Plan §4.6 + §1: a render function that compresses every berth marked
is_in_eoi_bundle=true on an interest into a compact range string
("A1-A3, B5-B7"), wired into both EOI generation paths (the Documenso
template-generate call and the in-app pdf-lib AcroForm fill).
- src/lib/templates/berth-range.ts: pure formatBerthRange() with the
full edge-case set from §4.6 - empty, single, run, gap, multiple
prefixes, sort/dedup, multi-letter prefixes, non-canonical
passthrough, long ranges. Sorts by (prefix, number); dedupes; passes
non-canonical inputs through with a logger warning.
- src/lib/templates/merge-fields.ts: new {{eoi.berthRange}} token
added to VALID_MERGE_TOKENS allow-list under a fresh `eoi` scope so
unknown-token validation at template creation time still rejects
typos.
- src/lib/services/eoi-context.ts: EoiContext gains eoiBerthRange.
Resolved by joining interest_berths (is_in_eoi_bundle=true) →
berths and feeding the mooring numbers through formatBerthRange.
- src/lib/services/documenso-payload.ts: formValues now includes
"Berth Range" alongside the legacy "Berth Number". Multi-berth EOIs
surface here; single-berth EOIs duplicate the primary.
- src/lib/pdf/fill-eoi-form.ts: in-app AcroForm fill mirrors the
Documenso payload by populating "Berth Range". Falls back silently
when older PDFs don't have the field (setText is no-op-on-missing).
15 unit tests on the formatter; existing EoiContext + Documenso
payload tests updated to assert the new field. 1022 -> 1037 passing.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Plan §4.4 + §13: pure SQL recommender, no AI. Single CTE chain
(feasible -> aggregates) + JS-side tier classification, fall-through
cooldown filter, heat scoring, and fit ranking. Per-port settings via
system_settings layered over global + DEFAULT_RECOMMENDER_SETTINGS.
Tier ladder (default):
A : no interest history
B : lost-only history (still recommendable + boosted by heat)
C : active interest in early stage (open..eoi_signed)
D : active interest at deposit_10pct or beyond (hidden by default)
Heat (only for tier B):
recency weight 30 full @ <=30 days, decays to 0 @ 365 days
furthest stage weight 40 full when prior reached deposit
interest count weight 15 saturates at 5+
EOI count weight 15 saturates at 3+
Multi-port isolation enforced (§14.10 critical): the SQL filters by
port_id AND the entry-point function rejects cross-port interest
lookups with an explicit error. Fall-through policy supports
immediate_with_heat (default), cooldown, and never_auto_recommend.
15 unit tests covering tier classification, heat saturation, weight
tuning, zero-weight guard. Smoke-tested end-to-end via
scripts/dev-recommender-smoke.ts.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Adds the read-only public-website data feed promised by plan §4.5 and
§7.3. The marketing site's `getBerths()` swap is now a one-line URL
change against the existing 5-min TTL behaviour.
- src/app/api/public/berths/route.ts: GET / unauth, returns the full
port-nimara berth list as { list, pageInfo } in the verbatim NocoDB
shape ("Mooring Number", "Side Pontoon", quoted-key fields). Cache:
s-maxage=300 + stale-while-revalidate=60. portSlug query param lets
future ports opt in.
- src/app/api/public/berths/[mooringNumber]/route.ts: GET single. Up-
front regex validation (^[A-Z]+\\d+$) rejects malformed lookups with
400 + cache-control:no-store before hitting the DB. 404 + no-store
when not found.
- src/app/api/public/health/route.ts: returns { status, env, appUrl,
timestamp } so the marketing site can refuse to start when its
CRM_PUBLIC_URL points at a different deployment env (§14.8 critical
env-mismatch protection).
- src/lib/services/public-berths.ts: pure mapper with derivePublicStatus
("sold" wins; otherwise specific-interest junction OR
status='under_offer' -> "Under Offer"; else "Available").
- 11 unit tests covering numeric coercion, status derivation,
archived-berth handling, missing-map-data omission, and the
status-precedence rule that "sold" trumps the specific-interest
signal.
Smoke-tested: /api/public/berths -> 117 rows, A1 correctly shows
"Under Offer" (has interest_berths.is_specific_interest=true link),
INVALID -> 400, Z99 -> 404. Total tests: 996 -> 1007.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Surfaces the recommender inputs added in Phase 2a (interests
.desired_length_ft / desired_width_ft / desired_draft_ft) on the
two interfaces reps actually use:
- /interests list: new "Berth size desired" column rendered as a
compact "60×18×6 ft" string. Cells with no dimensions show "-";
partial dimensions render "?" for the missing axis (recommender
treats null as "no constraint").
- New/Edit Interest form: three optional length/width/draft inputs
with explanatory subhead. Empty submissions collapse to undefined
so the API doesn't see "" -> numeric coercion errors.
- createInterestSchema gains the three optional desired-dim fields
with a shared transform that coerces strings/numbers to a positive
2-decimal numeric string for the postgres `numeric` column.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
The Phase 2b refactor (commit 6e3d910) added a junction table whose
berth_id has onDelete: 'restrict'. The vitest global teardown deletes
test-port berths but never explicitly clears interest_berths first,
so any test leaking junction rows (e.g. via the new createInterest
write path) leaves berths un-deletable and ports stranded.
Adds DELETE FROM interest_berths WHERE berth_id IN (test berths) to
the WITH-chain so cascading teardown completes cleanly.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Phase 2b of the berth-recommender refactor (plan §3.4). Every caller of
the legacy `interests.berth_id` column now reads / writes through the
`interest_berths` junction via the helper service introduced in Phase 2a;
the column itself is dropped in a final migration.
Service-layer changes
- interests.service: filter `?berthId=X` becomes EXISTS-against-junction;
list enrichment uses `getPrimaryBerthsForInterests`; create/update/
linkBerth/unlinkBerth all dispatch through the junction helpers, with
createInterest's row insert + junction write sharing a single transaction.
- clients / dashboard / report-generators / search: leftJoin chains pivot
through `interest_berths` filtered by `is_primary=true`.
- eoi-context / document-templates / berth-rules-engine / portal /
record-export / queue worker: read primary via `getPrimaryBerth(...)`.
- interest-scoring: berthLinked is now derived from any junction row count.
- dedup/migration-apply + public interest route: write a primary junction
row alongside the interest insert when a berth is provided.
API contract preserved: list/detail responses still emit `berthId` and
`berthMooringNumber`, derived from the primary junction row, so frontend
consumers (interest-form, interest-detail-header) need no changes.
Schema + migration
- Drop `interestsRelations.berth` and `idx_interests_berth`.
- Replace `berthsRelations.interests` with `interestBerths`.
- Migration 0029_puzzling_romulus drops `interests.berth_id` + the index.
- Tests that previously inserted `interests.berthId` now seed a primary
junction row alongside the interest.
Verified: vitest 995 passing (1 unrelated pre-existing flake in
maintenance-cleanup.test.ts), tsc clean.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Introduces the multi-berth interest model from plan §3.1: a junction
between interests and berths with three role flags so the same berth
can be linked as the primary deal target, an EOI-bundle inclusion,
or a "just exploring" link without conflating semantics.
- 0028 schema migration creates interest_berths with the unique
partial index "≤1 primary per interest", a unique compound on
(interest_id, berth_id), and indexes for the public-map "under
offer" lookup (where is_specific_interest=true).
- Same migration adds desired_length_ft / desired_width_ft /
desired_draft_ft to interests for the recommender.
- Same migration runs the Phase 2 data migration: every interest
with a non-null berth_id gets one junction row marked
is_primary=true, is_specific_interest=true, and is_in_eoi_bundle =
(eoi_status='signed'). Pre-flight check halts on dangling FKs
(§14.3 critical case).
- New service src/lib/services/interest-berths.service.ts owns reads
+ writes of the junction. getPrimaryBerth / getPrimaryBerthsForInterests
feed list pages; upsertInterestBerth demotes the prior primary in
the same transaction so the unique index is never violated.
- interests.berth_id stays in place this commit so existing callers
keep working; Phase 2b migrates them onto the helper service and a
later migration drops the column.
53 dev rows seeded into the junction; tests still green at 996.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Wire interests.yachtId -> yachts.name into the listInterests post-fetch
enrichment so the redesigned columns (Client · Yacht · Berth · Stage ·
EOI status · Source · Last activity) render the linked yacht.
- Add yachtId/yachtName to InterestRow.
- listInterests: fourth parallel join for yachts.name, Map merged
alongside the existing client/berth/tag/notes joins.
- interest-columns: add Yacht column (with link to /yachts/[id] when
the yacht has an id); replace Category with EOI status (badge
driven by interests.eoi_status); drop default-view Tags.
The "Berth size desired" column called out in §5.2 is deferred to
Phase 2 since the underlying desired_*_ft columns don't exist yet.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Wire primary email + primary phone into the /clients list service so
the redesigned columns (Name · Email · Phone · Country · Source ·
Latest stage · Created) actually have data. Picks the row marked
is_primary=true; falls back to most-recent created_at when the flag
is unset.
- 0026 schema migration: unique partial index
idx_cc_one_primary_per_channel on (client_id, channel) WHERE
is_primary=true. Prevents the §14.2 "multiple primaries" ambiguity.
- 0027 data migration: backfill clients.nationality_iso from the
primary phone's value_country. 218 -> 36 missing on dev. Idempotent.
- listClients: add a fifth parallel query for client_contacts; build
primaryEmailMap / primaryPhoneMap in-memory from the pre-sorted
result.
- client-columns: drop Yachts/Companies/Tags from the default view
per §5.1; add Email/Phone/Country/Latest-stage columns; rename
"Nationality" -> "Country" since phone country is a proxy (§14.2).
- client-card: prefer email, fall back to phone, for the line under
the name; replaces the old `contacts.find(isPrimary)` lookup.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Regenerates the 117-row berth seed via:
pnpm tsx scripts/import-berths-from-nocodb.ts --apply --update-snapshot
The JSON ordering matches the legacy seed-data.ts contract (idx 0..4
available, 5..9 under_offer, 10..11 sold, remainder by mooring number).
Mooring numbers are now in canonical form throughout.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Idempotent NocoDB Berths -> CRM `berths` import script with full
re-run safety. Re-running picks up NocoDB additions/edits without
clobbering CRM-side overrides (compares updated_at vs last_imported_at,
1-second tolerance for sub-second clock drift). --force overrides the
edit guard.
Mitigates the §14.1 critical/high cases:
- Mooring collisions: unique (port_id, mooring_number) on the table.
- Concurrent runs: pg_advisory_xact_lock on a stable BIGINT key.
- Numeric-with-units inputs: parseDecimalWithUnit() strips trailing
ft/m/kw/v/usd/$ markers before parsing.
- Metric drift: NocoDB's metric formula columns are ignored; metric
values recomputed from imperial via 0.3048 + round-to-2-decimals to
match NocoDB's `precision: 2` columns and avoid spurious diffs.
- Map Data shape: zod-validated; failures are skipped rather than
aborting the import.
- Status enum mapping: NocoDB display strings -> CRM snake_case.
- NocoDB row deleted: reported as "orphaned in CRM"; never auto-
deleted (rep decides via admin UI in a future phase).
Pure helpers (parseDecimalWithUnit, mapStatus, parseMapData,
extractNumerics, mapRow, buildPlan) live in
src/lib/services/berth-import.ts so vitest can exercise the mapping
logic without triggering the script's top-level db connection.
40 new unit tests (956 -> 996 passing).
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Adds the 5 pricing columns surfaced by the per-berth PDFs (Phase 6b
will populate them via the OCR parser) and the last_imported_at marker
the NocoDB import script (Phase 0c) uses to detect human edits and
skip overwriting them.
- weekly_rate_high_usd / weekly_rate_low_usd
- daily_rate_high_usd / daily_rate_low_usd
- pricing_valid_until (date) - drives the "stale pricing" chip on
the berth detail page when older than today
- last_imported_at - compared against updated_at so re-running the
import preserves CRM-side overrides
tenure_type comment widens to include 'fee_simple' and 'strata_lot'
to match the per-berth PDF tenure model; the column is plain text
so no DB-level enum change is required.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Sweep CRM mooring numbers from the legacy hyphen+padded form ("A-01")
to the canonical bare form ("A1") used by NocoDB, the public website,
the per-berth PDFs, and the Documenso EOI templates. Drift was
introduced by the original load-berths-to-port-nimara.ts seed; this
gates the Phase 3 public-website cutover where /berths/A1 URLs would
404 against a CRM still storing "A-01".
- 0024 data migration: idempotent regexp_replace + post-update sanity
check that surfaces any non-conforming rows for manual triage.
- Invert normalizeLegacyMooring in dedup/migration-apply: it now
canonicalizes ("D-32" -> "D32") instead of legacy-izing.
- Update tiptap-to-pdfme example tokens, EOI fixture moorings, and
smoke-test seed moorings.
- Refresh seed-data/berths.json to canonical form; drop the now-
redundant legacyMooringNumber field.
- Delete scripts/load-berths-to-port-nimara.ts (superseded in 0c).
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Replaces every em-dash and en-dash with regular ASCII hyphens
across comments, JSX strings, and dev-facing logs. Mostly cosmetic
but stops the inconsistent mix that crept in over the last few
months (some files used em-dashes in comments, others didn't,
some used both).
Bundles two small dashboard-layout tweaks that touch a couple of
already-modified files:
- (dashboard)/layout.tsx main padding goes from p-6 to pt-3 px-6
pb-6 so page content sits closer to the topbar.
- Sidebar now receives the ports list it needs for the footer
port switcher.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
migration-apply: residential client + interest inserts now wrap in
db.transaction so a partial failure can't leave an orphan client
row without its interest (or vice versa).
migration-transform: buildPlannedDocument returns null when there
are no signers so the apply pass doesn't try to send a Documenso
envelope without recipients. mapDocumentStatus gets an explicit
"Awaiting Further Details" branch that no longer auto-promotes via
stale sign-time fields. parseFlexibleDate handles ISO and DD-MM-YYYY
inputs uniformly.
backfill-legacy-lead-source: chunk UPDATE WHERE clause now
isNull(source) on top of the inArray match, so a re-run can't
overwrite a more accurate source written between batches.
Adds 235 lines of vitest coverage on migration-transform.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Adds /invoices/upload-receipts as the dedicated explainer for the
mobile scanner PWA: install instructions for iOS/Android, direct
deep-link button, and a walkthrough of the scan -> verify -> save
flow. Sidebar entry replaces the old "Scan receipt" tab so the
desktop side picks up the install steps before sending users to
the mobile-only surface.
Scanner layout moves PWA manifest + apple-* meta tags from inline
JSX into Next.js's metadata/viewport exports so the App Router
doesn't try to render a second <head>, fixing a hydration error
that surfaced as two console warnings on the scan page.
Scanner shell gains a centered Port Nimara logo header so the
standalone PWA looks branded when launched from the home screen
without the dashboard chrome.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
DateRangePicker grows a "Custom range" mode (From/To inputs capped
at today, mutually-bounded so From <= To). dashboard-shell threads
the range through to /api/v1/analytics, which validates calendar
dates via ISO round-trip and enforces a 365-day cap as a backstop
against the occupancy timeline N+1.
KpiCards now gates its query on currentPortId so the early
unhydrated-store fetch can't cache a zeroed/error response and
display "-" until staleTime expires.
MyRemindersRail drops xl:h-full so the rail no longer stretches
past its grid row and overlaps ActivityFeed below.
useRealtimeInvalidation switches to partial-prefix queryKeys so a
realtime mutation invalidates every cached range bucket at once
instead of just the one currently visible.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Replaces the topbar's separate AlertBell + NotificationBell with a
single Inbox popover that tabs between alerts and notifications.
NotificationBell keeps a popover-gate so it doesn't fire its list
fetch when Inbox is mounted alongside it.
Extracts the user dropdown into <UserMenu> and moves the port
switcher + role label + theme toggle into the sidebar footer so
the topbar can reclaim space for breadcrumbs and command search.
Adds dedicated Insights / Receipts nav sections in the sidebar
(scaffolds the website-analytics + upload-receipts entry points).
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Adds /[portSlug]/website-analytics dashboard page (pageviews, top
pages, top referrers) and a per-port admin config UI for the
Umami URL / website-ID / API token. Settings live in system_settings
keyed per-port so a future second port has its own Umami account.
Adds a website glance tile to the main dashboard, a server-side
test-credentials endpoint, and a stable cache key for the active-
visitor poll so React Query doesn't fragment the cache per range.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Adds website_submissions table + shared-secret POST endpoint so the
marketing site can dual-write inquiries alongside its NocoDB write.
Race-safe via INSERT ... ON CONFLICT, idempotent on submission_id,
refuses every request when WEBSITE_INTAKE_SECRET is unset. Also
repairs pre-existing 0020/0021/0022 prevId collision (renumbered +
journal re-sorted) so db:generate works again. 11 unit tests.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Two issues surfaced when applying the migration to dev:
1. Mooring number format mismatch
The legacy NocoDB Interests table writes bare mooring strings
("D32", "B16", "A4"), but the new berths table (mirroring the
NocoDB Berths snapshot) uses zero-padded dashed form ("D-32",
"B-16", "A-04"). The interest→berth lookup missed every reference.
migration-apply.ts now tries the literal value first, then falls
back to a normalized form via `normalizeLegacyMooring(raw)`:
"D32" -> "D-32"
"A4" -> "A-04"
"E18" -> "E-18"
Multi-mooring strings ("A3, D30") are left as-is so they surface in
the warnings list for human review rather than silently picking one.
2. port-nimara only had the 12 hand-rolled seed berths, not the 117-
berth NocoDB snapshot
The mobile-foundation seed only places those 12 in port-nimara; the
117-berth snapshot was added later but only seeded into Marina
Azzurra (the secondary test port). Migrated interests reference
moorings well beyond A-01..D-03, so most lookups failed.
New scripts/load-berths-to-port-nimara.ts: idempotently loads any
missing snapshot berths into port-nimara without disturbing the
existing 12 (skips moorings that already exist). Run once;
subsequent runs no-op.
Result of full migration run on dev:
237 clients inserted (out of 245 total — 8 from prior seed)
406 contacts, 52 addresses, 38 yachts, 252 interests
27 interest→berth links resolved (only 13 source rows had a Berth
field set in NocoDB to begin with — most legacy interests are early
inquiries with no berth assignment)
1 unresolved warning: source=277 has multi-mooring "A3, D30"
Verified in UI:
/port-nimara/clients shows real names (John-michael Seelye, Reza
Amjad, Etiennette Clamouze, …)
/port-nimara/clients/<id> renders contacts (gmail.com addresses,
E.164 phones), tab counts (Interests N, Yachts N), pipeline summary
Dashboard: 245 clients, 266 active interests, $46.5M pipeline value
Pipeline funnel chart now shows real distribution (180 Open, 45
EOI Signed, dropoff through stages)
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
A pre-import audit caught three places where outbound comms could escape
even with EMAIL_REDIRECT_TO set. Plugged each, added unit tests so the
behavior can't silently regress, and shipped a live smoke script the
operator can run before any production data import.
Leak 1: email-compose.service.ts (per-account user composer)
Built its own nodemailer transporter and called sendMail() directly,
bypassing the centralized sendEmail()'s redirect. Now mirrors the same
redirect: when EMAIL_REDIRECT_TO is set, "to" is rewritten, "cc" is
dropped, and the subject is prefixed with "[redirected from <orig>]".
Leak 2: documenso-client.sendDocument()
Tells Documenso to actually email the document. Recipient emails were
rerouted at create-time (in pass-3) but a document created BEFORE the
redirect was turned on could still trigger a real-client email. Now
short-circuited when the redirect is set — returns the existing doc
shape so downstream code doesn't see an unexpected null.
Leak 3: documenso-client.sendReminder()
Same shape as sendDocument: emails a stored recipient address that may
predate the redirect. Now short-circuits with a warn-level log.
Tests (tests/unit/comms-safety.test.ts):
- createDocument rewrites recipients
- generateDocumentFromTemplate rewrites both v1.13 formValues.*Email
keys AND v2.x recipients[] arrays
- sendDocument is short-circuited (no /send call)
- sendReminder is short-circuited (no /remind call)
- createDocument passes through unchanged when redirect unset
- sendEmail rewrites to + subject for single recipient
- sendEmail handles array of recipients (joined into subject prefix)
- sendEmail passes through unchanged when redirect unset
- Webhook worker reads process.env.EMAIL_REDIRECT_TO at dispatch time
(no module-level caching that could miss a runtime flip)
Live smoke (scripts/smoke-test-redirect.ts):
Monkey-patches nodemailer.createTransport, calls the real sendEmail()
with a fake real-client address, verifies the captured outbound has
the right "to" + subject. Run: `pnpm tsx scripts/smoke-test-redirect.ts`.
Exits non-zero if the redirect failed for any reason — drop-in for a
pre-deploy check.
Verification:
pnpm exec tsc --noEmit — 0 errors
pnpm exec vitest run — 936/936 (was 926, +10 new safety tests)
pnpm tsx scripts/smoke-test-redirect.ts — PASS
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Completes the migration script's apply phase, which was stubbed at
the P3 ship to defer until after the runtime surfaces (P2) and the
comms safety net were in place. Both prerequisites just landed on
main, so this unblocks the actual data import.
src/lib/dedup/migration-apply.ts (new):
Idempotent apply driver. Walks the MigrationPlan, inserting clients,
contacts, addresses, yacht stubs, and interests, threading every
insert through the migration_source_links ledger so re-runs against
the same data are safe. Per-entity transactions (not one giant
transaction) so partial-failure resumption is just "run again."
Per-entity behavior:
- clients: idempotent on (source_system, source_id, target_type=client)
across the entire dedup cluster — if any source row already maps
to a client, reuse that record.
- contacts: bulk insert, primary email + primary phone independent.
- addresses: bulk insert, port_id required (schema enforces it),
first address marked primary when multiple.
- yachts: minimal stub when the legacy interest had a yachtName,
currentOwnerType=client + currentOwnerId=migrated client. Linked
via migration_source_links target_type=yacht.
- interests: looks up berthId via mooring number, yachtId via the
stub above. Carries Documenso ID forward when present.
surnameToken from PlannedClient is dropped on insert (it's a dedup
blocking-index artifact; runtime dedup re-derives from fullName).
scripts/migrate-from-nocodb.ts:
- Removes the "not yet implemented" guard for --apply.
- Adds EMAIL_REDIRECT_TO precondition gate: --apply errors out unless
the env var is set, OR --unsafe-skip-redirect-check is also passed
(production cutover only). Refers to docs/operations/outbound-comms-safety.md.
- Re-fetches NocoDB at apply time (rather than reading a saved report
dir) so the data is always fresh. Re-running is safe via the
idempotency ledger.
- Resolves target port via --port-slug (or first port if omitted).
- Generates a UUID applyId tagged on every link, which pairs with a
future --rollback flag.
- Apply summary prints inserted/skipped counts per entity type plus
the first 20 warnings.
Verification: 0 tsc errors, 926/926 vitest passing, lint clean.
The actual end-to-end run requires NOCODB_URL + NOCODB_TOKEN in .env
which aren't configured in this checkout; that's the operator's next
step.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Both the residential-clients and residential-interests pages rendered
plain HTML <table>s with 5–6 columns directly. At 390px viewport the
header columns clipped at the right edge — "Sour..." for the clients
page, no header for the interests page either.
Adds a parallel mobile card list:
- <table> stays inside `hidden lg:block` (unchanged at lg+)
- new card list inside `lg:hidden` mirrors the row data:
- Clients: name + status pill on top, then email · phone ·
residence · source as a wrap-friendly meta row.
- Interests: stage label as headline, updated-at on the right,
preferences (line-clamp-2) and notes (line-clamp-1) below,
source small at the bottom.
- Each card is a Link to the detail page (matching the row click
target on desktop).
- Empty + loading states render as a centered card on mobile.
This is the same `hidden lg:block` / `lg:hidden` pattern used for the
main /clients and /interests pages. Doesn't refactor to the full
DataView primitive (would mean rebuilding the residential data layer
on TanStack Table) — keeps the change tightly scoped to the visible
output.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Closes a gap exposed by the comms safety audit: the existing
EMAIL_REDIRECT_TO env var only redirected outbound SMTP via the
sendEmail() bottleneck. Two channels still leaked when set:
1. Documenso e-signature recipients — Documenso's own server emails
them on our behalf, so SMTP redirect doesn't help. We were sending
real client emails to the Documenso REST API, which would then
deliver to the real client.
2. Outbound webhooks — fire from the BullMQ worker to user-configured
URLs. SSRF guard blocks internal hosts but doesn't pause production
endpoints.
Documenso (src/lib/services/documenso-client.ts):
- createDocument: rewrite every recipient.email to EMAIL_REDIRECT_TO
and prefix the recipient.name with the original email so the doc
is traceable.
- generateDocumentFromTemplate: same treatment for both v1.13
formValues.*Email keys and v2.x recipients[]. The redirect happens
BEFORE the API call, so even Documenso's own retry logic can't
reach the original recipient.
- Both paths log when they redirect so it's visible in dev.
Webhooks (src/lib/queue/workers/webhooks.ts):
- When EMAIL_REDIRECT_TO is set, short-circuit the dispatch and write
a `dead_letter` row with reason "Skipped: EMAIL_REDIRECT_TO is set,
outbound comms paused." so the attempt is still visible in the
deliveries listing.
Doc:
docs/operations/outbound-comms-safety.md catalogs every outbound
comms channel (email, Documenso, webhooks, WhatsApp/phone deep-links,
SMS-not-implemented) and explains how each one respects the env flag.
Includes a verification checklist to run before any production data
import + cutover steps for going live.
Single env var EMAIL_REDIRECT_TO now reliably pauses ALL automated
outbound comms. Unset for production.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Inquiry Settings + Business Rules cards used a flex-row layout that
crushed the label column into a narrow vertical stripe at 390px ("Inquiry
/ Contact / Email" wrapping one word per line) while the input took the
right side.
Stack label + helper text above the input on phone widths; restore the
side-by-side row from sm up. Same pattern as the other detail-edit rows
that were fixed in pass-2/pass-3.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
One-shot script that walks every phone / whatsapp contact with `value`
set but `value_e164` null and runs the raw value through libphonenumber-js
to produce the canonical E.164 form. Matches the existing dedup
phone-parser shape (script-safe wrapper that loads metadata as raw JSON
to dodge the Node 25 + tsx interop bug).
Two output buckets:
- parsed cleanly: e164 + country both resolved (33/36 in dev).
- parsed e164 only: e164 came back but country didn't (3/36 — the
UK +44 7700 900xxx fictional/reserved range that libphonenumber
refuses to assign a country to but still returns a canonical e164
for). Still safe to write — the e164 form is the canonical one.
Run dry-first, --apply to write:
pnpm tsx scripts/backfill-phone-e164.ts
pnpm tsx scripts/backfill-phone-e164.ts --apply
Applied to dev DB this session: 36 rows backfilled, 0 still missing.
Will need to be re-run after any future seed reload that introduces
unparsed phones.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Five small fixes from the third audit pass on previously-unchecked surfaces:
Yacht detail header (mobile):
- Stack the action cluster (Edit / Transfer / Archive) below the title
block on phone widths. Previously the three buttons crowded the right
side enough to truncate the status pill to "A..." and force the owner
name to wrap to two lines. Same fix that landed for berth / client /
company headers.
Company detail header (mobile):
- Same mobile stacking fix; legal-name + Tax-ID metadata no longer
wraps awkwardly.
Company detail Incorporation Date (all viewports):
- Strip the time portion of the ISO timestamp before passing to the
inline editor. Previously rendered the raw "2019-03-14T00:00:00.000Z"
Postgres-serialized form. Now reads "2019-03-14" and round-trips
through the YYYY-MM-DD inline editor cleanly.
Reminders list filter row:
- Allow flex-wrap on the My/All tabs + status filter + priority filter
cluster. At 390px, the priority filter dropdown was being pushed off
the right edge of the screen.
Client detail tab counts:
- Add interestCount + noteCount to getClientById response, surface as
badges on the Interests + Notes tabs. Brings them into parity with
Yachts/Companies/Reservations/Addresses which already showed counts;
Files + Activity are still stubs and don't get a count yet.
Verification: 0 tsc errors, 926/926 vitest passing, lint clean.
Out of scope (deferred):
- Residential clients / interests pages still render plain HTML tables
on phone widths (header columns clip at the right edge). Needs the
DataView card-on-mobile treatment that the main /clients and
/interests pages already have. Substantial separate work.
- Phone contacts in the legacy seed have value set but valueE164 NULL,
so InlinePhoneField shows "—" even though metadata is technically
populated. Fix is a one-time backfill via libphonenumber-js, not a
UI change.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Three small but high-leverage fixes from the second audit pass on main:
Admin index (src/app/(dashboard)/[portSlug]/admin/page.tsx):
- Grouped 21 sections into 7 categories: Access, Configuration, Content,
Data Quality, Operations, Tenancy, Integrations. Each group has a
one-line description so first-time admins can orient themselves
without reading every card.
- Added the missing Duplicates entry (links to /admin/duplicates from
the dedup-migration work) under Data Quality.
More sheet (mobile bottom-drawer nav):
- "Email" -> "Inbox". The page that opens is an email-inbox surface
(Inbox + Accounts tabs), not a generic email composer. The previous
label was ambiguous.
Interest detail header (Won / Lost outcome buttons):
- Added title="Mark as won" / "Close as lost" so the icon-only buttons
on mobile have a tooltip on long-press / desktop hover.
- Tightened mobile padding (px-2 vs px-2.5) so the full-text desktop
labels still fit on sm+ without re-introducing a regression where a
visible mobile "Won"/"Lost" inline label crowded the right cluster
enough to push Email/Call/WhatsApp action chips into a vertical
stack.
Verification: 0 tsc errors, 926/926 vitest passing, lint clean.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
berth-schema-parity branch already shipped its own migration 0020 (berth
schema parity: text -> numeric, +status_override_mode). Dedup's two
migrations need to land on top of that, not collide.
Renames:
0020_unusual_azazel.sql -> 0021_unusual_azazel.sql
0021_magenta_madame_hydra.sql -> 0022_magenta_madame_hydra.sql
meta/0020_snapshot.json -> meta/0021_snapshot.json
meta/0021_snapshot.json -> meta/0022_snapshot.json
_journal.json idx + tag fields updated to match.
Snapshot CONTENTS remain dedup-branch state (no berths-numeric awareness).
A `pnpm drizzle-kit generate` after main merges the berth changes will
produce a consistent forward path; until then the snapshots are slightly
out-of-sync with the post-merge live schema, which is harmless because
the dev DB applies migrations forward, not from snapshots.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Three categories added to .gitignore:
Tool caches / runtime state:
.claude/ — Claude Code session state + lock files
(scheduled_tasks.lock untracked here)
.serena/ — Serena MCP project cache
ruvector.db — RuVector AgentDB binary
Scratch screenshots:
/*.jpg — debug captures dropped at repo root
/.audit-screenshots/ — UX audit run output (regenerable)
Sister project:
/website/ — separate Nuxt marketing site, kept on disk
for reference but not tracked in this repo
The single tracked file in .claude/ (scheduled_tasks.lock) is removed
from the index here; future dev sessions won't bring it back.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Mounts a dev-only client component that syncs the react-grab debug
toolbar's pinned edge / collapsed state across viewport changes (so
the toolbar doesn't drift off-screen when resizing or rotating).
Render is gated by NODE_ENV === 'development' in src/app/layout.tsx;
production builds tree-shake the import out via process.env replacement.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Old order : Dashboard / Clients / Yachts / Berths / More
New order : Dashboard / Berths / Clients / Documents / More
Reasoning (also captured as in-file comments above each tab list):
- Yachts is asset-record traffic — rarely browsed standalone, almost
always reached from inside an interest or client. Pulled out of the
bottom row, kept available in the More sheet.
- Documents (signature tracking / EOI queue) earns a slot at the
bottom because reps chase signers as a daily activity.
- Interests is intentionally NOT in the bottom row: having both
Clients and Interests as peer tabs created a Clients-vs-Interests
confusion for sales reps. The new per-client Interests tab + the
bottom-sheet drawer (see ClientInterestsTab) cover the day-to-day
deal review without needing a dedicated bottom-nav peer.
- Clients moves to the center: it's the primary mental anchor for
"find this person", with everything else (yachts, companies,
interests) reached as a tab on the client detail page.
More-sheet reorder mirrors the new priority: Interests / Yachts /
Companies first (most-likely overflow targets), then financial
(Invoices, Expenses), then Email / Alerts / Reports / Reminders /
Settings / Admin. Documents removed from the More sheet (now in
the bottom row).
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
The Alerts rail and Reminders rail were using h-full unconditionally, which
worked at xl: where the dashboard grid pairs them with a sibling chart
column, but produced weirdly stretched empty cards in the single-column
mobile stack (no fixed-height context to fill).
alert-rail / my-reminders-rail: h-full -> xl:h-full
KPI tiles + skeleton rendered the same desktop padding (p-5) and font sizes
on phone, leaving the value cramped against a wide white frame. Tighter
mobile defaults that scale up at sm:
KPITile p-3 sm:p-5, label text-[10px] sm:text-xs,
value mt-1 text-lg sm:mt-2 sm:text-2xl, value truncates
KpiTileSkeleton: matching p-3 sm:p-5 + smaller skeleton bars on mobile
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
InlinePhoneField now lays the country picker + number on top, with Save +
Cancel buttons on a second line — the previous single-line cluster was
cramped at every viewport size and broke entirely below ~480px.
A new onEditingChange callback notifies the parent when the field enters
edit mode, so contact rows can react. ContactsEditor uses it to "dilate"
the row visually: lift out of the muted baseline with a soft primary
ring + slightly brighter surface + bumped padding. Single visual signal
replaces the need for any "now editing" label, and the dilation also
hides the noisy chip cluster (label / star / trash) that would otherwise
fight the editor for space.
Mobile improvements applied at the same time:
- Each row stacks value editor on top, action cluster below at <sm
- Action cluster ("Add tag" + Make-primary star + trash) uses
justify-end on the new row so it doesn't collide with the picker
- Trash icon stays opacity-0/group-hover on desktop but is always
visible on touch (no hover state on touch) — sm:opacity-0 +
sm:group-hover:opacity-100 instead of the prior unconditional fade
- NewContactForm wraps onto multiple lines below sm (basis-full on
the value field) so the channel picker, value, label, and buttons
each get usable width
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
When a user clicks an inline-edit affordance for country / timezone /
subdivision, the field flipped to its combobox trigger but the popover
didn't open — they had to click again. And if they dismissed the popover
without picking, the field stayed in edit mode showing a "Select country…"
trigger they couldn't get out of.
Combobox primitives (country / timezone / subdivision) now accept:
- defaultOpen — open on first render
- onOpenChange — fired on every open/close transition
InlineCountryField / InlineTimezoneField / and the country + subdivision
fields inside addresses-editor pass defaultOpen=true and use onOpenChange
to auto-exit edit mode when the popover closes without a selection. A
pickedRef gate prevents the close-handler from racing the commit() exit
when the user does pick a value.
Bonus: addresses-editor now renders a flag emoji next to the country name
in the read-only state (regional-indicator pair from the ISO code).
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Promotes interests from a stub tab to a first-class surface on the client
detail page, and surfaces pipeline activity in two more places:
UI:
- New ClientInterestsTab (475 lines) — table of every active interest
for the client with stage-stepper visualization, lead category, source,
last-activity timestamp, and a drawer-on-tap row preview.
- New OverviewTab pipeline-summary panel above the existing 2-column
layout, rendering ClientPipelineSummary (already on this branch) in
its panel variant. Reps see the live pipeline at a glance without
leaving Overview.
- Removes "Preferred Language" inline field from the Overview tab and
the create form — unused, and the field added noise without driving
any downstream behavior.
- Tab order: Overview / Interests / Yachts / Companies / ... (Interests
moves up from the back of the list, where it was a stub anyway).
Data:
- listClients now returns interestCount + latestInterest{stage, mooring}
per row, joined from interests + berths in two parallel queries.
ClientRow type updated to surface them; Client list views can now
render "3 interests · last on D-02 (EOI Signed)" without a per-row
fetch.
- Contact rows in client detail now expose valueE164 + valueCountry to
the UI (already returned by the API; just wasn't typed through the
detail-page contract).
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Replaces the inline "Source · email · phone" text strip with three primary
action chips and a smaller meta line:
Mail / Call / WhatsApp action buttons surface the most-used outbound
contacts on a single tap. WhatsApp deep-link strips the leading + from
the canonical E.164 number (or falls back to digit-only of the value).
Meta line now reads "Country · Added MMM d, yyyy" using nationalityIso
resolved through getCountryName(); date-fns formats createdAt.
Portal Invite + GDPR Export buttons remain available but only render
on sm+; on mobile they're reachable through the More sheet.
Archive / Restore is now a small icon button in the top-right corner
rather than a labeled button competing with the primary action chips.
Destructive intent stays out of the main action flow; hover swaps to
destructive color for archive (and stays neutral for restore).
The previous source/preferred-contact-method/preferred-language/timezone
fields no longer render in the header — they live on the Overview tab via
the inline editor pattern (see client-tabs.tsx).
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Followup to 886119c (refactor(sales): consolidate pipeline stages) — the
runtime enum was renamed but a few test fixtures and PDF report templates
still referenced the legacy names, leaving them broken at the type level
(36 tsc errors before this fix).
Renames in this commit:
visited -> in_communication (alerts test) / removed (PDF reports)
signed_eoi_nda -> eoi_signed
contract -> contract_signed (interests test) / contract_sent (factory)
Affected files: pipeline-report, revenue-report, makeCreateInterestInput
factory, alerts-engine, pipeline-transitions, interest-scoring.
Verification: tsc clean, 858/858 vitest passing.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Five mobile-UX issues caught in the 2026-05-03 audit, fixed in one pass:
1. SpecRow on berth detail clipped at right edge on phone widths.
"Length 49.21 ft / 15 r" (the "m" cut off). Mobile-first stack:
label on top, value full-width below; flex row only from sm up.
2. ResponsiveTabs collapsed to a Select on phone widths, which read like
a generic dropdown and obscured the existence of peer tabs. Replaced
with a horizontally-scrollable strip that auto-scrolls the active
trigger into view (so the user sees neighbors and gets a discovery
cue that more exists beyond the edge). Removes the phone-only Select
and unifies the tab UI across viewport sizes.
3. Documents page tab strip ("All / EOI queue / Awaiting them / ...")
overflowed the 390px viewport because the wrapper was a fixed flex
row. Same horizontal-scroll fix as (2); inherits because Documents
uses ResponsiveTabs.
4. Berth detail header: "Change Status" + "Edit" buttons crowded the
area subtitle on mobile, causing "North Pier" to wrap to two lines
("North" / "Pier"). Stacked vertically on phone widths; from sm up
the buttons sit beside the title block as before.
5. Empty contact rows on client detail rendered a stale "Add tag · star"
metadata strip even when the contact value was unset, which cluttered
the row and offered no useful action. The metadata block now only
shows when contact.value is non-empty; the trash icon stays visible
so users can clean up the empty placeholder.
Verification:
- pnpm exec vitest run: 858/858 passing
- pnpm exec tsc --noEmit: same 36 errors as baseline (all pre-existing
on feat/mobile-foundation, none introduced)
- lint clean
Defers:
- Mobile More sheet last-row alignment / "Email" label specificity
- Admin index grouping (Access / System / Configuration / Content)
- Interest detail header icon labels (trophy/X discoverability)
- Pipeline funnel x-axis label abbreviations
- Reminders rail width allocation on dashboard
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Berth detail (src/components/berths/berth-tabs.tsx):
- Numeric display polish, exposed by the new NocoDB-sourced seed:
- Power capacity now renders with kW unit (e.g. "330 kW")
- Voltage now renders with V unit (e.g. "480 V")
- All metric/imperial values rounded to <= 2 decimals
(was: "62.999112 m" -> now: "62.99 m")
- Nominal Boat Size shows full ft + m pair (was: ft only)
Seed ports (src/lib/db/seed.ts):
- Drop Marina Azzurra and Harbor Royale; install now seeds only:
- Port Nimara (the real install)
- Port Amador (secondary, for multi-tenant isolation tests / Panama
scaffolding)
- Existing dev DBs are not touched; this only affects fresh `pnpm db:seed`
runs. Users wanting to migrate should drop existing rows in the obsolete
ports manually before re-seeding.
Verification:
- lint clean, tsc unchanged from baseline (36 pre-existing errors), 858/858
vitest passing.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
The old seed only had 12 berths with made-up area names ("North Pier",
"Central Basin", etc.) and placeholder dimensions. Devs now get the real
117 berths exported from the legacy NocoDB Berths table — every editable
column populated with real production values.
What's in the snapshot (src/lib/db/seed-data/berths.json):
- 117 berths total (61 available / 45 under_offer / 11 sold)
- Areas A through E (matches NocoDB single-select)
- All numeric fields filled: length / width / draft (ft + m), water depth,
nominal boat size, power capacity (kW), voltage (V)
- All NocoDB single-selects filled where present: side pontoon,
mooring type, cleat/bollard type+capacity, access
- Bow facing, status_override_mode, berth_approved carried forward as-is
- Status normalized to lowercase snake_case ("Under Offer" -> "under_offer")
- Mooring numbers reformatted A1 -> A-01 to keep the existing "Letter-NN"
convention used elsewhere in the codebase
Pre-sorted to preserve seed semantics:
idx 0..4 -> 5 available (small) -- "open" / "details_sent" interests
idx 5..9 -> 5 under_offer (medium) -- "eoi_signed" / "deposit" / "contract"
idx 10..11 -> 2 sold (large) -- "completed" interests
This means existing interest/reservation seeds that index berthRows[0..11]
keep their semantic alignment without code changes.
End-to-end verified by clearing Marina Azzurra and re-seeding:
Port "Marina Azzurra" -- 117 berths, 8 clients, 3 companies, 12 yachts,
15 interests, 8 reservations
Future devs running `pnpm db:seed` on a fresh DB will now get realistic
berth data automatically.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Aligns the berths schema with the 117 production rows in NocoDB and exposes
every field for editing via the BerthForm sheet.
Schema (migration 0020):
- power_capacity / voltage / nominal_boat_size / nominal_boat_size_m: text -> numeric
(NocoDB stores plain numbers; text was wrong shape and broke filter/sort)
- ADD status_override_mode text (1/117 legacy rows have a value; carried
forward for parity but not yet wired into the UI)
- USING NULLIF(TRIM(...), '')::numeric so legacy whitespace and empty
strings convert cleanly
Validator + service:
- updateBerthSchema / createBerthSchema use z.coerce.number() for the
four numeric fields
- berths.service stringifies numeric values for Drizzle's numeric type
Form (src/components/berths/berth-form.tsx):
- adds: nominal boat size (ft/m), water depth (ft/m) + "is minimum" flag,
side pontoon, cleat type/capacity, bollard type/capacity, bow facing
- converts to typed selects (with NocoDB option lists in src/lib/constants):
area, side pontoon, mooring type, cleat type/capacity, bollard type/capacity,
access
- power capacity / voltage become numeric inputs (with kW / V hints)
Permissions (seed.ts + dev DB):
- sales_manager and sales_agent: berths.edit false -> true
("sales will sometimes have to update these and I cannot be the only one")
- super_admin / director already had it; viewer stays read-only
- dev DB updated in-place via UPDATE roles ... jsonb_set
Verification:
- pnpm exec vitest run: 858/858 passing
- pnpm exec tsc --noEmit: same 36 errors as baseline (all pre-existing
on feat/mobile-foundation, none introduced)
- lint clean
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Adds the live dedup pipeline on top of the P1 library + P3 migration
script. The new `client/interest` model now actively prevents duplicate
client records at creation time and gives admins a queue to triage
the borderline pairs the at-create check missed.
Three layers, per design §7:
Layer 1 — At-create suggestion
==============================
`GET /api/v1/clients/match-candidates`
Accepts free-text email / phone / name from the in-flight client
form, normalizes them via the dedup library, and returns scored
matches against the port's live client pool. Filters out
low-confidence noise (the background scoring queue picks those up
separately). Strict port scoping; never leaks across tenants.
`<DedupSuggestionPanel>` (`src/components/clients/dedup-suggestion-panel.tsx`)
Debounced React Query hook. Renders nothing for short inputs or
no useful match. On a high-confidence match it interrupts visually
with an amber-tinted card and a "Use this client" primary button.
Medium confidence falls back to a softer "possible match — check
before creating" treatment.
`<ClientForm>`
Renders the panel above the form (create path only — skipped on
edit). New `onUseExistingClient` callback fires when the user
picks the existing client; the form closes and the parent decides
what to do (typically: navigate to that client's detail page or
open the create-interest dialog pre-filled).
Layer 2 — Merge service
=======================
`mergeClients` (`src/lib/services/client-merge.service.ts`)
The atomic merge primitive that everything else calls. Single
transaction. Per §6 of the design:
- Locks both rows (FOR UPDATE) so concurrent merges of the same
loser fail with a clear error rather than racing.
- Snapshots the full loser state (contacts / addresses / notes /
tags / interest+reservation IDs / relationship rows) into the
`client_merge_log.merge_details` JSONB column for the eventual
undo flow.
- Reattaches every loser-side row to the winner: interests,
reservations, contacts (skipping duplicates by `(channel, value)`),
addresses, notes, tags (deduped), relationships.
- Optional `fieldChoices` — per-scalar overrides letting the user
keep the loser's value for fullName / nationality / preferences /
timezone / source.
- Marks the loser archived with `mergedIntoClientId` set (a redirect
pointer for stragglers; never hard-deleted within the undo window).
- Resolves any matching `client_merge_candidates` row to status='merged'.
- Writes audit log entry.
Schema additions:
- `clients.merged_into_client_id` (nullable text, indexed) — the
redirect pointer set on archive.
Tests: 6 cases against a real DB — happy path moves rows + writes log;
self-merge / cross-port / already-merged refused; duplicate-contact
deduped on reattach; fieldChoices copies loser values to winner.
Layer 3 — Admin review queue
============================
`GET /api/v1/admin/duplicates`
Pending merge candidates (status='pending') for the current port,
with both client summaries hydrated for side-by-side rendering.
Skips pairs where one side is already archived/merged.
`POST /api/v1/admin/duplicates/[id]/merge`
Confirms a candidate. Body picks the winner; the other side
becomes the loser. Calls into `mergeClients` — the only path that
writes `client_merge_log`.
`POST /api/v1/admin/duplicates/[id]/dismiss`
Marks the candidate dismissed. Future scoring runs skip the same
pair until a score change recreates the row.
`<DuplicatesReviewQueue>` (`/admin/duplicates`)
Side-by-side card UI for each pending pair. Click a card to pick
the winner; the other side is automatically the loser. Toolbar:
"Merge into selected" + "Dismiss". No per-field merge editor in
this PR — that's a future polish; the simple "pick the better row"
flow handles ~80% of cases.
Test coverage
=============
11 new integration tests (76 added in this branch total):
- 6 mergeClients (atomicity, refusal cases, contact dedup,
fieldChoices)
- 5 match-candidates API (shape, port scoping, confidence tiers,
Pattern F false-positive guard)
Full vitest: 926/926 passing (was 858 before the dedup branch).
Lint: clean. tsc: clean for new files (only pre-existing errors in
unrelated `tests/integration/` files remain, same as before this PR).
Out of scope, deferred
======================
- Background scoring cron that populates `client_merge_candidates`
(the queue is empty until this lands; manual seeding works for
now via the at-create flow).
- Side-by-side per-field merge editor with checkboxes (the simple
"pick the winner" UX shipped here covers ~80% of real cases).
- Admin settings UI for tuning the dedup thresholds. Defaults from
the design (90 / 50) are baked in for now.
- `unmergeClients` (the snapshot is captured in client_merge_log;
the undo endpoint just hasn't been wired yet).
These are all natural follow-up PRs that don't block shipping the
runtime UX.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
