R2-H1: smart-restore's berth_released auto-reversal was a no-op while
the wizard claimed success. Now uses the persisted interestId from
the decision detail to re-insert the interest_berths link and flip
the berth status back to under_offer. Verifies the interest still
exists and isn't archived before re-linking.
R2-H2: smart-archive berth status update had a TOCTOU race — read
outside tx, write inside without a lock. Now selects-for-update the
berths row inside the tx and re-checks status against the locked row
before flipping to available, preventing concurrent archive+sale
from un-selling a berth.
R2-H3: bulk-archive's berth→interest lookup fell back to
dossier.interests[0]?.interestId ?? '' which sent empty-string
interestIds that silently matched zero rows. Dossier now exposes
linkedInterestIds[] per berth (authoritative interest_berths join);
bulk + single-client wizard both use it and skip berths with no
linked interest. Affected:
- src/lib/services/client-archive-dossier.service.ts (DossierBerth)
- src/app/api/v1/clients/bulk/route.ts
- src/components/clients/smart-archive-dialog.tsx
R2-H4: external-EOI ran storage upload + 4 DB writes outside a
transaction. Now wraps file/document/event/interest writes in a
single tx; storage upload stays before the tx (S3 isn't
transactional), orphan-object on tx failure is acceptable.
R2-H5: bulk archive double-submit treated already-archived clients as
per-row failures. Bulk callback now early-returns success when the
dossier shows archivedAt is set, making the endpoint idempotent.
1175/1175 vitest passing.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
H1: hard-delete-request and bulk-hard-delete-request endpoints had no
rate limit; an admin's compromised account could email-bomb the
operator's inbox or use the endpoints as a client-id oracle. Added a
new `hardDeleteCode` limiter (5 per hour per user).
H3: hard-delete error messages distinguished "no code requested" from
"wrong code", letting an attacker brute-force the 4-digit space with
~5k attempts (vs the full 10k). Both single + bulk paths now return
the same 'Invalid or expired confirmation code' message.
H5: invalid Documenso webhook secret submissions are now rate-limited
per-IP (10 per 15min) and only audit-logged inside the cap, so a slow
enumeration can't fill the audit log silently. Real Documenso traffic
won't fail the secret check, so any traffic beyond the cap is
brute-force.
1175/1175 vitest passing.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
C1: src/worker.ts and src/server.ts only imported 5 of 10 BullMQ
workers. ai/bulk/maintenance/reports/webhooks were never started, so
in production: webhooks never delivered, no maintenance crons (DB
backups, session cleanup, retention sweeps, alerts, analytics refresh,
calendar sync), no scheduled reports, no AI features, no async bulk.
All 10 are now imported and held against GC.
R2-C1: Bulk archive's runBulk callback discarded the return value
from archiveClientWithDecisions, so Documenso envelopes marked for
void in the wizard were never queued and next-in-line notifications
never fired. Now we collect the per-archive (dossier, result) pairs
and replay the same post-commit fan-out the single-client route uses.
R2-C2: Archived-client header's Restore icon was hovering destructive-
red because an unconditional hover:text-foreground was overriding the
later conditional. Restore now hovers emerald; archive still hovers
red.
1175/1175 vitest passing.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Builds on the audit infra split (severity/source) by emitting events
from every place a security or operations review would want to see:
Sensitive data views (severity=warning):
- GDPR export download URL issued
- Audit log page opened (watch-the-watchers; first page only)
- CSV export of expenses
- Webhook secret regenerated
Authentication abuse (severity=warning, source=auth):
- Portal sign-in: success + failed-credentials + portal-disabled
- Portal password reset: unknown email + portal-disabled + bad token
- Portal activation: bad/expired token
Inbound webhook hardening:
- Documenso webhook with invalid X-Documenso-Secret now writes
webhook_failed instead of being silently logged
Background work (source=cron / job):
- New attachWorkerAudit() helper wires every BullMQ worker to emit
job_failed (severity=error) on .on('failed') and cron_run on
.on('completed') for any job whose name matches the recurring
scheduler list. Applied across all 10 workers.
1175/1175 vitest passing.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Audit log was previously silent on authentication and on background
work. This wires:
- Login (success + failed) and logout via a wrapper around better-auth's
[...all] handler. Failed logins are severity 'warning' and carry the
attempted email so brute-force attempts surface in the inspector.
- New severity (info|warning|error|critical) and source (user|auth|
system|webhook|cron|job) columns on audit_logs. permission_denied
defaults to 'warning', hard_delete to 'critical'.
- Webhook delivery success/failure/DLQ/retry now write audit rows
alongside the webhook_deliveries detail table.
- IP address is now visible as a column in the inspector (was already
captured at the helper level).
- Audit UI: severity badges per row, severity + source dropdowns, IP
column, expanded action filter covering hard-delete, webhook events,
job/cron events.
Migration 0044 adds the two columns + their port-scoped indexes.
1175/1175 vitest passing.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Splits seed bootstrap (ports/roles/profile) into a shared module so
two seed entry points can share it:
- pnpm db:seed realistic NocoDB-shaped fixture (existing)
- pnpm db:seed:synthetic 12 clients, one per pipeline stage + archive
variants (rich metadata for restore wizard)
scripts/db-reset.ts truncates all data tables (preserves migrations);
guarded by --confirm and a localhost host check. Companion npm scripts:
- pnpm db:reset
- pnpm db:reseed:realistic
- pnpm db:reseed:synthetic
scripts/dev-open-browser.ts launches a headed Chromium with no viewport
override (uses the host monitor's natural size), pre-fills the login
form for the requested role.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Walks the new dossier dialog end-to-end on a freshly-created client and
asserts the toast + list refresh. Companion test exercises the smart-
restore wizard.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Outbound webhook deliveries already retry with backoff, dead-letter
after maxAttempts, and notify super admins. This adds operator-level
replay: a per-row button on the deliveries log spawns a fresh pending
delivery + queues a new BullMQ job. The original failed row stays
intact so the response body remains for audit; the replay payload
carries retried_from/retried_at markers so receivers can deduplicate.
Inbound idempotency was already handled via the documentEvents
signatureHash unique index.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Replaces the single window.confirm() with a 3-stage wizard:
- preflight: counts auto/needs-reason/blocked (POST /bulk-archive-preflight)
- reasons: carousel through high-stakes clients capturing per-client
reason (≥5 chars) — bulk endpoint accepts reasonsByClientId map
- confirm: shows the final archivable count and submits
Low-stakes still auto-archives with safe defaults; blocked clients
are skipped with a per-row reason in the preflight summary.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Permanent client deletion is now reachable from:
- archived single-client detail page (icon button, gated by new
admin.permanently_delete_clients perm)
- archived clients list bulk action
Both flows are 2-stage: request a 4-digit code (sent to operator's
account email, 10min Redis TTL), then enter both code AND a typed
confirmation (client name single, "DELETE N CLIENTS" bulk). Cascade
strategy preserves audit trails: signed documents, email threads,
files and reminders are detached but retained; addresses, contacts,
notes, portal user, GDPR records, interests and reservations are
deleted via FK cascade or explicit tx delete.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Replaces the simple confirm-restore dialog with a wizard that reads the
persisted archive_metadata via /restore-dossier and surfaces:
- auto-reversed (e.g. berth still available → re-attached on restore)
- opt-in to undo (e.g. berth now under offer to another client)
- locked (e.g. yacht transferred and new owner has active interests)
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Post-archive side-effects now run with backpressure:
- Documenso envelope voids enqueue to BullMQ documents queue with retry/DLQ
- Released berths fan out a "next in line" notification to port users with
interests.change_stage; informational only, no auto stage transitions
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Sales reps need to file EOIs that were signed outside Documenso —
on paper, in person at a boat show, or via an alternate e-sign vendor.
Until now the EOI flow assumed Documenso was the only path.
- external-eoi.service.uploadExternallySignedEoi creates BOTH the
document row AND the signed-file record in one shot. Document is
marked isManualUpload=true with status=completed and signedFileId
set. Distinct from the existing uploadSignedManually which augments
a document row that came from the Documenso pathway.
- POST /api/v1/interests/[id]/external-eoi accepts multipart with the
PDF + optional title + signedAt date + comma-separated signer names
+ free-text notes. Gated on documents.upload_signed permission.
- Interest stage auto-advances to eoi_signed (only when the interest
is currently at or before eoi_sent — past that, just file the doc).
- The signing date, signer names, and any notes are captured into
document_events.eventData + the audit_log metadata so the audit
trail records who said the document was signed and when.
- ExternalEoiUploadDialog renders a small modal: file picker, title
override, signed-date (defaults to today), comma-separated signer
names, notes. Wired into interest-detail-header behind an Upload
icon button (gated on documents.upload_signed).
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Manual stage override
Sales reps need to skip canTransitionStage rules when the data was
entered out of order — e.g. recording a contract_signed deal whose
earlier stages were never tracked in the system.
- New permission flag interests.override_stage in RolePermissions.
Plumbed through the schema TS type, the role-editor UI, the seed
file's pre-built roles (super_admin/director/sales_manager get it,
sales_agent + viewer don't), and the test factories.
- changeStageSchema gains an optional `override` boolean and the
service checks it before evaluating canTransitionStage. When
override=true the reason field becomes required (min 5 chars) and
is recorded in the audit log.
- The route handler gates `override` on the new permission so a
sales_agent without it can't pass override=true and bypass.
- InterestStagePicker auto-detects when the requested transition is
blocked by the table and switches into "override mode" — shows an
amber warning, requires the reason, button label flips to
"Override stage". When the operator lacks the permission, the
warning is red and the button is disabled.
Residential Partner role
Per the smart-archive scoping conversation: external partners who
handle residential inquiries shouldn't see marina clients, yachts,
berths, or financials. The two residential_* permission groups
already exist; this commit just seeds a pre-built system role
("residential_partner") with those flags + minimal own-reminders, so
admins can invite a partner today via /admin/users without manually
building the permission set.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
The new smart-archive backend (d07f1ed) is now wired to the existing
bulk-archive endpoint. Previously /api/v1/clients/bulk just called the
legacy archiveClient — bypassing the dossier and the per-client
decisions. That's now a regression hazard: a power-user could bulk-
archive a client mid-deposit with no audit trail.
New behaviour:
- bulk action='archive' fetches the dossier per client.
- Low-stakes clients (open through eoi_signed) auto-archive with the
same default decisions the single-client modal would pick: release
available/under-offer berths, retain sold berths, cancel active
reservations, leave invoices, leave Documenso envelopes pending,
acknowledge signed documents inline.
- High-stakes clients (deposit_10pct and beyond) refuse with a clear
message: "open the client to confirm + supply a reason". The bulk
summary surfaces the failure per row so the user knows which clients
need individual handling.
- Pre-flight blocker check (e.g. active reservation on a sold berth)
also rejects with a per-row error instead of crashing.
The proper "bulk wizard" UI (per-high-stakes-client confirmation panel
with reason fields) is still TODO — this commit just makes the existing
button safe.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
UI side of the smart-archive backend that shipped in d07f1ed.
- SmartArchiveDialog renders the dossier as a sectioned modal:
Pipeline interests, Berths (with next-in-line listed), Yachts,
Active reservations, Outstanding invoices, In-flight Documenso
envelopes, Auto-handled summary. Each section has a per-row decision
dropdown with sensible defaults (release for available/under-offer
berths, retain for sold berths and yachts, cancel for active
reservations, leave for invoices and documents).
- High-stakes deals show an amber warning panel + require a reason in
the textarea before the Archive button enables. Signed-document
acknowledgment checkbox blocks submission until checked.
- Wires into client-detail-header in place of the previous dumb
ArchiveConfirmDialog (the simple confirm dialog is kept for the
restore case until the smart-restore wizard ships).
- Pre-flight blocker banner surfaces dossier.blockers (e.g. active
reservation on a sold berth) and disables the Archive button entirely.
Two side fixes from CSP rollout:
- next.config CSP allows unpkg.com in dev so the react-grab devtool
loads. Stripped in prod via the existing isProd flag.
- middleware whitelist now passes /manifest.json + icon-*.png +
apple-touch-icon through unauthenticated, so PWA installability
isn't blocked by the auth redirect.
Bulk variant + restore wizard + hard-delete-with-email-code land in
follow-on commits.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
The first slice of the smart-archive project. Replaces the dumb DELETE
client flow with a deliberate "look before you leap" pattern:
- New columns on clients: archived_by, archive_reason, archive_metadata
(jsonb capturing every decision made during archive, so restore can
attempt reversal). Migration 0043.
- client-archive-dossier.service builds a structured snapshot of "what's
at stake" for a given client: pipeline interests, berths under offer
(with next-in-line interests for the notification), yachts owned,
active reservations, outstanding invoices, signed/in-flight Documenso
envelopes, portal user, company memberships. Classifies the client as
low-stakes or high-stakes based on pipeline stage (HIGH_STAKES_STAGES
= deposit_10pct + later) so the bulk wizard knows which clients to
prompt individually.
- client-archive.service.archiveClientWithDecisions takes the operator's
decisions and applies them in a single transaction. Persists the
decision log into archive_metadata for restore. Auto-handles portal
user revocation + company membership end-dating; everything else is
caller-driven. Surfaces external cleanups (Documenso void) for the
caller to queue.
- client-restore.service.getRestoreDossier classifies each persisted
decision as autoReversible / reversibleWithPrompt / locked based on
the current state of the world (berth still available? new owner has
active interests on the yacht? etc). restoreClientWithSelections
applies reversals + un-archives the client.
- 4 API routes wire the services to HTTP. The existing /restore
endpoint is upgraded to use the smart restore but stays
backwards-compatible: clients archived before this feature have no
archive_metadata so the dossier returns empty, and a POST with no
body just un-archives them — same as before.
UI work + bulk variant + hard-delete + Documenso cleanup queueing land
in follow-on commits.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Two audit-pass-#3 prod-readiness gaps.
Security headers
next.config.ts now emits CSP, X-Frame-Options=DENY,
X-Content-Type-Options=nosniff, Referrer-Policy, Permissions-Policy
on every response, plus HSTS in production. CSP allows the small
set of inline-style/inline-script + unsafe-eval (dev-only) needed
by Tailwind, Radix, and Next dev HMR; img-src/connect-src kept
reasonably wide for s3.portnimara.com branding + Socket.IO. Verified
via curl -I that headers ship and that the dashboard route still
serves correctly.
website_submissions retention
Adds 'website-submissions-retention' case to the maintenance worker
with a 180-day window and schedules it at 07:00 daily. Raw inquiry
payloads include reCAPTCHA + IP + UA metadata; keeping them
indefinitely was a privacy + storage gap that audit-pass-#3 flagged.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Three audit-pass-#3 mobile findings, all in shared primitives so the
fix lands everywhere those primitives are used.
- Input defaults inputMode='decimal' when type='number' and the caller
hasn't overridden. Currency/dimension/price fields across invoices,
expenses, berth specs etc. now show iOS's numeric pad instead of full
QWERTY. Caller can still pass inputMode='numeric' for integer-only
fields.
- DialogContent: padding tightens to p-4 on mobile and restores p-6
at sm+ — the previous fixed p-6 ate ~48px of horizontal width on a
390px iPhone, crushing form-field space. Also adds a max-h-[100dvh]
+ overflow-y-auto so long modal forms scroll inside the dialog
instead of pushing the close button off-screen.
- MoreSheet (mobile bottom-tab "More" drawer): grid-cols-3 cells now
enforce min-h-[88px] so each Apple-HIG-sized 44pt touch target gets
reliable hit area. Icon size bumped from 6 to 7 for visual weight at
the larger cell.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Berth detail page was the last entity using read-only SpecRow widgets
while clients/yachts/companies all use the click-to-edit
InlineEditableField pattern. Marina staffers couldn't update
length/width/draft/price etc without exporting and re-importing.
- New EditableSpec wrapper preserves the SpecRow look + null-hiding
behaviour but defers the value to InlineEditableField with a per-
field PATCH callback.
- useBerthPatch hook hits PATCH /api/v1/berths/{id} (already shipped)
and invalidates the React Query cache for both the list and the
individual berth.
- Numeric helper handles the schema's NUMERIC-as-string convention:
empty input → null, non-numeric → throws, valid → coerced to number.
- 12 fields now editable: lengthFt, widthFt, draftFt, waterDepth,
mooringType, sidePontoon, bowFacing, access, powerCapacity, voltage,
cleatType, cleatCapacity, bollardType, bollardCapacity, price.
- Tags card uses InlineTagEditor instead of read-only badges, matching
the yacht/client pattern. The /api/v1/berths/[id]/tags endpoint was
already in place.
- Dropped the formatDim/formatPower/formatVoltage/price helpers that
inlined the metric column or currency suffix; the editable layout
shows ft/kW/V suffixes inline with the field labels instead. The
metric column is editable separately if needed.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
The /api/v1/companies/bulk endpoint shipped in the previous bulk
batch but the UI side was deferred. Mirrors the client-list /
yacht-list pattern: Add tag, Remove tag, Archive bulk actions with
a single TagPicker dialog for tag operations and a window.confirm
for the destructive archive.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Grab-bag of UX gaps from audit-pass-#2 + #3. Each one is a small,
focused fix; bundled because they touch different surfaces.
- Popover: collisionPadding={16} + responsive
w-[min(calc(100vw-2rem),18rem)] so popovers stop clipping past the
viewport on iPhone 12 portrait.
- public/manifest.json (was missing) + manifest reference in
layout.tsx — PWA installability now works; icons (192/512/512-
maskable) were already present.
- Admin webhooks page: 4 silent `// ignore` catches in load/delete/
toggle/regenerate replaced with toast.error / toast.success. Users
no longer see a stale list with no feedback when an op fails.
- Portal document-download button: blocking alert() → toast.error().
- src/app/(dashboard)/error.tsx: branded error boundary with retry +
back-to-dashboard, replacing Next.js's default uncaught-error UI.
- GDPR export modal: refetchInterval was a flat 5s while the modal was
open. Switched to a function that only polls (every 15s) when a job
is actually pending/building; settled exports stop polling entirely.
- client-yachts-tab empty state gains a CTA wired to the existing
Add-yacht dialog, instead of just saying "No yachts".
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Three audit-pass-#3 findings, all in the "wakes you at 3am" category.
- /api/public/health now runs DB SELECT 1 + Redis PING in parallel and
returns 503 + a degraded payload when either fails. Anonymous probes
(no X-Intake-Secret) still get a flat {status:'ok'} so generic uptime
monitors keep working; authenticated probes see the dep results.
- All worker entrypoints (ai, bulk, documents, email, export, import,
maintenance, notifications, reports, webhooks) and src/lib/redis.ts
now use env.REDIS_URL (Zod-validated at boot) instead of
process.env.REDIS_URL!. Previously a missing env let the app start
silently and fail at first job pickup.
- maintenance worker gains an `error-events-retention` case that
delete()s rows older than 90 days from error_events. scheduler.ts
registers it at 06:00 daily. Closes the contract from migration
0040 which declared the table "pruned at 90 days" but had no
implementation.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
audit-pass-#2 flagged that updateBrochure and archiveBrochure validate
portId in their preceding SELECT but omit it from the subsequent UPDATE
WHERE clause. Currently safe (the SELECT throws NotFoundError first),
but a refactor that drops the SELECT or a TOCTOU race would silently
allow a cross-tenant write.
Defense-in-depth: add and(eq(id), eq(portId)) to both UPDATE WHERE
clauses so the safety property doesn't depend on caller discipline.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Until now the only bulk action anywhere was Archive on the interests
list — implemented as parallel fan-out with no per-row failure
reporting. The bulk BullMQ worker was a TODO stub with no producers.
- bulk-helpers.runBulk wraps a per-row loop and returns
{results, summary} for the caller. Page-size capped at 100.
- New endpoints: /api/v1/{interests,clients,yachts,companies}/bulk
with a Zod discriminated union over the action. Interests support
change_stage + add_tag + remove_tag + archive; clients/yachts/companies
support archive + add_tag + remove_tag. Each action is permission-gated
individually (delete vs edit vs change_stage).
- interest-list, client-list, yacht-list expose the new actions in the
bulk-action toolbar with dialogs for stage / tag selection. Failure
summaries surface via window.confirm.
- bulkWorker stub gets a docblock explaining the v1 sync-only choice
and what the queue is reserved for (CSV imports, port-wide migrations,
bulk emails to >100 recipients).
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Closes the bulk of audit-pass-#1 admin gaps in one batch.
New admin pages:
- /admin/inquiries reads website_submissions with filter chips for
berth/residence/contact + payload viewer per row.
- /admin/sends reads document_sends with sent/failed filter chips and
expandable body markdown; failures surface errorReason and any
fallback-to-link reason from the SMTP retry.
- /admin/email-templates lets per-port admins override the subject of
each transactional template (8 templates catalogued in
template-catalog.ts). Body editing is a follow-on; portal_activation
+ portal_reset are wired to honor the override via loadSubjectOverride.
- /admin/reports replaces the "Coming in Layer 3" placeholder with a
KPI dashboard: 4 KPI tiles, pipeline funnel bars, berth occupancy
donut-bars, conversion %, refresh every 60s.
- backup/import/onboarding admin pages replace placeholders with
actionable guidance: backup posture + planned features, available CLI
imports + planned UI, ordered onboarding checklist linking to admin
pages.
Existing pages widened:
- settings-manager exposes the 9 berth-recommender tunables that were
previously code-only (recommender_*, heat_weight_*, fallthrough_*,
tier_ladder_hide_late_stage).
- role-form covers all 19 RolePermissions schema groups; previously
missing yachts/companies/memberships/reservations + missing
documents.edit + files.edit checkboxes. snake_case residential
labels replaced with friendly text.
portal-auth.service.ts now also writes audit_log rows for portal
invite, resend, activate, password-reset request, and reset (closes one
more audit-pass-#2 gap while we were touching the file).
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Until now only the global /admin/audit page surfaced audit_logs. Each
entity detail page either lacked the Activity tab entirely or rendered
"Activity log coming soon" text.
- entity-activity.service.loadEntityActivity wraps searchAuditLogs
with actor-email resolution; reused by all 5 endpoints.
- New endpoints: /api/v1/{clients,yachts,companies,berths,interests}/[id]/activity,
each gated on the per-entity .view permission and tenant-checked
against ctx.portId.
- EntityActivityFeed renders a timeline with action verb ("Updated",
"Archived"), actor name, relative time, and field old→new diff.
- client-tabs, yacht-tabs, company-tabs, berth-tabs now mount the feed
on their Activity tab. Interest already has the richer
InterestTimeline component.
- yacht-tabs YachtInterestsTab also gets a friendlier empty state with
guidance copy.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
The user-menu's Profile link previously 404'd, and CRM users had no way
to change their password from inside the app.
- /api/v1/me/password POST wraps better-auth changePassword, surfaces a
friendlier "Current password is incorrect" on the typical failure
mode, and writes an audit_log row with metadata.revokedOtherSessions.
- /{port}/settings/profile renders display name + email + change-password
card with current/new/confirm fields and a 'Sign out other devices'
toggle.
End-to-end verified: wrong current pw → 400 with mapped message;
correct → 200 + audit row; revert → 200.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
- AuditAction gains password_change, portal_invite/activate/reset
variants, send, view. AuditLogParams.ipAddress/userAgent now optional
so background jobs and internal helpers can log without faking values.
- alerts.service.dismissAlert/acknowledgeAlert now write
action='update' rows with metadata.kind so the audit log differentiates
the two state changes.
- expense-dedup.service.clearDuplicate/mergeDuplicate accept userId
and write action='update'/'merge' rows respectively. Routes pass
ctx.userId.
Audit gaps surfaced by audit-pass-#2: 6 services bypassed audit_logs
entirely. This commit closes 2 of them; portal-auth lands in a later
commit alongside the email-template-override work that already touches
the same file.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Two regressions from yesterday's audit-tier-0 work that broke the dev
server and every clients API call.
- baseListQuerySchema lived in route-helpers.ts, which was made
server-only by the rate-limit import. Every validator imported it,
pulling ioredis (and dns/net/tls/fs/node:async_hooks) into the client
bundle — every form/detail page returned 500 in dev. Extracted the
schema to src/lib/api/list-query.ts and updated all 14 validators.
- clients.service.listClients and email-compose used raw SQL
ANY(\${jsArray}) which Drizzle binds as JSON — Postgres rejects with
42809 "op ANY/ALL (array) requires array on right side". Switched to
the inArray helper.
GET /api/v1/clients now returns 200 again. Affects every form/detail
page that imports a validator.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
A focused review of every external integration surfaced six issues the
original audit missed. Fixed here.
HIGH
* Socket.IO had an unconditional 30-second idle disconnect on every
socket. The comment on the line acknowledged it was "for development
only, would be longer in prod" but no NODE_ENV guard existed, and the
`socket.onAny` listener only resets on inbound client events — every
dashboard connection that received only server-push events would have
been torn down every 30s in production. Removed the manual idle
timer entirely; Socket.IO's pingTimeout / pingInterval handles
dead-transport detection at the protocol level.
* SMTP transporters had no `connectionTimeout` / `greetingTimeout` /
`socketTimeout`. Nodemailer's defaults are 2 minutes for connect
and unlimited for socket — a hung SMTP server would have held a
BullMQ `email` worker concurrency slot for up to 10 min per job
(5 retries × 2 min). Set 10s/10s/30s on both the system transporter
in `src/lib/email/index.ts` and the user-account transporter in
`email-compose.service.ts`.
MEDIUM
* PostgreSQL pool had no `statement_timeout` /
`idle_in_transaction_session_timeout`. A slow query or transaction
held by a crashed handler would have eventually exhausted the
20-connection pool. 30s statement cap, 10s idle-in-tx cap, plus
`max_lifetime: 30min` to recycle connections.
* `umami_password` and `umami_api_token` were stored as plaintext in
`system_settings` (the SMTP and S3 secret paths use AES-GCM). The
reader now passes them through `readSecret()` which auto-detects
the encrypted `iv:cipher:tag` shape and decrypts, falling back to
legacy plaintext so operators can rotate without a flag-day.
* AI email-draft worker interpolated `additionalInstructions` (user-
controlled) directly into the OpenAI prompt — a hostile rep could
close the instructions block and inject prompt directives that
override the system prompt. Added `sanitizeForPrompt()` that
strips newlines + quote chars, caps at 500 chars, and the prompt
now wraps the value in a "treat as data not commands" preamble.
LOW
* Legacy `ensureBucket()` in `src/lib/minio/index.ts` was unguarded —
if any future code imported it (currently no callers), a misconfigured
prod deploy could mint a fresh empty bucket. Now matches the gate
used by the pluggable S3Backend (`MINIO_AUTO_CREATE_BUCKET=true`
required) so the legacy export and the new pluggable path agree.
Confirmed not-an-issue: BullMQ Workers create connections via
`{ url }` options object, and BullMQ sets `maxRetriesPerRequest: null`
internally for those — no fix needed. The shared `redis` singleton
that does keep `maxRetriesPerRequest: 3` is used only for direct
Redis ops (rate-limit sliding window, etc.), never for blocking
BullMQ commands, so the value is correct there.
Test status: 1175/1175 vitest, tsc clean.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Two parallel reviews of the Tier 0–6 work surfaced one CRITICAL
regression and a handful of remaining cross-tenant gaps that the
original audit didn't enumerate. All fixed here:
CRITICAL
* document-reminders.processReminderQueue — the new bulk-fetch
leftJoin to documentTemplates was scoped on `templateType` alone.
Templates of the same type exist in every port; the cartesian
explosion would have fired one Documenso reminder PER matching
template-row per cron tick (a 5-port deploy = 5 reminders to the
same signer per cycle). Added eq(documentTemplates.portId, portId)
to the join.
* All five remaining Documenso webhook handlers (RecipientSigned /
Completed / Opened / Rejected / Cancelled) accept and require an
optional portId now, with a shared resolveWebhookDocument() helper
that refuses to mutate when the lookup is ambiguous across tenants
without a resolved port. Tier 5's port-scoping was applied only to
Expired; the route now forwards the matched portId to every
handler. Tightens the WHERE clauses on subsequent UPDATEs to (id,
portId) for defense-in-depth.
HIGH
* verifyDocumensoSecret rejects when `expected` is empty —
timingSafeEqual(0-bytes, 0-bytes) was returning true, so a dev env
with a blank DOCUMENSO_WEBHOOK_SECRET would accept a request whose
X-Documenso-Secret header was also missing/empty.
listDocumensoWebhookSecrets skips the env entry when blank.
* /api/public/health — the website-intake-secret comparison was a
string `===` (not constant-time). Switched to timingSafeEqual via
Buffer.from().
MEDIUM
* server.ts SIGTERM ordering — Socket.io closes BEFORE the HTTP
drain so long-poll websockets stop holding the server open past
the compose stop_grace_period.
* /api/v1/me PATCH preferences merge — allow-list filter on the
merged JSONB so legacy rows from the old .passthrough() era stop
silently re-shipping their bloat to disk.
Migration fixes (deploy-blocking)
* 0041 referenced `port_role_overrides.permissions` (column is
`permission_overrides`) — overrides are partial JSONB and don't
need backfilling at all (deepMerge resolves edit from the base
role). Removed the override UPDATEs entirely.
* 0042 switched all FK + CHECK adds to NOT VALID + VALIDATE so the
brief table-lock phase is decoupled from the row-scan validation,
giving a cleaner abort-and-restart story if a constraint catches
dirty production data. Added a pre-cleanup UPDATE for
invoices.billing_entity_id = '' rows (backfills from clientName,
falls back to the row id) so the new non-empty CHECK passes on a
dirty table.
Test status: 1175/1175 vitest, tsc clean.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Final audit polish — closes the remaining LOW + MED items the previous
tiers didn't reach:
* Validation hardening: me.preferences uses .strict() + 8KB cap
instead of unbounded .passthrough(); files.uploadFile gains
magic-byte verification (jpeg/png/gif/webp/pdf/doc/xlsx); OCR scan
endpoint enforces 10MB cap + magic-byte check on receipt images;
port logoUrl + me.avatarUrl reject javascript:/data: schemes via
a shared httpUrl refinement.
* Permission gates: document-sends/{brochure,berth-pdf} now require
email.send (was withAuth-only); document-sends/{preview,list} on
email.view; ai/email-draft on email.send; documents/[id]/send
uses send_for_signing (was create); expenses/export/parent-company
flips from hard isSuperAdmin to expenses.export for parity;
admin/users/options gated on reminders.assign_others (was withAuth).
* Envelope hygiene: auth/set-password switches the third {message}
variant to errorResponse + {data: {email}}; ai/email-draft wraps
jobId in {data: {jobId}}.
* UI polish: reports-list.handleDownload surfaces failures via
toastError (was console-only).
* Ops/infra: pin pnpm@10.33.2 across all three Dockerfiles +
packageManager field in package.json; Dockerfile.worker re-orders
user creation BEFORE pnpm install so node_modules / .cache dirs
are worker-owned (fixes tesseract.js + sharp EACCES at first PDF
parse); add Redis-ping HEALTHCHECK to the worker container.
* Public health endpoint: returns full env+appUrl payload only when
the caller presents X-Intake-Secret, otherwise a minimal {status}
so generic uptime monitors still work but anonymous internet
doesn't get deployment fingerprints.
* Per-port Documenso webhook secret: new system_settings key
+ listDocumensoWebhookSecrets() helper. The webhook receiver
iterates every configured per-port secret with timing-safe
comparison + falls back to env, then forwards the resolved portId
into handleDocumentExpired so two ports sharing a documensoId
cannot cross-mutate.
Deferred (handled in dedicated follow-up PRs):
* Tier 5.1 — direct service tests for portal-auth / users /
email-accounts / document-sends / sales-email-config. MED, large
test-writing scope.
* The {ok: true} → {data: null} envelope migration across
alerts/expenses/admin-ocr-settings/storage routes. Mechanical but
needs coordinated client + test updates.
* CSP-nonce migration (drop unsafe-inline) — needs middleware-level
nonce generation that the Next 15 router has to thread through.
* Idempotency-Key header on Documenso createDocument. Requires
schema column on documents to persist the key; deferred so it
doesn't bundle a migration into this commit.
* The 16 better-auth user_id FKs — separate dedicated migration
with care (some columns are NOT NULL today and cascade decisions
matter).
* PermissionGate / Skeleton / EmptyState wraps across 5 admin lists
(auditor-H §§36–37) and the residential-clients filter bar.
Test status: 1175/1175 vitest, tsc clean.
Refs: docs/audit-comprehensive-2026-05-05.md MED §§28,29,30 + LOW §§32–43
+ HIGH §9 (Documenso secrets follow-up).
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Closes the highest-priority gaps from audit HIGH §19 + MED §§20–21:
* New tests/integration/documenso-webhook-route.test.ts exercises the
receiver route end-to-end: bad-secret rejection, valid-secret +
DOCUMENT_SIGNED writes a documentEvents row, dedup via signatureHash
refuses replays of the same body.
* tests/integration/documents-expired-webhook.test.ts gains a
cross-port assertion: two ports holding the same documenso_id, port
A receives the expired event, port B's document must NOT flip. Made
passing today by extending handleDocumentExpired to accept an
optional `portId` and refuse to mutate when the lookup is ambiguous
across multiple ports without one.
* tests/integration/custom-fields.test.ts gains a Cross-port Isolation
describe: definitions in port A invisible from port B,
setValues from port B with a port-A fieldId is rejected,
getValues for a port-A entity from port B is empty.
Deferred: Tier 5.1 (new test suites for portal-auth / users /
email-accounts / document-sends / sales-email-config) is a multi-hour
test-writing task best handled in a dedicated PR. Each service is
already covered indirectly via route + integration tests; the audit's
ask is direct service tests with cross-port negative paths, which
this commit doesn't address.
Test status: 1175/1175 vitest (was 1168), tsc clean.
Refs: docs/audit-comprehensive-2026-05-05.md HIGH §19 (auditor-J Issue 2)
+ MED §§20–21 (auditor-J Issues 3–4).
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Closes the audit's HIGH §10 + MED §§17–22 isolation footguns. None of
these are user-impactful TODAY — every site is preceded by a port-
scoped read or pre-validated by ctx.portId — but each is a future-
refactor accident waiting to happen, so the SQL itself now pins the
tenant boundary:
* mergeClients gains a callerPortId option; the route caller passes
ctx.portId. removeInterestBerth now requires portId and verifies
both the interest and the berth share it before deleting the
junction row. All three callers updated.
* Six service mutations now scope the WHERE to (id, portId):
form-templates update + delete, invoices.detectOverdue per-row
update, notifications.markRead, clients.deleteRelationship.
company-memberships uses an inArray sub-select against port
companies (no port_id column on the table itself), covering
updateMembership / endMembership / setPrimary.
* Port-scoped file lookups in portal.getDocumentDownloadUrl,
reports.getDownloadUrl (file presign), berth-reservations.activate
(contractFileId attach guard), and residential.getResidentialInterestById
(residentialClient join).
Test status: 1168/1168 vitest, tsc clean.
Refs: docs/audit-comprehensive-2026-05-05.md HIGH §10 + MED §§17–22
(auditor-B3 Issues 1–5,7).
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Replaces per-row fan-out with grouped queries / inArray pre-fetches
across the five dashboard + cron hotspots flagged in the audit
(MED §13 / HIGH §11–14):
* reminders.processFollowUpReminders — was 3 round trips per
enabled-and-due interest. Now: filter in JS, single clients
bulk-fetch, single reminders bulk-insert, single interests
bulk-update, one summary socket emit. 1k due interests: 6 round
trips total instead of 3000+.
* portal.getClientInvoices — was a full-table scan filtered in JS.
Now an inArray push-down on lower(billingEmail) + defensive
limit(100). After 12mo this would have been the worst portal
endpoint.
* interest-scoring.calculateBulkScores — was 6N round trips
(1 redis + 1 findFirst + 4 counts per interest). Now 4 grouped
count queries on the port's interest set + a single redis pipeline
to refresh the cache. 1k interests: ~7 round trips.
* document-reminders.processReminderQueue — was 5N round trips per
cron tick (port + template + lastReminder + pendingSigners + send
per doc). Now hoists port + per-type template map + grouped
lastReminder + bulk pendingSigners; per-row work collapses to a
Map.get and the documenso send. 500 docs: ~7 round trips.
* inquiry-notifications.sendInquiryNotifications — was sequential
createNotification + emailQueue.add per recipient inside a public
POST. Now Promise.all'd; a 20-user port stops blocking the public
inquiry POST on ~80 round trips.
Test status: 1168/1168 vitest, tsc clean.
Refs: docs/audit-comprehensive-2026-05-05.md HIGH §§11–14 (auditor-I
Issues 1–4) + MED §13 (auditor-I Issue 5).
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Two final waves of error-surface hygiene closing the audit's MED §12 +
HIGH §15 + HIGH §17 findings:
* 50 route files swept (61 sites): manual NextResponse.json({error,
status: 4xx|5xx}) early-returns replaced by typed throws +
errorResponse(err) at the catch.
- Super-admin gates (13 sites) use new requireSuperAdmin(ctx, action)
helper from src/lib/api/helpers.ts so denials hit the audit log.
- Path-param + body validation 400s become ValidationError throws.
- 404s become NotFoundError or CodedError('NOT_FOUND') for AI
feature-flag paths.
- 11 manual 5xx returns now re-throw so error_events captures the
request-id (the admin error inspector becomes usable from real
incidents).
- website-analytics 200-with-error anti-pattern flipped to 409 +
UMAMI_NOT_CONFIGURED. 502 upstream paths use UMAMI_UPSTREAM_ERROR.
- 11 sites intentionally preserved: storage/[token] anti-enumeration
token-failure paths, webhook-secret 401, "Unknown port" 400 in
public intake.
* 7 admin forms (roles, users, ports, webhooks, custom-fields,
document-templates, tags) gain a formatErrorBanner() helper from
src/lib/api/toast-error.ts that builds a multi-line "Error code / Reference ID"
banner — the rep can copy the request id when reporting a failed
save. Banners get whitespace-pre-line so newlines render.
Test status: 1168/1168 vitest, tsc clean.
Refs: docs/audit-comprehensive-2026-05-05.md MED §12 (auditor-F Issue 1)
+ HIGH §15 (auditor-F Issue 2) + HIGH §17 (auditor-H Issue 2).
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Two mechanical sweeps closing the audit's HIGH §16 + MED §11 findings:
* 38 client components / 56 toast.error sites converted to
toastError(err) so the new admin error inspector becomes usable from
user-reported issues — every failed inline-edit, save, send, archive,
upload, etc. now carries the request-id + error-code (Copy ID action).
* 26 service files / 62 bare-Error throws converted to CodedError or
the existing AppError subclasses. Adds new error codes:
DOCUMENSO_UPSTREAM_ERROR (502), DOCUMENSO_AUTH_FAILURE (502),
DOCUMENSO_TIMEOUT (504), OCR_UPSTREAM_ERROR (502),
IMAP_UPSTREAM_ERROR (502), UMAMI_UPSTREAM_ERROR (502),
UMAMI_NOT_CONFIGURED (409), and INSERT_RETURNING_EMPTY (500) for
post-insert returning-empty guards.
* Five vitest assertions updated to match the new user-facing wording
(client-merge "already been merged", expense/interest "couldn't find
that …", documenso "signing service didn't respond").
Test status: 1168/1168 vitest, tsc clean.
Refs: docs/audit-comprehensive-2026-05-05.md HIGH §16 (auditor-H Issue 1)
+ MED §11 (auditor-G Issue 1).
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Closes the second wave of HIGH-priority audit findings:
* fetchWithTimeout helper (new src/lib/fetch-with-timeout.ts) wraps
Documenso, OCR, currency, Umami, IMAP, etc. — a hung upstream can
no longer pin a worker concurrency slot indefinitely. OpenAI client
passes timeout: 30_000. ImapFlow gets socket / greeting / connection
timeouts.
* SIGTERM / SIGINT handler in src/server.ts drains in-flight HTTP,
closes Socket.io, and disconnects Redis before exit; compose
stop_grace_period bumped to 30s. Adds closeSocketServer() helper.
* env.ts gains zod-validated PORT and MULTI_NODE_DEPLOYMENT, and
filesystem.ts now reads from env (a typo can no longer silently
disable the multi-node guard).
* Per-port Documenso template + recipient IDs land in system_settings
with env fallback (PortDocumensoConfig now exposes eoiTemplateId,
clientRecipientId, developerRecipientId, approvalRecipientId).
document-templates.ts uses the per-port config and threads portId
into documensoGenerateFromTemplate().
* Migration 0042 wires the eleven HIGH-tier missing FK constraints
(documents/files/interests/reminders/berth_waiting_list/
form_submissions) plus polymorphic CHECK round 2
(yacht_ownership_history.owner_type, document_sends.document_kind),
invoices.billing_entity_id NOT EMPTY, and clients.merged_into self-FK.
Drizzle schema columns updated to .references(...) where possible
so the misleading "FK wired in relations.ts" comments are gone.
Test status: 1168/1168 vitest, tsc clean.
Refs: docs/audit-comprehensive-2026-05-05.md HIGH §§5,6,7,8,9,10 +
MED §§14,15,16,18.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Removes 12 direct minioClient.{put,get,remove}Object call sites that
bypassed the pluggable storage abstraction. Filesystem-mode deploys
(MULTI_NODE_DEPLOYMENT=false, storage_backend=filesystem) silently
broke at every site: GDPR export, invoice PDF, EOI generation, portal
download, file upload, folder create/rename/delete, signed PDF land,
maintenance cleanup, etc. Each site now resolves the active backend
and uses its put/get/delete + the new presignDownloadUrl() helper.
Folder marker objects in /files/folders/* keep the same on-the-wire
shape but route through the backend. A future refactor should move
folder bookkeeping to a DB-backed virtual-folder table (see audit
HIGH §3 follow-up note in the route file).
Sites left untouched: src/lib/services/system-monitoring.service.ts
and src/app/api/ready/route.ts use minioClient.bucketExists as an S3-
specific health probe — those are correctly mode-aware and stay.
Refs: docs/audit-comprehensive-2026-05-05.md HIGH §3 (auditor-D Issue 1)
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Closes the five highest-risk findings from
docs/audit-comprehensive-2026-05-05.md so the platform is not exposed
while the rest of the audit backlog (1 CRIT + 18 HIGH + 32 MED + 23 LOW)
is worked through:
* CVE-2025-29927 — bump next 15.1.0 → 15.2.9; nginx strips
X-Middleware-Subrequest at the edge as defense-in-depth.
* Cross-tenant role escalation — POST/PATCH/DELETE on /admin/roles now
require super-admin (was: any holder of admin.manage_users). Adds
shared `requireSuperAdmin(ctx)` helper.
* Silent-403 traps — `documents.edit` and `files.edit` keys added to
RolePermissions; seeded role values updated; migration 0041 backfills
the new keys on every existing roles+port_role_overrides JSONB. File
routes remap the dead `create` action to `upload` / `manage_folders`.
* Berth-PDF / brochure register endpoints — reject body.storageKey
unless it matches the namespace the matching presign endpoint issued
(prevents repointing a tenant's PDF at foreign-port bytes).
* Portal auth rate limits — sign-in 5/15min/(ip,email),
forgot-password 3/hr/IP, activate/reset/set-password 10/hr/IP. Adds
`enforcePublicRateLimit()` for non-`withAuth` routes.
Test status unchanged: 1168/1168 vitest, tsc clean.
Refs: docs/audit-comprehensive-2026-05-05.md (CRITICAL, HIGH §§1–4)
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
End-to-end error-handling overhaul. A user hitting any failure now sees
a plain-text message + stable error code + reference id. A super admin
can paste the id into /admin/errors/<id> for the full request shape,
sanitized body, error stack, and a heuristic likely-cause hint.
REQUEST CONTEXT (AsyncLocalStorage)
- src/lib/request-context.ts mints a per-request frame carrying
requestId + portId + userId + method + path + start timestamp.
- withAuth wraps every authenticated handler in runWithRequestContext
and accepts an upstream X-Request-Id header (validated shape) or
generates a fresh UUID. The id ALWAYS leaves on the X-Request-Id
response header, including early-return 401/403/4xx paths.
- Pino logger reads from the same context via mixin — every log
line emitted during the request automatically carries the ids
with no per-call threading.
ERROR CODE REGISTRY
- src/lib/error-codes.ts defines stable DOMAIN_REASON codes with
HTTP status + plain-text user-facing message (no jargon, written
for the rep on the phone with a customer).
- New CodedError class wraps a registered code + optional
internalMessage (admin-only — never sent to client).
- Existing AppError subclasses got plain-text default rewrites so
legacy throw sites improve immediately without migration.
- High-impact services migrated to specific codes:
expenses (RECEIPT_REQUIRED, INVOICE_LINKED), interest-berths
(CROSS_PORT_LINK_REJECTED), berth-pdf (PDF_MAGIC_BYTE / PDF_EMPTY /
PDF_TOO_LARGE / VERSION_ALREADY_CURRENT), recommender
(INTEREST_PORT_MISMATCH).
ERROR ENVELOPE
- errorResponse always sets X-Request-Id header + requestId field.
- 5xx responses include a "Quote error ID …" friendly line.
- 4xx kept clean (validation, permission, not-found don't pollute
the inspector — they're already in audit log).
PERSISTENCE (error_events table, migration 0040)
- One row per 5xx, keyed on requestId, with method/path/status/error
name+message/stack head (4KB cap)/sanitized body excerpt (1KB cap;
password/token/secret/etc keys redacted)/duration/IP/UA/metadata.
- captureErrorEvent extracts Postgres SQLSTATE/severity/cause.code
so the classifier can recognize FK / unique / NOT NULL / schema-
drift violations.
- Failure to persist is logged-not-thrown.
LIKELY-CULPRIT CLASSIFIER (src/lib/error-classifier.ts)
- 4-pass heuristic (first match wins):
1. Postgres SQLSTATE → human reason (23503 FK, 23505 unique,
42703 schema drift, 53300 connection limit, …)
2. Error class name (AbortError, TimeoutError, FetchError,
ZodError)
3. Stack-path patterns (/lib/storage/, /lib/email/, documenso,
openai|claude, /queue/workers/)
4. Free-text message keywords (econnrefused, rate limit, timeout,
unauthorized|invalid api key)
- Returns { label, hint, subsystem } for the inspector badge.
CLIENT SIDE
- apiFetch throws structured ApiError with message + code + requestId
+ details + retryAfter.
- toastError() helper renders the standard 3-line toast:
plain message / Error code: X / Reference ID: Y [Copy ID].
ADMIN INSPECTOR
- /<port>/admin/errors lists captured 5xx with status badge + path +
likely-culprit badge + truncated message + reference id. Filter by
status code; auto-refresh via TanStack Query.
- /<port>/admin/errors/<requestId> deep-dive: request shape, full
error name+message+stack, sanitized body excerpt, raw metadata,
registered-code lookup (so admin can compare to what user saw),
likely-culprit hint with subsystem tag.
- /<port>/admin/errors/codes is the in-app code reference page —
every registered code grouped by domain prefix, searchable, with
HTTP status + user message inline. Linked from inspector header
so admins can flip to it while triaging.
- Permission: admin.view_audit_log. Super admins see all ports;
regular admins port-scoped.
- system-monitoring dashboard now surfaces error_events alongside
permission_denied audit + queue failed jobs (RecentError gains
source: 'request' variant).
DOCS
- docs/error-handling.md walks through coded errors, plain-text
message guidelines, client toasting, admin inspector usage,
persistence rules, classifier internals, pruning, and the
legacy → CodedError migration path.
MIGRATION SAFETY
- Audit confirmed all 41 migrations (0000-0040) apply cleanly in
journal order against an empty DB. 0040 references ports(id)
which exists from 0000. 0035/0038 don't deadlock under sequential
psql -f. Removed redundant idx_ds_sent_by from 0038 (created in
0037).
Tests: 1168/1168 vitest passing. tsc clean.
- security-error-responses tests updated for plain-text messages
+ new optional response keys (code/requestId/message).
- berth-pdf-versions tests assert stable error codes via
toMatchObject({ code }) rather than message regex.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Per the trips/events design discussion: instead of building a full
events domain (table + CRUD UI + calendar) for the 6–12 yacht shows
a year, ship the cheap version that covers the actual asks.
Expenses — `tripLabel` free-text:
- New `expenses.trip_label` text column (migration 0039) + index for
filter / autocomplete lookup.
- Validator: createExpenseShape + listExpensesSchema +
exportExpensePdfSchema.filter all accept tripLabel.
- Service: createExpense + updateExpense persist; listExpenses filters;
new `listTripLabels(portId, search?)` returns distinct values
ordered by most-recent expenseDate so the autocomplete surfaces
recently-used labels first.
- New `GET /api/v1/expenses/trip-labels` endpoint (gated by
expenses.view) backs the autocomplete.
- Form dialog: native `<datalist>` powered by the autocomplete query
so reps don't end up with "Palm Beach 2026" / "palm-beach 2026"
fragmented across two PDF sections.
- Expense list: new "Trip" column (badge) + free-text filter.
- Detail page: trip label rendered alongside Category / Payer.
- PDF export: GroupBy gains 'trip'; filter.tripLabel narrows the
export. Untagged rows fall under "(no trip)".
- Trim/normalize on write so " Palm Beach 2026 " === "Palm Beach 2026".
Interests — event tagging via existing tag system:
- Reps can tag interests with an event tag (e.g. "Palm Beach 2026")
via the existing InlineTagEditor on the detail page; tags are
port-scoped and reusable.
- Interest list now has a TagPicker filter rendered next to the
FilterBar so reps can sort prospects by event attended ("show me
every lead from Palm Beach"). Hidden 'relation'-typed
FilterDefinition for tagIds wires URL round-trip + saved-views
capture without rendering inside the FilterBar.
- FilterBar deserializer now handles `relation` types as comma-joined
arrays on URL load.
Why a free-text trip label and not a trips table:
- 6–12 events/year doesn't justify a domain. The CRUD UI cost would
be most of the engineering, and reps already have the events on
their personal calendars.
- If usage proves demand for per-event ROI dashboards or richer
attribution, promote to a real `trips` table later. Migration
path: trip_label → tripId is a backfill+swap.
Test status: 1168/1168 vitest. tsc clean. Migration 0039 applied
in dev (also caught + fixed an unrelated audit-v3 follow-up: 0037
had `idx_br_interest` colliding with the existing
`berth_recommendations.idx_br_interest`; renamed to
`idx_brr_interest` / `idx_brr_contract_file`).
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Working through the audit-v2 deferred backlog. Each round was tested
(typecheck + 1168/1168 vitest) before moving on.
Round 1 — DB performance + AI cost visibility:
- Add missing FK indexes Postgres doesn't auto-create on
berth_reservations.{interest_id, contract_file_id},
documents.{file_id, signed_file_id}, document_events.signer_id,
document_templates.source_file_id, form_submissions.{form_template_id,
client_id}, document_sends.{brochure_id, brochure_version_id,
sent_by_user_id}. Without these, RESTRICT-checks on parent delete +
reverse-lookups walk the child tables fully. Migration 0037.
- AI worker now writes one ai_usage_ledger row per OpenAI call so admins
can audit spend per port/user/feature and future per-port budgets have
history to read from. Failure to write is logged-not-thrown so the
user-facing email draft is unaffected.
Round 2 — Boot-time + transport hardening:
- S3 backend verifies the bucket exists at startup (or auto-creates
when MINIO_AUTO_CREATE_BUCKET=true). A typo'd bucket name now
surfaces with a clear boot error instead of a vague Minio error
inside the first user-facing request.
- Documenso v1 placeFields: 3-attempt exponential-backoff retry on 5xx
+ network errors, fail-fast on 4xx. Stops one transient flake from
leaving a document with a partial field set.
- FilesystemBackend logs a structured warn-once at boot when the dev
HMAC fallback is in effect, so two processes started with different
BETTER_AUTH_SECRET values are observable (random 401s on file
downloads otherwise).
- Logger redact paths extended to cover *.headers.{authorization,
cookie}, *.config.headers.authorization, encrypted-credential blobs
(secretKeyEncrypted, smtpPassEncrypted, etc.), the Documenso
X-Documenso-Secret header, and 2-level nested forms.
Round 3 — UI feedback + permission gates:
- Storage admin migrate dialog: success toast with row count + error
toast on both dryRun and migrate mutations.
- Invoice detail Send + Record-payment buttons wrapped in
PermissionGate (invoices.send / invoices.record_payment); both
mutations now toast on success/error.
- Admin user list Edit button wrapped in PermissionGate(admin.manage_users).
- Scan-receipt page surfaces an amber warning when OCR fails so reps
know they can fill the form manually instead of staring at a stalled
spinner; the editable form now also opens on scanMutation.isError
/ uploadedFile, not only on success.
- Email threads list now renders skeleton rows during load + shared
EmptyState for the empty case (was a single "Loading…" line).
Round 4 — Service / route correctness:
- documentSends.sent_by_user_id was a free-text NOT NULL column with no
FK. Now nullable + FK to user(id) ON DELETE SET NULL so the audit row
survives a user being hard-deleted. Migration 0038 with a defensive
null-out for any orphan ids before attaching the constraint.
- Saved-views route: documented why withAuth alone is correct (the
service strictly filters by (portId, userId) — owner-only by design).
- Public-interests audit log: replaced "userId: null as unknown as
string" cast with userId: null; AuditLogParams already accepts null
for system-generated events.
- EOI in-app PDF fill: extracted setBerthRange() that, when the
AcroForm field is missing AND the context has a non-empty range
string, logs a structured warn so the deployment gap (live Documenso
template needs the field) is observable instead of silently dropping
the multi-berth range.
Test status: 1168/1168 vitest. tsc clean. Two new migrations
(0037/0038) need pnpm db:push (or migration apply) on the dev DB.
Deferred-doc updated with the remaining open items (bigger refactors).
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Five-domain audit (security, routes, DB, integrations, UI/UX) ran after
the cf37d09 merge. Critical + high-impact items landed here; deferred
medium/low items indexed in docs/audit-final-deferred.md (now organised
into a "Audit-final v2" section).
Security:
- Storage proxy tokens now bind to op (`'get'` vs `'put'`). A long-lived
download URL minted by `presignDownload` for an emailed brochure can no
longer be replayed against the proxy PUT to overwrite the original
storage object. `verifyProxyToken` requires `expectedOp` and rejects
mismatches; legacy tokens missing `op` fail-closed. Regression tests
added.
- Markdown email merge values are now markdown-escaped (`[`, `]`, `(`,
`)`, `*`, `_`, `\`, backticks, braces) before substitution into the
rep-authored body. A malicious value like `[click here](https://evil)`
stored in `client.fullName` no longer survives `escapeHtml` to render
as a real `<a href>` in the outbound email. Phishing-via-merge-field
closed; regression tests added.
- Middleware now performs an Origin/Referer check on
POST/PUT/PATCH/DELETE to `/api/v1/**`. Defense-in-depth on top of
better-auth's SameSite=Lax cookie. Webhooks/public/auth/portal routes
exempt as they don't carry the session cookie.
Routes:
- Template management routes were calling `withPermission('documents',
'manage', ...)` — but `documents` doesn't have a `manage` action. The
registry has `document_templates.manage`. Every non-superadmin was
getting 403'd on the seven template endpoints. Fixed across the
/admin/templates surface.
- Custom-fields permission resource is hardcoded to `clients` regardless
of which entity (yacht/company/etc.) the values belong to. Documented
as deferred (requires per-entity routes).
DB:
- documentSends: every parent FK (client_id, interest_id, berth_id,
brochure_id, brochure_version_id) now uses ON DELETE SET NULL so the
audit trail outlasts hard-deletes. The denormalized columns
(recipient_email, document_kind, body_markdown, from_address) were
added precisely for this. Migration 0035.
- Polymorphic discriminators on yachts.current_owner_type and
invoices.billing_entity_type now have CHECK constraints — typos like
`'clients'` vs `'client'` were silently inserting unreachable rows
before. Migration 0036.
Integrations:
- Email attachment resolution (`src/lib/email/index.ts`) was importing
MinIO directly instead of `getStorageBackend()`. Filesystem-backend
deployments would have broken every email-with-attachment send. Now
routes through the pluggable abstraction per CLAUDE.md.
- Documenso DOCUMENT_OPENED webhook filter relaxed: v2 may omit
`readStatus` or send lowercase, so an event that was the SIGNAL of an
open was being silently dropped. Now treats any recipient on a
DOCUMENT_OPENED event as opened.
UI/UX:
- Expense detail used to render `receiptFileIds` as opaque UUID badges —
reps couldn't view the receipt they uploaded. Now renders an image
thumbnail (via `/api/v1/files/[id]/preview`) plus a Download link for
PDFs. Closed the "where's my receipt?" loop in the expense flow.
- Expense detail Edit + Archive buttons now `<PermissionGate>` and the
archive mutation surfaces success/error toasts instead of silent 403s.
- Brochures admin: setDefault/archive/create mutations now have onError
toasts (only onSuccess existed before).
- Removed broken bulk-upload link in scan/page (route doesn't exist;
used a raw `<a>` triggering a full reload to a 404).
Test status: 1168/1168 vitest passing. tsc clean.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Post-merge security review caught a cross-tenant authorization bypass
in the per-berth PDF endpoints (HIGH severity, confidence 10):
GET /api/v1/berths/[id]/pdf-versions
POST /api/v1/berths/[id]/pdf-versions
POST /api/v1/berths/[id]/pdf-upload-url
POST /api/v1/berths/[id]/pdf-versions/[versionId]/rollback
POST /api/v1/berths/[id]/pdf-versions/parse-results/apply
Each handler looked up the target berth by id only — `eq(berths.id, ...)`.
withAuth resolves ctx.portId from the user-controlled X-Port-Id header
(only verifying the user has SOME role on that port), and
withPermission('berths', 'view'|'edit', ...) is a coarse capability
check, not a row-level grant. A rep with berths:edit on Port A could
supply a Port B berth UUID and:
- list + receive 15-min presigned download URLs to every PDF version
- mint an upload URL targeting `berths/<port-B-id>/uploads/...`
- POST a new version (overwriting current_pdf_version_id on foreign berth)
- rollback to any prior version on a foreign berth
- apply rep-confirmed parse-result fields onto a foreign berth's columns
Sibling routes (waiting-list etc.) already pair the id filter with
`eq(berths.portId, ctx.portId)`, so this was an omission, not design.
Fix:
- Push `portId: string` into uploadBerthPdf, listBerthPdfVersions,
rollbackToVersion, applyParseResults, reconcilePdfWithBerth.
- Each function now filters the berth lookup with
`and(eq(berths.id, ...), eq(berths.portId, portId))` and throws
NotFoundError on mismatch (no foreign-port disclosure).
- Inline the same `and(...)` filter in the pdf-upload-url handler.
- Every handler passes ctx.portId through.
Coverage:
- New `cross-port tenant guard` test exercises every entrypoint with a
foreign-port id and asserts NotFoundError.
- 1164/1164 vitest passing. Typecheck clean.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Multi-phase work bundle (24 commits, 159 files, ~127k LOC) covering plan
docs/berth-recommender-and-pdf-plan.md:
Phase 0 — NocoDB berth import + mooring normalization (A-01 → A1)
Phase 1 — /clients + /interests list-column redesign (contacts/yachts join)
Phase 2 — M:M interest_berths junction with role flags
Phase 3 — Public berths API + /api/public/health
Phase 4 — Berth recommender (SQL ranking, tier ladder, heat scoring)
Phase 5 — Multi-berth EOI bundle + range formatter
Phase 6 — Pluggable storage backend + per-berth PDF parser
Phase 7 — Sales send-outs + brochures + email-from settings
Phase 8 — CLAUDE.md conventions update
Plus a memory-efficient streaming expense PDF export (replaces a legacy
implementation that OOM'd on hundreds of receipts), receipt-less expense
flag with PDF warning annotations, receipt upload UI in the expense
form dialog, and the scan-receipt page accepting device-uploaded photos
in parallel with the OCR scan.
Four audit passes (audit-1 → audit-final, mostly Opus 4.7 reviewers in
parallel) drove progressive hardening: ~50 findings landed; the last
audit's 5 critical / 12 high items are fixed in 180912b. Medium/low
items are deferred and indexed in docs/audit-final-deferred.md.
Tests: 1163/1163 vitest passing. tsc clean. 12 new migrations applied
in dev (0023..0034), three of which (0028/0029, 0034) involve careful
backfills.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Final audit pass on feat/berth-recommender (3 parallel Opus agents)
caught 5 critical and ~12 high-severity findings. All addressed in-branch;
medium/low items deferred to docs/audit-final-deferred.md.
Critical:
- Add filesystem-backend PUT handler at /api/storage/[token] so
presigned uploads stop 405-ing in filesystem mode (every browser-driven
berth-PDF + brochure upload was broken). Same token-verify + replay
protection as GET, plus magic-byte gate when c=application/pdf.
- Forward req.signal into streamExpensePdf so an aborted 1000-receipt
export no longer keeps grinding for minutes.
- Strengthen Content-Disposition filename sanitization: \s matches CR/LF
which would let documentName forge headers; restrict to [\w. -]+ and
add filename* RFC 5987 fallback.
- Lock public berths feed behind an explicit slug allowlist instead of
?portSlug= enumeration.
- Reject cross-port interest_berths upserts (defense-in-depth on top of
the recommender SQL port filter).
High:
- Recommender: width-only feasibility now caps length via L/W ratio so a
200ft berth doesn't surface for a 30ft beam request; total_interest_count
filters out junction rows whose interest is in another port.
- Mooring normalization follow-up migration (0034) catches un-hyphenated
padded forms (A01) the original 0024 WHERE missed.
- Send-out rate limit moved AFTER validation and scoped per-(port, user)
so typos don't burn a slot and a multi-port rep can't be DoS'd by
another tenant.
- Default-brochure path now blocks an archived row from sneaking through
the partial unique index.
- NocoDB import --update-snapshot honoured under --dry-run so reps can
refresh the seed JSON without committing DB writes.
- PDF export: orderBy desc(expenseDate); apply isNull(archivedAt) when
expenseIds are passed (was bypassed); flag rate-unavailable rows with
an amber footer instead of silently treating them as 1:1; skip the
USD->EUR chain when source already matches target.
- expense-form-dialog: revokeObjectURL captures the URL in the closure
instead of revoking the still-displayed one; reset upload state on
close.
- scan/page: handleClearReceipt resets in-flight scan/upload mutations;
Save disabled while upload pending.
- updateExpense re-asserts receipt-or-acknowledgement at the merged
row so PATCH can't slip past the create-time refine.
Plus the in-progress receipt upload UI for the expense form dialog
(receipt picker + "I have no receipt" checkbox + warning banner) and
a noReceiptAcknowledged flag on ExpenseRow for edit-mode hydration.
Includes the canonical plan doc (referenced in CLAUDE.md), the handoff
prompt, and a deferred-findings index for follow-up issues.
1163/1163 vitest passing. Typecheck clean.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Replaces the legacy text-only expense PDF (was just dumping rows into a
single pdfme text field — no images, no pagination) with a proper
streaming export modelled on the legacy Nuxt client-portal but
re-architected for memory safety. The legacy implementation OOM'd on
hundreds of receipts because it:
- buffered every receipt image into memory simultaneously
- accumulated PDF chunks into an array, concat'd at end
- base64-encoded the whole PDF into a JSON response (3x peak memory)
- had no image downscaling
The new design:
- `streamExpensePdf()` (src/lib/services/expense-pdf.service.ts):
pdfkit pipes bytes directly to the HTTP response (no Buffer
accumulation). Receipts are processed serially so peak heap is one
image at a time. Sharp downscales any receipt > 500 KB or > 1500 px
to JPEG q80 — typical 8 MB phone photo collapses to ~250 KB. For a
500-receipt export, peak RSS stays under ~100 MB; legacy needed >2
GB for the same input.
- Pages: cover summary box (count, totals, currency equiv, optional
processing fee), grouped expense table (groupBy=none|payer|category|
date), one-page-per-receipt with header (establishment, amount,
date, payer, category, file name) and full-bleed image.
- Storage backend abstraction — receipts stream from
`getStorageBackend().get(storageKey)`, works on MinIO/S3/filesystem.
- Route: POST /api/v1/expenses/export/pdf streams binary
application/pdf with cache-control:no-store. Validator caps
expenseIds at 1000 to prevent runaway loops.
Receipt-less expense flow (per user request):
- Schema: 0033 migration adds `expenses.no_receipt_acknowledged`
boolean (default false).
- Validator: createExpenseSchema requires either receiptFileIds OR
noReceiptAcknowledged=true; the .refine() error message tells the
rep exactly what to do. updateExpenseSchema is partial and skips
the rule (existing rows can be edited without re-acknowledging).
- PDF: receiptless expenses get an inline red "(no receipt)" tag in
the establishment cell + a red footer warning in the summary box
showing the count and at-risk amount.
- The legacy parent-company reimbursement queue may refuse to pay
receiptless expenses, so the warning is load-bearing for ops.
Audit-3 fixes piggy-backed:
- 🔴 Tesseract OCR runtime now races a 30s timeout (CPU-bomb DoS
protection — a crafted PDF rasterizing to high-res noise could
pin the worker indefinitely).
- 🟠 brochures.service.ts:listBrochures dropped a wasted query (the
legacy single-brochure fast-path was discarding its result on the
multi-brochure branch).
- 🟠 berth-pdf.service.ts:listBerthPdfVersions now Promise.all's the
presignDownload calls instead of awaiting each in a for-loop —
20-version berths went from 20× round-trip to 1×.
- 🟡 public berths route no longer logs the full `row` object on
enum drift (was dumping price + amenity columns into ops logs).
- 🟡 dropped the dead `void sql` import from public berths route.
Tests still 1163/1163. tsc clean.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Two reviewer agents did a second-pass deep audit of the 21-commit
refactor. Eight findings; four fixed here (one was deferred with a
schema comment, three were 🟡 nice-to-haves left for follow-up).
Integration regressions (🟠 high):
- Outbound webhook `interest.berth_linked` now fires from the new
junction-add handler. Was emitting a socket-only event, leaving
external integrations silent post-refactor.
- Two new webhook events `interest.berth_unlinked` and
`interest.berth_link_updated` added to WEBHOOK_EVENTS +
INTERNAL_TO_WEBHOOK_MAP. PATCH and DELETE handlers now dispatch them
alongside the existing socket emits — lifecycle parity restored.
- BerthInterestPulse adds useRealtimeInvalidation for berth-link
events. The query key was berth-scoped while the linked-berths
dialog invalidates interest-scoped keys (no prefix match), so the
pulse went stale. Bridges via the realtime hook now.
Recommender semantic fix (🟠 medium-high):
- aggregates CTE: active_interest_count now filters on
`ib.is_specific_interest = true`, matching the public-map "Under
Offer" derivation. EOI-bundle-only links no longer demote a berth
to Tier C for other reps. Smoke test confirms previously-all-Tier-C
results now correctly classify as Tier A.
- Same CTE: `total_interest_count` uses COUNT(ib.berth_id) instead of
COUNT(*) so a berth with no junction rows reports 0 (not 1 from
the LEFT JOIN's NULL-right-side row). Prevents heat over-counting.
Data integrity (🟠):
- AcroForm tier rejects negative numerics in coerceFieldValue (was
letting through `length_ft="-50"` which would poison the
recommender feasibility filter on apply).
- FilesystemBackend.resolveHmacSecret throws in production when
storage_proxy_hmac_secret_encrypted is null. Dev still derives from
BETTER_AUTH_SECRET for ergonomics; prod must explicitly configure.
- Documented the circular FK between berths.current_pdf_version_id
and berth_pdf_versions.id. Drizzle's `.references()` can't express
the cycle so the schema column is plain text + a comment; the FK
is authoritatively maintained by migration 0030.
Tests still 1163/1163. tsc clean.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
