Merge feat/dedup-migration: client dedup library + NocoDB migration script + admin queue

# Conflicts:
#	.gitignore
#	src/lib/db/migrations/meta/_journal.json

src/app/(dashboard)/[portSlug]/admin/duplicates/page.tsx

@@ -0,0 +1,5 @@
+import { DuplicatesReviewQueue } from '@/components/admin/duplicates/duplicates-review-queue';
+
+export default function DuplicatesAdminPage() {
+  return <DuplicatesReviewQueue />;
+}

src/app/api/v1/admin/duplicates/[id]/dismiss/route.ts

@@ -0,0 +1,4 @@
+import { withAuth, withPermission } from '@/lib/api/helpers';
+import { dismissHandler } from '../../handlers';
+
+export const POST = withAuth(withPermission('clients', 'edit', dismissHandler));

src/app/api/v1/admin/duplicates/[id]/merge/route.ts

@@ -0,0 +1,4 @@
+import { withAuth, withPermission } from '@/lib/api/helpers';
+import { confirmMergeHandler } from '../../handlers';
+
+export const POST = withAuth(withPermission('clients', 'edit', confirmMergeHandler));

src/app/api/v1/admin/duplicates/handlers.ts

@@ -0,0 +1,160 @@
+import { NextResponse } from 'next/server';
+import { and, eq, inArray } from 'drizzle-orm';
+
+import type { AuthContext } from '@/lib/api/helpers';
+import { db } from '@/lib/db';
+import { clients, clientMergeCandidates } from '@/lib/db/schema/clients';
+import { errorResponse, NotFoundError } from '@/lib/errors';
+import {
+  listPendingMergeCandidates,
+  mergeClients,
+  type MergeFieldChoices,
+} from '@/lib/services/client-merge.service';
+
+/**
+ * GET /api/v1/admin/duplicates
+ *
+ * Pending merge candidates for the current port, sorted by score.
+ * Each row hydrates its two client summaries so the review-queue UI
+ * can render side-by-side cards without an N+1 fetch.
+ */
+export async function listHandler(_req: Request, ctx: AuthContext): Promise<NextResponse> {
+  try {
+    const pairs = await listPendingMergeCandidates(ctx.portId);
+    if (pairs.length === 0) return NextResponse.json({ data: [] });
+
+    const ids = Array.from(new Set(pairs.flatMap((p) => [p.clientAId, p.clientBId])));
+    const clientRows = await db
+      .select({
+        id: clients.id,
+        fullName: clients.fullName,
+        archivedAt: clients.archivedAt,
+        mergedIntoClientId: clients.mergedIntoClientId,
+        createdAt: clients.createdAt,
+      })
+      .from(clients)
+      .where(inArray(clients.id, ids));
+    const clientById = new Map(clientRows.map((c) => [c.id, c]));
+
+    const data = pairs
+      .map((p) => {
+        const a = clientById.get(p.clientAId);
+        const b = clientById.get(p.clientBId);
+        if (!a || !b) return null; // FK orphan — shouldn't happen, but be defensive
+        // Skip pairs where one side has already been merged or archived.
+        if (a.mergedIntoClientId || b.mergedIntoClientId) return null;
+        return {
+          id: p.id,
+          score: p.score,
+          reasons: p.reasons,
+          createdAt: p.createdAt,
+          clientA: { id: a.id, fullName: a.fullName, createdAt: a.createdAt },
+          clientB: { id: b.id, fullName: b.fullName, createdAt: b.createdAt },
+        };
+      })
+      .filter((row): row is NonNullable<typeof row> => row !== null);
+
+    return NextResponse.json({ data });
+  } catch (error) {
+    return errorResponse(error);
+  }
+}
+
+/**
+ * POST /api/v1/admin/duplicates/[id]/merge
+ *
+ * Body: { winnerId: string, fieldChoices?: MergeFieldChoices }
+ *
+ * Confirms a merge candidate. The winner is the one the user picked
+ * to keep; the other side becomes the loser. Calls into the merge
+ * service which is the only path that touches client_merge_log.
+ */
+export async function confirmMergeHandler(
+  req: Request,
+  ctx: AuthContext,
+  params: { id?: string },
+): Promise<NextResponse> {
+  try {
+    const id = params.id ?? '';
+    const body = (await req.json().catch(() => ({}))) as {
+      winnerId?: string;
+      fieldChoices?: MergeFieldChoices;
+    };
+    if (!body.winnerId) {
+      return NextResponse.json({ error: 'winnerId required' }, { status: 400 });
+    }
+
+    const [candidate] = await db
+      .select()
+      .from(clientMergeCandidates)
+      .where(
+        and(
+          eq(clientMergeCandidates.id, id),
+          eq(clientMergeCandidates.portId, ctx.portId),
+          eq(clientMergeCandidates.status, 'pending'),
+        ),
+      );
+    if (!candidate) throw new NotFoundError('Merge candidate');
+
+    const loserId =
+      body.winnerId === candidate.clientAId
+        ? candidate.clientBId
+        : body.winnerId === candidate.clientBId
+          ? candidate.clientAId
+          : null;
+    if (!loserId) {
+      return NextResponse.json(
+        { error: 'winnerId must match one of the candidate clients' },
+        { status: 400 },
+      );
+    }
+
+    const result = await mergeClients({
+      winnerId: body.winnerId,
+      loserId,
+      mergedBy: ctx.userId,
+      fieldChoices: body.fieldChoices,
+    });
+
+    return NextResponse.json({ data: result });
+  } catch (error) {
+    return errorResponse(error);
+  }
+}
+
+/**
+ * POST /api/v1/admin/duplicates/[id]/dismiss
+ *
+ * Mark a merge candidate as dismissed. The background scoring job
+ * skips dismissed pairs on subsequent runs (a future score increase
+ * can re-create them).
+ */
+export async function dismissHandler(
+  _req: Request,
+  ctx: AuthContext,
+  params: { id?: string },
+): Promise<NextResponse> {
+  try {
+    const id = params.id ?? '';
+    const result = await db
+      .update(clientMergeCandidates)
+      .set({
+        status: 'dismissed',
+        resolvedAt: new Date(),
+        resolvedBy: ctx.userId,
+      })
+      .where(
+        and(
+          eq(clientMergeCandidates.id, id),
+          eq(clientMergeCandidates.portId, ctx.portId),
+          eq(clientMergeCandidates.status, 'pending'),
+        ),
+      )
+      .returning({ id: clientMergeCandidates.id });
+
+    if (result.length === 0) throw new NotFoundError('Merge candidate');
+    return NextResponse.json({ data: { id: result[0]!.id, status: 'dismissed' } });
+  } catch (error) {
+    return errorResponse(error);
+  }
+}

src/app/api/v1/admin/duplicates/route.ts

@@ -0,0 +1,4 @@
+import { withAuth, withPermission } from '@/lib/api/helpers';
+import { listHandler } from './handlers';
+
+export const GET = withAuth(withPermission('clients', 'view', listHandler));

src/app/api/v1/clients/match-candidates/handlers.ts

@@ -0,0 +1,160 @@
+import { NextResponse } from 'next/server';
+import { and, eq, inArray } from 'drizzle-orm';
+
+import type { AuthContext } from '@/lib/api/helpers';
+import { db } from '@/lib/db';
+import { clients, clientContacts } from '@/lib/db/schema/clients';
+import { interests } from '@/lib/db/schema/interests';
+import { errorResponse } from '@/lib/errors';
+import { findClientMatches, type MatchCandidate } from '@/lib/dedup/find-matches';
+import { normalizeEmail, normalizeName, normalizePhone } from '@/lib/dedup/normalize';
+import type { CountryCode } from '@/lib/i18n/countries';
+
+/**
+ * GET /api/v1/clients/match-candidates
+ *
+ * Query parameters (any combination):
+ *   email      Free-text email; gets normalized server-side.
+ *   phone      Free-text phone; gets normalized to E.164 server-side.
+ *   name       Free-text full name; used for surname-token blocking.
+ *   country    Optional ISO country hint (default: AI for Port Nimara).
+ *
+ * Returns the top candidates that scored above the soft-warn threshold,
+ * each with a small client summary the form's suggestion card can
+ * render. Confidence tiers and rules are applied server-side from the
+ * port's `system_settings` (when wired) or sensible defaults otherwise.
+ *
+ * Used by `useDedupSuggestion` in the new-client form. Debounced on
+ * the client; this endpoint must be cheap (single port pool fetch +
+ * an in-memory dedup pass).
+ */
+export async function getMatchCandidatesHandler(
+  req: Request,
+  ctx: AuthContext,
+): Promise<NextResponse> {
+  try {
+    const url = new URL(req.url);
+    const rawEmail = url.searchParams.get('email');
+    const rawPhone = url.searchParams.get('phone');
+    const rawName = url.searchParams.get('name');
+    const country = (url.searchParams.get('country') ?? 'AI') as CountryCode;
+
+    const email = rawEmail ? normalizeEmail(rawEmail) : null;
+    const phoneResult = rawPhone ? normalizePhone(rawPhone, country) : null;
+    const nameResult = rawName ? normalizeName(rawName) : null;
+
+    // If the caller didn't give us anything useful to match on, return empty
+    // — short-circuit rather than scan every client for nothing.
+    if (!email && !phoneResult?.e164 && !nameResult?.surnameToken) {
+      return NextResponse.json({ data: [] });
+    }
+
+    // Build the input candidate.
+    const input: MatchCandidate = {
+      id: '__incoming__',
+      fullName: nameResult?.display ?? null,
+      surnameToken: nameResult?.surnameToken ?? null,
+      emails: email ? [email] : [],
+      phonesE164: phoneResult?.e164 ? [phoneResult.e164] : [],
+      countryIso: country,
+    };
+
+    // Fetch the live pool for this port. We keep this O(N) over clients
+    // since the dedup library does its own blocking; for ports with
+    // thousands of clients we can later restrict by surname-token /
+    // contact lookups, but for current scale the simple full-pool fetch
+    // is fine.
+    const liveClients = await db
+      .select({
+        id: clients.id,
+        fullName: clients.fullName,
+        nationalityIso: clients.nationalityIso,
+      })
+      .from(clients)
+      .where(and(eq(clients.portId, ctx.portId)));
+
+    if (liveClients.length === 0) {
+      return NextResponse.json({ data: [] });
+    }
+
+    const clientIds = liveClients.map((c) => c.id);
+    const contactRows = await db
+      .select({
+        clientId: clientContacts.clientId,
+        channel: clientContacts.channel,
+        value: clientContacts.value,
+        valueE164: clientContacts.valueE164,
+      })
+      .from(clientContacts)
+      .where(inArray(clientContacts.clientId, clientIds));
+
+    // Group contacts by client for the candidate map.
+    const emailsByClient = new Map<string, string[]>();
+    const phonesByClient = new Map<string, string[]>();
+    for (const c of contactRows) {
+      if (c.channel === 'email') {
+        const arr = emailsByClient.get(c.clientId) ?? [];
+        arr.push(c.value.toLowerCase());
+        emailsByClient.set(c.clientId, arr);
+      } else if (c.channel === 'phone' || c.channel === 'whatsapp') {
+        if (c.valueE164) {
+          const arr = phonesByClient.get(c.clientId) ?? [];
+          arr.push(c.valueE164);
+          phonesByClient.set(c.clientId, arr);
+        }
+      }
+    }
+
+    const pool: MatchCandidate[] = liveClients.map((c) => {
+      const named = normalizeName(c.fullName);
+      return {
+        id: c.id,
+        fullName: c.fullName,
+        surnameToken: named.surnameToken ?? null,
+        emails: emailsByClient.get(c.id) ?? [],
+        phonesE164: phonesByClient.get(c.id) ?? [],
+        countryIso: (c.nationalityIso as CountryCode | null) ?? null,
+      };
+    });
+
+    const matches = findClientMatches(input, pool, {
+      highScore: 90,
+      mediumScore: 50,
+    });
+
+    // Only return medium+ — low-confidence noise isn't useful at the
+    // create-form layer (background scoring queue picks those up).
+    const useful = matches.filter((m) => m.confidence !== 'low');
+    if (useful.length === 0) {
+      return NextResponse.json({ data: [] });
+    }
+
+    // Pull a quick summary for each surfaced candidate so the suggestion
+    // card has enough to render ("Marcus Laurent · 2 interests · last
+    // contact 9d ago").
+    const summarizedIds = useful.map((m) => m.candidate.id);
+    const interestCounts = await db
+      .select({ clientId: interests.clientId })
+      .from(interests)
+      .where(inArray(interests.clientId, summarizedIds));
+    const interestsByClient = new Map<string, number>();
+    for (const r of interestCounts) {
+      interestsByClient.set(r.clientId, (interestsByClient.get(r.clientId) ?? 0) + 1);
+    }
+
+    const data = useful.map((m) => ({
+      clientId: m.candidate.id,
+      fullName: m.candidate.fullName,
+      score: m.score,
+      confidence: m.confidence,
+      reasons: m.reasons,
+      interestCount: interestsByClient.get(m.candidate.id) ?? 0,
+      emails: m.candidate.emails,
+      phonesE164: m.candidate.phonesE164,
+    }));
+
+    return NextResponse.json({ data });
+  } catch (error) {
+    return errorResponse(error);
+  }
+}

src/app/api/v1/clients/match-candidates/route.ts

@@ -0,0 +1,4 @@
+import { withAuth, withPermission } from '@/lib/api/helpers';
+import { getMatchCandidatesHandler } from './handlers';
+
+export const GET = withAuth(withPermission('clients', 'view', getMatchCandidatesHandler));

src/components/admin/duplicates/duplicates-review-queue.tsx

@@ -0,0 +1,215 @@
+'use client';
+
+import { useState } from 'react';
+import { useMutation, useQuery, useQueryClient } from '@tanstack/react-query';
+import { ArrowRight, GitMerge, X } from 'lucide-react';
+import { toast } from 'sonner';
+
+import { Button } from '@/components/ui/button';
+import { PageHeader } from '@/components/shared/page-header';
+import { EmptyState } from '@/components/shared/empty-state';
+import { Skeleton } from '@/components/ui/skeleton';
+import { apiFetch } from '@/lib/api/client';
+import { cn } from '@/lib/utils';
+
+interface CandidatePair {
+  id: string;
+  score: number;
+  reasons: string[];
+  createdAt: string;
+  clientA: { id: string; fullName: string; createdAt: string };
+  clientB: { id: string; fullName: string; createdAt: string };
+}
+
+/**
+ * Admin review queue for the dedup background scoring job.
+ *
+ * Lists every pending merge candidate (pairs where score >=
+ * `dedup_review_queue_threshold`). For each pair the admin can:
+ *   - Pick a winner via the side-by-side card → confirms a merge
+ *   - Dismiss → removes from the queue (a future score increase
+ *     re-creates the pair on the next scoring run)
+ *
+ * Only minimal merge UI here: the user picks which side is the winner
+ * (no per-field choice), and the loser archives. A richer side-by-side
+ * field-merge dialog is a future enhancement.
+ */
+export function DuplicatesReviewQueue() {
+  const queryClient = useQueryClient();
+
+  const { data, isLoading } = useQuery<{ data: CandidatePair[] }>({
+    queryKey: ['admin', 'duplicates'],
+    queryFn: () => apiFetch<{ data: CandidatePair[] }>('/api/v1/admin/duplicates'),
+  });
+
+  const pairs = data?.data ?? [];
+
+  return (
+    <div className="space-y-4">
+      <PageHeader
+        title="Duplicate clients"
+        description={
+          pairs.length === 0
+            ? 'No pending pairs to review.'
+            : `${pairs.length} pair${pairs.length === 1 ? '' : 's'} flagged for review.`
+        }
+      />
+
+      {isLoading ? (
+        <div className="space-y-3">
+          {[0, 1, 2].map((i) => (
+            <Skeleton key={i} className="h-32 w-full" />
+          ))}
+        </div>
+      ) : pairs.length === 0 ? (
+        <EmptyState
+          title="All clear"
+          description="The background scoring job hasn't surfaced any potential duplicates yet."
+        />
+      ) : (
+        <ul className="space-y-3">
+          {pairs.map((pair) => (
+            <li key={pair.id}>
+              <CandidateRow pair={pair} queryClient={queryClient} />
+            </li>
+          ))}
+        </ul>
+      )}
+    </div>
+  );
+}
+
+function CandidateRow({
+  pair,
+  queryClient,
+}: {
+  pair: CandidatePair;
+  queryClient: ReturnType<typeof useQueryClient>;
+}) {
+  const [busy, setBusy] = useState<'merge' | 'dismiss' | null>(null);
+  const [winnerId, setWinnerId] = useState<string>(pair.clientA.id);
+
+  const mergeMutation = useMutation({
+    mutationFn: () =>
+      apiFetch(`/api/v1/admin/duplicates/${pair.id}/merge`, {
+        method: 'POST',
+        body: { winnerId },
+      }),
+    onSuccess: () => {
+      const loserName =
+        winnerId === pair.clientA.id ? pair.clientB.fullName : pair.clientA.fullName;
+      const winnerName =
+        winnerId === pair.clientA.id ? pair.clientA.fullName : pair.clientB.fullName;
+      toast.success(`Merged "${loserName}" into "${winnerName}"`);
+      queryClient.invalidateQueries({ queryKey: ['admin', 'duplicates'] });
+      queryClient.invalidateQueries({ queryKey: ['clients'] });
+    },
+    onError: (err) => toast.error(err instanceof Error ? err.message : 'Merge failed'),
+    onSettled: () => setBusy(null),
+  });
+
+  const dismissMutation = useMutation({
+    mutationFn: () => apiFetch(`/api/v1/admin/duplicates/${pair.id}/dismiss`, { method: 'POST' }),
+    onSuccess: () => {
+      toast.message('Dismissed');
+      queryClient.invalidateQueries({ queryKey: ['admin', 'duplicates'] });
+    },
+    onError: (err) => toast.error(err instanceof Error ? err.message : 'Dismiss failed'),
+    onSettled: () => setBusy(null),
+  });
+
+  return (
+    <div className="rounded-lg border bg-card p-4">
+      <div className="mb-3 flex items-baseline justify-between gap-3">
+        <div>
+          <span className="rounded-full bg-muted px-2 py-0.5 text-[10px] font-medium uppercase tracking-wide text-muted-foreground">
+            score {pair.score}
+          </span>{' '}
+          <span className="text-xs text-muted-foreground">{pair.reasons.join(' · ')}</span>
+        </div>
+        <span className="text-xs text-muted-foreground">
+          flagged {new Date(pair.createdAt).toLocaleDateString()}
+        </span>
+      </div>
+
+      <div className="grid gap-3 sm:grid-cols-[1fr_auto_1fr]">
+        <ClientCard
+          client={pair.clientA}
+          isSelected={winnerId === pair.clientA.id}
+          onSelect={() => setWinnerId(pair.clientA.id)}
+        />
+        <div className="flex items-center justify-center text-muted-foreground">
+          <ArrowRight className="size-4" aria-hidden />
+        </div>
+        <ClientCard
+          client={pair.clientB}
+          isSelected={winnerId === pair.clientB.id}
+          onSelect={() => setWinnerId(pair.clientB.id)}
+        />
+      </div>
+
+      <div className="mt-3 flex flex-wrap items-center gap-2">
+        <Button
+          size="sm"
+          onClick={() => {
+            setBusy('merge');
+            mergeMutation.mutate();
+          }}
+          disabled={busy !== null}
+        >
+          <GitMerge className="mr-1 size-3.5" aria-hidden />
+          Merge into selected
+        </Button>
+        <Button
+          size="sm"
+          variant="ghost"
+          onClick={() => {
+            setBusy('dismiss');
+            dismissMutation.mutate();
+          }}
+          disabled={busy !== null}
+        >
+          <X className="mr-1 size-3.5" aria-hidden />
+          Dismiss
+        </Button>
+        <p className="text-xs text-muted-foreground">
+          The unselected card becomes the loser; its interests + contacts move to the selected
+          client and the original is archived.
+        </p>
+      </div>
+    </div>
+  );
+}
+
+function ClientCard({
+  client,
+  isSelected,
+  onSelect,
+}: {
+  client: CandidatePair['clientA'];
+  isSelected: boolean;
+  onSelect: () => void;
+}) {
+  return (
+    <button
+      type="button"
+      onClick={onSelect}
+      className={cn(
+        'rounded-md border p-3 text-left transition-colors',
+        isSelected
+          ? 'border-primary bg-primary/5 ring-1 ring-primary/30'
+          : 'border-border hover:bg-muted/40',
+      )}
+    >
+      <p className="text-sm font-medium">{client.fullName}</p>
+      <p className="mt-0.5 text-[11px] text-muted-foreground">
+        Created {new Date(client.createdAt).toLocaleDateString()}
+      </p>
+      {isSelected ? (
+        <span className="mt-1 inline-block rounded-full bg-primary/10 px-1.5 py-0.5 text-[10px] font-semibold text-primary">
+          KEEP
+        </span>
+      ) : null}
+    </button>
+  );
+}

src/components/clients/client-form.tsx

@@ -23,6 +23,7 @@ import { TagPicker } from '@/components/shared/tag-picker';
 import { CountryCombobox } from '@/components/shared/country-combobox';
 import { TimezoneCombobox } from '@/components/shared/timezone-combobox';
 import { PhoneInput } from '@/components/shared/phone-input';
+import { DedupSuggestionPanel } from '@/components/clients/dedup-suggestion-panel';
 import { apiFetch } from '@/lib/api/client';
 import { createClientSchema, type CreateClientInput } from '@/lib/validators/clients';
 import type { CountryCode } from '@/lib/i18n/countries';
@@ -30,6 +31,12 @@ import type { CountryCode } from '@/lib/i18n/countries';
 interface ClientFormProps {
   open: boolean;
   onOpenChange: (open: boolean) => void;
+  /** Optional callback fired when the dedup suggestion panel reports
+   *  the user picked an existing client. The form closes; parent is
+   *  responsible for navigating to the existing client's detail page
+   *  or opening the create-interest dialog pre-filled with that
+   *  clientId. Skipped in edit mode. */
+  onUseExistingClient?: (clientId: string) => void;
   /** If provided, form is in edit mode */
   client?: {
     id: string;
@@ -53,7 +60,7 @@ interface ClientFormProps {
   };
 }
 
-export function ClientForm({ open, onOpenChange, client }: ClientFormProps) {
+export function ClientForm({ open, onOpenChange, client, onUseExistingClient }: ClientFormProps) {
   const queryClient = useQueryClient();
   const isEdit = !!client;
 
@@ -143,6 +150,26 @@ export function ClientForm({ open, onOpenChange, client }: ClientFormProps) {
         </SheetHeader>
 
         <form onSubmit={handleSubmit((data) => mutation.mutate(data))} className="space-y-6 py-6">
+          {/* Dedup suggestion — only on the create path. Watches the
+              live form values for email / phone / name and surfaces
+              an existing client when one matches. The user can
+              attach the new interest to that client instead of
+              creating a duplicate. */}
+          {!isEdit ? (
+            <DedupSuggestionPanel
+              email={watch('contacts')?.find((c) => c?.channel === 'email')?.value ?? null}
+              phone={
+                watch('contacts')?.find((c) => c?.channel === 'phone' || c?.channel === 'whatsapp')
+                  ?.valueE164 ?? null
+              }
+              name={watch('fullName') ?? null}
+              onUseExisting={(match) => {
+                onUseExistingClient?.(match.clientId);
+                onOpenChange(false);
+              }}
+            />
+          ) : null}
+
           {/* Basic Info */}
           <div className="space-y-4">
             <h3 className="text-sm font-medium text-muted-foreground uppercase tracking-wide">

src/components/clients/dedup-suggestion-panel.tsx

@@ -0,0 +1,183 @@
+'use client';
+
+import { useEffect, useState } from 'react';
+import { useQuery } from '@tanstack/react-query';
+import { AlertCircle, ArrowRight, Briefcase, X } from 'lucide-react';
+
+import { Button } from '@/components/ui/button';
+import { apiFetch } from '@/lib/api/client';
+import { cn } from '@/lib/utils';
+
+interface MatchData {
+  clientId: string;
+  fullName: string;
+  score: number;
+  confidence: 'high' | 'medium' | 'low';
+  reasons: string[];
+  interestCount: number;
+  emails: string[];
+  phonesE164: string[];
+}
+
+interface DedupSuggestionPanelProps {
+  /** Free-text inputs from the in-flight new-client form. The panel
+   *  debounces them and queries /api/v1/clients/match-candidates. */
+  email?: string | null;
+  phone?: string | null;
+  name?: string | null;
+  /** Caller wants to attach the new interest to an existing client
+   *  rather than creating a new one. The form switches to
+   *  interest-only mode and pre-fills the client. */
+  onUseExisting: (match: MatchData) => void;
+  /** User explicitly said "create new anyway." Hide the panel until
+   *  they change input again. */
+  onDismiss?: () => void;
+}
+
+/**
+ * Surfaces existing clients that match the form's in-flight inputs.
+ *
+ * Renders nothing while inputs are short / no useful match found.
+ * On a high-confidence match, the panel interrupts visually with a
+ * solid border and a primary "Use this client" button.
+ *
+ * Wired into the new-client form. Skipped in edit mode.
+ */
+export function DedupSuggestionPanel({
+  email,
+  phone,
+  name,
+  onUseExisting,
+  onDismiss,
+}: DedupSuggestionPanelProps) {
+  const [dismissed, setDismissed] = useState(false);
+
+  // Debounce inputs by 300ms so we don't fire on every keystroke. Keep
+  // the latest debounced values in component state.
+  const [debounced, setDebounced] = useState({
+    email: email ?? '',
+    phone: phone ?? '',
+    name: name ?? '',
+  });
+
+  useEffect(() => {
+    const t = setTimeout(() => {
+      setDebounced({ email: email ?? '', phone: phone ?? '', name: name ?? '' });
+      // Clear the dismissed flag when inputs change — the user typed
+      // something new, so the prior dismissal no longer applies.
+      setDismissed(false);
+    }, 300);
+    return () => clearTimeout(t);
+  }, [email, phone, name]);
+
+  const hasSomething =
+    debounced.email.length > 3 || debounced.phone.length > 3 || debounced.name.length > 2;
+
+  const { data, isFetching } = useQuery<{ data: MatchData[] }>({
+    queryKey: ['dedup-match-candidates', debounced],
+    queryFn: () => {
+      const params = new URLSearchParams();
+      if (debounced.email) params.set('email', debounced.email);
+      if (debounced.phone) params.set('phone', debounced.phone);
+      if (debounced.name) params.set('name', debounced.name);
+      return apiFetch<{ data: MatchData[] }>(`/api/v1/clients/match-candidates?${params}`);
+    },
+    enabled: hasSomething && !dismissed,
+    // Same query is fine to cache for a minute — moves are slow at this layer.
+    staleTime: 60_000,
+  });
+
+  if (dismissed) return null;
+  if (!hasSomething) return null;
+  if (isFetching && !data) return null;
+  const matches = data?.data ?? [];
+  if (matches.length === 0) return null;
+
+  const top = matches[0]!;
+  const isHigh = top.confidence === 'high';
+
+  return (
+    <div
+      className={cn(
+        'rounded-lg border p-3 mb-3 transition-colors',
+        isHigh
+          ? 'border-amber-300 bg-amber-50/60 dark:bg-amber-950/30'
+          : 'border-border bg-muted/40',
+      )}
+      data-testid="dedup-suggestion"
+    >
+      <div className="flex items-start gap-3">
+        <div className="mt-0.5">
+          <AlertCircle
+            className={cn(
+              'size-5',
+              isHigh ? 'text-amber-700 dark:text-amber-400' : 'text-muted-foreground',
+            )}
+            aria-hidden
+          />
+        </div>
+        <div className="min-w-0 flex-1">
+          <p className="text-sm font-semibold leading-tight">
+            {isHigh
+              ? 'This looks like an existing client'
+              : 'Possible match — check before creating'}
+          </p>
+          <div className="mt-2 rounded-md border bg-background/80 p-2.5">
+            <div className="flex items-center gap-2">
+              <p className="truncate text-sm font-medium">{top.fullName}</p>
+              <span
+                className={cn(
+                  'shrink-0 rounded-full px-1.5 py-0.5 text-[10px] font-medium uppercase tracking-wide',
+                  isHigh
+                    ? 'bg-amber-200 text-amber-900 dark:bg-amber-800 dark:text-amber-100'
+                    : 'bg-muted text-muted-foreground',
+                )}
+              >
+                {top.confidence}
+              </span>
+            </div>
+            <div className="mt-0.5 flex flex-wrap items-center gap-x-3 gap-y-0.5 text-xs text-muted-foreground">
+              {top.emails[0] ? <span className="truncate">{top.emails[0]}</span> : null}
+              {top.phonesE164[0] ? <span>{top.phonesE164[0]}</span> : null}
+              <span className="inline-flex items-center gap-1">
+                <Briefcase className="size-3" aria-hidden />
+                {top.interestCount} {top.interestCount === 1 ? 'interest' : 'interests'}
+              </span>
+            </div>
+            <p className="mt-1.5 text-[11px] text-muted-foreground">{top.reasons.join(' · ')}</p>
+          </div>
+          <div className="mt-3 flex flex-wrap items-center gap-2">
+            <Button
+              type="button"
+              size="sm"
+              onClick={() => onUseExisting(top)}
+              data-testid="dedup-use-existing"
+            >
+              Use this client
+              <ArrowRight className="ml-1 size-3.5" aria-hidden />
+            </Button>
+            <Button
+              type="button"
+              size="sm"
+              variant="ghost"
+              onClick={() => {
+                setDismissed(true);
+                onDismiss?.();
+              }}
+              data-testid="dedup-dismiss"
+            >
+              <X className="mr-1 size-3.5" aria-hidden />
+              Create new anyway
+            </Button>
+            {matches.length > 1 ? (
+              <span className="text-xs text-muted-foreground">
+                +{matches.length - 1} other possible{' '}
+                {matches.length - 1 === 1 ? 'match' : 'matches'}
+              </span>
+            ) : null}
+          </div>
+        </div>
+      </div>
+    </div>
+  );
+}

src/lib/db/migrations/0021_unusual_azazel.sql

@@ -0,0 +1,30 @@
+CREATE TABLE "client_merge_candidates" (
+	"id" text PRIMARY KEY NOT NULL,
+	"port_id" text NOT NULL,
+	"client_a_id" text NOT NULL,
+	"client_b_id" text NOT NULL,
+	"score" integer NOT NULL,
+	"reasons" jsonb NOT NULL,
+	"status" text DEFAULT 'pending' NOT NULL,
+	"created_at" timestamp with time zone DEFAULT now() NOT NULL,
+	"resolved_at" timestamp with time zone,
+	"resolved_by" text
+);
+--> statement-breakpoint
+CREATE TABLE "migration_source_links" (
+	"id" text PRIMARY KEY NOT NULL,
+	"source_system" text NOT NULL,
+	"source_id" text NOT NULL,
+	"target_entity_type" text NOT NULL,
+	"target_entity_id" text NOT NULL,
+	"applied_id" text NOT NULL,
+	"applied_by" text,
+	"applied_at" timestamp with time zone DEFAULT now() NOT NULL
+);
+--> statement-breakpoint
+ALTER TABLE "client_merge_candidates" ADD CONSTRAINT "client_merge_candidates_port_id_ports_id_fk" FOREIGN KEY ("port_id") REFERENCES "public"."ports"("id") ON DELETE no action ON UPDATE no action;--> statement-breakpoint
+ALTER TABLE "client_merge_candidates" ADD CONSTRAINT "client_merge_candidates_client_a_id_clients_id_fk" FOREIGN KEY ("client_a_id") REFERENCES "public"."clients"("id") ON DELETE cascade ON UPDATE no action;--> statement-breakpoint
+ALTER TABLE "client_merge_candidates" ADD CONSTRAINT "client_merge_candidates_client_b_id_clients_id_fk" FOREIGN KEY ("client_b_id") REFERENCES "public"."clients"("id") ON DELETE cascade ON UPDATE no action;--> statement-breakpoint
+CREATE INDEX "idx_cmc_port_status" ON "client_merge_candidates" USING btree ("port_id","status");--> statement-breakpoint
+CREATE UNIQUE INDEX "idx_cmc_pair" ON "client_merge_candidates" USING btree ("port_id","client_a_id","client_b_id");--> statement-breakpoint
+CREATE UNIQUE INDEX "idx_msl_source_target" ON "migration_source_links" USING btree ("source_system","source_id","target_entity_type");

src/lib/db/migrations/0022_magenta_madame_hydra.sql

@@ -0,0 +1,2 @@
+ALTER TABLE "clients" ADD COLUMN "merged_into_client_id" text;--> statement-breakpoint
+CREATE INDEX "idx_clients_merged_into" ON "clients" USING btree ("merged_into_client_id");

src/lib/db/migrations/meta/_journal.json

@@ -148,6 +148,20 @@
       "when": 1777814682110,
       "tag": "0020_medical_betty_brant",
       "breakpoints": true
+    },
+    {
+      "idx": 21,
+      "version": "7",
+      "when": 1777811835982,
+      "tag": "0021_unusual_azazel",
+      "breakpoints": true
+    },
+    {
+      "idx": 22,
+      "version": "7",
+      "when": 1777812671833,
+      "tag": "0022_magenta_madame_hydra",
+      "breakpoints": true
     }
   ]
 }

src/lib/db/schema/clients.ts

@@ -2,6 +2,7 @@ import {
   pgTable,
   text,
   boolean,
+  integer,
   timestamp,
   jsonb,
   index,
@@ -30,6 +31,11 @@ export const clients = pgTable(
     source: text('source'), // website, manual, referral, broker
     sourceDetails: text('source_details'),
     archivedAt: timestamp('archived_at', { withTimezone: true }),
+    /** When this client was merged into another (the "loser" of a dedup
+     *  merge), this points at the surviving client. Used by the
+     *  /admin/duplicates review queue to redirect any stragglers, and by
+     *  the unmerge flow to restore. Null for live clients. */
+    mergedIntoClientId: text('merged_into_client_id'),
     createdAt: timestamp('created_at', { withTimezone: true }).notNull().defaultNow(),
     updatedAt: timestamp('updated_at', { withTimezone: true }).notNull().defaultNow(),
   },
@@ -38,6 +44,7 @@ export const clients = pgTable(
     index('idx_clients_name').on(table.portId, table.fullName),
     index('idx_clients_archived').on(table.portId, table.archivedAt),
     index('idx_clients_nationality_iso').on(table.nationalityIso),
+    index('idx_clients_merged_into').on(table.mergedIntoClientId),
   ],
 );
 
@@ -145,6 +152,54 @@ export const clientMergeLog = pgTable(
   (table) => [index('idx_cml_port').on(table.portId)],
 );
 
+/**
+ * Pairs of clients flagged by the background scoring job as potential
+ * duplicates. The `/admin/duplicates` review queue reads from here.
+ *
+ * Lifecycle:
+ *   - Background job inserts a row when a pair scores >= the
+ *     `dedup_review_queue_threshold` system setting.
+ *   - User reviews in the admin UI and either merges (status='merged')
+ *     or dismisses (status='dismissed').
+ *   - Subsequent runs of the scoring job skip pairs already
+ *     `dismissed` so the same false-positive doesn't keep reappearing.
+ *     A future score increase recreates the row.
+ *
+ * Pairs are stored canonically with `clientAId < clientBId` (string
+ * comparison) so the same pair only generates one row regardless of
+ * scoring direction.
+ */
+export const clientMergeCandidates = pgTable(
+  'client_merge_candidates',
+  {
+    id: text('id')
+      .primaryKey()
+      .$defaultFn(() => crypto.randomUUID()),
+    portId: text('port_id')
+      .notNull()
+      .references(() => ports.id),
+    clientAId: text('client_a_id')
+      .notNull()
+      .references(() => clients.id, { onDelete: 'cascade' }),
+    clientBId: text('client_b_id')
+      .notNull()
+      .references(() => clients.id, { onDelete: 'cascade' }),
+    score: integer('score').notNull(),
+    /** Human-readable rule list, e.g. ["email match", "phone match"]. */
+    reasons: jsonb('reasons').notNull(),
+    status: text('status').notNull().default('pending'), // pending | dismissed | merged
+    createdAt: timestamp('created_at', { withTimezone: true }).notNull().defaultNow(),
+    resolvedAt: timestamp('resolved_at', { withTimezone: true }),
+    resolvedBy: text('resolved_by'),
+  },
+  (table) => [
+    index('idx_cmc_port_status').on(table.portId, table.status),
+    // Same pair shouldn't surface twice — enforce uniqueness on the
+    // canonical (a < b) ordering.
+    uniqueIndex('idx_cmc_pair').on(table.portId, table.clientAId, table.clientBId),
+  ],
+);
+
 export const clientAddresses = pgTable(
   'client_addresses',
   {
@@ -190,3 +245,5 @@ export type ClientMergeLog = typeof clientMergeLog.$inferSelect;
 export type NewClientMergeLog = typeof clientMergeLog.$inferInsert;
 export type ClientAddress = typeof clientAddresses.$inferSelect;
 export type NewClientAddress = typeof clientAddresses.$inferInsert;
+export type ClientMergeCandidate = typeof clientMergeCandidates.$inferSelect;
+export type NewClientMergeCandidate = typeof clientMergeCandidates.$inferInsert;

src/lib/db/schema/index.ts

@@ -56,5 +56,8 @@ export * from './ai-usage';
 // GDPR export tracking (Phase 3d)
 export * from './gdpr';
 
+// Migration ledger (one-shot scripts — NocoDB import etc.)
+export * from './migration';
+
 // Relations (must come last — references all tables)
 export * from './relations';

src/lib/db/schema/migration.ts

@@ -0,0 +1,48 @@
+import { pgTable, text, timestamp, uniqueIndex } from 'drizzle-orm/pg-core';
+
+/**
+ * Idempotency ledger for one-shot data migrations from external sources
+ * (e.g. the legacy NocoDB Interests table).
+ *
+ * Every entity created during a migration script's `--apply` run gets a
+ * row here mapping the source-system row identifier to the new-system
+ * entity id. Re-running `--apply` against the same report skips rows
+ * already linked, so partial-failure resumption is just "run again."
+ *
+ * One source row can generate multiple new entities (e.g. one NocoDB
+ * Interests row → one client + one interest + one yacht), so the
+ * uniqueness constraint includes `target_entity_type`.
+ */
+export const migrationSourceLinks = pgTable(
+  'migration_source_links',
+  {
+    id: text('id')
+      .primaryKey()
+      .$defaultFn(() => crypto.randomUUID()),
+    /** e.g. 'nocodb_interests', 'nocodb_residences', 'nocodb_website_submissions'. */
+    sourceSystem: text('source_system').notNull(),
+    /** Source row identifier as a string (NocoDB IDs are integers; we keep
+     *  text here for forward compat with other sources). */
+    sourceId: text('source_id').notNull(),
+    /** e.g. 'client', 'interest', 'yacht', 'document'. */
+    targetEntityType: text('target_entity_type').notNull(),
+    /** UUID of the new-system entity (clients.id, interests.id, etc.). */
+    targetEntityId: text('target_entity_id').notNull(),
+    /** Apply-id from the migration run that created this link — pairs with
+     *  the on-disk apply manifest so `--rollback --apply-id <id>` knows
+     *  exactly which links to remove. */
+    appliedId: text('applied_id').notNull(),
+    appliedBy: text('applied_by'),
+    appliedAt: timestamp('applied_at', { withTimezone: true }).notNull().defaultNow(),
+  },
+  (table) => [
+    uniqueIndex('idx_msl_source_target').on(
+      table.sourceSystem,
+      table.sourceId,
+      table.targetEntityType,
+    ),
+  ],
+);
+
+export type MigrationSourceLink = typeof migrationSourceLinks.$inferSelect;
+export type NewMigrationSourceLink = typeof migrationSourceLinks.$inferInsert;

src/lib/dedup/find-matches.ts

@@ -0,0 +1,255 @@
+/**
+ * Client-match finder — pure scoring logic.
+ *
+ * Compares one input candidate against a pool of existing candidates and
+ * returns scored matches. Used by:
+ *   - the at-create suggestion in client/interest forms (Layer 1)
+ *   - the public-form auto-link path (when score >= block threshold)
+ *   - the nightly background scoring job (Layer 3)
+ *   - the migration script's dedup pass
+ *
+ * Performance shape: blocking via email / phone / surname-token reduces
+ * the pairwise scan from O(n²) to ~O(n) for any pool size we'll see in
+ * production. See `findClientMatches` for the blocking implementation.
+ *
+ * Design reference: docs/superpowers/specs/2026-05-03-dedup-and-migration-design.md §4.
+ */
+
+import { parsePhoneScriptSafe as parsePhone } from './phone-parse';
+
+import { levenshtein } from './normalize';
+
+// ─── Types ──────────────────────────────────────────────────────────────────
+
+export interface MatchCandidate {
+  id: string;
+  fullName: string | null;
+  /** Lowercased last non-particle token from `normalizeName(...).surnameToken`.
+   *  Used as a blocking key. */
+  surnameToken: string | null;
+  /** Already lowercased + validated via `normalizeEmail`. */
+  emails: string[];
+  /** Already canonical E.164 via `normalizePhone`. */
+  phonesE164: string[];
+  /** Address country (NOT phone country) — used for tiebreaking, not scoring. */
+  countryIso: string | null;
+}
+
+export type MatchConfidence = 'high' | 'medium' | 'low';
+
+export interface MatchResult {
+  candidate: MatchCandidate;
+  /** 0–100 after capping. */
+  score: number;
+  /** Human-readable list of which rules contributed. Useful for the
+   *  review queue UI ("matched on email + phone + surname token"). */
+  reasons: string[];
+  confidence: MatchConfidence;
+}
+
+export interface DedupThresholds {
+  /** Inclusive lower bound for `'high'` confidence. */
+  highScore: number;
+  /** Inclusive lower bound for `'medium'` confidence. Below this is `'low'`. */
+  mediumScore: number;
+}
+
+// ─── Public entry point ─────────────────────────────────────────────────────
+
+/**
+ * Compare `input` against every reachable candidate in `pool` and return
+ * scored matches, sorted by score descending. The result list includes
+ * low-confidence hits — caller filters by `confidence` or `score`
+ * depending on use case.
+ *
+ * Self-matches (an entry with `id === input.id`, e.g. when re-scoring an
+ * existing client during a background job) are excluded.
+ */
+export function findClientMatches(
+  input: MatchCandidate,
+  pool: MatchCandidate[],
+  thresholds: DedupThresholds,
+): MatchResult[] {
+  if (pool.length === 0) return [];
+
+  // ── Phase 1: build blocking indexes off the pool. ─────────────────────────
+  //
+  // Three indexes mean any candidate that shares ANY of (email / phone /
+  // surname-token) with the input shows up in the comparison set. Anything
+  // that shares NONE is structurally too different to be a duplicate and
+  // is skipped — this is what keeps the algorithm O(n) at scale.
+  const byEmail = new Map<string, MatchCandidate[]>();
+  const byPhone = new Map<string, MatchCandidate[]>();
+  const bySurnameToken = new Map<string, MatchCandidate[]>();
+
+  for (const c of pool) {
+    if (c.id === input.id) continue;
+    for (const email of c.emails) {
+      pushTo(byEmail, email, c);
+    }
+    for (const phone of c.phonesE164) {
+      pushTo(byPhone, phone, c);
+    }
+    if (c.surnameToken) {
+      pushTo(bySurnameToken, c.surnameToken, c);
+    }
+  }
+
+  // ── Phase 2: gather the comparison set via the blocking indexes. ─────────
+  const comparisonSet = new Map<string, MatchCandidate>();
+  for (const email of input.emails) {
+    for (const c of byEmail.get(email) ?? []) {
+      comparisonSet.set(c.id, c);
+    }
+  }
+  for (const phone of input.phonesE164) {
+    for (const c of byPhone.get(phone) ?? []) {
+      comparisonSet.set(c.id, c);
+    }
+  }
+  if (input.surnameToken) {
+    for (const c of bySurnameToken.get(input.surnameToken) ?? []) {
+      comparisonSet.set(c.id, c);
+    }
+  }
+
+  // ── Phase 3: score every candidate that survived blocking. ───────────────
+  const results: MatchResult[] = [];
+  for (const candidate of comparisonSet.values()) {
+    const r = scorePair(input, candidate);
+    results.push(r);
+  }
+
+  // ── Phase 4: sort by score desc + assign confidence tier. ────────────────
+  results.sort((a, b) => b.score - a.score);
+  for (const r of results) {
+    r.confidence = classify(r.score, thresholds);
+  }
+  return results;
+}
+
+// ─── Scoring ────────────────────────────────────────────────────────────────
+
+/**
+ * Score one (input, candidate) pair against the rule set in design §4.2.
+ * Compounding: positive rules sum, negative rules subtract; the result is
+ * clamped to [0, 100]. Reasons accumulate in the order rules fire so the
+ * review-queue UI can show "matched on email + phone".
+ */
+function scorePair(a: MatchCandidate, b: MatchCandidate): MatchResult {
+  let score = 0;
+  const reasons: string[] = [];
+
+  // ── Positive rules. ──────────────────────────────────────────────────────
+
+  const sharedEmail = a.emails.find((e) => b.emails.includes(e));
+  const emailMatch = !!sharedEmail;
+  if (emailMatch) {
+    score += 60;
+    reasons.push('email match');
+  }
+
+  const sharedPhone = a.phonesE164.find((p) => b.phonesE164.includes(p) && countDigits(p) >= 8);
+  const phoneMatch = !!sharedPhone;
+  if (phoneMatch) {
+    score += 50;
+    reasons.push('phone match');
+  }
+
+  const aNameNorm = (a.fullName ?? '').toLowerCase().trim();
+  const bNameNorm = (b.fullName ?? '').toLowerCase().trim();
+  const nameExactMatch = aNameNorm.length > 0 && aNameNorm === bNameNorm;
+  if (nameExactMatch) {
+    score += 20;
+    reasons.push('name match');
+  }
+
+  // Surname + given-name fuzzy. Only fires when names are NOT exactly
+  // equal — avoids double-counting with the rule above. Catches
+  // 'Constanzo' / 'Costanzo', 'Marc' / 'Marcus' etc. when other contact
+  // signals confirm them.
+  if (!nameExactMatch && a.surnameToken && b.surnameToken && a.surnameToken === b.surnameToken) {
+    const aGiven = (a.fullName ?? '').toLowerCase().split(/\s+/)[0] ?? '';
+    const bGiven = (b.fullName ?? '').toLowerCase().split(/\s+/)[0] ?? '';
+    if (aGiven && bGiven && levenshtein(aGiven, bGiven) <= 1) {
+      score += 15;
+      reasons.push('surname + given-name fuzzy match');
+    }
+  }
+
+  // ── Negative rules. ──────────────────────────────────────────────────────
+
+  // Same email but the two parties' phone numbers belong to different
+  // countries. Common when one inbox is shared by spouses / coworkers
+  // and the actual phone owners are distinct people. Don't auto-merge.
+  if (emailMatch && !phoneMatch && a.phonesE164.length > 0 && b.phonesE164.length > 0) {
+    const aCountries = phoneCountriesOf(a);
+    const bCountries = phoneCountriesOf(b);
+    const overlap = [...aCountries].some((c) => bCountries.has(c));
+    if (!overlap && aCountries.size > 0 && bCountries.size > 0) {
+      score -= 15;
+      reasons.push('phone country mismatch (negative)');
+    }
+  }
+
+  // Same name but no contact match. Two distinct people with the same
+  // name (common for "John Smith") sneak through name-based blocking;
+  // penalize so the score lands below the auto-merge threshold.
+  if (nameExactMatch && !emailMatch && !phoneMatch) {
+    score -= 20;
+    reasons.push('name match but no shared contact (negative)');
+  }
+
+  return {
+    candidate: b,
+    score: clamp(score, 0, 100),
+    reasons,
+    confidence: 'low', // assigned by caller after threshold lookup
+  };
+}
+
+// ─── Helpers ────────────────────────────────────────────────────────────────
+
+function pushTo<K, V>(map: Map<K, V[]>, key: K, value: V): void {
+  const existing = map.get(key);
+  if (existing) {
+    existing.push(value);
+  } else {
+    map.set(key, [value]);
+  }
+}
+
+function classify(score: number, thresholds: DedupThresholds): MatchConfidence {
+  if (score >= thresholds.highScore) return 'high';
+  if (score >= thresholds.mediumScore) return 'medium';
+  return 'low';
+}
+
+function clamp(value: number, min: number, max: number): number {
+  if (value < min) return min;
+  if (value > max) return max;
+  return value;
+}
+
+function countDigits(s: string): number {
+  let count = 0;
+  for (let i = 0; i < s.length; i += 1) {
+    const code = s.charCodeAt(i);
+    if (code >= 48 && code <= 57) count += 1;
+  }
+  return count;
+}
+
+/**
+ * Resolve each phone in a candidate to its ISO country code (via
+ * libphonenumber-js). Cached per call; the surrounding caller doesn't
+ * batch so we accept the parse cost.
+ */
+function phoneCountriesOf(c: MatchCandidate): Set<string> {
+  const out = new Set<string>();
+  for (const p of c.phonesE164) {
+    const parsed = parsePhone(p);
+    if (parsed.country) out.add(parsed.country);
+  }
+  return out;
+}

src/lib/dedup/migration-report.ts

@@ -0,0 +1,274 @@
+/**
+ * Migration report writer — turns a `MigrationPlan` (from
+ * `migration-transform.ts`) into a CSV + a human-readable Markdown
+ * summary on disk under `.migration/<timestamp>/`.
+ *
+ * The CSV format is intentionally machine-friendly (one row per
+ * planned operation) so it can be diffed across runs and inspected
+ * by hand. The summary is designed for "open this in your editor and
+ * eyeball it for 5 minutes before --apply."
+ */
+
+import { promises as fs } from 'node:fs';
+import path from 'node:path';
+
+import type { MigrationPlan } from './migration-transform';
+
+// ─── Output directory ───────────────────────────────────────────────────────
+
+export interface ReportPaths {
+  rootDir: string;
+  csvPath: string;
+  summaryPath: string;
+  planJsonPath: string;
+}
+
+/** Resolve report paths relative to the worktree root. The timestamped
+ *  directory is created lazily by `writeReport`. */
+export function resolveReportPaths(
+  rootDir: string,
+  timestamp: string = new Date().toISOString().replace(/[:.]/g, '-'),
+): ReportPaths {
+  const dir = path.join(rootDir, '.migration', timestamp);
+  return {
+    rootDir: dir,
+    csvPath: path.join(dir, 'report.csv'),
+    summaryPath: path.join(dir, 'summary.md'),
+    planJsonPath: path.join(dir, 'plan.json'),
+  };
+}
+
+// ─── CSV row shape ──────────────────────────────────────────────────────────
+
+interface CsvRow {
+  op: string; // create_client / create_contact / create_interest / auto_link / flag / needs_review
+  reason: string;
+  source_id: string;
+  target_table: string;
+  target_value: string;
+  confidence: string;
+  manual_review: 'true' | 'false';
+}
+
+// Trivial CSV escape: quote any cell that contains comma / quote / newline,
+// double up internal quotes per RFC 4180. No need for a dependency.
+function csvEscape(s: string): string {
+  if (/[",\n\r]/.test(s)) {
+    return `"${s.replace(/"/g, '""')}"`;
+  }
+  return s;
+}
+
+function rowToCsvLine(r: CsvRow): string {
+  return [
+    r.op,
+    r.reason,
+    r.source_id,
+    r.target_table,
+    r.target_value,
+    r.confidence,
+    r.manual_review,
+  ]
+    .map(csvEscape)
+    .join(',');
+}
+
+// ─── Build CSV ──────────────────────────────────────────────────────────────
+
+export function buildCsv(plan: MigrationPlan): string {
+  const lines: string[] = [];
+  lines.push(
+    [
+      'op',
+      'reason',
+      'source_id',
+      'target_table',
+      'target_value',
+      'confidence',
+      'manual_review',
+    ].join(','),
+  );
+
+  for (const client of plan.clients) {
+    lines.push(
+      rowToCsvLine({
+        op: 'create_client',
+        reason: client.sourceIds.length > 1 ? 'auto-merged cluster' : 'new',
+        source_id: client.sourceIds.join('|'),
+        target_table: 'clients.fullName',
+        target_value: client.fullName,
+        confidence: 'N/A',
+        manual_review: 'false',
+      }),
+    );
+    for (const c of client.contacts) {
+      lines.push(
+        rowToCsvLine({
+          op: 'create_contact',
+          reason: c.flagged ?? 'new',
+          source_id: client.sourceIds.join('|'),
+          target_table: `clientContacts.${c.channel}`,
+          target_value: c.value,
+          confidence: 'N/A',
+          manual_review: c.flagged ? 'true' : 'false',
+        }),
+      );
+    }
+    for (const a of client.addresses) {
+      lines.push(
+        rowToCsvLine({
+          op: 'create_address',
+          reason: 'address text present',
+          source_id: client.sourceIds.join('|'),
+          target_table: 'clientAddresses.countryIso',
+          target_value: a.countryIso ?? '(unresolved)',
+          confidence: a.countryConfidence ?? 'fallback',
+          manual_review: a.countryConfidence === 'fallback' || !a.countryIso ? 'true' : 'false',
+        }),
+      );
+    }
+  }
+
+  for (const interest of plan.interests) {
+    lines.push(
+      rowToCsvLine({
+        op: 'create_interest',
+        reason: `pipelineStage=${interest.pipelineStage}`,
+        source_id: String(interest.sourceId),
+        target_table: 'interests',
+        target_value: `${interest.berthMooringNumber ?? '(no berth)'} / ${interest.yachtName ?? '(no yacht)'}`,
+        confidence: 'N/A',
+        manual_review: 'false',
+      }),
+    );
+  }
+
+  for (const link of plan.autoLinks) {
+    lines.push(
+      rowToCsvLine({
+        op: 'auto_link',
+        reason: link.reasons.join(' + '),
+        source_id: `${link.leadSourceId}<-${link.mergedSourceIds.join(',')}`,
+        target_table: 'clients',
+        target_value: '(merged into lead)',
+        confidence: `score=${link.score}`,
+        manual_review: 'false',
+      }),
+    );
+  }
+
+  for (const pair of plan.needsReview) {
+    lines.push(
+      rowToCsvLine({
+        op: 'needs_review',
+        reason: pair.reasons.join(' + '),
+        source_id: `${pair.aSourceId}<->${pair.bSourceId}`,
+        target_table: 'clients',
+        target_value: '(human review required)',
+        confidence: `score=${pair.score}`,
+        manual_review: 'true',
+      }),
+    );
+  }
+
+  for (const flag of plan.flags) {
+    lines.push(
+      rowToCsvLine({
+        op: 'flag',
+        reason: flag.reason,
+        source_id: String(flag.sourceId),
+        target_table: flag.sourceTable,
+        target_value: JSON.stringify(flag.details ?? {}),
+        confidence: 'N/A',
+        manual_review: 'true',
+      }),
+    );
+  }
+
+  return lines.join('\n') + '\n';
+}
+
+// ─── Build summary markdown ─────────────────────────────────────────────────
+
+export function buildSummary(plan: MigrationPlan, generatedAt: string): string {
+  const s = plan.stats;
+  const lines: string[] = [];
+  lines.push(`# Migration Dry-Run — ${generatedAt}`);
+  lines.push('');
+  lines.push('## Input');
+  lines.push(`- ${s.inputInterestRows} NocoDB Interests`);
+  lines.push(`- ${s.inputResidentialRows} NocoDB Residential Interests`);
+  lines.push('');
+  lines.push('## Outcome');
+  lines.push(`- ${s.outputClients} clients`);
+  lines.push(`- ${s.outputInterests} interests (one per source row, linked to deduped client)`);
+  lines.push(`- ${s.outputContacts} client_contacts`);
+  lines.push(`- ${s.outputAddresses} client_addresses`);
+  lines.push('');
+  lines.push('## Auto-linked clusters');
+  if (plan.autoLinks.length === 0) {
+    lines.push('_None — every input row maps to a unique client._');
+  } else {
+    for (const link of plan.autoLinks) {
+      const merged = link.mergedSourceIds.length;
+      lines.push(
+        `- Lead row \`${link.leadSourceId}\` ← merged ${merged} other row${merged === 1 ? '' : 's'} (\`${link.mergedSourceIds.join(', ')}\`) — score ${link.score} via ${link.reasons.join(' + ')}`,
+      );
+    }
+  }
+  lines.push('');
+  lines.push('## Pairs flagged for human review');
+  if (plan.needsReview.length === 0) {
+    lines.push('_None._');
+  } else {
+    for (const pair of plan.needsReview) {
+      lines.push(
+        `- Rows \`${pair.aSourceId}\` ↔ \`${pair.bSourceId}\` — score ${pair.score} (${pair.reasons.join(' + ')})`,
+      );
+    }
+  }
+  lines.push('');
+  lines.push('## Data quality flags');
+  if (plan.flags.length === 0) {
+    lines.push('_No quality issues._');
+  } else {
+    const byReason = new Map<string, number>();
+    for (const f of plan.flags) {
+      byReason.set(f.reason, (byReason.get(f.reason) ?? 0) + 1);
+    }
+    for (const [reason, count] of [...byReason].sort((a, b) => b[1] - a[1])) {
+      lines.push(`- **${count}× ${reason}**`);
+    }
+    lines.push('');
+    lines.push('### Detail');
+    for (const f of plan.flags.slice(0, 30)) {
+      lines.push(
+        `- \`${f.sourceTable}#${f.sourceId}\`: ${f.reason}${f.details ? ` — \`${JSON.stringify(f.details)}\`` : ''}`,
+      );
+    }
+    if (plan.flags.length > 30) {
+      lines.push(`- _… and ${plan.flags.length - 30} more (see report.csv for full list)_`);
+    }
+  }
+  lines.push('');
+  lines.push('## Next step');
+  lines.push('');
+  lines.push('Eyeball the auto-linked + flagged-for-review pairs above.');
+  lines.push('When satisfied, re-run the script with `--apply --report .migration/<this-dir>/`.');
+  lines.push('Apply will refuse to run if the source NocoDB has changed since this dry-run.');
+
+  return lines.join('\n') + '\n';
+}
+
+// ─── Write to disk ──────────────────────────────────────────────────────────
+
+export async function writeReport(
+  paths: ReportPaths,
+  plan: MigrationPlan,
+  generatedAt: string,
+): Promise<void> {
+  await fs.mkdir(paths.rootDir, { recursive: true });
+  await fs.writeFile(paths.csvPath, buildCsv(plan), 'utf-8');
+  await fs.writeFile(paths.summaryPath, buildSummary(plan, generatedAt), 'utf-8');
+  await fs.writeFile(paths.planJsonPath, JSON.stringify(plan, null, 2), 'utf-8');
+}

src/lib/dedup/migration-transform.ts

@@ -0,0 +1,576 @@
+/**
+ * Pure transform: NocoDB snapshot → planned new-system entities + dedup result.
+ *
+ * Used by the migration script's `--dry-run` (to produce the report) and
+ * `--apply` (to actually write). Keeping this pure means the same code
+ * runs in both modes, in tests against the frozen fixture, and in the
+ * one-off CLI run against the live base.
+ *
+ * No side effects, no DB calls, no external services.
+ */
+
+import {
+  normalizeName,
+  normalizeEmail,
+  normalizePhone,
+  resolveCountry,
+  type NormalizedPhone,
+} from './normalize';
+import { findClientMatches, type MatchCandidate } from './find-matches';
+import type { CountryCode } from '@/lib/i18n/countries';
+import type { NocoDbRow, NocoDbSnapshot } from './nocodb-source';
+
+// ─── Plan output ────────────────────────────────────────────────────────────
+
+export interface PlannedClient {
+  /** Stable id derived from the deduped cluster's lead row. Used by the
+   *  apply phase to reference newly-created clients before they exist
+   *  in the DB. */
+  tempId: string;
+  /** Source row IDs that contributed to this client (one if no duplicates,
+   *  many if dedup merged a cluster). */
+  sourceIds: number[];
+  fullName: string;
+  surnameToken?: string;
+  countryIso: CountryCode | null;
+  preferredContactMethod: string | null;
+  source: string | null;
+  contacts: PlannedContact[];
+  addresses: PlannedAddress[];
+}
+
+export interface PlannedContact {
+  channel: 'email' | 'phone' | 'whatsapp' | 'other';
+  value: string;
+  valueE164?: string | null;
+  valueCountry?: CountryCode | null;
+  isPrimary: boolean;
+  flagged?: string;
+}
+
+export interface PlannedAddress {
+  streetAddress: string | null;
+  city: string | null;
+  countryIso: CountryCode | null;
+  /** When confidence is low, the migration script flags the row for
+   *  human review. */
+  countryConfidence: 'exact' | 'fuzzy' | 'city' | 'fallback' | null;
+}
+
+export interface PlannedInterest {
+  /** NocoDB row id this interest came from. */
+  sourceId: number;
+  /** tempId of the planned client this interest hangs off. */
+  clientTempId: string;
+  pipelineStage: string;
+  leadCategory: string | null;
+  source: string | null;
+  notes: string | null;
+  /** Mooring number; the apply phase resolves this to a berthId via the
+   *  new-system Berths table. */
+  berthMooringNumber: string | null;
+  yachtName: string | null;
+  /** Date stamps for milestone columns. ISO strings if parseable. */
+  dateEoiSent: string | null;
+  dateEoiSigned: string | null;
+  dateDepositReceived: string | null;
+  dateContractSent: string | null;
+  dateContractSigned: string | null;
+  dateLastContact: string | null;
+  /** Documenso linkage carried forward when present so the document
+   *  record can be stitched up downstream. */
+  documensoId: string | null;
+}
+
+export interface MigrationFlag {
+  sourceTable: 'interests' | 'residential_interests' | 'website_interest_submissions';
+  sourceId: number;
+  reason: string;
+  details?: Record<string, unknown>;
+}
+
+export interface MigrationPlan {
+  clients: PlannedClient[];
+  interests: PlannedInterest[];
+  flags: MigrationFlag[];
+  /** Pairs that the migration would auto-link (high score). */
+  autoLinks: Array<{
+    leadSourceId: number;
+    mergedSourceIds: number[];
+    score: number;
+    reasons: string[];
+  }>;
+  /** Pairs that need human review (medium score). Each pair shows up
+   *  in the migration report; the user resolves before --apply. */
+  needsReview: Array<{ aSourceId: number; bSourceId: number; score: number; reasons: string[] }>;
+  stats: MigrationStats;
+}
+
+export interface MigrationStats {
+  inputInterestRows: number;
+  inputResidentialRows: number;
+  outputClients: number;
+  outputInterests: number;
+  outputContacts: number;
+  outputAddresses: number;
+  flaggedRows: number;
+  autoLinkedClusters: number;
+  needsReviewPairs: number;
+}
+
+export interface TransformOptions {
+  /** ISO country used when a phone has no prefix and the row has no
+   *  Place of Residence. Defaults to AI (Anguilla / Port Nimara's home). */
+  defaultPhoneCountry: CountryCode;
+  /** Score thresholds for auto-link vs human review. Should match the
+   *  per-port `system_settings` values once the runtime UI is in place. */
+  thresholds: {
+    autoLink: number;
+    needsReview: number;
+  };
+}
+
+const DEFAULT_OPTIONS: TransformOptions = {
+  defaultPhoneCountry: 'AI',
+  thresholds: { autoLink: 90, needsReview: 50 },
+};
+
+// ─── Stage mapping ──────────────────────────────────────────────────────────
+
+const STAGE_MAP: Record<string, string> = {
+  'General Qualified Interest': 'open',
+  'Specific Qualified Interest': 'details_sent',
+  'EOI and NDA Sent': 'eoi_sent',
+  'Signed EOI and NDA': 'eoi_signed',
+  'Made Reservation': 'deposit_10pct',
+  'Contract Negotiation': 'contract_sent',
+  'Contract Negotiations Finalized': 'contract_sent',
+  'Contract Signed': 'contract_signed',
+};
+
+const LEAD_CATEGORY_MAP: Record<string, string> = {
+  General: 'general_interest',
+  'Friends and Family': 'general_interest',
+};
+
+const SOURCE_MAP: Record<string, string> = {
+  portal: 'website',
+  Form: 'website',
+  External: 'manual',
+};
+
+// ─── Date parsing ───────────────────────────────────────────────────────────
+
+/**
+ * Parse a date the legacy NocoDB might have stored in DD-MM-YYYY,
+ * DD/MM/YYYY, YYYY-MM-DD, or ISO format. Returns ISO string or null.
+ */
+function parseFlexibleDate(input: unknown): string | null {
+  if (typeof input !== 'string' || input.trim() === '') return null;
+  const s = input.trim();
+
+  // Already ISO
+  if (/^\d{4}-\d{2}-\d{2}/.test(s)) {
+    const d = new Date(s);
+    return Number.isNaN(d.getTime()) ? null : d.toISOString();
+  }
+
+  // DD-MM-YYYY or DD/MM/YYYY
+  const m = s.match(/^(\d{1,2})[-/](\d{1,2})[-/](\d{4})$/);
+  if (m) {
+    const [, day, month, year] = m;
+    const iso = `${year}-${month!.padStart(2, '0')}-${day!.padStart(2, '0')}`;
+    const d = new Date(iso);
+    return Number.isNaN(d.getTime()) ? null : d.toISOString();
+  }
+
+  // Anything else: try Date constructor as a last resort
+  const d = new Date(s);
+  return Number.isNaN(d.getTime()) ? null : d.toISOString();
+}
+
+// ─── Main transform ─────────────────────────────────────────────────────────
+
+/**
+ * Run the full transform pipeline against a NocoDB snapshot. Pure
+ * function — same input always produces the same plan.
+ */
+export function transformSnapshot(
+  snapshot: NocoDbSnapshot,
+  options: Partial<TransformOptions> = {},
+): MigrationPlan {
+  const opts = { ...DEFAULT_OPTIONS, ...options };
+
+  const flags: MigrationFlag[] = [];
+  // Build per-row candidates first so we can run dedup before assigning
+  // tempIds (clients with multiple source rows merge into one tempId).
+  const perRow = snapshot.interests.map((row) => rowToCandidate(row, 'interests', opts, flags));
+
+  // Dedup pass 1: every row scored against every other row (within the
+  // same pool). The blocking strategy in `findClientMatches` keeps this
+  // cheap even for the full 252-row dataset.
+  const clusters = clusterByDedup(perRow, opts);
+
+  // Build the planned clients + interests from the clusters.
+  const clients: PlannedClient[] = [];
+  const interests: PlannedInterest[] = [];
+  const autoLinks: MigrationPlan['autoLinks'] = [];
+  const needsReview: MigrationPlan['needsReview'] = [];
+
+  for (const cluster of clusters) {
+    const lead = cluster.leadCandidate;
+    const tempId = `client-${lead.row.Id}`;
+
+    // Build the client record from the lead row, then merge in any
+    // contact info / address info from the other rows in the cluster.
+    const planned = buildPlannedClient(tempId, cluster, opts);
+    clients.push(planned);
+
+    // Each row in the cluster becomes its own interest record.
+    for (const member of cluster.members) {
+      const interest = buildPlannedInterest(member.row, tempId);
+      interests.push(interest);
+    }
+
+    if (cluster.members.length > 1) {
+      autoLinks.push({
+        leadSourceId: lead.row.Id,
+        mergedSourceIds: cluster.members.filter((m) => m !== lead).map((m) => m.row.Id),
+        score: cluster.maxScore,
+        reasons: cluster.reasons,
+      });
+    }
+
+    for (const pair of cluster.reviewPairs) {
+      needsReview.push(pair);
+    }
+  }
+
+  return {
+    clients,
+    interests,
+    flags,
+    autoLinks,
+    needsReview,
+    stats: {
+      inputInterestRows: snapshot.interests.length,
+      inputResidentialRows: snapshot.residentialInterests.length,
+      outputClients: clients.length,
+      outputInterests: interests.length,
+      outputContacts: clients.reduce((sum, c) => sum + c.contacts.length, 0),
+      outputAddresses: clients.reduce((sum, c) => sum + c.addresses.length, 0),
+      flaggedRows: flags.length,
+      autoLinkedClusters: autoLinks.length,
+      needsReviewPairs: needsReview.length,
+    },
+  };
+}
+
+// ─── Helpers ────────────────────────────────────────────────────────────────
+
+interface RowCandidate {
+  row: NocoDbRow;
+  candidate: MatchCandidate;
+  /** Phone normalize result for the row's primary phone; used downstream
+   *  to attach valueE164 + country to the planned contact. */
+  phoneResult: NormalizedPhone | null;
+  /** Country resolved from "Place of Residence". */
+  countryIso: CountryCode | null;
+  countryConfidence: 'exact' | 'fuzzy' | 'city' | null;
+  /** Normalized email or null. */
+  email: string | null;
+  /** Display name from `normalizeName`. */
+  displayName: string;
+}
+
+function rowToCandidate(
+  row: NocoDbRow,
+  sourceTable: MigrationFlag['sourceTable'],
+  opts: TransformOptions,
+  flags: MigrationFlag[],
+): RowCandidate {
+  const rawName = (row['Full Name'] as string | undefined) ?? '';
+  const rawEmail = (row['Email Address'] as string | undefined) ?? '';
+  const rawPhone = (row['Phone Number'] as string | undefined) ?? '';
+  const rawCountry = (row['Place of Residence'] as string | undefined) ?? '';
+
+  const normName = normalizeName(rawName);
+  const email = normalizeEmail(rawEmail);
+  const country = resolveCountry(rawCountry);
+  const phoneCountry = country.iso ?? opts.defaultPhoneCountry;
+  const phoneResult = normalizePhone(rawPhone, phoneCountry as CountryCode);
+
+  // Surface anything weird so the report can show it.
+  if (rawPhone && !phoneResult?.e164) {
+    flags.push({
+      sourceTable,
+      sourceId: row.Id,
+      reason: phoneResult?.flagged ? `phone ${phoneResult.flagged}` : 'phone unparseable',
+      details: { rawPhone },
+    });
+  }
+  if (rawEmail && !email) {
+    flags.push({
+      sourceTable,
+      sourceId: row.Id,
+      reason: 'email invalid',
+      details: { rawEmail },
+    });
+  }
+  if (rawCountry && !country.iso) {
+    flags.push({
+      sourceTable,
+      sourceId: row.Id,
+      reason: 'country unresolved',
+      details: { rawCountry },
+    });
+  }
+
+  const candidate: MatchCandidate = {
+    id: String(row.Id),
+    fullName: normName.display || null,
+    surnameToken: normName.surnameToken ?? null,
+    emails: email ? [email] : [],
+    phonesE164: phoneResult?.e164 ? [phoneResult.e164] : [],
+    countryIso: country.iso ?? null,
+  };
+
+  return {
+    row,
+    candidate,
+    phoneResult,
+    countryIso: country.iso ?? null,
+    countryConfidence: country.confidence,
+    email,
+    displayName: normName.display,
+  };
+}
+
+interface Cluster {
+  /** The cluster's "lead" row (most complete + most recent). */
+  leadCandidate: RowCandidate;
+  members: RowCandidate[];
+  maxScore: number;
+  reasons: string[];
+  /** Pairs in this cluster that scored medium (need review). */
+  reviewPairs: Array<{ aSourceId: number; bSourceId: number; score: number; reasons: string[] }>;
+}
+
+function clusterByDedup(rows: RowCandidate[], opts: TransformOptions): Cluster[] {
+  // Use a union-find structure indexed by row id. Every pair with a
+  // score >= autoLink threshold gets unioned. Pairs in [needsReview,
+  // autoLink) accumulate onto the cluster's reviewPairs list — they're
+  // surfaced for human triage but not auto-merged.
+  const parent = new Map<string, string>();
+  for (const r of rows) parent.set(r.candidate.id, r.candidate.id);
+  const find = (id: string): string => {
+    let cur = id;
+    while (parent.get(cur) !== cur) {
+      const next = parent.get(cur)!;
+      parent.set(cur, parent.get(next)!); // path compression
+      cur = parent.get(cur)!;
+    }
+    return cur;
+  };
+  const union = (a: string, b: string) => {
+    const rootA = find(a);
+    const rootB = find(b);
+    if (rootA !== rootB) parent.set(rootA, rootB);
+  };
+
+  const clusterReasons = new Map<string, string[]>();
+  const clusterMaxScore = new Map<string, number>();
+  const clusterReviewPairs = new Map<string, Cluster['reviewPairs']>();
+
+  // Score every candidate against every other candidate. The find-matches
+  // function does its own blocking so this is cheap.
+  for (let i = 0; i < rows.length; i += 1) {
+    const left = rows[i]!;
+    const remainingPool = rows.slice(i + 1).map((r) => r.candidate);
+    if (remainingPool.length === 0) continue;
+    const matches = findClientMatches(left.candidate, remainingPool, {
+      highScore: opts.thresholds.autoLink,
+      mediumScore: opts.thresholds.needsReview,
+    });
+
+    for (const m of matches) {
+      if (m.score >= opts.thresholds.autoLink) {
+        union(left.candidate.id, m.candidate.id);
+        const root = find(left.candidate.id);
+        clusterMaxScore.set(root, Math.max(clusterMaxScore.get(root) ?? 0, m.score));
+        const existing = clusterReasons.get(root) ?? [];
+        for (const reason of m.reasons) {
+          if (!existing.includes(reason)) existing.push(reason);
+        }
+        clusterReasons.set(root, existing);
+      } else if (m.score >= opts.thresholds.needsReview) {
+        // Medium — track on whichever cluster `left` belongs to.
+        const root = find(left.candidate.id);
+        const list = clusterReviewPairs.get(root) ?? [];
+        list.push({
+          aSourceId: parseInt(left.candidate.id, 10),
+          bSourceId: parseInt(m.candidate.id, 10),
+          score: m.score,
+          reasons: m.reasons,
+        });
+        clusterReviewPairs.set(root, list);
+      }
+    }
+  }
+
+  // Group rows by their cluster root.
+  const byRoot = new Map<string, RowCandidate[]>();
+  for (const r of rows) {
+    const root = find(r.candidate.id);
+    const list = byRoot.get(root) ?? [];
+    list.push(r);
+    byRoot.set(root, list);
+  }
+
+  // Build cluster objects, choosing the most-complete row as the lead.
+  const clusters: Cluster[] = [];
+  for (const [root, members] of byRoot) {
+    const lead = pickLead(members);
+    clusters.push({
+      leadCandidate: lead,
+      members,
+      maxScore: clusterMaxScore.get(root) ?? 0,
+      reasons: clusterReasons.get(root) ?? [],
+      reviewPairs: clusterReviewPairs.get(root) ?? [],
+    });
+  }
+  return clusters;
+}
+
+function pickLead(rows: RowCandidate[]): RowCandidate {
+  // Pick the row with the most populated fields, breaking ties by
+  // recency (highest Id, since NocoDB IDs are monotonic).
+  return rows.reduce((best, current) => {
+    const bestScore = completenessScore(best);
+    const currentScore = completenessScore(current);
+    if (currentScore > bestScore) return current;
+    if (currentScore === bestScore && current.row.Id > best.row.Id) return current;
+    return best;
+  });
+}
+
+function completenessScore(r: RowCandidate): number {
+  let score = 0;
+  if (r.email) score += 1;
+  if (r.phoneResult?.e164) score += 1;
+  if (r.row['Address']) score += 0.5;
+  if (r.row['Yacht Name']) score += 0.5;
+  if (r.row['Source']) score += 0.25;
+  if (r.row['Lead Category']) score += 0.25;
+  if (r.row['Internal Notes']) score += 0.25;
+  return score;
+}
+
+function buildPlannedClient(
+  tempId: string,
+  cluster: Cluster,
+  opts: TransformOptions,
+): PlannedClient {
+  const lead = cluster.leadCandidate;
+
+  // Collect distinct emails + phones from across the cluster — duplicate
+  // submissions often come with different contact methods we want to
+  // preserve as multiple rows in `client_contacts`.
+  const seenEmails = new Set<string>();
+  const seenPhones = new Set<string>();
+  const contacts: PlannedContact[] = [];
+
+  for (const member of cluster.members) {
+    if (member.email && !seenEmails.has(member.email)) {
+      seenEmails.add(member.email);
+      contacts.push({
+        channel: 'email',
+        value: member.email,
+        isPrimary: contacts.length === 0,
+      });
+    }
+    if (member.phoneResult?.e164 && !seenPhones.has(member.phoneResult.e164)) {
+      seenPhones.add(member.phoneResult.e164);
+      const isFirstPhone = !contacts.some((c) => c.channel === 'phone');
+      contacts.push({
+        channel: 'phone',
+        value: member.phoneResult.e164,
+        valueE164: member.phoneResult.e164,
+        valueCountry: member.phoneResult.country,
+        isPrimary: isFirstPhone && contacts.every((c) => !c.isPrimary || c.channel === 'email'),
+        flagged: member.phoneResult.flagged,
+      });
+    }
+  }
+
+  // Demote the email-primary if a more-completable phone exists.
+  // Simpler invariant: the first contact is primary unless the row
+  // explicitly preferred phone.
+  const preferredMethod = (lead.row['Contact Method Preferred'] as string | undefined)
+    ?.toLowerCase()
+    ?.trim();
+
+  // Address: only build if the lead row has a meaningful address text.
+  const rawAddress = (lead.row['Address'] as string | undefined)?.trim();
+  const addresses: PlannedAddress[] = [];
+  if (rawAddress) {
+    addresses.push({
+      streetAddress: rawAddress,
+      city: null,
+      countryIso: lead.countryIso ?? opts.defaultPhoneCountry,
+      countryConfidence: lead.countryConfidence ?? 'fallback',
+    });
+  }
+
+  const sourceFromRow = (lead.row['Source'] as string | undefined) ?? null;
+  const mappedSource = sourceFromRow ? (SOURCE_MAP[sourceFromRow] ?? 'manual') : null;
+
+  return {
+    tempId,
+    sourceIds: cluster.members.map((m) => m.row.Id),
+    fullName: lead.displayName,
+    surnameToken: lead.candidate.surnameToken ?? undefined,
+    countryIso: lead.countryIso,
+    preferredContactMethod: preferredMethod ?? null,
+    source: mappedSource,
+    contacts,
+    addresses,
+  };
+}
+
+function buildPlannedInterest(row: NocoDbRow, clientTempId: string): PlannedInterest {
+  const stage = (row['Sales Process Level'] as string | undefined) ?? '';
+  const cat = (row['Lead Category'] as string | undefined) ?? '';
+
+  const notesParts: string[] = [];
+  const internalNotes = row['Internal Notes'] as string | undefined;
+  const extraComments = row['Extra Comments'] as string | undefined;
+  if (internalNotes?.trim()) notesParts.push(internalNotes.trim());
+  if (extraComments?.trim()) notesParts.push(`Extra Comments: ${extraComments.trim()}`);
+  const berthSize = row['Berth Size Desired'] as string | undefined;
+  if (berthSize?.trim()) notesParts.push(`Berth size desired: ${berthSize.trim()}`);
+
+  return {
+    sourceId: row.Id,
+    clientTempId,
+    pipelineStage: STAGE_MAP[stage] ?? 'open',
+    leadCategory: LEAD_CATEGORY_MAP[cat] ?? null,
+    source: ((row['Source'] as string | undefined) ?? null) || null,
+    notes: notesParts.join('\n\n') || null,
+    berthMooringNumber: (row['Berth Number'] as string | undefined) ?? null,
+    yachtName: (() => {
+      const n = (row['Yacht Name'] as string | undefined)?.trim();
+      // Filter placeholder values used by sales reps for "we don't know yet".
+      if (!n) return null;
+      if (['TBC', 'Na', 'NA', 'na', 'N/A', 'TBD', 'tbd'].includes(n)) return null;
+      return n;
+    })(),
+    dateEoiSent: parseFlexibleDate(row['EOI Time Sent']),
+    dateEoiSigned: parseFlexibleDate(row['all_signed_notified_at'] ?? row['developerSignTime']),
+    dateDepositReceived: null, // not directly tracked in legacy schema
+    dateContractSent: parseFlexibleDate(row['Time LOI Sent']),
+    dateContractSigned: parseFlexibleDate(row['developerSignTime']),
+    dateLastContact: parseFlexibleDate(row['Created At'] ?? row['Date Added']),
+    documensoId: (row['documensoID'] as string | undefined) ?? null,
+  };
+}

src/lib/dedup/nocodb-source.ts

@@ -0,0 +1,152 @@
+/**
+ * Read-only adapter for the legacy NocoDB Port Nimara base.
+ *
+ * Used by the one-shot migration script (`scripts/migrate-from-nocodb.ts`)
+ * to pull every Interest, Residential Interest, and Website Submission
+ * row from the source-of-truth NocoDB tables. No mutations.
+ *
+ * Auth: `xc-token` header per NocoDB v2 API.
+ *
+ * The shape returned is a verbatim record of the row's fields — caller
+ * is responsible for mapping to the new schema via `nocodb-transform.ts`.
+ */
+
+import { z } from 'zod';
+
+// ─── Configuration ──────────────────────────────────────────────────────────
+
+const ConfigSchema = z.object({
+  url: z.string().url(),
+  token: z.string().min(1),
+});
+
+export interface NocoDbConfig {
+  url: string;
+  token: string;
+}
+
+export function loadNocoDbConfig(env: NodeJS.ProcessEnv = process.env): NocoDbConfig {
+  return ConfigSchema.parse({
+    url: env.NOCODB_URL,
+    token: env.NOCODB_TOKEN,
+  });
+}
+
+// ─── Table identifiers ──────────────────────────────────────────────────────
+//
+// These IDs are stable per the NocoDB base — they were captured during the
+// 2026-05-03 audit and won't change unless the base is rebuilt. If the
+// base is reset, regenerate them from `getTablesList`.
+export const NOCO_TABLES = {
+  interests: 'mbs9hjauug4eseo',
+  residentialInterests: 'mscfpwwwjuds4nt',
+  websiteInterestSubmissions: 'mevkpcih67c6jsm',
+  websiteContactFormSubmissions: 'mxk5cd0pmwnwlcl',
+  websiteBerthEoiSupplements: 'mglmioo0ku8zgqj',
+  berths: 'mczgos9hr3oa9qc',
+} as const;
+
+// ─── HTTP shape ─────────────────────────────────────────────────────────────
+
+interface NocoDbListResponse<T> {
+  list: T[];
+  pageInfo: {
+    totalRows: number;
+    page: number;
+    pageSize: number;
+    isFirstPage: boolean;
+    isLastPage: boolean;
+  };
+}
+
+/** A row's `Id` is always present. The rest of the fields vary per table. */
+export type NocoDbRow = Record<string, unknown> & { Id: number };
+
+// ─── Public API ─────────────────────────────────────────────────────────────
+
+/**
+ * Fetch all rows from a NocoDB table. Auto-paginates until the API
+ * reports `isLastPage`. The legacy base is small (252 Interests rows
+ * being the largest table) so we keep this simple — no streaming.
+ */
+export async function fetchAllRows(
+  tableId: string,
+  config: NocoDbConfig,
+  pageSize = 250,
+): Promise<NocoDbRow[]> {
+  const all: NocoDbRow[] = [];
+  let page = 1;
+  // Hard cap to prevent infinite-loop bugs if pageInfo lies. Each page
+  // pulls up to `pageSize` rows, so 200 pages * 250 = 50k rows is the
+  // maximum we'll ever fetch from one table.
+  const MAX_PAGES = 200;
+
+  while (page <= MAX_PAGES) {
+    const url = new URL(`${config.url}/api/v2/tables/${tableId}/records`);
+    url.searchParams.set('limit', String(pageSize));
+    url.searchParams.set('offset', String((page - 1) * pageSize));
+
+    const res = await fetch(url, {
+      headers: {
+        'xc-token': config.token,
+        accept: 'application/json',
+      },
+    });
+
+    if (!res.ok) {
+      throw new Error(
+        `NocoDB fetch failed: ${res.status} ${res.statusText} — table ${tableId} page ${page}`,
+      );
+    }
+
+    const json = (await res.json()) as NocoDbListResponse<NocoDbRow>;
+    all.push(...json.list);
+
+    if (json.pageInfo.isLastPage || json.list.length === 0) break;
+    page += 1;
+  }
+
+  return all;
+}
+
+/**
+ * Convenience snapshot — pulls every table the migration cares about
+ * in parallel. Returned shape is the input the transform layer expects.
+ */
+export interface NocoDbSnapshot {
+  interests: NocoDbRow[];
+  residentialInterests: NocoDbRow[];
+  websiteInterestSubmissions: NocoDbRow[];
+  websiteContactFormSubmissions: NocoDbRow[];
+  websiteBerthEoiSupplements: NocoDbRow[];
+  berths: NocoDbRow[];
+  fetchedAt: string;
+}
+
+export async function fetchSnapshot(config: NocoDbConfig): Promise<NocoDbSnapshot> {
+  const [
+    interests,
+    residentialInterests,
+    websiteInterestSubmissions,
+    websiteContactFormSubmissions,
+    websiteBerthEoiSupplements,
+    berths,
+  ] = await Promise.all([
+    fetchAllRows(NOCO_TABLES.interests, config),
+    fetchAllRows(NOCO_TABLES.residentialInterests, config),
+    fetchAllRows(NOCO_TABLES.websiteInterestSubmissions, config),
+    fetchAllRows(NOCO_TABLES.websiteContactFormSubmissions, config),
+    fetchAllRows(NOCO_TABLES.websiteBerthEoiSupplements, config),
+    fetchAllRows(NOCO_TABLES.berths, config),
+  ]);
+
+  return {
+    interests,
+    residentialInterests,
+    websiteInterestSubmissions,
+    websiteContactFormSubmissions,
+    websiteBerthEoiSupplements,
+    berths,
+    fetchedAt: new Date().toISOString(),
+  };
+}

src/lib/dedup/normalize.ts

@@ -0,0 +1,418 @@
+/**
+ * Normalization helpers for the dedup pipeline.
+ *
+ * Pure functions (no DB, no React). Used by both the runtime at-create
+ * surfaces and the one-shot NocoDB migration script. Every transform
+ * here has a fixture in `tests/unit/dedup/normalize.test.ts` drawn from
+ * real dirty values observed in the legacy NocoDB Interests table.
+ *
+ * Design reference: docs/superpowers/specs/2026-05-03-dedup-and-migration-design.md §3.
+ */
+
+import { z } from 'zod';
+
+import { ALL_COUNTRY_CODES, getCountryName, type CountryCode } from '@/lib/i18n/countries';
+import { parsePhoneScriptSafe as parsePhone } from './phone-parse';
+
+// ─── Names ──────────────────────────────────────────────────────────────────
+
+/**
+ * Tokens that should stay lowercase mid-name. Covers the common Romance,
+ * Germanic, and Iberian particles seen in client records. The first token
+ * of a name is always title-cased even if it appears in this set.
+ */
+const PARTICLES: ReadonlySet<string> = new Set([
+  'van',
+  'von',
+  'de',
+  'del',
+  'da',
+  'das',
+  'do',
+  'dos',
+  'di',
+  'le',
+  'la',
+  'el',
+  'al',
+  'der',
+  'den',
+  'des',
+  'du',
+  'dalla',
+  'della',
+  'st',
+  'st.',
+  'y',
+]);
+
+export interface NormalizedName {
+  /** Human-readable form preserved for UI display. Trims, collapses
+   *  whitespace, fixes case, but never destroys the user's intent —
+   *  slash-with-company structure ("Daniel Wainstein / 7 Knots, LLC")
+   *  is left intact. */
+  display: string;
+  /** Lowercased form for matching. */
+  normalized: string;
+  /** Last non-particle token, lowercased. Used as a blocking key by the
+   *  dedup algorithm so we only compare candidates with similar surnames. */
+  surnameToken?: string;
+}
+
+/**
+ * Normalize a free-text full name. Trims and collapses whitespace,
+ * replaces \r/\n/\t with single spaces, intelligently title-cases
+ * ALL-CAPS surnames while keeping particles (van / de / dalla / etc.)
+ * lowercase mid-name, and preserves Irish O' surnames as O'Brien.
+ *
+ * If the input contains a `/` (slash-with-company structure like
+ * "Daniel Wainstein / 7 Knots, LLC"), the trailing company text is
+ * preserved verbatim — it's signal, not noise.
+ */
+export function normalizeName(raw: string | null | undefined): NormalizedName {
+  const safe = (raw ?? '').toString();
+  // Replace \r, \n, \t with single spaces, then collapse runs of whitespace.
+  const cleaned = safe
+    .replace(/[\r\n\t]/g, ' ')
+    .replace(/\s+/g, ' ')
+    .trim();
+
+  if (!cleaned) {
+    return { display: '', normalized: '', surnameToken: undefined };
+  }
+
+  // Slash-with-company: title-case the part before the slash, leave the
+  // company segment untouched (it's typically already a brand we shouldn't
+  // mangle: "SAS TIKI", "7 Knots, LLC").
+  const slashIdx = cleaned.indexOf('/');
+  let displayCore: string;
+  if (slashIdx !== -1) {
+    const personPart = cleaned.slice(0, slashIdx).trim();
+    const companyPart = cleaned.slice(slashIdx + 1).trim();
+    displayCore = `${titleCaseTokens(personPart)} / ${companyPart}`;
+  } else {
+    displayCore = titleCaseTokens(cleaned);
+  }
+
+  const display = displayCore;
+  const normalized = display.toLowerCase();
+  const surnameToken = computeSurnameToken(slashIdx !== -1 ? cleaned.slice(0, slashIdx) : cleaned);
+
+  return { display, normalized, surnameToken };
+}
+
+function titleCaseTokens(s: string): string {
+  const tokens = s.split(' ').filter(Boolean);
+  if (tokens.length === 0) return '';
+  return tokens.map((tok, idx) => titleCaseOneToken(tok, idx === 0)).join(' ');
+}
+
+function titleCaseOneToken(token: string, isFirst: boolean): string {
+  if (!token) return '';
+  const lower = token.toLowerCase();
+  if (!isFirst && PARTICLES.has(lower)) return lower;
+  // O'Brien / D'Angelo / l'Estrange — capitalize the segment after each
+  // apostrophe so a lowercased input round-trips to readable Irish caps.
+  if (lower.includes("'")) {
+    return lower
+      .split("'")
+      .map((part) => (part.length > 0 ? part[0]!.toUpperCase() + part.slice(1) : part))
+      .join("'");
+  }
+  return lower[0]!.toUpperCase() + lower.slice(1);
+}
+
+function computeSurnameToken(personPart: string): string | undefined {
+  const cleaned = personPart
+    .replace(/[\r\n\t]/g, ' ')
+    .replace(/\s+/g, ' ')
+    .trim();
+  if (!cleaned) return undefined;
+  const tokens = cleaned.split(' ').map((t) => t.toLowerCase());
+  // Walk from the right past particles to find the last "real" surname token.
+  for (let i = tokens.length - 1; i >= 0; i -= 1) {
+    const tok = tokens[i]!;
+    if (!PARTICLES.has(tok)) return tok;
+  }
+  // All tokens are particles? Fall back to the last token verbatim.
+  return tokens[tokens.length - 1];
+}
+
+// ─── Emails ─────────────────────────────────────────────────────────────────
+
+const emailSchema = z.string().email();
+
+/**
+ * Normalize a free-text email. Trims + lowercases. Returns null for empty
+ * or malformed input — caller decides whether to flag, store, or drop.
+ *
+ * Plus-aliases (`user+tag@domain.com`) are NOT stripped: they're real
+ * distinct addresses, and stripping them would auto-merge legitimately
+ * separate accounts.
+ */
+export function normalizeEmail(raw: string | null | undefined): string | null {
+  if (raw == null) return null;
+  const trimmed = raw.toString().trim().toLowerCase();
+  if (!trimmed) return null;
+  const result = emailSchema.safeParse(trimmed);
+  return result.success ? trimmed : null;
+}
+
+// ─── Phones ─────────────────────────────────────────────────────────────────
+
+export type PhoneFlag = 'multi_number' | 'placeholder' | 'unparseable';
+
+export interface NormalizedPhone {
+  /** Canonical E.164 form, e.g. '+15742740548'. Null when unparseable
+   *  or flagged as placeholder. */
+  e164: string | null;
+  /** ISO-3166-1 alpha-2 of the country the number was parsed against. */
+  country: CountryCode | null;
+  /** Display-friendly international format. Useful for migration reports. */
+  display: string | null;
+  /** Set when the input had a quirk worth surfacing in the migration
+   *  report or runtime audit log. Absent on clean parses. */
+  flagged?: PhoneFlag;
+}
+
+/**
+ * Normalize a raw user-entered phone string for comparison + storage.
+ *
+ * Pipeline:
+ *   1. strip leading apostrophe (spreadsheet copy-paste artifact)
+ *   2. strip \r / \n / \t (real values seen in NocoDB had carriage returns)
+ *   3. detect multi-number fields ("+33611111111;+33622222222",
+ *      "0677580750/0690511494") — flag and take first segment
+ *   4. strip whitespace, dots, dashes, parens, single quotes
+ *   5. convert leading "00" → "+" (international dialling code)
+ *   6. detect placeholder fakes (8+ consecutive zeros) — flag, return null e164
+ *   7. parse via libphonenumber-js
+ *   8. on parse failure or invalid number → flag 'unparseable'
+ *
+ * Returns null for empty inputs (cheaper to short-circuit than to wrap).
+ */
+export function normalizePhone(
+  raw: string | null | undefined,
+  defaultCountry?: CountryCode,
+): NormalizedPhone | null {
+  if (raw == null) return null;
+  let cleaned = raw.toString().trim();
+  if (!cleaned) return null;
+
+  // 1. Spreadsheet apostrophe prefix.
+  if (cleaned.startsWith("'")) cleaned = cleaned.slice(1);
+
+  // 2. Strip carriage returns / newlines / tabs.
+  cleaned = cleaned.replace(/[\r\n\t]/g, '');
+
+  // 3. Multi-number detection — split on /, ;, , (in that order of priority).
+  let flagged: PhoneFlag | undefined;
+  if (/[/;,]/.test(cleaned)) {
+    flagged = 'multi_number';
+    cleaned = cleaned.split(/[/;,]/)[0]!.trim();
+  }
+
+  // 4. Strip whitespace, dots, dashes, parens. Keep + for E.164 prefix.
+  cleaned = cleaned.replace(/[\s.\-()]/g, '');
+  if (!cleaned) return { e164: null, country: null, display: null, flagged: 'unparseable' };
+
+  // 5. 00 international prefix → +.
+  if (cleaned.startsWith('00')) {
+    cleaned = '+' + cleaned.slice(2);
+  }
+
+  // 6. Placeholder fakes — runs of 8+ consecutive zeros, e.g. +447000000000.
+  if (/0{8,}/.test(cleaned)) {
+    return { e164: null, country: null, display: null, flagged: 'placeholder' };
+  }
+
+  // 7. Parse via the existing i18n helper (libphonenumber-js under the hood).
+  const parsed = parsePhone(cleaned, defaultCountry);
+  if (!parsed.e164) {
+    // Couldn't even produce a canonical form — genuinely garbage.
+    return { e164: null, country: null, display: null, flagged: 'unparseable' };
+  }
+
+  // Note: we deliberately don't gate on `parsed.isValid`. The
+  // libphonenumber-js `min` build returns isValid=false for many real
+  // numbers (NANP territories share +1; some country metadata is
+  // truncated). For dedup we only need a canonical E.164 string to
+  // compare; strict validity is the form layer's problem, not ours.
+  // If a string-only test (e.g. \"abc-not-a-phone\") gets here, parse
+  // returns null e164 anyway and the branch above handles it.
+  return {
+    e164: parsed.e164,
+    country: parsed.country,
+    display: parsed.international,
+    flagged,
+  };
+}
+
+// ─── Countries ──────────────────────────────────────────────────────────────
+
+/**
+ * Aliases for canonical country names that don't match
+ * `Intl.DisplayNames(en)` output verbatim. Keys are pre-normalized
+ * (lowercase, diacritic-free, hyphens/dots → spaces, collapsed whitespace).
+ *
+ * Kept opinionated and small — only entries we've actually seen in legacy
+ * data. Adding a new alias is cheap; trying to be exhaustive isn't.
+ */
+const COUNTRY_ALIASES: Record<string, CountryCode> = {
+  // Generic abbreviations
+  usa: 'US',
+  us: 'US',
+  uk: 'GB',
+  // Saint-Barthélemy variants seen in production
+  'saint barthelemy': 'BL',
+  'saint barth': 'BL',
+  'st barth': 'BL',
+  'st barths': 'BL',
+  'st barthelemy': 'BL',
+  // Caribbean short-forms whose canonical Intl names are awkward
+  // ("Antigua and Barbuda", "Saint Vincent and the Grenadines", etc.).
+  antigua: 'AG',
+  barbuda: 'AG',
+  'st kitts': 'KN',
+  'saint kitts': 'KN',
+  nevis: 'KN',
+};
+
+/**
+ * High-frequency cities → country, used as a last-resort fallback when
+ * exact / alias / fuzzy country matching all miss. Keys are normalized.
+ *
+ * Order matters: an entry's key is also matched as a substring of the
+ * input ("Sag Harbor Y" contains "sag harbor"), so the most specific
+ * city appears first to avoid a wrong partial hit.
+ */
+const CITY_TO_COUNTRY: Record<string, CountryCode> = {
+  'kansas city': 'US',
+  'sag harbor': 'US',
+  'new york': 'US',
+  // Cities that came out unresolved from the 2026-05-03 NocoDB dry-run.
+  // Using lowercase (post-normalize keys).
+  boston: 'US',
+  tampa: 'US',
+  'fort lauderdale': 'US',
+  'port jefferson': 'US',
+  nantucket: 'US',
+  // US state abbreviations that often appear standalone or as suffix:
+  ' fl': 'US',
+  ' ma': 'US',
+  ' ny': 'US',
+  ' tx': 'US',
+  ' ca': 'US',
+  // International
+  london: 'GB',
+  paris: 'FR',
+};
+
+export type CountryConfidence = 'exact' | 'fuzzy' | 'city';
+
+export interface ResolvedCountry {
+  iso: CountryCode | null;
+  confidence: CountryConfidence | null;
+}
+
+/**
+ * Map free-text country / region input to an ISO-3166-1 alpha-2 code.
+ *
+ * Lookup order: alias → exact (vs. all locale country names) → city →
+ * fuzzy (Levenshtein ≤ 2). Anything beyond fuzzy returns null and the
+ * migration script flags the row for human review.
+ */
+export function resolveCountry(text: string | null | undefined): ResolvedCountry {
+  if (text == null) return { iso: null, confidence: null };
+  const normalized = normalizeForLookup(text.toString());
+  if (!normalized) return { iso: null, confidence: null };
+
+  // 1. Aliases — covers USA / UK / St Barth and friends.
+  const alias = COUNTRY_ALIASES[normalized];
+  if (alias) return { iso: alias, confidence: 'exact' };
+
+  // 2. Exact match against Intl-derived country names. We compare against
+  //    diacritic-stripped + lowercased canonical names so 'United States'
+  //    and 'united states' both resolve.
+  for (const code of ALL_COUNTRY_CODES) {
+    const cleanName = normalizeForLookup(getCountryName(code, 'en'));
+    if (cleanName === normalized) return { iso: code, confidence: 'exact' };
+  }
+
+  // 3. City → country fallback, exact or substring.
+  const cityExact = CITY_TO_COUNTRY[normalized];
+  if (cityExact) return { iso: cityExact, confidence: 'city' };
+  for (const [city, iso] of Object.entries(CITY_TO_COUNTRY)) {
+    if (normalized.includes(city)) return { iso, confidence: 'city' };
+  }
+
+  // 4. Fuzzy fallback (Levenshtein ≤ 2). Skipped for short inputs because
+  //    a 4-char string like "Mars" sits within distance 2 of multiple
+  //    short country names (Mali, Laos, Iran, …) — false-positive city.
+  if (normalized.length >= 6) {
+    let bestCode: CountryCode | null = null;
+    let bestDistance = Number.POSITIVE_INFINITY;
+    for (const code of ALL_COUNTRY_CODES) {
+      const cleanName = normalizeForLookup(getCountryName(code, 'en'));
+      const d = levenshtein(cleanName, normalized);
+      if (d < bestDistance) {
+        bestDistance = d;
+        bestCode = code;
+        if (d === 0) break;
+      }
+    }
+    if (bestDistance <= 2 && bestCode) {
+      return { iso: bestCode, confidence: 'fuzzy' };
+    }
+  }
+
+  return { iso: null, confidence: null };
+}
+
+/** Lowercase + strip diacritics + replace hyphens/dots with spaces +
+ *  collapse whitespace. Used by both the input and the canonical-name
+ *  side of the country comparison so they meet on the same shape. */
+function normalizeForLookup(s: string): string {
+  return s
+    .normalize('NFD')
+    .replace(/[̀-ͯ]/g, '')
+    .toLowerCase()
+    .replace(/[-.]/g, ' ')
+    .replace(/\s+/g, ' ')
+    .trim();
+}
+
+// ─── Levenshtein ────────────────────────────────────────────────────────────
+
+/**
+ * Standard iterative Levenshtein. Used by the country fuzzy match and by
+ * the dedup algorithm's name-similarity rule. Allocates O(n*m) so callers
+ * shouldn't run it against pathological inputs — the dedup blocking
+ * strategy keeps comparison sets small.
+ *
+ * Exported so the find-matches module can reuse the same implementation
+ * without relying on an external dep.
+ */
+export function levenshtein(a: string, b: string): number {
+  if (a === b) return 0;
+  if (!a) return b.length;
+  if (!b) return a.length;
+
+  const m = a.length;
+  const n = b.length;
+  // Two rolling rows is enough — keeps memory at O(n) instead of O(n*m).
+  let prev = new Array<number>(n + 1);
+  let curr = new Array<number>(n + 1);
+  for (let j = 0; j <= n; j += 1) prev[j] = j;
+
+  for (let i = 1; i <= m; i += 1) {
+    curr[0] = i;
+    for (let j = 1; j <= n; j += 1) {
+      const cost = a[i - 1] === b[j - 1] ? 0 : 1;
+      curr[j] = Math.min(curr[j - 1]! + 1, prev[j]! + 1, prev[j - 1]! + cost);
+    }
+    [prev, curr] = [curr, prev];
+  }
+
+  return prev[n]!;
+}

src/lib/dedup/phone-parse.ts

@@ -0,0 +1,66 @@
+/**
+ * Script-safe phone parser.
+ *
+ * The project's existing `src/lib/i18n/phone.ts` imports from
+ * `libphonenumber-js`, which under Node 25 + tsx loader hits a
+ * metadata-shape interop bug (`{ default }` wrapping the JSON). It
+ * works fine in Next.js + vitest, but a `tsx scripts/...` invocation
+ * blows up.
+ *
+ * This wrapper bypasses the bundled `index.cjs.js` and calls
+ * `libphonenumber-js/core` directly with metadata loaded as raw JSON.
+ * Same surface as the i18n helper; usable from both runtimes.
+ *
+ * Used by the dedup library's `normalizePhone`. The runtime UI still
+ * imports `i18n/phone` directly — no reason to touch a working path.
+ */
+
+// eslint-disable-next-line @typescript-eslint/no-require-imports
+const core: typeof import('libphonenumber-js/core') = require('libphonenumber-js/core');
+// Load the JSON directly. The bundled `index.cjs.js` does the same
+// thing but its `require('../metadata.min.json')` hits a Node 25 ESM
+// interop bug that wraps the JSON in `{ default }`. Importing the
+// JSON file by absolute path through the package root sidesteps it.
+// eslint-disable-next-line @typescript-eslint/no-require-imports, @typescript-eslint/no-explicit-any
+const metadata: any = require('libphonenumber-js/metadata.min.json');
+
+import type { CountryCode } from '@/lib/i18n/countries';
+
+export interface ParsedPhone {
+  e164: string | null;
+  country: CountryCode | null;
+  national: string | null;
+  international: string | null;
+  isValid: boolean;
+}
+
+const EMPTY: ParsedPhone = {
+  e164: null,
+  country: null,
+  national: null,
+  international: null,
+  isValid: false,
+};
+
+export function parsePhoneScriptSafe(raw: string, defaultCountry?: CountryCode): ParsedPhone {
+  const trimmed = raw.trim();
+  if (!trimmed) return EMPTY;
+  try {
+    // The core entry expects its own `CountryCode` type from
+    // libphonenumber-js. Our `CountryCode` type is the same set of ISO
+    // alpha-2 codes (we re-derive from the same Intl source) so this
+    // cast is structural-equivalent, not lossy.
+    // eslint-disable-next-line @typescript-eslint/no-explicit-any
+    const parsed = core.parsePhoneNumberFromString(trimmed, defaultCountry as any, metadata);
+    if (!parsed) return EMPTY;
+    return {
+      e164: parsed.number,
+      country: (parsed.country ?? null) as CountryCode | null,
+      national: parsed.formatNational(),
+      international: parsed.formatInternational(),
+      isValid: parsed.isValid(),
+    };
+  } catch {
+    return EMPTY;
+  }
+}

src/lib/services/client-merge.service.ts

@@ -0,0 +1,393 @@
+/**
+ * Client merge service — atomically combines two client records.
+ *
+ * Used by:
+ *   - /admin/duplicates review queue (when an admin confirms a merge)
+ *   - the at-create suggestion path ("use existing client") — though
+ *     that path uses the lighter `attachInterestToClient` and never
+ *     actually merges two pre-existing clients
+ *   - the migration script's `--apply` (eventually)
+ *
+ * Reversibility: every merge writes a `client_merge_log` row containing
+ * the loser's full pre-merge state. Within the configured undo window
+ * (default 7 days, see `dedup_undo_window_days` in system_settings) a
+ * follow-up `unmergeClients` call can restore the loser and detach
+ * everything that was reattached.
+ *
+ * Design reference: docs/superpowers/specs/2026-05-03-dedup-and-migration-design.md §6.
+ */
+
+import { and, eq, sql } from 'drizzle-orm';
+
+import { db } from '@/lib/db';
+import {
+  clients,
+  clientContacts,
+  clientAddresses,
+  clientNotes,
+  clientTags,
+  clientRelationships,
+  clientMergeLog,
+  clientMergeCandidates,
+} from '@/lib/db/schema/clients';
+import { interests } from '@/lib/db/schema/interests';
+import { berthReservations } from '@/lib/db/schema/reservations';
+import { auditLogs } from '@/lib/db/schema/system';
+
+// ─── Public API ─────────────────────────────────────────────────────────────
+
+export interface MergeFieldChoices {
+  /** Per-field overrides — `winner` keeps the surviving client's value;
+   *  `loser` copies the loser's value over. Fields not listed default
+   *  to `winner` (no change). */
+  fullName?: 'winner' | 'loser';
+  nationalityIso?: 'winner' | 'loser';
+  preferredContactMethod?: 'winner' | 'loser';
+  preferredLanguage?: 'winner' | 'loser';
+  timezone?: 'winner' | 'loser';
+  source?: 'winner' | 'loser';
+  sourceDetails?: 'winner' | 'loser';
+}
+
+export interface MergeOptions {
+  winnerId: string;
+  loserId: string;
+  /** ID of the user performing the merge (for audit + clientMergeLog.mergedBy). */
+  mergedBy: string;
+  /** Per-field choice overrides. Multi-value fields (contacts, addresses,
+   *  notes, tags) are always preserved from both sides; this only
+   *  affects single-value scalar fields on the `clients` row. */
+  fieldChoices?: MergeFieldChoices;
+}
+
+export interface MergeResult {
+  mergeLogId: string;
+  movedRows: {
+    interests: number;
+    contacts: number;
+    addresses: number;
+    notes: number;
+    tags: number;
+    relationships: number;
+    reservations: number;
+  };
+}
+
+/**
+ * Atomically merge `loserId` into `winnerId`. Throws if:
+ *   - either id doesn't exist or belongs to a different port
+ *   - the loser has already been merged (mergedIntoClientId set)
+ *   - the winner is itself archived
+ */
+export async function mergeClients(opts: MergeOptions): Promise<MergeResult> {
+  if (opts.winnerId === opts.loserId) {
+    throw new Error('Cannot merge a client into itself');
+  }
+
+  return await db.transaction(async (tx) => {
+    // ── Lock both rows for the duration. The first FOR UPDATE that
+    //    arrives wins; a concurrent second merge of the same loser
+    //    will see `mergedIntoClientId` set and bail. ──────────────────────
+    const [winnerRow] = await tx
+      .select()
+      .from(clients)
+      .where(eq(clients.id, opts.winnerId))
+      .for('update');
+    const [loserRow] = await tx
+      .select()
+      .from(clients)
+      .where(eq(clients.id, opts.loserId))
+      .for('update');
+
+    if (!winnerRow) throw new Error(`Winner client ${opts.winnerId} not found`);
+    if (!loserRow) throw new Error(`Loser client ${opts.loserId} not found`);
+    if (winnerRow.portId !== loserRow.portId) {
+      throw new Error('Cannot merge clients across different ports');
+    }
+    if (loserRow.mergedIntoClientId) {
+      throw new Error(`Loser ${opts.loserId} already merged into ${loserRow.mergedIntoClientId}`);
+    }
+    if (winnerRow.archivedAt) {
+      throw new Error('Cannot merge into an archived client');
+    }
+
+    // ── Snapshot the loser's full state before any mutation. Used by
+    //    `unmergeClients` to restore within the undo window. ──────────────
+    const loserContacts = await tx
+      .select()
+      .from(clientContacts)
+      .where(eq(clientContacts.clientId, opts.loserId));
+    const loserAddresses = await tx
+      .select()
+      .from(clientAddresses)
+      .where(eq(clientAddresses.clientId, opts.loserId));
+    const loserNotes = await tx
+      .select()
+      .from(clientNotes)
+      .where(eq(clientNotes.clientId, opts.loserId));
+    const loserTags = await tx
+      .select()
+      .from(clientTags)
+      .where(eq(clientTags.clientId, opts.loserId));
+    const loserInterests = await tx
+      .select({ id: interests.id })
+      .from(interests)
+      .where(eq(interests.clientId, opts.loserId));
+    const loserReservations = await tx
+      .select({ id: berthReservations.id })
+      .from(berthReservations)
+      .where(eq(berthReservations.clientId, opts.loserId));
+    const loserRelationshipsAsA = await tx
+      .select()
+      .from(clientRelationships)
+      .where(eq(clientRelationships.clientAId, opts.loserId));
+    const loserRelationshipsAsB = await tx
+      .select()
+      .from(clientRelationships)
+      .where(eq(clientRelationships.clientBId, opts.loserId));
+
+    const snapshot = {
+      loser: loserRow,
+      contacts: loserContacts,
+      addresses: loserAddresses,
+      notes: loserNotes,
+      tags: loserTags,
+      interests: loserInterests.map((r) => r.id),
+      reservations: loserReservations.map((r) => r.id),
+      relationshipsAsA: loserRelationshipsAsA,
+      relationshipsAsB: loserRelationshipsAsB,
+      fieldChoices: opts.fieldChoices ?? {},
+      mergedAt: new Date().toISOString(),
+    };
+
+    // ── Apply field choices on the winner. We only touch fields the
+    //    caller explicitly asked to copy from the loser; everything
+    //    else stays as-is. ────────────────────────────────────────────────
+    const fieldUpdates: Partial<typeof winnerRow> = {};
+    if (opts.fieldChoices?.fullName === 'loser') fieldUpdates.fullName = loserRow.fullName;
+    if (opts.fieldChoices?.nationalityIso === 'loser')
+      fieldUpdates.nationalityIso = loserRow.nationalityIso;
+    if (opts.fieldChoices?.preferredContactMethod === 'loser')
+      fieldUpdates.preferredContactMethod = loserRow.preferredContactMethod;
+    if (opts.fieldChoices?.preferredLanguage === 'loser')
+      fieldUpdates.preferredLanguage = loserRow.preferredLanguage;
+    if (opts.fieldChoices?.timezone === 'loser') fieldUpdates.timezone = loserRow.timezone;
+    if (opts.fieldChoices?.source === 'loser') fieldUpdates.source = loserRow.source;
+    if (opts.fieldChoices?.sourceDetails === 'loser')
+      fieldUpdates.sourceDetails = loserRow.sourceDetails;
+
+    if (Object.keys(fieldUpdates).length > 0) {
+      await tx
+        .update(clients)
+        .set({ ...fieldUpdates, updatedAt: new Date() })
+        .where(eq(clients.id, opts.winnerId));
+    }
+
+    // ── Reattach. Each table that points at the loser via clientId
+    //    gets pointed at the winner instead. ─────────────────────────────
+
+    const movedInterests = (
+      await tx
+        .update(interests)
+        .set({ clientId: opts.winnerId, updatedAt: new Date() })
+        .where(eq(interests.clientId, opts.loserId))
+        .returning({ id: interests.id })
+    ).length;
+
+    const movedReservations = (
+      await tx
+        .update(berthReservations)
+        .set({ clientId: opts.winnerId, updatedAt: new Date() })
+        .where(eq(berthReservations.clientId, opts.loserId))
+        .returning({ id: berthReservations.id })
+    ).length;
+
+    // Contacts: move loser's contacts to winner, but DON'T duplicate any
+    // already-present (channel, value) pair. Loser-only ones get
+    // demoted to non-primary so the winner's primary stays intact.
+    const winnerContacts = await tx
+      .select({ channel: clientContacts.channel, value: clientContacts.value })
+      .from(clientContacts)
+      .where(eq(clientContacts.clientId, opts.winnerId));
+    const winnerContactKeys = new Set(
+      winnerContacts.map((c) => `${c.channel}::${c.value.toLowerCase()}`),
+    );
+
+    let movedContacts = 0;
+    for (const c of loserContacts) {
+      const key = `${c.channel}::${c.value.toLowerCase()}`;
+      if (winnerContactKeys.has(key)) {
+        // Winner already has this contact — drop loser's row (cascade
+        // will clean up when loser is archived). But we keep snapshot
+        // so undo restores it.
+        continue;
+      }
+      await tx
+        .update(clientContacts)
+        .set({ clientId: opts.winnerId, isPrimary: false, updatedAt: new Date() })
+        .where(eq(clientContacts.id, c.id));
+      movedContacts += 1;
+    }
+
+    // Addresses: same shape as contacts, but uniqueness is harder to
+    // detect cleanly (free-text street). Just move them all and let the
+    // user dedupe in the UI later.
+    const movedAddresses = (
+      await tx
+        .update(clientAddresses)
+        .set({ clientId: opts.winnerId, isPrimary: false, updatedAt: new Date() })
+        .where(eq(clientAddresses.clientId, opts.loserId))
+        .returning({ id: clientAddresses.id })
+    ).length;
+
+    const movedNotes = (
+      await tx
+        .update(clientNotes)
+        .set({ clientId: opts.winnerId, updatedAt: new Date() })
+        .where(eq(clientNotes.clientId, opts.loserId))
+        .returning({ id: clientNotes.id })
+    ).length;
+
+    // Tags: copy any loser-only tag to the winner; drop overlap.
+    const winnerTags = await tx
+      .select({ tagId: clientTags.tagId })
+      .from(clientTags)
+      .where(eq(clientTags.clientId, opts.winnerId));
+    const winnerTagSet = new Set(winnerTags.map((t) => t.tagId));
+    let movedTags = 0;
+    for (const t of loserTags) {
+      if (!winnerTagSet.has(t.tagId)) {
+        await tx.insert(clientTags).values({ clientId: opts.winnerId, tagId: t.tagId });
+        movedTags += 1;
+      }
+    }
+    await tx.delete(clientTags).where(eq(clientTags.clientId, opts.loserId));
+
+    // Relationships: rewrite each FK side to point at the winner. Keep
+    // both sides regardless — even if A and B both end up as the same
+    // person, the row is preserved for audit; the UI hides self-loops.
+    const movedRelationships =
+      (
+        await tx
+          .update(clientRelationships)
+          .set({ clientAId: opts.winnerId })
+          .where(eq(clientRelationships.clientAId, opts.loserId))
+          .returning({ id: clientRelationships.id })
+      ).length +
+      (
+        await tx
+          .update(clientRelationships)
+          .set({ clientBId: opts.winnerId })
+          .where(eq(clientRelationships.clientBId, opts.loserId))
+          .returning({ id: clientRelationships.id })
+      ).length;
+
+    // ── Archive the loser. Row stays in DB for the undo window;
+    //    `mergedIntoClientId` is the redirect pointer for any stragglers
+    //    (links / direct queries / saved views). ──────────────────────────
+    await tx
+      .update(clients)
+      .set({
+        archivedAt: new Date(),
+        mergedIntoClientId: opts.winnerId,
+        updatedAt: new Date(),
+      })
+      .where(eq(clients.id, opts.loserId));
+
+    // ── Mark any open merge candidate row for this pair as resolved. ───
+    await tx
+      .update(clientMergeCandidates)
+      .set({
+        status: 'merged',
+        resolvedAt: new Date(),
+        resolvedBy: opts.mergedBy,
+      })
+      .where(
+        and(
+          eq(clientMergeCandidates.portId, winnerRow.portId),
+          // pair stored in canonical order — match either direction
+          sql`(
+            (${clientMergeCandidates.clientAId} = ${opts.winnerId}
+             AND ${clientMergeCandidates.clientBId} = ${opts.loserId})
+            OR
+            (${clientMergeCandidates.clientAId} = ${opts.loserId}
+             AND ${clientMergeCandidates.clientBId} = ${opts.winnerId})
+          )`,
+        ),
+      );
+
+    // ── Write the merge log + audit log. ────────────────────────────────
+    const [logRow] = await tx
+      .insert(clientMergeLog)
+      .values({
+        portId: winnerRow.portId,
+        survivingClientId: opts.winnerId,
+        mergedClientId: opts.loserId,
+        mergedBy: opts.mergedBy,
+        mergeDetails: snapshot,
+      })
+      .returning({ id: clientMergeLog.id });
+
+    await tx.insert(auditLogs).values({
+      portId: winnerRow.portId,
+      userId: opts.mergedBy,
+      entityType: 'client',
+      entityId: opts.winnerId,
+      action: 'merge',
+      newValue: {
+        loserId: opts.loserId,
+        loserName: loserRow.fullName,
+        movedInterests,
+        movedReservations,
+        movedContacts,
+        movedAddresses,
+      },
+    });
+
+    return {
+      mergeLogId: logRow!.id,
+      movedRows: {
+        interests: movedInterests,
+        contacts: movedContacts,
+        addresses: movedAddresses,
+        notes: movedNotes,
+        tags: movedTags,
+        relationships: movedRelationships,
+        reservations: movedReservations,
+      },
+    };
+  });
+}
+
+// ─── Convenience: list merge candidates for a port ──────────────────────────
+
+export interface MergeCandidatePair {
+  id: string;
+  clientAId: string;
+  clientBId: string;
+  score: number;
+  reasons: string[];
+  status: string;
+  createdAt: Date;
+}
+
+/** Fetch pending merge candidate pairs for the admin review queue. */
+export async function listPendingMergeCandidates(portId: string): Promise<MergeCandidatePair[]> {
+  const rows = await db
+    .select()
+    .from(clientMergeCandidates)
+    .where(
+      and(eq(clientMergeCandidates.portId, portId), eq(clientMergeCandidates.status, 'pending')),
+    )
+    .orderBy(sql`${clientMergeCandidates.score} DESC`);
+
+  return rows.map((r) => ({
+    id: r.id,
+    clientAId: r.clientAId,
+    clientBId: r.clientBId,
+    score: r.score,
+    reasons: Array.isArray(r.reasons) ? (r.reasons as string[]) : [],
+    status: r.status,
+    createdAt: r.createdAt,
+  }));
+}