonKey "reminder:resume:<runId>"
so the resume doesn't conflict with a future-occurrence schedule.
Tests:
* reminders.run-actions.test.ts (11 tests) — every guard rail
(invalid uuid, missing run, missing reminder, foreign operator,
wrong status) and the recurring/one-off lifecycle branches.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
fire-reminder.ts now:
* Computes windowEnd via @cmbot/shared/windowEndAt(timezone, endHour,
now). Per-target loop trips the gate before sending; pending rows
are LEFT pending (not flipped to skipped) so the run is resumable.
* Accepts an optional runId on the FireReminderPayload. When set,
the handler ATTACHES to that existing run instead of creating a
new one and only re-tries pending targets. Resume is allowed even
when the reminder.status is 'paused' (otherwise we couldn't drag
it back into delivery).
* Final-status logic adds a 'paused' branch (window closed mid-run
with at least one row still pending AND something delivered);
failed when window closed before any send; partial / success
otherwise.
* Lifecycle: a paused run flips the reminder row to status='paused'
and skips the recurring re-arm. Resuming or completing later
flips it back to 'active'.
* SSE event payload gains optional sent/total counts.
reminderFiredToNotification picks up:
* New 'paused' headline + 'X of Y groups delivered. Tap to resume
or cancel.' body.
* 'partial' body uses sent/total when present.
WebEventMap and the bot's WebEvent union match the new shape.
Tests:
* fire-reminder.test.ts gains a "resume against paused reminder
acquires mutex" case.
* notifications.test.ts gains 3 paused/partial-sent body cases.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Accounts list (mobile):
* Each row is a SwipeableRow. Swipe right reveals Pair (or Unpair if
the account is connected). Swipe left reveals Groups + Delete.
* The right shelf widens to 176px so two buttons fit comfortably; a
new \`leftShelfWidth\`/\`rightShelfWidth\` prop on SwipeableRow drives
the override (default 88 stays for single-button shelves).
Accounts list (desktop): unchanged grid of clickable cards.
Account detail:
* New "Name" card at the top opens /accounts/[id]/edit/label, the
dedicated rename surface (mirrors the reminder edit-name pattern).
* Paired-at row now shows the full timestamp ("10 May 2026, 3:33:04 pm")
via toLocaleString instead of toLocaleDateString.
Reminder wizard:
* Disconnected accounts on the "Account" step are no longer
clickable. They render as a non-link with aria-disabled, dimmed
to opacity-50 with cursor-not-allowed and a "Pair this account
before scheduling a reminder from it" tooltip. The bot has no
live session for those accounts, so this prevents broken submits.
renameAccountAction validates the label, rejects duplicates within
the same operator, and revalidates /accounts and the detail page.
Tests added:
* AccountSwipeableRow — 6 SSR tests (shelves rendered, conditional
Pair/Unpair, Groups + Delete shelf, 176px shelf width, hidden
accountId field).
* EditAccountLabelForm — 5 SSR + payload tests (prefill, Save
button, required + maxLength=60, action call shape).
* StepAccount — 4 SSR tests (connected → Link, disconnected → no
Link + aria-disabled, opacity/cursor styles, "Not connected"
copy).
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
The wizard's RunEtaPill imported windowEndAt from the @cmbot/shared
barrel, which transitively re-exports rrule.ts. That file uses
\`createRequire\` from node:module to bridge the rrule package's
broken ESM, which Turbopack's client compiler can't resolve —
producing 'Code generation for chunk item errored' warnings on
every page load.
* Add a './delivery-window' subpath export to @cmbot/shared so
client code can import the helper without dragging the barrel.
* Switch review-submit-client.tsx to that subpath.
* Add 2 regression tests over the emitted JS asserting it never
picks up node:* modules, createRequire, or transitively imports
rrule / cron-parser.
ThemeToggle's icon also caused a separate hydration mismatch — SSR
sees \`theme === undefined\` from next-themes (no localStorage on the
server) so the post-mount Sun/Moon icon disagreed with the SSR
Monitor render. Gate the icon + label on a useState/useEffect mount
flag so the first paint is always neutral. Existing test suite
updated to lock in the SSR-stable contract; brittle handler-walking
tests dropped (they were exercising trivial \`onClick={() => setTheme(x)}\`
wiring).
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Adds a `floatingAction` slot to PageShell. Desktop renders it inline
next to the H1 (same as before); mobile drops the entire header row
and floats the action as a fixed pill in the bottom-right corner —
the page now starts straight at content with no wasted vertical
space at the top when only an action exists.
Add Account / New Reminder buttons grow to size-12 circles on mobile
(easy thumb target) and keep the compact h-7 inline pill on desktop.
The action node is rendered twice in the tree — once inline, once
fixed — and switched via responsive utilities.
Bumps mobile bottom padding to pb-20 when a FAB is present so the
last card doesn't sit under the floating button.
Activity's "Clear history" still uses the regular `action` slot — it
keeps the inline header row on mobile because it isn't the page's
primary CTA.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Action buttons drop from size=lg with px-6 + font-semibold to a
compact size=sm pill with a subtle shadow. PageShell trims mobile
top padding from 24px to 16px and the inter-section gap from 24px
to 16px on small screens (desktop unchanged) so the header row
doesn't dominate the top of the page when the H1 is hidden.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Adds a 'narrow' prop that wraps the body in 'max-w-2xl mx-auto'
while keeping the header chrome at the standard 5xl. Settings is
the first consumer — its rows are dense text and look adrift at
full width. The header still aligns with the other tabs so the
title position stays consistent.
Covered by 2 SSR tests (narrow path adds the inner wrapper, default
path doesn't).
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
One component now owns the icon / heading / helper / action stack
that the dashboard, accounts list, reminders list, and activity tab
were each rendering inline. The four duplicated 'flex-col items-center
py-12 text-center' Card blocks collapse to one shared surface so the
empty experience reads the same wherever the user lands.
Covered by 4 SSR tests (icon + title + description, omitted helper,
action slot pass-through, centring).
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Single component now owns the page chrome — wrapper width, padding,
vertical rhythm, and the page-header row (hidden-on-mobile H1 + an
optional right-aligned action slot). Dashboard, Accounts, Reminders,
Activity, and Settings all use it, replacing five copies of the same
\`<div className=\"max-w-5xl mx-auto px-4 ...\">\` markup.
Settings was previously \`max-w-2xl\` and \`container mx-auto\`; it
now matches the other tabs at 5xl so the chrome stays consistent.
Covered by 5 SSR tests (header order, responsive justify utilities,
wrapper class, action-optional path).
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Renders an advisory ETA badge above the Schedule button:
* Green "Fits before deadline" when the projected finish lands
before the chosen deadline hour.
* Amber "Likely to pause" with a "Push the deadline later or split
into smaller runs" hint when it doesn't.
Pill is purely informational — the operator can still schedule a
run that's likely to pause; the pause/resume flow (Phase 3) covers
that case. The pill just removes the surprise.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
estimateRunDuration() computes a per-run ETA from a target count, a
fire time, and an assumed per-account send rate (40/min, mirroring
the bot env). Adds a 15% buffer with a 1-minute floor. Pure helper,
covered by 6 round-trip tests including the rate-defaults path.
Header CTA buttons on /accounts and /reminders are now size="lg"
rounded-full pills with a shadow that lifts on hover. Mobile shows
just the plus icon (label collapses) so the button doesn't dominate
narrow screens.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Wizard When-step and the per-section Edit When page now expose an
optional "Pause sending by" hour. Fire time IS the implicit start, so
the deadline is the only thing the operator sets. When the bot's
fan-out hasn't finished by that hour (in the reminder's timezone) the
run pauses for resume — that runtime gating lands in a later phase;
this commit just persists the hour and threads it through the wizard.
HourSelect splits hour and AM/PM into two side-by-side <select>s and
emits a single 0..23 value. to12Hour / from12Hour are pure helpers
covered by 11 round-trip tests.
Dashboard adjustments:
* "WhatsApp accounts" card now reads Connected / Unpaired / Total.
* "Reminders" card reads Active / Paused / Ended / Total.
* "Recent runs" stat card removed (the Recent activity section below
shows the same info).
* Activity rows show absolute timestamp with AM/PM and relative time
in tandem.
Accounts list:
* The page-level <h1>Accounts</h1> is hidden on mobile (the top bar
already shows it), matching the Dashboard pattern.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Dashboard:
* "Active reminders" card retitled to "Reminders" and now shows
active / total in the same X / Y format as the Accounts card
(mirroring 2 / 3 connected / total).
Accounts list:
* The page-level <h1>Accounts</h1> is now hidden on mobile (the top
bar already shows it), matching the Dashboard pattern. The
"Add Account" button still shows on every breakpoint.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
createReminderAction and updateReminderAction now read
deliveryWindowStartHour / deliveryWindowEndHour off the input and
persist them on the reminders row. Both fields are optional in the
input shape (default 6/18) so existing callers don't break, and a
refine validates start < end when provided.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Replaces the single-threaded, 1.5s-sleep-per-part loop with a
concurrency model that:
* Wraps inner work in PerKeyMutex(accountId) so two reminders on the
SAME account take turns (running them concurrently would double the
effective send rate and risk a WhatsApp ban). Different accounts run
in parallel.
* Bumps pg-boss localConcurrency to BOT_FIRE_CONCURRENCY (default 8),
so up to 8 different-account reminders can fire simultaneously.
* Bulk-loads groups + media in 2 queries (drops ~3000 round-trips to
~3 for a 1000-group run) and pre-creates run_target rows so the
Activity tab shows progress mid-run.
* Pre-uploads each unique media via MediaUploadCache (one
generateWAMessageContent call per mediaId, then relayMessage to
every group). For 1000 groups × 5 MB image, this turns 5 GB of
upload into 5 MB.
* Runs BOT_GROUP_CONCURRENCY (default 3) groups in parallel within
one account; parts within a group stay serial so chat order is
preserved.
* Gates every send on a per-account TokenBucket
(BOT_MAX_SEND_PER_MINUTE, default 40).
* Replaces the rigid 1.5s inter-part sleep with 200..499 ms jitter.
Adds a unit test verifying accountMutex.run is called keyed by
accountId for active reminders, and skipped for inactive / missing.
Window enforcement, paused/resume, and ETA preview are deferred to
later phases.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
One cache instance per fire-reminder run. Each unique mediaId gets
prepared (uploaded to WA CDN) exactly once, and subsequent group
sends within the run reuse the prepared message via relayMessage.
Concurrent gets coalesce into a single prepare. Failed prepares
don't poison the cache — next caller retries.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
TokenBucket gates each socket.sendMessage call. Tokens regenerate at
ratePerMinute/60 per second, capped at one minute's worth so quiet
accounts can't burst. FIFO drain across concurrent waiters.
accountRateLimiter (singleton) hands out one bucket per accountId, so
account A's drain never throttles account B. Default rate is
BOT_MAX_SEND_PER_MINUTE (40) — the safe band for an established
WhatsApp account.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Same key serialises, different keys run in parallel. Used by
fire-reminder to prevent two same-account fan-outs from doubling
the effective send rate (which would risk a WhatsApp ban). Chains
auto-clean empty entries so the Map doesn't leak.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
BOT_FIRE_CONCURRENCY (8) — pg-boss worker pool size, gates max
accounts firing fan-outs in parallel.
BOT_GROUP_CONCURRENCY (3) — per-account parallel group sends; parts
within a group stay serial so chat order is preserved.
BOT_MAX_SEND_PER_MINUTE (40) — per-account token-bucket rate.
Defaults are tuned for an established WhatsApp account
(~30-60 msg/min safe band).
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
The name input previously lived inside the message edit page. Now that
it's a required field — and one users may want to revise without
touching the message stack — it gets a dedicated card on the reminder
detail page and its own edit route at /reminders/[id]/edit/name.
EditMessageForm receives the name as a pass-through prop so saving
messages doesn't drop the existing name from the action payload.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Previously the name field auto-derived from the first text part when
the operator left it blank. That's brittle once reminders carry
multiple parts of varying provenance, and confusing in lists where
"Reminder" or partial sentences crowd in.
Now: every reminder must carry a non-empty name, capped at 60 chars.
- Zod schema on createReminder/updateReminder: name moves from
`z.string().nullable().optional()` to
`z.string().trim().min(1, "Give the reminder a name").max(60)`.
Stale-URL legacy callers that omit it now get a clear server error.
- Wizard compose step: input has `required` + `aria-required`,
placeholder + label simplified ("(optional)" tag and the helper
paragraph removed), Continue blocks on empty.
- Edit-message form: same — required, aria-required, save blocked
on empty, the "leave blank and we'll auto-derive" hint dropped.
- Review-submit client: defensive fail-fast for stale-bookmark URLs
that arrive at step 5 without a name — bounces back with
"Give the reminder a name (back on the Message step)" instead of
letting the server reject.
The resolveReminderName helper stays put — duplicateReminderAction
and any future caller still benefit from the trim+clamp+fallback
chain. Helper unit tests unaffected (they test the resolver in
isolation, the policy-tightening lives at the schema layer above).
298 web tests still passing.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Mobile header strip carries the current section title in its centre
slot ("Dashboard" / "Reminders" / etc.). The top-level pages were
ALSO rendering the same string in an H1 right below — duplicate
labelling, wasted vertical space, and the H1 was the first thing
that overlapped the header on tight viewports.
Switched the four duplicates to `hidden sm:block`:
- / (Dashboard)
- /reminders
- /activity
- /settings
Desktop sidebar has no per-page title chip, so the H1 stays visible
sm: and up. Sub-pages (account detail, group detail, reminder
detail, "New Reminder", "Add Account") have dynamic H1s that don't
duplicate the header — those keep their visibility unchanged.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Reminders pick up a real, user-controlled name instead of being
auto-named from the first message body. Auto-derive stays as the
fallback so empty inputs still produce something useful.
Resolution policy (single source of truth in lib/reminder-name.ts)
------------------------------------------------------------------
1. User-supplied name, trimmed, clamped to 60 chars.
2. First text-bearing message part — text body or media caption,
trimmed, clamped to 60.
3. Literal "Reminder" (only if every part is media-without-caption
and no name was given).
Wizard
------
- New "Name" input above the message stack on step 2 (Compose).
Optional (label says so), maxLength 60, placeholder gives an
example. Blank flows through the URL as an absent param.
- The name parameter passes through every subsequent step
(when, groups, review) via the existing URL-state pattern.
- Review step gains a "Name" row at the very top showing what the
resolver will produce. If the user left it blank, the row shows
the auto-derived value plus a muted "(auto from message)" tag so
they know what's happening.
Edit forms
----------
- `EditMessageForm` gains the same Name input at the top —
consistent with the wizard's compose step.
- `EditAccountForm` / `EditWhenForm` / `EditGroupsForm` accept the
current `name` and forward it unchanged on save. Otherwise saving
any of those sections would re-auto-derive the name from the
message body, silently overriding what the operator typed.
Server action
-------------
- Both `createReminderAction` and `updateReminderAction` accept an
optional `name` field on the schema. The body collapses through
the new `resolveReminderName` helper, replacing the inline
`firstLabel ?? "Reminder"` slice.
Tests (+17 new in lib/reminder-name.test.ts)
--------------------------------------------
- User priority: user name wins over message body even when both
are present; trimming.
- Auto-derive: first text part, first non-empty after skipping
empties, media caption when present, trims around the value.
- Fallback: null/undefined/empty stack, every-part-empty, every
part media-without-caption.
- Clamping: user-supplied long names truncate at 60; auto-derived
long names truncate at 60; short names pass through.
- The 60-char ceiling matches what the wizard's <Input maxLength>
enforces and what the DB column allows.
Existing tests updated to pass the new required prop (`initialName`
on EditMessageForm, `name` on EditAccountForm/EditGroupsForm SSR
fixtures, plus a couple in no-render-warnings.test.tsx).
Total: 298 web + 31 shared + 26 bot = 355 passing (was 338).
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
In-tab notification bridge so the operator gets a system notification
when a reminder fires successfully (or partly / fails) and when a
send-test message lands. Foundation for true background push later
(VAPID + service-worker subscription); this lands the wiring so
behaviour is testable today.
Pieces
------
- `lib/notifications.ts` — pure helper module:
* notificationSupport / getPermission — feature detection that
treats the SSR / unsupported-browser case as "denied" so callers
don't have to handle a third state.
* isOptedIn / setOptedIn — localStorage-backed opt-in flag
(key `cmbot.notifications.optedIn`). Survives gracefully when
window is missing or storage throws (private mode / quota).
* showNotification(opts) — gated dispatch returning a discriminated
result ({ ok: true, tag } | { ok: false, reason }) so callers
can fall back to a UI toast on opt-out / unsupported / error.
* reminderFiredToNotification + sendTestDoneToNotification —
pure mappers from the bot's SSE events into notification args.
Skips bookkeeping noise (status === "skipped") and failures
that the in-page toast already shows verbatim.
- `components/notification-manager.tsx` — client component mounted
once at the app shell. Subscribes to `reminder.fired` and
`send_test.done` via useEvents and forwards each through the pure
mappers. Renders no DOM.
- `components/notifications-toggle.tsx` — settings-page card with
three states (unsupported / not-granted / granted+opted-in).
"Send test" button fires a sample notification so the operator
can verify the wiring without waiting for a real reminder. The
blocked-by-browser path points them at site settings instead of
silently doing nothing.
- `app/settings/page.tsx` — new "Notifications" card sits above
the Appearance card.
- `app/layout.tsx` — `<NotificationManager />` rendered alongside
`<Toaster />` inside ThemeProvider so the SSE subscription is
active across all routes.
Bot side
--------
- `apps/bot/src/scheduler/fire-reminder.ts` — emits
`pgNotifyWeb({ type: "reminder.fired", reminderId, runId, status })`
after every run regardless of success/partial/failed. The web
side decides whether to surface it as a notification (skipped is
filtered out client-side).
- send_test.done was already emitted by `ipc/send-test-handler.ts`.
PWA service-worker tests (the original ask before this thread)
--------------------------------------------------------------
- Extracted the Serwist config into `pwa/config.ts` so the choices
(skipWaiting, clientsClaim, navigationPreload, runtimeCaching,
precacheEntries) are pinnable without booting a worker scope.
- 6 tests in `pwa/config.test.ts` lock the surface (no extra keys
appear silently, the manifest passes through unchanged, the
pinned booleans stay where production expects them).
- 6 tests in `app/manifest.webmanifest/route.test.ts` cover the
manifest contract (display=standalone, start_url=/, dark theme
colors match the OS, both icons are PNG + maskable, paths
match committed PNGs in public/).
Test counts
-----------
281 web + 31 shared + 26 bot = 338 total (was 306).
- +6 pwa/config (service-worker config pinning)
- +6 app/manifest.webmanifest (PWA manifest contract)
- +20 lib/notifications (full coverage of mappers + dispatch
gates + SSR / unsupported / blocked / opted-out paths)
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
The web app is now installable on a phone home screen with offline
fallback for static assets and the navigation shell.
Pieces
------
- `src/app/manifest.webmanifest/route.ts` — dynamic manifest route.
Standalone display mode, portrait orientation, dark theme matching
the app, "any maskable" icons so the same PNG works for both
regular launchers and Android adaptive icons.
- `src/pwa/sw.ts` — service worker entry. Uses serwist's stock
recipe: skipWaiting + clientsClaim so a new worker takes over on
the next navigation, navigationPreload to race the network with
the worker boot, and `defaultCache` for HTML-network-first /
static-cache-first / image+font cache TTLs.
- `next.config.ts` — wraps the existing config with `withSerwistInit`.
Disabled in development (`NODE_ENV !== "production"`) because a
service worker on every dev reload makes hot-reload extremely
flaky.
- `package.json` build script switched to `next build --webpack`.
`@serwist/next` doesn't yet support Turbopack (it logs a warning
and silently skips emitting `sw.js`), and Next 16 defaults the
build to Turbopack. The dev server still uses Turbopack — only
production builds switch to webpack.
- `src/app/layout.tsx` metadata gains `manifest`, `icons.icon` (192
+ 512 PNG), and `icons.apple` (180 PNG). The existing
`appleWebApp.capable` already opts iOS into standalone mode.
Icons
-----
Generated by a tiny one-shot script (`scripts/gen-pwa-icons.ts`)
that uses the workspace's already-installed sharp to render an SVG
wordmark at 512 / 192 / 180 px. Placeholder branding (dark square
with "cm" wordmark) — swap in real artwork later by editing the SVG
in the script and re-running `pnpm --filter @cmbot/web run gen:icons`.
Build artefacts
---------------
- `apps/web/public/icon-512.png`, `icon-192.png`,
`apple-touch-icon.png` ARE committed (stable input).
- `apps/web/public/sw.js` and `swe-worker-*.js` are NOT — they're
regenerated on every production build. Added to `.gitignore`.
Verification
------------
- Production build emits `[serwist] Bundling the service worker
script with the URL '/sw.js' and the scope '/'...` and `sw.js`
shows up in `public/`.
- `/manifest.webmanifest` is in the build's static-route table.
- 249 web tests still passing.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Old behaviour: HEIC/AVIF photos, .mov / .webm / .mkv videos, and niche
audio (FLAC, etc.) got rejected outright at upload with "Images are
not supported" / "Videos are not supported" errors. Strict but
unfriendly — recipients could still receive these as a downloadable
file via WhatsApp's document path; we just weren't using it.
New behaviour: anything not playable inline gets routed through the
document path automatically. The recipient downloads the file and
opens it in their default app. The 100 MB document cap applies
instead of the inline 5 / 16 / 16 MB caps. Only oversized uploads
get rejected.
Where the policy lives
----------------------
The classifier moved into a new `@cmbot/shared/whatsapp-media`
module so the web upload validator AND the bot's fire-reminder send
path share one source of truth:
- resolveDeliveryKind(mime, bytes?) → "image" | "video" | "audio"
| "document". Native types stay as-is; HEIF / AVIF / QuickTime /
WebM / Matroska / non-MP3-or-M4A audio all collapse to "document".
- Bytes argument is optional but recommended — sniffing the first
12 bytes of the file catches iOS Safari's habit of labelling
a HEIC as image/jpeg or a .mov as video/mp4. Bytes win when they
disagree with the mime.
Web side
--------
- `lib/whatsapp-media.ts` re-exports the shared helpers and keeps
only the validator + byte-formatter. `validateForWhatsApp` calls
resolveDeliveryKind internally; the size cap it returns is for the
RESOLVED kind (so a HEIC routes to document and gets the 100 MB
cap). The "Images are not supported" / "Videos are not supported"
rejection messages are gone — there's no format rejection anymore.
- `actions/media.ts` collapses the previous explicit-mime + byte-sniff
pair into a single `validateForWhatsApp(mime, size, bytes)` call.
- Compose-step upload-zone hint updated to spell out the per-kind
caps: "JPEG/PNG up to 5 MB · MP4/3GP up to 16 MB · MP3/M4A/OGG
up to 16 MB · documents up to 100 MB".
Bot side
--------
- `fire-reminder.ts` reads the first 12 bytes of the file before
dispatching and calls `resolveDeliveryKind(mimeType, head)` to
pick the senderKind. So a HEIC on disk (whose mime claims
image/jpeg) gets sent via Baileys' document path — no failed
thumbnail extraction, message arrives as a downloadable .heic.
- New `readHeadBytes(filePath, n)` helper opens, reads N bytes,
closes — no full-file slurp.
Tests
-----
249 web + 31 shared + 26 bot = 306 passing total.
Web (`lib/whatsapp-media.test.ts`):
- "HEIC at 30 MB allowed: routes to document (100 MB cap)"
- "HEIC at 110 MB rejects: exceeds the document cap"
- "MOV at 50 MB allowed (would be 16 MB cap as video, 100 MB as
document)"
- "MOV pretending to be mp4 demotes to document (50 MB allowed)"
- "FLAC audio routes to document path"
- "genuine MP4 byte-sniff path keeps it as video"
Shared (`packages/shared/src/whatsapp-media.test.ts`, new):
- The cross-package contract: 11 tests covering size limits,
classifyMediaKind, resolveDeliveryKind for native + demoted +
byte-sniff cases, plus the underlying helpers.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Three threads from the recent UX iteration:
1. Reminder list / detail no longer shows raw "Cron: 32 11 * * *"
----------------------------------------------------------------
`describeRecurrence` for a kind=cron spec used to emit
"Cron: <expr>" verbatim, which is unreadable on the list row's
recurrence line.
New pure helper `describeCronRule(rule)` parses the cron shapes
the recurrence picker produces and renders them as natural
sentences:
"0 9 * * *" → "Every day at 09:00"
"0 9 * * 1-5" → "Every week on Mon, Tue, Wed, Thu, Fri at 09:00"
"0 9 * * 1,3,5" → "Every week on Mon, Wed, Fri at 09:00"
"0 9 1,15 * *" → "Every month on days 1, 15 at 09:00"
"0 9 13 5 *" → "Every year in May on day 13 at 09:00"
"30 17 1,15 1,4,7,10 *" → "Every year in Jan, Apr, Jul, Oct on days 1, 15 at 17:30"
Multi-line rules ("0 9 * * 1\n0 17 * * 5") join the per-line
descriptions with " · " for compactness in the list density.
Long DOM lists (>6 days) collapse with a "+N more" tail to keep
the line short; same convention the picker's per-row preview uses.
Unrecognised shapes (e.g. "*/5 * * * *") fall back to the raw
expression — better than swallowing entirely.
2. HEIC/AVIF magic-byte sniffing at upload
----------------------------------------------------------------
The mime-only check we shipped earlier missed iOS Safari's
habit of uploading HEIC photos with Content-Type: image/jpeg.
The file then made it to the bot, where Sharp's HEIF decoder
plugin is missing, the thumbnail extraction failed, and the
message went out without a working preview — read by the user
as "image still not send".
New helper `sniffUnsupportedImage(bytes)` reads bytes 4..11 of
the upload and looks for the ISOBMFF "ftyp" marker followed by
one of the brands Sharp can't decode (HEIF: heic / heix / hevc
/ heim / heis / mif1 / msf1; AVIF: avif / avis). Brand match is
case-insensitive. Plain JPEG / PNG / unrelated ftyp brands like
mp4 are not flagged.
`uploadMediaAction` now runs the sniff against the buffered
bytes before persisting, returning the same "Images are not
supported, please re-upload images" error as the mime path.
3. Sidebar brand link → dashboard tests
----------------------------------------------------------------
Asserts the desktop <aside> contains an <a href="/" aria-label=
"Go to dashboard"> at the top, scoped via a new extractSidebar
helper so it can't accidentally match the mobile-header brand
link (which uses aria-label="Go home"). A second test confirms
the two aria-labels stay distinct.
22 web test files / 232 passing (was 212):
- +12 cron-description cases in lib/recurrence.test.ts
- +6 magic-byte sniff cases in lib/whatsapp-media.test.ts
- +2 sidebar-brand-link cases in app-shell.test.tsx
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Two new SSR tests in `app-shell.test.tsx`:
- Desktop sidebar's brand header is now a link, not a static <div>.
Asserts the <aside> contains an <a href="/" aria-label="Go to
dashboard"> at the top. The slice is scoped to the <aside> via a
new `extractSidebar` helper so it can't accidentally match the
mobile-header brand link.
- Mobile and desktop brand links carry distinct aria-labels
("Go home" vs "Go to dashboard"). On a wide window where the
desktop sidebar is visible alongside the (sm:hidden) mobile
header — which technically can't happen at any one breakpoint
but is a useful invariant for screen readers in split-screen /
zoom contexts — the two announcements stay distinguishable.
Total: 212 web tests passing (was 210).
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Three small follow-ups:
1. HEIC/HEIF/AVIF uploads now rejected at the door
----------------------------------------------------
Symptom: an iPhone-shot image uploaded fine but came through on
WhatsApp without a thumbnail. Bot logs:
failed to obtain extra info
heif: Error while loading plugin: Support for this compression
format has not been built in
Cause: the bot container's Sharp ships without a HEIF/AVIF
decoder, so the thumbnail-extraction step Baileys runs throws and
the message is sent without a preview.
Fix: the upload validator (`validateForWhatsApp`) now rejects the
HEIF family before the file ever reaches the action body. Error
message: "Images are not supported, please re-upload images".
New tests in `lib/whatsapp-media.test.ts`:
- `isUnsupportedImageMime` recognises image/heic, image/heif,
image/heic-sequence, image/avif (case-insensitive).
- `isUnsupportedImageMime` does NOT flag jpeg/png/webp/gif.
- `validateForWhatsApp` rejects a HEIC upload regardless of size,
even below the 5 MB image cap.
2. Desktop sidebar brand is now a link to /
----------------------------------------------------
The mobile header brand pill was already a link to /; the desktop
sidebar version was a static <div>, so clicking the "cm WhatsApp
Bot" header in the sidebar did nothing. Wrapped in <Link href="/">
with `aria-label="Go to dashboard"` and a hover background to
make the affordance obvious.
3. Activity tab strip switched from full-width to scrollable
----------------------------------------------------
The activity page has six tabs (All / Success / Partial / Failed
/ Skipped / Archived) — packing them into a `w-full` row at h-8
left every label squeezed to ~50px on mobile. Wrapped the
<TabsList> in an `overflow-x-auto` scroller (with negative
horizontal margins so the strip extends to the page edges and the
first/last tabs aren't clipped) so each tab keeps a comfortable
touch target on phones; on desktop the row fits naturally and no
scroll bar appears.
Reminders page kept its full-width layout — only 4 tabs there,
they don't crowd.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
12 new SSR tests in app-shell.test.tsx:
Mobile header
- Fixed top header is rendered with `sm:hidden` so it disappears on
the desktop breakpoint.
- Brand mark on the left links home and carries `aria-label="Go home"`.
- Page title in the centre is derived from usePathname:
* "/" → "Dashboard"
* "/accounts/123" → "Accounts" (sub-route falls back to parent label)
* unknown route → generic "WhatsApp Bot"
- Menu button on the right is labelled `aria-label="Open menu"`.
Menu drawer (Sheet primitives mocked transparent so SSR shows content)
- Renders one nav link per NAV_ITEM, in declared order.
- The active route's link gets `aria-current="page"`; others don't.
- Dashboard ("/") matches by exact equality, not by `startsWith`, so
every page doesn't get marked Dashboard.
- The drawer does NOT include the theme toggle — it lives only in
the desktop sidebar footer per the recent product call.
- Drawer header carries the brand wording and the SR-only nav-menu
description.
Desktop sidebar
- Renders with `hidden sm:flex` (mobile-hidden, desktop-visible).
- All NAV_ITEMS appear.
- Theme toggle is present in the sidebar footer.
Plus the small follow-up the user pointed at:
UI: status tabs span the full row
- The shadcn `<TabsList>` defaults to `inline-flex w-fit`, which
packed Active/Ended/Paused into a tight cluster on the left of
the reminders + activity pages. Added `w-full` to both
`<TabsList>` invocations so the tabs distribute evenly across
the available row width (`flex-1` on each `<TabsTrigger>` already
handles even widths once the parent stretches).
Total: 206 web tests passing (was 194; +12 from app-shell.test.tsx).
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Several user-reported bugs and UX nits fixed in one cut:
1. Editing account / when / groups silently dropped messages 2..N
--------------------------------------------------------------
Symptom: a reminder with 3 message parts came back with 1 after
the user edited any section other than the message itself.
Cause: the three section forms were still on the legacy
{text, mediaId, caption} prop shape. The parent pages pulled only
messages[0] from the DB, reduced it to those three fields, and
the form posted them through to updateReminderAction. The action
then folded the legacy fields into a single MessagePart and
replaced the whole reminder_messages row set — wiping parts 2..N
even though the user only meant to change the schedule.
Fix: each form (edit-account / edit-when / edit-groups) now takes
the full `messages: MessagePart[]` and forwards it unchanged. The
three parent pages load the full stack (sorted by position) and
pass it through.
Test: new edit-section-forms.test.tsx asserts a 3-part stack
reaches updateReminderAction intact for both the account-form and
groups-form code paths, plus a sanity test that the legacy
single-message payload shape (without `messages`) is what a
future regression would look like.
2. Reminders list: removed the Group filter
--------------------------------------------------------------
Per request — Account + Search already cover the use cases the
Group filter was supposed to. Search even matches group names
directly, so the dropdown was redundant. Page no longer fetches
the groups table for its filter bar at all.
3. Mobile chrome: bottom nav → top header w/ menu drawer
--------------------------------------------------------------
Removed the bottom tab bar. Mobile now has a single-row top
header:
┌──┐ ┌────┐
│cm│ <current page title> │menu│
└──┘ └────┘
- Brand mark on the left links home.
- Current page title sits in the middle so the user always knows
where they are.
- Menu icon on the right opens a right-side Sheet (radix Dialog)
containing the full nav list. Active item highlighted; the
drawer auto-closes when a nav item is clicked (effect on the
pathname change).
- Theme toggle stays only in the desktop sidebar footer per the
follow-up ask.
Main content padding adjusted: pt-16 (mobile) for the h-14
header, no bottom padding now.
4. Cleaned up the now-unused legacy props
--------------------------------------------------------------
`text` / `mediaId` / `caption` removed from the three section
form prop types. The wizard's URL-state pass-through still
accepts the legacy fields and folds them into the new
`messages` shape on entry, so old bookmarked /reminders/new
URLs still work.
194 passing web tests (was 194; net 0 — the new edit-section-forms
tests replaced coverage we lost when the legacy props went away).
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Symptom
-------
The upload action rejected anything over 50 MB with a flat
"File too large (>50MB)" — a number that was both too generous for
images (WA caps at 5 MB) and too restrictive for documents (WA
allows 100 MB). And anything over 1 MB was being rejected even
earlier by Next's default Server Action body limit, with a much
less actionable error.
Fix
---
1. New `lib/whatsapp-media.ts` resolves an uploaded file's MIME type
to a WhatsApp delivery kind and validates it against the
per-kind cap that WA actually enforces:
image → 5 MB image/* except sticker-mode
video → 16 MB video/*
audio → 16 MB audio/*
document → 100 MB anything else (PDFs, office docs, …)
Anything not recognised as image/video/audio falls through to
"document", which is also the Baileys sender path the bot uses
to deliver it. So a .zip or .csv ends up correctly classified
AND correctly limited to the document cap.
Error messages now name the kind and show both the actual size
and the cap: "Image too large (5.2 MB > 5.0 MB limit on
WhatsApp)".
2. `next.config.ts` lifts the Server Action body limit from the 1 MB
default to 100 MB, so document uploads actually reach the action
instead of getting bounced at the framework boundary. The WA
per-kind validator inside the action enforces the real limit
from there.
3. The compose-step upload zone hint now reflects the per-kind caps
("Image up to 5 MB · video / audio up to 16 MB · document up to
100 MB") instead of the wrong flat "up to 50 MB" value.
Tests (17 new cases, total 189)
-------------------------------
- classifyMediaKind: image/video/audio prefix routing, fall-through
to document for unknown / empty / octet-stream / text/plain.
- validateForWhatsApp: at-cap, just-under-cap, just-over-cap for
image (5 MB) / video (16 MB) / audio (16 MB) / document (100 MB);
zero-byte rejected; unknown-mime 60 MB upload accepted as document.
- WA_MAX_BYTES sanity: equals the document cap and is >= every other
per-kind limit (so it's safe to use as the framework body cap).
- formatBytes: bytes / KB (no decimals) / MB (one decimal) rendering.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Three small bugs the user surfaced after the swipe rollout:
1. Editing a paused/ended one-off reminder threw "Time is in the past"
----------------------------------------------------------------------
The four edit-section pages (account / message / groups / when) all
POST through `updateReminderAction`. The action's "scheduledAt
must be in the future" check fires on every submit — including the
three section pages that don't change the time and just pass the
original `scheduledAt` straight through. So a user editing the
message body of a reminder they paused yesterday saw their save
rejected with "Time is in the past".
New pure helper `validateUpdateScheduledAt` in lib/reminder-update.ts
keeps the future-time check in place for active reminders that are
actually changing the time, but allows past timestamps when:
- the reminder is paused or ended (won't fire while in those
states regardless of what the row says about scheduledAt), OR
- the submitted timestamp matches the existing one within a
second of rounding (the form is a passthrough).
Tests: 10 cases in `lib/reminder-update.test.ts` covering active
future, active past, paused passthrough, ended passthrough, paused
with deliberate change, sub-second drift tolerance, exact-NOW edge,
null existing scheduledAt, malformed ISO.
Also (drive-by, related): `updateReminderAction` no longer force-
sets `status: "active"` on save. Editing a paused reminder's
message shouldn't silently un-pause it. The user uses Restart for
that.
2. Reminder list reshuffled after Pause/Restart
----------------------------------------------------------------------
The list defaulted to `sort=scheduled_desc`, so clicking Restart on
row N (which moves scheduledAt forward to the next occurrence) flipped
the row to row 0. Felt like the wrong action ran. Fixed:
- Page now hard-codes `sort = "created_desc"` (created_at never
changes, so a row stays where it is).
- Sort dropdown removed from `<ReminderFilterBar>` since it has
nothing to drive anymore. Account + Group filters and the
search box stay.
3. Swipe shelf stayed open after the action ran
----------------------------------------------------------------------
`SwipeableRow` keeps its offset in component state. When a shelf
button submits the form, the page revalidates and re-renders, but
React keeps the same row instance (matched by `key={reminder.id}`),
so the open offset stuck around. Now both row sites encode the
"row state" into the key:
- reminders: `key={\`${reminder.id}-${reminder.status}\`}`
- activity: `key={\`${run.id}-${run.archivedAt ? "1" : "0"}\`}`
Status flip → key change → React unmounts/remounts → offset back
to 0 → shelf closed. Costs nothing (these rows are cheap).
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Two follow-ups on the activity-row swipe work:
1. SwipeableRow now supports BOTH directions
----------------------------------------
The component grew a `leftActions` slot alongside the existing
right shelf. Drag the row right to pull the left shelf into view
(non-destructive action: Archive, Pause, etc.); drag left to pull
the right shelf into view (destructive: Delete). Past
REVEAL_THRESHOLD (60 px) the corresponding shelf locks open;
below it, snaps closed. Each shelf is opt-in — omit a slot and
the row only swipes one direction.
- `computeSwipeNext` and the new `snapPosition` helper take a
`{ leftActions, rightActions }` flag pair so the math knows
which directions are valid. Drags toward a missing shelf get
clamped to 0 instead of fully open.
- Activity rows wired as iOS-Mail-style: leading edge (right
swipe) = Archive/Restore (amber), trailing edge (left swipe)
= Delete (destructive red).
- Tests grew to 16 cases covering: snap-to-closed below threshold
either way, snap-to-open at/past threshold either way, clamps
don't escape the shelf width, missing-shelf rows don't snap to
a non-existent open state, baseOffset-aware reverse-drag math,
and SSR markup contracts (data-testid, data-state="closed",
translateX(0px), aria-hidden=true on closed shelves, no
orphaned shelf wrapper when only one slot is provided).
Also fixed a `-0` slip in the clamp branch (`-maxRight` is `-0`
when maxRight is 0) so call-site equality checks behave.
2. Reminders list rows are swipeable too
----------------------------------------
/reminders page now wraps each row in SwipeableRow:
- Left swipe → Delete (always available, destructive).
- Right swipe → Pause (when status is "active") OR Restart
(when "paused" or "ended"). Other lifecycle states (failed)
omit the right shelf entirely; the row only swipes one way.
Each shelf button is a tiny `<form>` posting to the existing
server action (delete / pause / restart) — no client-side state
beyond the swipe gesture. Page revalidates after the action,
list re-renders, row redraws in its new state.
Reused the same shelf-button visual language as the activity
tab (color-coded action, icon + label, dark-mode pairs) via a
tiny inline `ReminderShelfButton` helper.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Two unrelated bits the user asked for in the same breath:
1. Activity row swipe-to-reveal actions
----------------------------------------
On the mobile activity tab, drag a row left to reveal an Archive
button (Restore when already archived) and a Delete button. Past a
60 px threshold the shelf locks open; below that it springs back.
Tapping anywhere outside an open row closes it. Desktop keeps a
table layout but gains the same two row-level icon-buttons in a
new Actions column, since hover-then-discover is more natural with
a mouse than a swipe.
- New `<SwipeableRow>` (apps/web/src/components/swipeable-row.tsx)
— pointer-events only (no third-party gesture lib), 130 lines.
The drag math lives in a pure helper `computeSwipeNext` so it's
unit-testable without a DOM.
- Migration 0007 adds `reminder_runs.archived_at timestamptz`
(null = visible by default, non-null = archived). Soft-archive
keeps the row queryable under a new "Archived" filter tab; hard
Delete drops the row entirely (run_targets cascade via FK).
- Server actions: `archiveRunAction` / `unarchiveRunAction` /
`deleteRunAction`. Each rate-limits to 30/min/IP. Ownership
check piggybacks on the same operator-or-orphan rule the
activity query already uses.
- `listActivityRuns(operatorId, { archived })` extended to filter
in or out of the archived window. Default is archived: false so
the existing tabs (All / Success / Partial / Failed / Skipped)
keep showing only live runs.
- Tests
* `swipeable-row.test.tsx` — 6 unit tests covering the drag math
(clamp at 0 / -SHELF_WIDTH, snap-to-closed below threshold,
snap-to-open at or past threshold, snap math respects the
previous offset) plus 2 SSR markup contracts (data-testid /
aria-hidden / starts at translateX(0px) / data-state="closed").
* Total web suite: 154 passing (was 146).
2. Send-test toast text trim
----------------------------------------
"Sent ✓ — check the WhatsApp group." → "Sent ✓". The trailing
note told the user something they could already see (they're the
one who clicked Send Test on a specific group). Less noise.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Reminders can now deliver a stack of message parts in send order. The
DB and bot's fire-reminder loop already supported this — only the UI
and the server action's input shape were single-message. This change
makes the whole flow stack-aware end-to-end.
What's new
----------
A reminder is now a list of MessagePart objects:
{ kind: "text", textContent: "Hi", mediaId: null }
{ kind: "media", textContent: "cap", mediaId: uuid }
{ kind: "media", textContent: null, mediaId: uuid }
The bot fires them in order with ~1.5 s spacing (already the case in
fire-reminder.ts).
Cap of 10 parts per reminder. Anything more clutters the URL beyond
the 2KB practical budget for the wizard's encoded `messages=…` param.
Where this shows up
-------------------
1. `<MessageStack>` — new shared component (apps/web/src/components/
message-stack.tsx). Each block is either a text Textarea or a
media block (file picker + preview + caption Input). Per-block
move-up / move-down / delete buttons. "+ Add text" / "+ Add file"
buttons at the bottom. Reused by both the wizard's compose step
AND the per-section Edit Message page.
2. Edit Message page — was a single Textarea + read-only attachment
indicator with a "Replacing it isn't supported" note. Now uses
MessageStack and lets the operator add/remove/reorder parts AND
swap the file on a media block, fixing
the asked-for "should let user change media files too" gap.
3. Wizard — Compose / When / Groups / Review pass a single
`messages=<urlencoded JSON>` param instead of three separate
text/mediaId/caption fields. The Review step renders one row per
part, with file names resolved from the DB so users see "menu.pdf"
not an opaque uuid. Every step accepts the legacy fields too and
folds them into the new shape on entry, so older bookmarked URLs
keep working.
4. Server actions (createReminder / updateReminder) accept either:
- The new `messages: MessagePart[]` field, OR
- The legacy `text` / `mediaId` / `caption` triple,
and resolve to a flat parts list before doing anything else. Both
actions then write one row per part into `reminder_messages` with
a sequential `position` column, replacing the old "always 1 row"
logic in updateReminderAction.
5. The reminder name (visible in lists, detail header, etc.) is
sourced from the first part with a non-empty text body — falling
back to the literal "Reminder" if every part is media-without-
caption. Capped at 50 chars to fit the existing column.
Wire-format helpers
-------------------
New `lib/reminder-messages.ts`:
- `MessagePart` interface (the canonical shape)
- `isValidMessagePart` — reject empty texts and orphan-mediaId rows
- `encodeMessages` / `decodeMessages` — URI-encoded JSON, drops
invalid entries, returns null when nothing valid is left
- `legacyMessageToParts` — synthesise a one-element stack from the
old text/mediaId/caption fields (used by step pages on entry)
Tests (15 + 5 = 20 new; 146 total, was 132 + adjustment)
--------------------------------------------------------
- `lib/reminder-messages.test.ts`: round-trip a non-trivial stack;
survive URL-unsafe characters in text (\\n, & = % #); reject
null / empty / garbage; drop invalid entries; legacy-fallback paths.
- `edit-message-form.test.tsx`: rewrites for the new prop shape
(initialMessages instead of initialText/initialMediaId/initialCaption);
asserts the form renders one block per initial part and that media
filename appears in the SSR markup.
- `no-render-warnings.test.tsx`: same prop-shape update for the two
EditMessageForm hydration / button-nesting guards.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Two small additions:
1. Accounts list "Last connected" line: bumped from a date-only string
("10 May 2026") to a full timestamp with day, month, year, hour
(12-hour with AM/PM), minute, second — same KL timezone, en-MY
locale. Useful for diagnosing recent disconnects vs old ones at a
glance.
2. New \`duplicateReminderAction\` server action plus a fourth card on
the reminder detail's ActionsBar (Pause / Restart / Duplicate /
Delete). The action copies the source reminder's account, groups,
message parts, and schedule (rrule unchanged). The new row starts
\`paused\` so it doesn't fire on top of the original — operator
tweaks the schedule from the detail page and Restarts when ready.
Name is suffixed with " (copy)" (capped at 60 chars).
ActionsBar grid bumped from 3-column to 4-column at lg, with a 2x2
fallback at sm so it doesn't get cramped on narrower screens.
Test mock for actions-bar.test.tsx widened to include the new action.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Symptom
-------
Click Unpair on a connected account (or Delete Account). The web
action runs:
DELETE FROM whatsapp_groups WHERE account_id = ?
and Postgres rejects it:
error: update or delete on table "whatsapp_groups" violates
foreign key constraint
"reminder_run_targets_group_id_whatsapp_groups_id_fk"
on table "reminder_run_targets"
Cause
-----
\`reminder_run_targets.group_id\` had a non-null FK to
whatsapp_groups.id with no ON DELETE rule (defaults to NO ACTION /
RESTRICT). So any reminder that had ever fired pinned the group rows
in place. Unpair couldn't wipe the synced groups, the action threw,
and the row never reached \`status='unpaired'\`.
Fix
---
Mirror the pattern \`reminder_runs.reminder_id\` already uses
(migration 0005): nullable column + ON DELETE SET NULL + a
denormalised label snapshot, so historical fan-out records survive a
group wipe but stay readable.
Migration 0006:
- Drop the composite \`(run_id, group_id)\` PK; add a surrogate
\`id uuid pk default gen_random_uuid()\` since \`group_id\` can no
longer be part of the PK once it's nullable.
- Make \`group_id\` nullable.
- Re-create the FK with ON DELETE SET NULL.
- Add \`group_label text\` for the snapshot.
fire-reminder.ts now writes the group's name into \`group_label\`
on every insert path (success / failed / skipped /
account-not-connected / group-missing) so the Activity tab can keep
showing "Sent to <Group Name>" even after the group is gone.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Symptom
-------
Click "Unpair" on a connected account. The web action sets
\`status='unpaired'\`, but the account detail page often still shows
"Disconnected" — and on accounts that had been previously connected,
the QR pair flow restarts a few seconds later all on its own.
Cause
-----
Two races inside the session manager:
1. The web's \`unpairAccountAction\` notifies the bot via \`pg_notify\`
and then writes \`status='unpaired'\` to the row. The bot's
\`handleUnpair\` calls \`sessionManager.stop()\` which closes the
Baileys socket; Baileys eventually fires a \`connection: close\`
event which the manager's \`handleEvent\` translates into a
\`status='disconnected'\` UPDATE. Whichever write lands second wins.
The user clicks Unpair and sees Disconnected.
2. The same close-event handler schedules a 5-second
\`stop().then(start())\` reconnect for accounts whose
\`lastConnectedAt\` is set. Five seconds after unpair, the bot
silently re-opens the socket, the row flips to \`pending\`, and the
QR carousel restarts.
Fix
---
\`stop(accountId, { intentional: true })\` marks the account in a new
\`intentionalStops\` Set. When the close event lands, \`handleEvent\`
drains the flag (with \`Set.delete()\` returning whether the key was
present, so it's exactly-once and a stale flag can't bleed into a
later session) and skips both the DB UPDATE and the reconnect
schedule. The caller — only \`handleUnpair\` for now — is the one
choosing the row's next state, so we step out of its way.
The flag is set ONLY when callers ask for it. Internal recoveries
(restartRequired auto re-open, ephemeral-close back-off) keep the
default behaviour and continue to write \`disconnected\` + reschedule.
Drive-bys
---------
- Refresh the stale "the row is gone by the time we run" comment in
unpair-handler — the row stays alive now (the operator can re-pair
without retyping the label). Look up the account first so the
audit log carries the real \`operatorId\` instead of \`null\`. The
delete-account flow really does delete the row before notifying us;
the lookup tolerates that and falls back to \`null\`.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
DialogFooter with showCloseButton now lays out the auto-rendered Close
and the caller's primary action (typically a <form>-wrapped Submit) in
a 2-column grid that's identical at every viewport. Both buttons are
sized "sm" with w-full so they fill their column and match in height.
The trick to making this transparent for callers: \`[&>form]:contents\`
collapses the form box so its <Button> child becomes a real grid item
sibling of the Close button, not a single grid cell containing the
button. \`[&>form>button]:w-full\` then sizes the submit button to
match the Close button's column width.
Five existing call sites pick this up automatically — no changes
needed at the call site:
- reminder pause/restart/delete (actions-bar)
- account unpair / delete
- dashboard "Clear history"
- activity tab "Clear history"
Also: dashboard "Recent activity" now shows the 3 most recent runs
instead of 10. The "Recent runs" stat card description updates to
match ("3 most recent runs"), points to /activity, and a "View all"
ghost link sits beside the section heading so you can jump to the
full history without hunting.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Yearly was a single Month dropdown + a Day number input — one Month and
one Day per rule. That meant "every quarter on the 1st" needed four
separate schedule rows.
Now Yearly mirrors Monthly's grid pattern but with two grids:
Months [Jan][Feb][Mar][Apr][May][Jun]
[Jul][Aug][Sep][Oct][Nov][Dec]
Days [ 1][ 2][ 3]...[31] (7×5 grid)
Both grids are multi-select. Cron output uses the comma-list form on
both DOM and month positions:
months: [1,4,7,10] + days: [1] → "0 9 1 1,4,7,10 *"
months: [12] + days: [24,25,31] → "0 9 24,25,31 12 *"
The cron field is a Cartesian product — every selected day fires in
every selected month. So "every quarter on the 1st" is now one rule.
Round-trip: parser accepts comma-lists for both DOM and month, with
single-element shapes (the old "0 9 13 5 *") still loading fine.
Migration of saved data: old yearly rules with one DOM + one month
parse into monthDays=[X], months=[Y] — identical visual selection in
the new grid, identical cron output. No DB changes needed.
Renamed `Draft.month` to `Draft.months: number[]`. The "Single
day-of-month for yearly" field is gone — yearly now reads
`monthDays` (same as monthly).
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Monthly tab is now a 7-column grid of buttons for days 1-31. Tap any
combination to fire the reminder on those days every month — picks
multiple. The summary line reads, e.g.:
"Every month on days 1, 15 at 09:00"
Days that don't exist in some months (29-31) skip naturally — that's
just how the cron DOM field works, no extra plumbing needed.
Cron output uses the comma-list form:
Selected days [1, 15, 28] → "0 9 1,15,28 * *"
The parser now accepts a comma-separated DOM list on the way back in,
so the picker round-trips a saved monthly rule with all selected days
restored. Pre-existing single-day monthly rules (e.g. "0 9 15 * *")
still load fine — the same regex handles both.
Empty-state copy: rewrote the verbose
"Doesn't repeat — fires once at the date and time above."
down to just "No Repeats". The label "Repeats" above the box plus the
"Add a recurring schedule" button below already explain the behaviour;
the long sentence was pure noise.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
The Daily tab had two radios: "Every day" vs "Every weekday (Mon–Fri)".
That's confusing — Mon-Fri-only is a weekly pattern, not a daily one,
and it overlapped exactly with what the Weekly tab can already do
(select Mon, Tue, Wed, Thu, Fri).
So Daily now means literally every day. The tab body is just the time
picker plus a one-liner ("Fires once a day at the time below.").
Legacy reminders that stored "MM HH * * 1-5" still load fine — the
parser maps any DOW list (including the 1-5 range) onto a Weekly draft
with Mon-Fri pre-selected. So a saved "weekday" daily reminder shows
up as Weekly with the right days checked, no data loss.
RadioRow component went unused after this — removed.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Each schedule row in the recurrence picker now owns its own HH:MM
fire time. Previously every row inherited the time from the date+time
inputs above, which meant you couldn't say "every Monday at 09:00 AND
every Friday at 17:00" — both rules shared whatever the form-level
time was.
The Daily / Weekly / Monthly / Yearly tabs each render a small
"Fires at" time field. The picker still seeds new rules with the
form's date+time so the most common case (one rule, time matches the
first fire) doesn't need extra clicks.
Round-trip: the cron line `35 17 * * 1` now restores hour=17, minute=35
on a weekly draft. The parser pulls MM HH off the front of every
expression and feeds the rest of the pattern matchers as before.
Also clean up two pieces of duplicated/obsolete UI feedback in both
the wizard When step and the edit When form:
- Removed the `<p>` showing "Cron: 0 9 * * *" — the per-row
description ("Every day at 09:00") already says it, in human form.
- Removed the standalone "Times are in Asia/Kuala_Lumpur" footer.
The timezone is shown elsewhere (header) and the cron output is
always evaluated in the configured zone — telling the user "times
are in <tz>" inline in the picker is noise.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Two changes in one cut, both per the user's redesign asks:
1. Bring the recurrence picker INLINE into the When form section.
The dialog is gone — the type tabs and per-type config now live
directly under the date+time inputs:
[ Starts on ] [ Time ]
Repeats
┌──────────────────────────────────────────────────┐
│ Schedule 1 [✕] │
│ [Daily] [Weekly] [Monthly] [Yearly] │
│ <per-tab config> │
│ Every weekday at 09:00 │
├──────────────────────────────────────────────────┤
│ Schedule 2 [✕] │
│ [Daily] [Weekly] [Monthly] [Yearly] │
│ <per-tab config> │
│ Every Friday at 17:00 │
└──────────────────────────────────────────────────┘
[+ Add another schedule]
2. Allow multiple recurrence rules per reminder. Each row is its own
tab strip + config; the picker compiles them down to a single
newline-joined CRON: rule. Empty list = "Don't repeat" (one-off).
MAX_RULES is 8.
Storage stays the same (`reminders.rrule`, `CRON:` sentinel). The
multi-rule format is just newline-separated cron expressions:
CRON:0 9 * * 1
0 17 * * 5
`@cmbot/shared` updates to support that:
- nextOccurrence: splits on newline, computes the next match for
each rule independently, returns the earliest. Malformed lines
are skipped (so one bad rule doesn't kill the whole schedule).
- validateMinInterval: validates every line; any single line firing
more often than the 5-min minimum fails the whole rule.
Removed: the standalone modal Dialog wrapper, Reset/Cancel/Save
buttons, and the saved-vs-draft synchronisation. The picker now
edits state directly and the parent form's Save commits everything
at once (consistent with the date+time inputs that have always
behaved that way).
Tests (+3 in shared rrule.test.ts; total 20 shared + 26 bot + 132 web
= 178)
- nextOccurrence on a multi-line cron picks the earliest:
* "0 9 * * 1\n0 17 * * 5" starting Saturday → Mon 09:00 KL
* Same rule starting Tuesday → Fri 17:00 KL
- nextOccurrence ignores malformed lines and still returns the next
match from the valid ones.
- validateMinInterval: passes a clean two-line rule; rejects a rule
containing a too-frequent line.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
The previous flat radio list with N-minutes / N-hours / Custom-cron
options is gone. Per the Temenos UUX `date-recurrence-picker` pattern
(developer.temenos.com/uux/docs/components/date-recurrence-picker), the
form now shows a single read-only trigger field summarising the saved
rule:
┌────────────────────────────────────────────────────┐
│ 📅 Don't repeat ▾ │
└────────────────────────────────────────────────────┘
Clicking the trigger opens a modal with the recurrence types as a
tab strip and per-type config swapped in below:
┌──────────────────────────────────────────────────┐
│ Repeat schedule ✕ │
├──────────────────────────────────────────────────┤
│ [Don't repeat] [Daily] [Weekly] [Monthly] [Yearly] │
│ │
│ <per-type config> │
│ │
│ Fires: <plain-text confirmation> │
│ │
│ [Reset] [Cancel] [Save] │
└──────────────────────────────────────────────────┘
Per-tab config:
- Don't repeat — informational text only
- Daily — radio: "Every day" / "Every weekday (Mon–Fri)"
- Weekly — Mon..Sun chip multi-select
- Monthly — day-of-month input (1-31)
- Yearly — month select + day input
The "Fires: …" sentence updates live as the user edits and reflects
the outer time-picker's HH:MM. Save commits, Cancel discards.
Removed:
- Every N minutes
- Every N hours
- Custom cron expression…
- The standalone helpers `flowToCron` / `flowFromCron` /
`freqChoices` / `defaultFlowState` / `FlowState` / `FreqChoice`
in `lib/recurrence.ts`. Their job (compile a UI state to a cron
string and parse one back) now lives privately inside the picker.
Storage / runtime
- Output is still a `CRON:` prefixed rule in `reminders.rrule`. The
bot's `nextOccurrence` already dispatches cron rules through
cron-parser, so no schema or scheduler changes were needed.
Tests (132 web)
- recurrence.test.ts trimmed to keep only what survives: CRON-rule
round-trip via buildRrule + specFromRrule, and the ISO→cron
weekday helper.
- Existing wizard / edit-when-form integration tests are unaffected
because the picker exports the same `<RecurrencePicker>` props.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
