feat(dedup): runtime surfaces — merge service, at-create suggestion, admin queue (P2)

Adds the live dedup pipeline on top of the P1 library + P3 migration
script. The new `client/interest` model now actively prevents duplicate
client records at creation time and gives admins a queue to triage
the borderline pairs the at-create check missed.

Three layers, per design §7:

Layer 1 — At-create suggestion
==============================

`GET /api/v1/clients/match-candidates`
  Accepts free-text email / phone / name from the in-flight client
  form, normalizes them via the dedup library, and returns scored
  matches against the port's live client pool. Filters out
  low-confidence noise (the background scoring queue picks those up
  separately). Strict port scoping; never leaks across tenants.

`<DedupSuggestionPanel>` (`src/components/clients/dedup-suggestion-panel.tsx`)
  Debounced React Query hook. Renders nothing for short inputs or
  no useful match. On a high-confidence match it interrupts visually
  with an amber-tinted card and a "Use this client" primary button.
  Medium confidence falls back to a softer "possible match — check
  before creating" treatment.

`<ClientForm>`
  Renders the panel above the form (create path only — skipped on
  edit). New `onUseExistingClient` callback fires when the user
  picks the existing client; the form closes and the parent decides
  what to do (typically: navigate to that client's detail page or
  open the create-interest dialog pre-filled).

Layer 2 — Merge service
=======================

`mergeClients` (`src/lib/services/client-merge.service.ts`)
  The atomic merge primitive that everything else calls. Single
  transaction. Per §6 of the design:

  - Locks both rows (FOR UPDATE) so concurrent merges of the same
    loser fail with a clear error rather than racing.
  - Snapshots the full loser state (contacts / addresses / notes /
    tags / interest+reservation IDs / relationship rows) into the
    `client_merge_log.merge_details` JSONB column for the eventual
    undo flow.
  - Reattaches every loser-side row to the winner: interests,
    reservations, contacts (skipping duplicates by `(channel, value)`),
    addresses, notes, tags (deduped), relationships.
  - Optional `fieldChoices` — per-scalar overrides letting the user
    keep the loser's value for fullName / nationality / preferences /
    timezone / source.
  - Marks the loser archived with `mergedIntoClientId` set (a redirect
    pointer for stragglers; never hard-deleted within the undo window).
  - Resolves any matching `client_merge_candidates` row to status='merged'.
  - Writes audit log entry.

Schema additions:
  - `clients.merged_into_client_id` (nullable text, indexed) — the
    redirect pointer set on archive.

Tests: 6 cases against a real DB — happy path moves rows + writes log;
self-merge / cross-port / already-merged refused; duplicate-contact
deduped on reattach; fieldChoices copies loser values to winner.

Layer 3 — Admin review queue
============================

`GET /api/v1/admin/duplicates`
  Pending merge candidates (status='pending') for the current port,
  with both client summaries hydrated for side-by-side rendering.
  Skips pairs where one side is already archived/merged.

`POST /api/v1/admin/duplicates/[id]/merge`
  Confirms a candidate. Body picks the winner; the other side
  becomes the loser. Calls into `mergeClients` — the only path that
  writes `client_merge_log`.

`POST /api/v1/admin/duplicates/[id]/dismiss`
  Marks the candidate dismissed. Future scoring runs skip the same
  pair until a score change recreates the row.

`<DuplicatesReviewQueue>` (`/admin/duplicates`)
  Side-by-side card UI for each pending pair. Click a card to pick
  the winner; the other side is automatically the loser. Toolbar:
  "Merge into selected" + "Dismiss". No per-field merge editor in
  this PR — that's a future polish; the simple "pick the better row"
  flow handles ~80% of cases.

Test coverage
=============

11 new integration tests (76 added in this branch total):
  - 6 mergeClients (atomicity, refusal cases, contact dedup,
    fieldChoices)
  - 5 match-candidates API (shape, port scoping, confidence tiers,
    Pattern F false-positive guard)

Full vitest: 926/926 passing (was 858 before the dedup branch).
Lint: clean. tsc: clean for new files (only pre-existing errors in
unrelated `tests/integration/` files remain, same as before this PR).

Out of scope, deferred
======================

- Background scoring cron that populates `client_merge_candidates`
  (the queue is empty until this lands; manual seeding works for
  now via the at-create flow).
- Side-by-side per-field merge editor with checkboxes (the simple
  "pick the winner" UX shipped here covers ~80% of real cases).
- Admin settings UI for tuning the dedup thresholds. Defaults from
  the design (90 / 50) are baked in for now.
- `unmergeClients` (the snapshot is captured in client_merge_log;
  the undo endpoint just hasn't been wired yet).

These are all natural follow-up PRs that don't block shipping the
runtime UX.

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
This commit is contained in:
Matt Ciaccio
2026-05-03 14:59:04 +02:00
parent 18e5c124b0
commit 4bcc7f8be6
17 changed files with 12018 additions and 1 deletions

View File

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

View File

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

View File

@@ -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);
}
}

View File

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

View File

@@ -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);
}
}

View File

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