chore(release): v1.0.21

Cut by rmancinas via the "Cut release" workflow. Pushing the tag triggers build.yml; deploy separately with tag=1.0.21.

.gitea/workflows/deploy-galactus.yml

@@ -17,6 +17,12 @@
 #   3. prisma migrate deploy  forward-only. Prisma has no down-migrations; see
 #                           docs/DEPLOY_AND_MIGRATIONS.md — expand/contract is
 #                           the rule, the backup is the emergency lever.
+#                           Done HERE so the schema moves while the OLD code is
+#                           still serving. The api container ALSO migrates at
+#                           start (docker/api-entrypoint.sh); `migrate deploy`
+#                           is idempotent, so the second run is a no-op and the
+#                           container is what covers a restart that never goes
+#                           through this workflow at all.
 #   4. app (api + web)      the new images.
 #   5. verify               ask the running API what it actually is.
 #
@@ -55,10 +61,10 @@
 #                                     uses until somebody saves them there
 #    These are NOT galactus-specific (no _GALACTUS suffix) — one SES identity
 #    serves every deployment.
-#  - The runner (which lives on cubex) must be able to reach BOTH
-#    galactus:9443 (Portainer) and galactus:3306 (MySQL, for migrate deploy).
-#    If it cannot reach 3306, run the migration by hand from a host that can
-#    and dispatch with skip_migrate=true.
+#  - The runner (which lives on cubex) must be able to reach galactus:9443
+#    (Portainer). It should also reach galactus:3306 for step 3, but that is
+#    no longer load-bearing: dispatch with skip_migrate=true and the api
+#    container applies the migrations itself at start.
 #  - ONE-TIME, on a database that predates migration history (i.e. one built
 #    with `prisma db push`): baseline it before the first run, or step 3 fails
 #    with P3005 "database schema is not empty":
@@ -88,7 +94,7 @@ on:
         required: false
         default: false
       skip_migrate:
-        description: "Skip prisma migrate deploy (use when the runner cannot reach MySQL and you migrated by hand)"
+        description: "Skip the runner-side migrate step (safe: the api container migrates at start)"
         type: boolean
         required: false
         default: false
@@ -237,6 +243,10 @@ jobs:
         run: node deploy/scripts/pre-migrate-backup.mjs
 
       # --- schema, forward-only ---------------------------------------------
+      # Belt to the container's braces: this runs while the OLD code is still
+      # serving, which is the order expand/contract is designed around. The
+      # api container repeats it at start for the paths this step cannot
+      # reach (skip_migrate, a host reboot, a stack re-applied by hand).
       - name: Apply database migrations
         if: ${{ github.event.inputs.skip_migrate != 'true' }}
         env:

.gitea/workflows/deploy.yml

@@ -12,6 +12,7 @@
 # git tag v1.2.3 into image tag 1.2.3. Tag v1.2.3, dispatch 1.2.3.
 #
 # Order: db+minio (full only) -> pre-migrate backup -> prisma migrate deploy ->
+# (the api container also migrates at start; see docker/api-entrypoint.sh)
 # app -> verify the API reports the version you asked for. Rollback = dispatch
 # an older tag; that rolls back CODE only, never the schema, which is why every
 # schema change must be expand/contract. See docs/DEPLOY_AND_MIGRATIONS.md.
@@ -47,9 +48,11 @@
 #      # Database stack (full only)
 #      MYSQL_PASSWORD             app-user password (matches DATABASE_URL)
 #      MYSQL_ROOT_PASSWORD        mysql root password
-#  - the runner must reach BOTH Portainer (9443) and MySQL (3306) — the
-#    migration step connects to the database directly. If it cannot reach 3306,
-#    migrate by hand and dispatch with skip_migrate=true.
+#  - the runner must reach Portainer (9443). It should also reach MySQL (3306)
+#    for the migrate step, but that is no longer load-bearing: dispatch with
+#    skip_migrate=true and the api container applies the migrations itself at
+#    start (docker/api-entrypoint.sh). `migrate deploy` is idempotent, so the
+#    two never conflict.
 #  - ONE-TIME on a database built with `prisma db push` (i.e. every database
 #    that exists today): baseline it before the first run, or the migrate step
 #    fails with P3005 "database schema is not empty":
@@ -79,7 +82,7 @@ on:
         required: false
         default: false
       skip_migrate:
-        description: "Skip prisma migrate deploy (use when the runner cannot reach MySQL and you migrated by hand)"
+        description: "Skip the runner-side migrate step (safe: the api container migrates at start)"
         type: boolean
         required: false
         default: false

apps/api/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@jorgecuadros/api",
-  "version": "1.0.18",
+  "version": "1.0.21",
   "private": true,
   "scripts": {
     "build": "nest build",

apps/api/src/policies/lookup-delete-guard.spec.ts

@@ -0,0 +1,94 @@
+import { BadRequestException } from "@nestjs/common";
+import { PoliciesService } from "./policies.service";
+
+/**
+ * Deleting a lookup row that policies still reference used to succeed and
+ * silently blank the field on every one of them, because both FKs are
+ * `ON DELETE SET NULL` (`0000_init`). That is not a hypothetical: it is how
+ * the `M_EMPR` policy type disappeared from the dev database and left 5
+ * policies with a null `policyTypeId`, found only by querying months later.
+ *
+ * These tests pin the refusal. They drive the service with a stub client
+ * rather than a database because what is being asserted is the guard, not
+ * Prisma — and a test that needed a live MySQL would not run in CI.
+ */
+function serviceWith(counts: {
+  policies?: number;
+  claims?: number;
+}): { service: PoliciesService; deleted: string[] } {
+  const deleted: string[] = [];
+  const prisma = {
+    policy: { count: async () => counts.policies ?? 0 },
+    claim: { count: async () => counts.claims ?? 0 },
+    insuranceProvider: {
+      findUnique: async () => ({ id: "p1", name: "ANA SEGUROS" }),
+      delete: async () => {
+        deleted.push("provider");
+        return { id: "p1" };
+      },
+    },
+    policyType: {
+      findUnique: async () => ({ id: "t1", name: "M_EMPR" }),
+      delete: async () => {
+        deleted.push("policyType");
+        return { id: "t1" };
+      },
+    },
+    adjuster: {
+      findUnique: async () => ({ id: "a1", name: "JUAN PEREZ" }),
+      delete: async () => {
+        deleted.push("adjuster");
+        return { id: "a1" };
+      },
+    },
+  };
+  const storage = {} as never;
+  return {
+    service: new PoliciesService(prisma as never, storage),
+    deleted,
+  };
+}
+
+describe("lookup deletes refuse while the row is in use", () => {
+  it("refuses a policy type that policies still carry, and names the count", () => {
+    const { service, deleted } = serviceWith({ policies: 5 });
+    return service.removePolicyType("t1").then(
+      () => {
+        throw new Error("expected the delete to be refused");
+      },
+      (err: unknown) => {
+        expect(err).toBeInstanceOf(BadRequestException);
+        // The operator has to be told WHICH row and HOW MANY, or the message
+        // is not actionable.
+        expect((err as Error).message).toContain("M_EMPR");
+        expect((err as Error).message).toContain("5");
+        expect(deleted).toEqual([]);
+      },
+    );
+  });
+
+  it("refuses a carrier that policies still carry", async () => {
+    const { service, deleted } = serviceWith({ policies: 738 });
+    await expect(service.removeProvider("p1")).rejects.toBeInstanceOf(
+      BadRequestException,
+    );
+    expect(deleted).toEqual([]);
+  });
+
+  it("refuses an adjuster still assigned to claims", async () => {
+    // Same `ON DELETE SET NULL` trap, on `claims.adjusterId`.
+    const { service, deleted } = serviceWith({ claims: 2 });
+    await expect(service.removeAdjuster("a1")).rejects.toBeInstanceOf(
+      BadRequestException,
+    );
+    expect(deleted).toEqual([]);
+  });
+
+  it("allows the delete once nothing references the row", async () => {
+    const { service, deleted } = serviceWith({ policies: 0, claims: 0 });
+    await service.removePolicyType("t1");
+    await service.removeProvider("p1");
+    await service.removeAdjuster("a1");
+    expect(deleted).toEqual(["policyType", "provider", "adjuster"]);
+  });
+});

apps/api/src/policies/policies.service.ts

@@ -1,4 +1,4 @@
-import { Injectable, NotFoundException } from "@nestjs/common";
+import { BadRequestException, Injectable, NotFoundException } from "@nestjs/common";
 import { randomUUID } from "node:crypto";
 import { Prisma } from "@jorgecuadros/database";
 import { PrismaService } from "../prisma/prisma.service";
@@ -554,7 +554,40 @@ export class PoliciesService {
   updateProvider(id: string, dto: UpdateProviderDto) {
     return this.prisma.insuranceProvider.update({ where: { id }, data: dto });
   }
-  removeProvider(id: string) {
+  /**
+   * Deleting a lookup row that policies still point at is silent data loss.
+   *
+   * Both FKs are `ON DELETE SET NULL` (see `0000_init`), so the delete
+   * succeeds, returns 200, and blanks the field on every policy that used it
+   * — with no error and nothing in the UI to suggest anything happened. That
+   * is how the `M_EMPR` policy type disappeared and left 5 policies with a
+   * null `policyTypeId`, only found later by querying.
+   *
+   * Refusing is the whole fix. There is no "are you sure": the operator
+   * reassigns those policies first, which is work the app cannot do for them
+   * because only they know which type is correct.
+   */
+  private async assertLookupUnused(
+    kind: "provider" | "policyType",
+    id: string,
+  ): Promise<void> {
+    const where = kind === "provider" ? { insuranceProviderId: id } : { policyTypeId: id };
+    const count = await this.prisma.policy.count({ where });
+    if (count === 0) return;
+
+    const label =
+      kind === "provider"
+        ? (await this.prisma.insuranceProvider.findUnique({ where: { id } }))?.name
+        : (await this.prisma.policyType.findUnique({ where: { id } }))?.name;
+    const noun = kind === "provider" ? "La aseguradora" : "El tipo de póliza";
+    throw new BadRequestException(
+      `${noun} «${label ?? id}» está en uso por ${count} póliza(s). ` +
+        "Reasígnelas antes de eliminarlo.",
+    );
+  }
+
+  async removeProvider(id: string) {
+    await this.assertLookupUnused("provider", id);
     return this.prisma.insuranceProvider.delete({ where: { id } });
   }
 
@@ -564,7 +597,8 @@ export class PoliciesService {
   updatePolicyType(id: string, dto: UpdatePolicyTypeDto) {
     return this.prisma.policyType.update({ where: { id }, data: dto });
   }
-  removePolicyType(id: string) {
+  async removePolicyType(id: string) {
+    await this.assertLookupUnused("policyType", id);
     return this.prisma.policyType.delete({ where: { id } });
   }
 
@@ -574,7 +608,17 @@ export class PoliciesService {
   updateAdjuster(id: string, dto: UpdateAdjusterDto) {
     return this.prisma.adjuster.update({ where: { id }, data: dto });
   }
-  removeAdjuster(id: string) {
+  /** Same `ON DELETE SET NULL` trap as the two above, on `claims.adjusterId`:
+   *  deleting a busy adjuster would quietly strip them off their claims. */
+  async removeAdjuster(id: string) {
+    const count = await this.prisma.claim.count({ where: { adjusterId: id } });
+    if (count > 0) {
+      const row = await this.prisma.adjuster.findUnique({ where: { id } });
+      throw new BadRequestException(
+        `El ajustador «${row?.name ?? id}» está asignado a ${count} siniestro(s). ` +
+          "Reasígnelos antes de eliminarlo.",
+      );
+    }
     return this.prisma.adjuster.delete({ where: { id } });
   }
 }

apps/api/src/policy-ocr/name-matcher.spec.ts

@@ -0,0 +1,161 @@
+import {
+  nameTokens,
+  suggestCustomersByName,
+  suggestionNote,
+  type CustomerNameRow,
+} from "./name-matcher";
+
+/**
+ * Every row here is a real name out of the customer book (1536 rows, dev
+ * mirror of production), chosen because it is one of the shapes that breaks
+ * naive matching: surname-first ordering, a middle initial, a Spanish double
+ * surname, a joint account, a missing comma, and the `(SIN NOMBRE)`
+ * placeholder the migration left for customers whose DATGRAL row had no name.
+ */
+const BOOK: CustomerNameRow[] = [
+  { id: "c1", name: "WAGONER, PAMELA" },
+  { id: "c2", name: "MCWILLIAMS, BRIAN MICHAEL" },
+  { id: "c3", name: "MCWILLIAMS, BRIAN" },
+  { id: "c4", name: "WEAKLAND, RICHARD E." },
+  { id: "c5", name: "ESTRADA, JERRY & MARILYN" },
+  { id: "c6", name: "CABALLERO PRIETO, GUILLERMO" },
+  { id: "c7", name: "GREENE STEPHANIE" },
+  { id: "c8", name: "(SIN NOMBRE)" },
+  { id: "c9", name: "MUÑOZ, LUIS ALBERTO" },
+  { id: "c10", name: "SMITH, DANIEL" },
+  { id: "c11", name: "SMITH, JOHN" },
+];
+
+describe("nameTokens", () => {
+  it("makes the two orderings the same set", () => {
+    expect(nameTokens("PAMELA WAGONER").sort()).toEqual(
+      nameTokens("WAGONER, PAMELA").sort(),
+    );
+  });
+
+  it("drops initials, particles and corporate suffixes", () => {
+    expect(nameTokens("WEAKLAND, RICHARD E.")).toEqual(["WEAKLAND", "RICHARD"]);
+    expect(nameTokens("GARCIA DE LA TORRE, ANA")).toEqual(["GARCIA", "TORRE", "ANA"]);
+    expect(nameTokens("CONSTRUCTORA BAJA S.A. DE C.V.")).toEqual([
+      "CONSTRUCTORA",
+      "BAJA",
+    ]);
+  });
+
+  it("folds accents so OCR's MUNOZ reaches the book's MUÑOZ", () => {
+    expect(nameTokens("MUÑOZ")).toEqual(["MUNOZ"]);
+  });
+
+  it("drops the phone number ANA prints against the insured name", () => {
+    // Observed verbatim from the ANA automobile face.
+    expect(nameTokens("MARIA GARCIA Ph.3102001538")).toEqual([
+      "MARIA",
+      "GARCIA",
+      "PH",
+    ]);
+  });
+});
+
+describe("suggestCustomersByName", () => {
+  it("matches the reversed name exactly", () => {
+    const [top] = suggestCustomersByName("PAMELA WAGONER", BOOK);
+    expect(top).toMatchObject({ customerId: "c1", tier: "EXACT", score: 1 });
+  });
+
+  it("treats a printed middle name the book lacks as a partial hit", () => {
+    const hits = suggestCustomersByName("PAMELA DENISE WAGONER", BOOK);
+    expect(hits[0]).toMatchObject({ customerId: "c1", tier: "PARTIAL" });
+    expect(hits[0].score).toBeCloseTo(2 / 3);
+  });
+
+  it("ranks the exact row above the row that merely contains it", () => {
+    // Both MCWILLIAMS rows are reachable from this name; the one that holds
+    // the middle name is the exact set and must come first.
+    const hits = suggestCustomersByName("BRIAN MICHAEL MCWILLIAMS", BOOK);
+    expect(hits.map((h) => h.customerId)).toEqual(["c2", "c3"]);
+    expect(hits[0].tier).toBe("EXACT");
+    expect(hits[1].tier).toBe("PARTIAL");
+  });
+
+  it("reaches a joint account from the one spouse the carrier printed", () => {
+    const hits = suggestCustomersByName("JERRY ESTRADA", BOOK);
+    expect(hits[0]).toMatchObject({ customerId: "c5", tier: "PARTIAL" });
+  });
+
+  it("will not reach a joint account on given names alone", () => {
+    // No surname printed: `JERRY MARILYN` overlaps ESTRADA, JERRY & MARILYN
+    // on two tokens, and matching on that would book a stranger's policy.
+    expect(suggestCustomersByName("JERRY MARILYN", BOOK)).toEqual([]);
+  });
+
+  it("matches a Spanish double surname regardless of where the comma fell", () => {
+    const [top] = suggestCustomersByName("GUILLERMO CABALLERO PRIETO", BOOK);
+    expect(top).toMatchObject({ customerId: "c6", tier: "EXACT" });
+  });
+
+  it("still matches a book row that has no comma", () => {
+    const [top] = suggestCustomersByName("STEPHANIE GREENE", BOOK);
+    expect(top).toMatchObject({ customerId: "c7", tier: "EXACT" });
+  });
+
+  it("never suggests the (SIN NOMBRE) placeholder", () => {
+    expect(suggestCustomersByName("SIN NOMBRE", BOOK)).toEqual([]);
+    expect(suggestCustomersByName("NOMBRE DEL ASEGURADO", BOOK)).toEqual([]);
+  });
+
+  it("returns nothing on a shared surname alone", () => {
+    // 185 surnames are shared by 524 customers; one token is not evidence.
+    expect(suggestCustomersByName("SMITH", BOOK)).toEqual([]);
+  });
+
+  it("returns nothing for a different person with the same surname", () => {
+    expect(suggestCustomersByName("ROBERT SMITH", BOOK)).toEqual([]);
+  });
+
+  it("refuses a page-sized blob", () => {
+    // GMX's especificación has no field labels and the parser has handed its
+    // whole first page over as the insured name.
+    const blob =
+      "ESPECIFICACION DE LA POLIZA DE SEGURO DE RESPONSABILIDAD CIVIL " +
+      "EXPEDIDA A FAVOR DE PAMELA WAGONER CON VIGENCIA DEL 01 DE ENERO";
+    expect(suggestCustomersByName(blob, BOOK)).toEqual([]);
+  });
+
+  it("caps the list", () => {
+    expect(suggestCustomersByName("BRIAN MICHAEL MCWILLIAMS", BOOK, 1)).toHaveLength(1);
+  });
+
+  it("handles a null insured name", () => {
+    expect(suggestCustomersByName(null, BOOK)).toEqual([]);
+  });
+});
+
+describe("suggestionNote", () => {
+  it("says nothing when there is nothing", () => {
+    expect(suggestionNote([])).toBeNull();
+  });
+
+  it("names a single exact hit", () => {
+    expect(suggestionNote(suggestCustomersByName("PAMELA WAGONER", BOOK))).toBe(
+      "posible cliente por nombre: WAGONER, PAMELA",
+    );
+  });
+
+  it("reports a tie rather than picking one", () => {
+    // The book really does hold EMERY, LAURA twice and KIRCHHOFF, CINDY
+    // three times.
+    const dupes: CustomerNameRow[] = [
+      { id: "d1", name: "EMERY, LAURA" },
+      { id: "d2", name: "EMERY, LAURA" },
+    ];
+    expect(suggestionNote(suggestCustomersByName("LAURA EMERY", dupes))).toBe(
+      "2 clientes tienen ese mismo nombre; elija cuál",
+    );
+  });
+
+  it("lists partial hits", () => {
+    expect(suggestionNote(suggestCustomersByName("PAMELA DENISE WAGONER", BOOK))).toBe(
+      "posibles clientes por nombre: WAGONER, PAMELA",
+    );
+  });
+});

apps/api/src/policy-ocr/name-matcher.ts

@@ -0,0 +1,199 @@
+/**
+ * Suggests which existing customer a printed insured name belongs to.
+ *
+ * The office books customers surname-first ("WAGONER, PAMELA") and carriers
+ * print them given-name-first ("PAMELA DENISE WAGONER"), so a string compare
+ * never hits. Comparing *token sets* does, and it is order-insensitive by
+ * construction — which is the whole trick.
+ *
+ * **These are suggestions, never matches.** Nothing here sets
+ * `matchedCustomerId` or `confident`; the review screen offers the ranked
+ * names and a human picks. That line is not caution, it is what the book
+ * measures out to: of 1536 customers, 1487 have a distinct normalized token
+ * set — but loosen the rule to surname + first given name only and 131 of
+ * them (8.5%) collide, because the book holds `MCWILLIAMS, BRIAN MICHAEL`
+ * *and* `MCWILLIAMS, BRIAN`, and `CUADROS, JORGE JR` alongside three
+ * `CUADROS, JORGE H.`. 185 surnames are shared by 524 customers, so a
+ * surname alone carries no information at all.
+ *
+ * The two tiers below are drawn at the two places that measurement puts a
+ * cliff: full token-set equality, where cross-person collisions are
+ * effectively zero, and strict containment, where they are common enough
+ * that the result can only ever be a hint.
+ */
+
+/** A customer row as the matcher needs it — id and the book's name. */
+export interface CustomerNameRow {
+  id: string;
+  name: string;
+}
+
+export type NameMatchTier = "EXACT" | "PARTIAL";
+
+export interface CustomerNameSuggestion {
+  customerId: string;
+  customerName: string;
+  /**
+   * `EXACT` — the two names carry the same tokens, in any order.
+   * `PARTIAL` — one name's tokens are all present in the other's, plus the
+   * surname. A printed middle name the book does not hold, or a joint
+   * account where the carrier named one spouse, both land here.
+   */
+  tier: NameMatchTier;
+  /** Shared tokens over the longer name's token count, 0..1. */
+  score: number;
+}
+
+/**
+ * Words that carry no identity. Spanish particles and the ampersand joining
+ * a couple are noise; the corporate suffixes are dropped so `S.A. DE C.V.`
+ * does not make every company look alike.
+ */
+const NOISE = new Set([
+  "DE", "DEL", "LA", "LAS", "LOS", "Y", "AND", "VDA",
+  "JR", "SR", "II", "III", "IV",
+  "SA", "CV", "SAPI", "SRL", "RL", "SC", "INC", "LLC", "LTD", "CORP", "CO",
+]);
+
+/**
+ * Placeholder rows the migration left behind. Fourteen customers are named
+ * literally `(SIN NOMBRE)`; without this they would be one 14-way tie on
+ * every unreadable name.
+ */
+const PLACEHOLDER = new Set(["SIN NOMBRE", "NOMBRE SIN"]);
+
+/**
+ * A name blob longer than this is not a name. GMX's PVL especificación has
+ * no field labels, and the parser has been seen handing its entire first
+ * page over as `insuredName`; matching that against the book would find
+ * a surname somewhere in the prose and suggest a stranger.
+ */
+const MAX_TOKENS = 8;
+const MAX_CHARS = 80;
+
+/**
+ * Splits a name into comparable tokens.
+ *
+ * Accents go first, and deliberately in both directions: the book holds
+ * `MUÑOZ` where OCR routinely reads `MUNOZ`, and folding both to the same
+ * ASCII makes that a hit rather than a miss.
+ *
+ * Tokens containing digits are dropped outright. ANA's automobile face
+ * prints the phone number hard against the insured name — the parser has
+ * emitted `MARIA GARCIA Ph.3102001538` — and the digits would otherwise
+ * be an extra token forever blocking `EXACT`.
+ *
+ * Single letters are dropped as initials: the book is full of
+ * `WEAKLAND, RICHARD E.`, and a carrier that prints the middle name in
+ * full should still match the row that abbreviates it.
+ */
+export function nameTokens(raw: string): string[] {
+  const cleaned = raw
+    .normalize("NFD")
+    .replace(/[\u0300-\u036f]/g, "")
+    .toUpperCase()
+    .replace(/[^A-Z0-9]+/g, " ")
+    .trim();
+
+  const tokens = cleaned
+    .split(" ")
+    .filter((t) => t.length > 1 && !/\d/.test(t) && !NOISE.has(t));
+
+  return [...new Set(tokens)];
+}
+
+/** The surname tokens — everything before the comma the book writes. */
+function surnameTokens(bookName: string): string[] {
+  const comma = bookName.indexOf(",");
+  // 54 of 1536 rows have no comma at all ("GREENE STEPHANIE",
+  // "FAROOQ VAKIL"), and which half is the surname is unknowable. Requiring
+  // a surname we cannot identify would silently exclude those rows, so they
+  // fall back to requiring nothing beyond the containment rule.
+  if (comma < 0) return [];
+  return nameTokens(bookName.slice(0, comma));
+}
+
+function isPlaceholder(tokens: string[]): boolean {
+  return tokens.length === 0 || PLACEHOLDER.has([...tokens].sort().join(" "));
+}
+
+function containsAll(haystack: Set<string>, needles: string[]): boolean {
+  return needles.every((n) => haystack.has(n));
+}
+
+/**
+ * Ranks the book against one printed name.
+ *
+ * Returns at most `limit` suggestions, `EXACT` before `PARTIAL` and higher
+ * score first. An empty array means the printed name was unusable (too
+ * long, too few real tokens) or nothing in the book came close — both of
+ * which leave the review screen exactly as it is today.
+ */
+export function suggestCustomersByName(
+  printedName: string | null | undefined,
+  customers: CustomerNameRow[],
+  limit = 3,
+): CustomerNameSuggestion[] {
+  if (!printedName || printedName.length > MAX_CHARS) return [];
+
+  const printed = nameTokens(printedName);
+  // One usable token is a surname or a given name on its own, and 34% of the
+  // book shares a surname with someone. Nothing useful can come of it.
+  if (printed.length < 2 || printed.length > MAX_TOKENS) return [];
+
+  const printedSet = new Set(printed);
+  const out: CustomerNameSuggestion[] = [];
+
+  for (const c of customers) {
+    const book = nameTokens(c.name);
+    if (isPlaceholder(book) || book.length < 2) continue;
+
+    const bookSet = new Set(book);
+    const overlap = printed.filter((t) => bookSet.has(t)).length;
+    // Two shared tokens is the floor: one is a bare surname collision.
+    if (overlap < 2) continue;
+
+    const bookInPrinted = containsAll(printedSet, book);
+    const printedInBook = containsAll(bookSet, printed);
+    if (!bookInPrinted && !printedInBook) continue;
+
+    // When the book's name is the shorter one, containment already proves
+    // the surname was printed. When the printed name is shorter — the book
+    // holds a middle name or a second spouse the carrier omitted — the
+    // surname must be there explicitly, or `JERRY MARILYN` would match
+    // `ESTRADA, JERRY & MARILYN` on given names alone.
+    if (!bookInPrinted && !containsAll(printedSet, surnameTokens(c.name))) continue;
+
+    out.push({
+      customerId: c.id,
+      customerName: c.name,
+      tier: bookInPrinted && printedInBook ? "EXACT" : "PARTIAL",
+      score: overlap / Math.max(book.length, printed.length),
+    });
+  }
+
+  out.sort((a, b) => {
+    if (a.tier !== b.tier) return a.tier === "EXACT" ? -1 : 1;
+    if (b.score !== a.score) return b.score - a.score;
+    return a.customerName.localeCompare(b.customerName);
+  });
+
+  return out.slice(0, limit);
+}
+
+/** Review-queue wording for what the suggestions amount to. */
+export function suggestionNote(suggestions: CustomerNameSuggestion[]): string | null {
+  if (suggestions.length === 0) return null;
+
+  const exact = suggestions.filter((s) => s.tier === "EXACT");
+  // More than one exact hit is the duplicate-customer case the book really
+  // has (`EMERY, LAURA` twice, `KIRCHHOFF, CINDY` three times). Saying so is
+  // more useful than naming whichever one sorted first.
+  if (exact.length > 1) {
+    return `${exact.length} clientes tienen ese mismo nombre; elija cuál`;
+  }
+  if (exact.length === 1) {
+    return `posible cliente por nombre: ${exact[0].customerName}`;
+  }
+  return `posibles clientes por nombre: ${suggestions.map((s) => s.customerName).join(", ")}`;
+}

apps/api/src/policy-ocr/parsers/policy-parser.spec.ts

@@ -15,6 +15,11 @@ function page(text: string): OcrPage {
   return { text, words: [], confidence: 0.95 };
 }
 
+/** Coverages keyed by their risk label, so an assertion names the coverage
+ *  it is about instead of an array index that shifts when one is added. */
+const byRisk = (p: ReturnType<typeof parsePolicy>): Record<string, ParsedCoverage> =>
+  Object.fromEntries(p.coverages.map((c) => [c.risk, c]));
+
 describe("detectPolicyProvider", () => {
   it("claims GMX from the brand wordmark on the letterhead", () => {
     expect(
@@ -111,6 +116,13 @@ describe("parsePolicy / GMX", () => {
     expect(p.coverages.length).toBeGreaterThan(10);
   });
 
+  it("names the product MULT for confirm to resolve", () => {
+    // The caratula's own header reads "Multiple Policy / Home". MULT is the
+    // legacy discriminator for that multi-line home policy; INCENDIO is
+    // fire-only and no policy in the book has ever used it.
+    expect(parsePolicy(GMX_FULL).policyTypeName).toBe("MULT");
+  });
+
   it("leaves premium fields null on the certificate page and notes it", () => {
     const p = parsePolicy(GMX_FULL);
     expect(p.netPremium).toBeNull();
@@ -144,4 +156,764 @@ describe("parsePolicy / GMX", () => {
     expect(eqTyped.deductible).toContain("sum insured");
     expect(eqTyped.lossParticipation).toBe("20%");
   });
-});
+});
+
+/**
+ * The second GMX document family: the Spanish PVL "especificación" the office
+ * receives as `…-CondicionesParticulares.pdf`. Verbatim excerpts from
+ * `007_LGS-HGMX_07006957_01_0-CondicionesParticulares.pdf` through
+ * `pdftotext -layout`, indentation included — the column positions and the
+ * blank lines between blocks are what the parser reads, so a cleaned-up
+ * fixture would test nothing.
+ */
+describe("parsePolicy / GMX especificación (PVL Hogar)", () => {
+  const HEADER =
+    "                    ESPECIFICACIÓN QUE SE ADHIERE Y FORMA PARTE INTEGRANTE DE LA PÓLIZA\n" +
+    "                                           07-037-07006957-00000-01\n" +
+    "\n";
+
+  const GMX_ESPEC = page(
+    HEADER +
+      "\n" +
+      "                                            Nombre del asegurado EMMER . KATHLEEN\n" +
+      "\n" +
+      "                                          Tipo Persona Asegurada Propietario\n" +
+      "\n" +
+      "                                               Ubicación del riesgo LOS PELICANOS ESTE NO. 98 Col. LAS GAVIOTAS PLAYAS\n" +
+      "                                                                    DE ROSARITO BAJA CALIFORNIA 22713\n" +
+      "\n" +
+      "                                      Características del Inmueble Casa Tipo constructivo Combinado: Macizo y Madera.\n" +
+      "                                                                   Consta de 2 pisos incluyendo sótanos y planta baja.\n" +
+      "\n" +
+      "                                             -500 mts.cuerpo agua SI\n" +
+      "\n" +
+      "                                              Asegurado Adicional\n" +
+      "\n" +
+      "PVL Hogar - GMX Seguros                                                     Página: 1 de 10\n" +
+      HEADER +
+      "\n" +
+      " SECCIÓN INCENDIO EDIFICIO Y CONTENIDOS\n" +
+      "\n" +
+      " EDIFICIO\n" +
+      "\n" +
+      " Límite Máximo de Responsabilidad:\n" +
+      " $200,000.00 USD\n" +
+      "\n" +
+      " Quedan amparados los muros de contención y bardas, así como puertas y portones, hasta un sublimite de $ 50,000.00 M.N. o su\n" +
+      " equivalente en dólares americanos, o hasta el 10% de la suma asegurada de la sección de Edificio, lo que resulte menor.\n" +
+      "\n" +
+      "\n" +
+      " CONTENIDOS\n" +
+      "\n" +
+      " Límite Máximo de Responsabilidad:\n" +
+      " $20,000.00 USD\n" +
+      "\n" +
+      "    2. Terremoto o erupción volcánica: Sección Edificio EXCLUIDO, Sección Contenidos EXCLUIDO\n" +
+      "\n" +
+      "    3. Fenómenos hidrometeorológicos: Sección Edificio $200,000.00 USD, Sección Contenidos $20,000.00 USD\n" +
+      "\n" +
+      " Riesgos adicionales.\n" +
+      "\n" +
+      " Remoción de escombros\n" +
+      "\n" +
+      "    Límite Máximo de Responsabilidad:\n" +
+      "    Edificio\n" +
+      "    $20,000.00 USD\n" +
+      "    Contenidos\n" +
+      "    $2,000.00 USD\n" +
+      "\n" +
+      " Gastos extraordinarios para casa habitación\n" +
+      "\n" +
+      " En caso de siniestro por los riesgos cubiertos en esta póliza, GMX Seguros pagará la renta de casa o departamento, casa de\n" +
+      " huéspedes u hotel cuando se asegure el inmueble, así como los gastos de mudanza, seguro de transporte del menaje de casa y\n" +
+      " efectuados.\n" +
+      "\n" +
+      "    Límite Máximo de Responsabilidad:\n" +
+      "    $22,000.00 USD\n" +
+      "    Periodo de indemnización: 4 meses.\n" +
+      "\n" +
+      " Bienes a la Intemperie:\n" +
+      "\n" +
+      "\n" +
+      "     5 POR CIENTO SOBRE SUMA ASEGURADA, 20 PORCIENTO DE PARTICIPACIÓN A CARGO DEL ASEGURADO DE TODA\n" +
+      "     Y CADA PÉRDIDA.\n" +
+      "\n" +
+      "\n" +
+      "     Límite Máximo de Responsabilidad: $10,000.00 USD\n" +
+      "\n" +
+      " DEDUCIBLES:\n" +
+      "\n" +
+      " El procedimiento que se seguirá para la aplicación de deducibles en caso de que la póliza cuente con cláusula inflacionaria en todas\n" +
+      " y/o en algunas de sus coberturas será como sigue:\n" +
+      "\n" +
+      " Fenómenos hidrometeorológicos\n" +
+      " Zona: A2\n" +
+      " Deducible\n" +
+      "     Edificio: 1 POR CIENTO SOBRE SUMA ASEGURADA\n" +
+      "     Coaseguro:\n" +
+      "     Zona 1: (INTERIOR) Participación a cargo del asegurado del 10% de toda y cada pérdida.\n" +
+      "     Zona 2: Participación a cargo del asegurado del 10% de toda y cada pérdida.\n" +
+      "\n" +
+      "     Deducible\n" +
+      "     Contenidos: 1 POR CIENTO SOBRE SUMA ASEGURADA\n" +
+      "\n" +
+      " II.- SECCIÓN DIVERSOS MISCELÁNEOS\n" +
+      "\n" +
+      " ROBO DE CONTENIDOS\n" +
+      "\n" +
+      " Límite de Responsabilidad:\n" +
+      " $4,000.00 USD\n" +
+      "\n" +
+      "\n" +
+      " Deducible:\n" +
+      " Sin deducible\n" +
+      "\n" +
+      "\n" +
+      " Sublímites:\n" +
+      " Joyas, artículos de oro y plata, armas, relojes, pieles, piedras preciosas montadas, colecciones, obras de arte y demás que por su\n" +
+      "\n" +
+      "PVL Hogar - GMX Seguros                                                     Página: 7 de 10\n" +
+      HEADER +
+      "\n" +
+      " naturaleza se consideran como objetos de difícil o imposible reposición\n" +
+      "\n" +
+      "\n" +
+      " Límite de Responsabilidad:\n" +
+      "\n" +
+      " $2,000.00 USD\n" +
+      "\n" +
+      "\n" +
+      " Deducible:\n" +
+      " Sin deducible\n" +
+      "\n" +
+      " Las condiciones generales que forman parte de la presente póliza son las identificadas bajo el nombre:\n" +
+      " W_HogarGMX_12.11.2025.pdf\n" +
+      "\n" +
+      "PVL Hogar - GMX Seguros                                                    Página: 10 de 10\n",
+  );
+
+  it("reads a policy number whose groups are not the caratula's widths", () => {
+    // 2-3-8-5-2 here vs 3-3-8-4-2 on the English caratula. Pinning the widths
+    // reads one family and returns null on the other.
+    expect(parsePolicy(GMX_ESPEC).policyNumber).toBe("07-037-07006957-00000-01");
+  });
+
+  it("reads the insured, the risk location across its wrapped line, and the ZIP", () => {
+    const p = parsePolicy(GMX_ESPEC);
+    expect(p.provider).toBe("GMX");
+    expect(p.insuredName).toBe("EMMER . KATHLEEN");
+    expect(p.legalAddress).toBe(
+      "LOS PELICANOS ESTE NO. 98 Col. LAS GAVIOTAS PLAYAS DE ROSARITO BAJA CALIFORNIA 22713",
+    );
+    expect(p.zip).toBe("22713");
+    // The cell is printed but empty on this policy — an empty label must not
+    // capture the next line of the form.
+    expect(p.additionalInsured).toBeNull();
+  });
+
+  it("leaves the fields this document does not carry null, and says so", () => {
+    const p = parsePolicy(GMX_ESPEC);
+    expect(p.policyFrom).toBeNull();
+    expect(p.policyTo).toBeNull();
+    expect(p.policyDate).toBeNull();
+    expect(p.agentName).toBeNull();
+    expect(p.netPremium).toBeNull();
+    expect(p.total).toBeNull();
+    // The note must tell the reviewer to key them in — those three are
+    // captured by hand on this layout — and must say what silently breaks if
+    // the vigencia is left empty.
+    const notes = p.notes.join(" ");
+    expect(notes).toMatch(/no trae vigencia, agente ni prima/i);
+    expect(notes).toMatch(/captúrelos a mano/i);
+    expect(notes).toMatch(/avisos de renovación/i);
+  });
+
+  it("takes the currency from the printed limits, not from the M.N. sublimits", () => {
+    // The body prose quotes sublimits in pesos ("$ 50,000.00 M.N."); every
+    // limit is in USD, and only the limits vote.
+    expect(parsePolicy(GMX_ESPEC).currency).toBe("USD");
+  });
+
+  it("reads each coverage under its own heading", () => {
+    const c = byRisk(parsePolicy(GMX_ESPEC));
+    expect(c.EDIFICIO?.insuredAmount).toBe(200000);
+    expect(c.CONTENIDOS?.insuredAmount).toBe(20000);
+    expect(c["ROBO DE CONTENIDOS"]?.insuredAmount).toBe(4000);
+    expect(c["ROBO DE CONTENIDOS"]?.deductible).toBe("Sin deducible");
+  });
+
+  it("splits a limit printed under Edificio / Contenidos sub-labels", () => {
+    const c = byRisk(parsePolicy(GMX_ESPEC));
+    expect(c["Remoción de escombros — Edificio"]?.insuredAmount).toBe(20000);
+    expect(c["Remoción de escombros — Contenidos"]?.insuredAmount).toBe(2000);
+  });
+
+  it("names a coverage after its heading, not after the wrapped tail of the prose above it", () => {
+    // Walking back from the limit hits "efectuados." — short, and the only
+    // thing separating it from a heading is that it is not preceded by a
+    // blank line.
+    const c = byRisk(parsePolicy(GMX_ESPEC));
+    expect(c["Gastos extraordinarios para casa habitación"]?.insuredAmount).toBe(22000);
+    expect(c["efectuados."]).toBeUndefined();
+  });
+
+  it("reads a limit printed on the label's own line", () => {
+    const c = byRisk(parsePolicy(GMX_ESPEC));
+    expect(c["Bienes a la Intemperie"]?.insuredAmount).toBe(10000);
+  });
+
+  it("reads a deductible stated as a sentence above the limit", () => {
+    const c = byRisk(parsePolicy(GMX_ESPEC));
+    expect(c["Bienes a la Intemperie"]?.deductible).toBe(
+      "5 POR CIENTO SOBRE SUMA ASEGURADA, 20 PORCIENTO DE PARTICIPACIÓN A CARGO DEL ASEGURADO DE TODA Y CADA PÉRDIDA.",
+    );
+  });
+
+  it("never borrows a neighbouring coverage's prose as a deductible", () => {
+    // "…o hasta el 10% de la suma asegurada de la sección de Edificio" is a
+    // sublimit rule for EDIFICIO, printed two paragraphs above CONTENIDOS.
+    const c = byRisk(parsePolicy(GMX_ESPEC));
+    expect(c.CONTENIDOS?.deductible).toBeNull();
+    expect(c.EDIFICIO?.deductible).toBeNull();
+  });
+
+  it("does not read the page-level DEDUCIBLES paragraph as a deductible", () => {
+    const p = parsePolicy(GMX_ESPEC);
+    expect(
+      p.coverages.some((c) => (c.deductible ?? "").includes("cláusula inflacionaria")),
+    ).toBe(false);
+  });
+
+  it("reads a sublimit block as a sublimit OF the coverage above it", () => {
+    // The amount sits after a blank line AND a page break, and the block's
+    // own heading ("Sublímites:") names no risk.
+    const c = byRisk(parsePolicy(GMX_ESPEC));
+    expect(c["ROBO DE CONTENIDOS — sublímite"]?.insuredAmount).toBe(2000);
+  });
+
+  it("records an excluded catastrophic risk as excluded, never as zero", () => {
+    const p = parsePolicy(GMX_ESPEC);
+    const quake = p.coverages.filter((c) => /Terremoto/i.test(c.risk));
+    expect(quake).toHaveLength(2);
+    for (const c of quake) {
+      expect(c.risk).toMatch(/EXCLUIDO/);
+      // A coverage insured for $0 and an excluded coverage are the same
+      // number and very different facts.
+      expect(c.insuredAmount).toBeNull();
+    }
+  });
+
+  it("attaches the hydrometeorological deductible and coinsurance from its own block", () => {
+    const c = byRisk(parsePolicy(GMX_ESPEC));
+    const building = c["Fenómenos hidrometeorológicos — Sección Edificio"];
+    expect(building?.insuredAmount).toBe(200000);
+    expect(building?.deductible).toBe("1 POR CIENTO SOBRE SUMA ASEGURADA");
+    expect(building?.lossParticipation).toBe("10%");
+    expect(c["Fenómenos hidrometeorológicos — Sección Contenidos"]?.insuredAmount).toBe(20000);
+  });
+
+  it("names the same product as the caratula — one policy, two artifacts", () => {
+    expect(parsePolicy(GMX_ESPEC).policyTypeName).toBe("MULT");
+  });
+
+  it("carries the underwriting context the fields have no home for", () => {
+    const notes = parsePolicy(GMX_ESPEC).notes.join(" | ");
+    expect(notes).toMatch(/tipo de persona asegurada: Propietario/);
+    expect(notes).toMatch(/características del inmueble: Casa/);
+    expect(notes).toMatch(/cuerpo de agua/);
+    expect(notes).toMatch(/zona catastrófica declarada: A2/);
+    expect(notes).toMatch(/W_HogarGMX_12\.11\.2025\.pdf/);
+  });
+});
+
+/* ------------------------------------------------------------------ ANA */
+
+/**
+ * Verbatim `pdftotext -layout` output of the PDFs A.N.A.'s portal produced
+ * for three real policies, cut at the end of the risk table (the legal
+ * boilerplate and the repeated AGENT COPY below it are not parsed, and the
+ * repeats are covered by their own test).
+ *
+ * The column padding is load-bearing on the driver's policy, which
+ * distinguishes SUM INSURED from PREMIUM by horizontal position alone — do
+ * not reflow these strings.
+ */
+
+const ANA_AUTO_AMPLIA = page(`A.N.A. COMPAÑIA DE SEGUROS SA DE CV
+LUIS CABRERA #2033 INT. 201, Col. ZONA URBANA RIO TIJUANA
+C.P. 22010 MUNICIPIO DE TIJUANA, BAJA CALIFORNIA
+www.anaseguros.com.mx
+AUTOMOBILE
+                                       ALL CLAIMS MUST BE REPORTED BEFORE LEAVING MEXICO
+U.S. CELL PHONES TRY      + 011-52-55-5322-82-66 MEXICAN CELL PHONES 800-911-911-9                                                   SPECIAL POLICY FOR TOURISTS
+TOLL-FREE FROM THE U.S.A. 888-335-7072           BELIZE CELL PHONES 00-52-55-5322-8266
+WHATSAPP                  + 52-55-80-50-3633
+                                                                                                                            No.           700489651
+                   ISSUED BY:                               DATE ISSUED                                   TERM OF INSURANCE
+                                                                                                                                                                  DAYS
+JORGE HUMBERTO CUADROS                                     DAY MONTH YEAR                                DAY    MONTH    YEAR                        TIME
+BENITO JUAREZ 25 No.50 INT 38 CENTRO
+                                                            04         08      2026 FROM                  07             08            2026          12:01
+                                                                                                                                                                     365
+ROSARITO, BAJA CALIFORNIA 22710
+.                                                70175                                 TO                 07             08            2027          12:01
+         DISCOUNT                   PREMIUM                     POLICY FEE                        TAX                     LOCAL TAX                        TOTAL
+              -                       298.61                        30.00                        26.29                          0.00                       354.90
+
+    INSURED                  RAY DEAN II AND SUSAN ROCKHOLD
+                                                                                                                     LICENSE           P0066762
+    ADDRESS                  10308 DONNA AVE                                                                         EMAIL             PROLABSALE@AOL.COM
+    CITY & STATE             NORTHRIDGE, CA                                                             91326        TELEPHONE 8184453524
+                                                                                                                                                  PAYMENT DEADLINE
+  INSURANCE COMPANY                                                              LIEN HOLDER
+                                                                                                                                                        IMMEDIATE
+
+      ITEM                   YEAR             MAKE                                      BODY                                    SERIAL No.                     PLATES
+VEHICLE                      2017           CHRYSLER                                  PACIFICA                              2C4RC1DG7HR654698                  8BPX206
+TRAILER                                                                                                                             .                             .
+TOWING                                                                                                                              .                             .
+*** VALUE STATED MUST NOT EXCEED MARKET VALUE ***
+***VEHICLES THAT HAVE BEEN ACQUIRED AS SALVAGE, REBUILT, OR HAVE BEEN USED PREVIOUSLY AS A TAXI WILL BE CONSIDERED WITH A REDUCED VALUE OF 35% (thirty-five percent), TAKING
+AS A BASE THE VALUE OF A SIMILAR NORMAL VEHICLE, THAT IS, ONE THAT HAS NOT BEEN ACQUIRED AS SALVAGE AND ITS PREVIOUS USE HAS NOT BEEN AS A TAXI OR REBUILT. IT WILL BE THE
+SOLE OBLIGATION AND RESPONSIBILITY OF THE INSURED TO DECLARATE THIS WHEN ACQUIRING THE POLICY.
+SECTION                                     SPECIFICATION OF RISKS                                                                              LIMIT OF LIABILITY
+            MATERIAL DAMAGE WITH MANDATORY DEDUCTIBLE                                                            COVERED/EXCLUDED VEHICLE                   8,000.00 DLLS.
+    1       DEDUCTIBLE: WITH MINIMUM OF $500.00 ON AUTOS
+                                                                                                                                              TRAILER
+            (SEDANS, COUPES, CONVERTIBLES AND STATION WAGONS)                                                         COVERED
+            AND $500.00 ON ALL OTHERS (PICK UPS, VANS, SUV´s AND MOTOR HOMES).                                                                                  0.00 DLLS.
+
+            TOTAL THEFT WITH MANDATORY DEDUCTIBLE                                                                COVERED/EXCLUDED TOWING
+    2       DEDUCTIBLE: WITH MINIMUM OF $1,000.00 ON AUTOS                                                                                                      0.00 DLLS.
+            (SEDANS, COUPES, CONVERTIBLES AND STATION WAGONS)                                                         COVERED
+            AND $1,000.00 ON ALL OTHERS (PICK UPS, VANS, SUV´s AND MOTOR HOMES).
+            LIABILITY FOR PROPERTY DAMAGE TO THIRD PARTIES
+    3                                                                                                                                               100,000.00         DLLS.
+
+
+            BODILY INJURY LIABILITY                                                               PER                               PER
+    4                                                                                             PERSON         100,000.00         ACCIDENT        200,000.00         DLLS.
+
+            MEDICAL EXPENSES                                                                      PER                               PER
+    5                                                                                             PERSON         5,000.00           ACCIDENT        25,000.00          DLLS.
+
+                                                                                                                 COVERED/EXCLUDED                       PREMIUM
+    6       A.N.A.'s LEGAL AID
+                                                                                                                      COVERED                             40.00
+                                                                                                                 COVERED/EXCLUDED                       PREMIUM
+     7      A.N.A.'s ROADSIDE ASSISTANCE
+                                                                                                                      COVERED                             40.00
+            CATASTROPHIC LIABILITY FOR DEATH OF THIRD                                                                                                   PREMIUM
+     8                                                                                                          EXCLUDED
+            PARTIES                                                                                                                    DLLS.               0.00
+            ELITE OR ELITE PLUS WITH MANDATORY DEDUCTIBLE                                                          COVERED/EXCLUDED
+     9     PARTIAL THEFT (LIMIT 0.00 DLLS.WITH DEDUCTIBLE: 0.00 DLLS. PER EVENT)                                                                           0.00
+           VANDALISM (LIMIT 0.00 DLLS.WITH DEDUCTIBLE: 0.00 DLLS. PER EVENT)                                           EXCLUDED
+
+
+ISSUED ONLINE`);
+
+const ANA_AUTO_RC_DIAS = page(`A.N.A. COMPAÑIA DE SEGUROS SA DE CV
+LUIS CABRERA #2033 INT. 201, Col. ZONA URBANA RIO TIJUANA
+C.P. 22010 MUNICIPIO DE TIJUANA, BAJA CALIFORNIA
+www.anaseguros.com.mx
+AUTOMOBILE
+                                       ALL CLAIMS MUST BE REPORTED BEFORE LEAVING MEXICO
+U.S. CELL PHONES TRY      + 011-52-55-5322-82-66 MEXICAN CELL PHONES 800-911-911-9                                                   SPECIAL POLICY FOR TOURISTS
+TOLL-FREE FROM THE U.S.A. 888-335-7072           BELIZE CELL PHONES 00-52-55-5322-8266
+WHATSAPP                  + 52-55-80-50-3633
+                                                                                                                            No.           700487807
+                   ISSUED BY:                               DATE ISSUED                                   TERM OF INSURANCE
+                                                                                                                                                                         DAYS
+JORGE HUMBERTO CUADROS DIARIA                              DAY MONTH YEAR                                DAY    MONTH    YEAR                            TIME
+BENITO JUAREZ 25 NO50 INT 38 COL CENTRO
+                                                            22         07      2026 FROM                  23             07            2026              12:01
+                                                                                                                                                                            3
+ROSARITO BAJA CALIFORNIA 22710
+(661) 612 12 55                                  70175                                 TO                 26             07            2026              12:01
+         DISCOUNT                   PREMIUM                     POLICY FEE                        TAX                     LOCAL TAX                             TOTAL
+              -                        10.77                        25.00                         2.86                          0.00                              38.63
+
+    INSURED                  STEPHEN RUPAN SHATAFIAN
+                                                                                                                     LICENSE           C1394198
+    ADDRESS                  13181 CROSSROADS PARKWAY NORTH STE 300                                                  EMAIL             sshatafian@lee-associates.com
+
+
+    CITY & STATE             CITY OF INDUSTRY, CA                                                        91746       TELEPHONE 7143221072
+                                                                                                                                                     PAYMENT DEADLINE
+  INSURANCE COMPANY                                                              LIEN HOLDER
+                                                                                                                                                            IMMEDIATE
+
+      ITEM                   YEAR               MAKE                                    BODY                                    SERIAL No.                             PLATES
+VEHICLE                      2022               FORD                                   TRANSIT                              1FBAX2CG3NKA69091                          EC46T99
+TRAILER                                                                                                                             .                                     .
+TOWING                                                                                                                              .                                     .
+*** VALUE STATED MUST NOT EXCEED MARKET VALUE ***
+***VEHICLES THAT HAVE BEEN ACQUIRED AS SALVAGE, REBUILT, OR HAVE BEEN USED PREVIOUSLY AS A TAXI WILL BE CONSIDERED WITH A REDUCED VALUE OF 35% (thirty-five percent), TAKING
+AS A BASE THE VALUE OF A SIMILAR NORMAL VEHICLE, THAT IS, ONE THAT HAS NOT BEEN ACQUIRED AS SALVAGE AND ITS PREVIOUS USE HAS NOT BEEN AS A TAXI OR REBUILT. IT WILL BE THE
+SOLE OBLIGATION AND RESPONSIBILITY OF THE INSURED TO DECLARATE THIS WHEN ACQUIRING THE POLICY.
+SECTION                                     SPECIFICATION OF RISKS                                                                                LIMIT OF LIABILITY
+            MATERIAL DAMAGE WITH MANDATORY DEDUCTIBLE                                                            COVERED/EXCLUDED VEHICLE                              0.00 DLLS.
+    1       DEDUCTIBLE: ON AUTOS (SEDANS, COUPES, CONVERTIBLES AND
+                                                                                                                                               TRAILER
+            STATION WAGONS) AND OTHERS (PICK UPS, VANS,                                                              EXCLUDED
+            SUV´s AND MOTOR HOMES).                                                                                                                                    0.00 DLLS.
+
+            TOTAL THEFT WITH MANDATORY DEDUCTIBLE                                                                COVERED/EXCLUDED TOWING
+    2       DEDUCTIBLE: ON AUTOS (SEDANS, COUPES, CONVERTIBLES AND                                                                                                     0.00 DLLS.
+            STATION WAGONS) AND OTHERS (PICK UPS, VANS,                                                              EXCLUDED
+            SUV´s AND MOTOR HOMES).
+            LIABILITY FOR PROPERTY DAMAGE TO THIRD PARTIES
+    3                                                                                                                                                  100,000.00            DLLS.
+
+
+            BODILY INJURY LIABILITY                                                               PER                               PER
+    4                                                                                             PERSON         100,000.00         ACCIDENT           200,000.00            DLLS.
+
+            MEDICAL EXPENSES                                                                      PER                               PER
+    5                                                                                             PERSON         5,000.00           ACCIDENT           25,000.00             DLLS.
+
+                                                                                                                 COVERED/EXCLUDED                            PREMIUM
+    6       A.N.A.'s LEGAL AID
+                                                                                                                      COVERED                                    2.25
+                                                                                                                 COVERED/EXCLUDED                            PREMIUM
+     7      A.N.A.'s ROADSIDE ASSISTANCE
+                                                                                                                      COVERED                                    2.25
+            CATASTROPHIC LIABILITY FOR DEATH OF THIRD                                                                                                        PREMIUM
+     8                                                                                                           EXCLUDED
+            PARTIES                                                                                                                    DLLS.                     0.00
+            ELITE OR ELITE PLUS WITH MANDATORY DEDUCTIBLE                                                          COVERED/EXCLUDED
+     9     PARTIAL THEFT (LIMIT 0.00 DLLS.WITH DEDUCTIBLE: 0.00 DLLS. PER EVENT)                                                                                 0.00
+           VANDALISM (LIMIT 0.00 DLLS.WITH DEDUCTIBLE: 0.00 DLLS. PER EVENT)                                           EXCLUDED
+
+
+ISSUED ONLINE`);
+
+const ANA_LICENCIA = page(`A.N.A. COMPAÑIA DE SEGUROS SA DE CV
+LUIS CABRERA #2033 INT. 201, Col.4 ZONA URBANA RIO TIJUANA
+C.P. 22010 MUNICIPIO DE TIJUANA, BAJA CALIFORNIA
+www.anaseguros.com.mx
+                                                 DRIVER´S POLICY FOR AUTOMOBILE
+                                       ALL CLAIMS MUST BE REPORTED BEFORE LEAVING MEXICO
+U.S. CELL PHONES TRY      + 011-52-55-5322-82-66 MEXICAN CELL PHONES 800-911-911-9
+                                                                                                                                          SPECIAL POLICY FOR TOURISTS
+TOLL-FREE FROM THE U.S.A. 888-335-7072           BELIZE CELL PHONES 00-52-55-5322-8266
+WHATSAPP                  + 52-55-80-50-3633                                                                                     No.            700489616
+                   ISSUED BY:                              DATE ISSUED & TIME                                 TERM OF INSURANCE
+JORGE HUMBERTO CUADROS
+                                                                                                                                                                          DAYS
+                                                            DAY MONTH YEAR                                   DAY    MONTH    YEAR                           TIME
+BENITO JUAREZ 25 No.50 INT 38 CENTRO                          04         08      2026 FROM                     06             08            2026            12:01
+                                                                                                                                                                            365
+ROSARITO, BAJA CALIFORNIA 22710                                                           TO                   06             08            2027            12:01
+.                                                 70175
+      DISCOUNT                      PREMIUM                      POLICY FEE                           TAX                      LOCAL TAX                         TOTAL
+             -                        142.78                           30.00                         13.82                           0.00                         186.60
+
+   LICENSE           N0017668                                    EMAIL          PWAGONER49@AOL.COM                        TELEPHONE 3102001538
+       POLICY HOLDER
+      1. NAME          :                   PAMELA DENISE WAGONER        Ph.3102001538
+        ADDRESS        :                   49305 HIGHWAY 74 SPC 10, PALM DESERT, CA, 92260,
+        DRIVER LICENSE :                   N0017668
+      2. NAME          :
+        ADDRESS        :
+        DRIVER LICENSE :
+                                             NONE
+      3. NAME          :
+        ADDRESS        :
+        DRIVER LICENSE :
+                                             NONE
+      4. NAME          :
+        ADDRESS        :
+        DRIVER LICENSE :                     NONE
+      5. NAME          :
+        ADDRESS        :
+        DRIVER LICENSE :                     NONE
+     SPECIFICATION OF RISKS                                                                                    SUM INSURED                                     PREMIUM
+     LIABILITY FOR PROPERTY DAMAGE TO THIRD PARTIES                                                     100,000.00 usd.                                      18.70         usd.
+     BODILY INJURY LIABILITY ( EXCLUDING OCCUPANTS OF THE VEHICLE )                                     100,000.00 usd. Per Person
+                                                                                                                                                             54.27         usd.
+                                                                                                        200,000.00 usd. Per Accident
+     CATASTROPHIC LIABILITY FOR DEATH OF THIRD PARTIES                                                 0.00        usd.                                        0.00        usd.
+
+     MEDICAL EXPENSES                                                                                     4,000.00 usd. Per Person
+                                                                                                                                                               9.81        usd.
+                                                                                                         20,000.00 usd. Per Accident
+                                                                                                                      COVERED/EXCLUDED                         PREMIUM
+       LEGAL AID
+                                                                                                                           COVERED                           30.00          usd.
+                                                                                                                      COVERED/EXCLUDED                         PREMIUM
+       AUTOMOBILE ASSISTANCE
+                                                                                                                           COVERED                           30.00          usd.
+
+     The following risks are excluded            Collision, overtuning and glass breakage, fire, total theft and natural disasters, partial theft and vandalism.`);
+
+
+describe("detectPolicyProvider / ANA", () => {
+  it("claims ANA from the letterhead", () => {
+    expect(
+      detectPolicyProvider("A.N.A. COMPAÑIA DE SEGUROS SA DE CV\nwww.anaseguros.com.mx"),
+    ).toBe("ANA");
+  });
+
+  it("does not let GMX's layout rules claim an ANA page", () => {
+    // Both books print "MATERIAL DAMAGE"-ish headings; the brand pass runs
+    // before any layout rule precisely so this can't go the other way.
+    expect(detectPolicyProvider(ANA_AUTO_AMPLIA.text)).toBe("ANA");
+    expect(detectPolicyProvider(ANA_LICENCIA.text)).toBe("ANA");
+  });
+});
+
+describe("parsePolicy / ANA automobile", () => {
+  const p = parsePolicy(ANA_AUTO_AMPLIA);
+
+  it("reads the header band", () => {
+    expect(p.provider).toBe("ANA");
+    expect(p.policyNumber).toBe("700489651");
+    expect(p.insuredName).toBe("RAY DEAN II AND SUSAN ROCKHOLD");
+    expect(p.agentName).toBe("JORGE HUMBERTO CUADROS");
+    expect(p.legalAddress).toBe("10308 DONNA AVE, NORTHRIDGE, CA 91326");
+    expect(p.zip).toBe("91326");
+    expect(p.currency).toBe("USD");
+    expect(p.premiumPayment).toBe("IMMEDIATE");
+  });
+
+  it("reads DD MM YYYY out of the three date column cells", () => {
+    expect(p.policyDate?.toISOString().slice(0, 10)).toBe("2026-08-04");
+    expect(p.policyFrom?.toISOString().slice(0, 10)).toBe("2026-08-07");
+    expect(p.policyTo?.toISOString().slice(0, 10)).toBe("2027-08-07");
+  });
+
+  it("maps the six money cells positionally, not by finding six amounts", () => {
+    // DISCOUNT prints as a bare "-" here. A "take the amounts in order"
+    // reading would shift every value one column left.
+    expect(p.netPremium).toBe(298.61);
+    expect(p.policyFee).toBe(30);
+    expect(p.total).toBe(354.9);
+    expect(p.notes.join(" | ")).toMatch(/impuesto: 26\.29/);
+  });
+
+  it("reads the vehicle by token role, not by column", () => {
+    expect(p.vehicles).toHaveLength(1);
+    expect(p.vehicles[0]).toEqual({
+      item: "VEHICLE",
+      modelYear: "2017",
+      make: "CHRYSLER",
+      bodyType: "PACIFICA",
+      vinNumber: "2C4RC1DG7HR654698",
+      licensePlate: "8BPX206",
+    });
+  });
+
+  it("reads a two-word BODY cell without losing the VIN", () => {
+    // "GENESIS SEDAN" is two tokens where "PACIFICA" is one — the VIN shape
+    // is the anchor, not the token count.
+    const v = parsePolicy(ANA_AUTO_RC_DIAS).vehicles[0];
+    expect(v.make).toBe("FORD");
+    expect(v.vinNumber).toBe("1FBAX2CG3NKA69091");
+    expect(v.licensePlate).toBe("EC46T99");
+  });
+
+  it("skips the empty TRAILER and TOWING slots", () => {
+    // Both print a "." per cell rather than being absent.
+    expect(p.vehicles.map((v) => v.item)).toEqual(["VEHICLE"]);
+  });
+
+  it("records the insured as a named driver with their licence", () => {
+    expect(p.drivers).toHaveLength(1);
+    expect(p.drivers[0].fullName).toBe("RAY DEAN II AND SUSAN ROCKHOLD");
+    expect(p.drivers[0].licenseNumber).toBe("P0066762");
+    expect(p.drivers[0].email).toBe("PROLABSALE@AOL.COM");
+  });
+
+  it("does not read the agent's own street number as the policy number", () => {
+    // "BENITO JUAREZ 25 No.50 INT 38" sits three lines above the No. cell.
+    expect(p.policyNumber).not.toBe("50");
+    expect(p.notes.join(" | ")).not.toMatch(/formas/);
+  });
+
+  it("reads the agent clave without picking up their postal code", () => {
+    // "ROSARITO, BAJA CALIFORNIA 22710" is five digits in the same band.
+    expect(p.notes.join(" | ")).toMatch(/clave de agente: 70175/);
+    expect(p.notes.join(" | ")).not.toMatch(/22710/);
+  });
+
+  it("labels the declared values by their printed item slot", () => {
+    const c = byRisk(p);
+    expect(c["MATERIAL DAMAGE — VEHICLE"]?.insuredAmount).toBe(8000);
+    expect(c["MATERIAL DAMAGE — TRAILER"]?.insuredAmount).toBe(0);
+    expect(c["TOTAL THEFT — TOWING"]?.insuredAmount).toBe(0);
+  });
+
+  it("keeps the deductible sentence out of the value columns", () => {
+    const c = byRisk(p);
+    expect(c["MATERIAL DAMAGE — VEHICLE"]?.deductible).toBe(
+      "WITH MINIMUM OF $500.00 ON AUTOS (SEDANS, COUPES, CONVERTIBLES AND " +
+        "STATION WAGONS) AND $500.00 ON ALL OTHERS (PICK UPS, VANS, SUV´s AND " +
+        "MOTOR HOMES).",
+    );
+  });
+
+  it("does not mistake the $500.00 inside the deductible for a sum insured", () => {
+    // It is the one amount in the block not suffixed "DLLS.".
+    const amounts = p.coverages.map((c) => c.insuredAmount);
+    expect(amounts).not.toContain(500);
+  });
+
+  it("splits the per-person and per-accident limits", () => {
+    const c = byRisk(p);
+    expect(c["BODILY INJURY LIABILITY — POR PERSONA"]?.insuredAmount).toBe(100000);
+    expect(c["BODILY INJURY LIABILITY — POR EVENTO"]?.insuredAmount).toBe(200000);
+    expect(c["MEDICAL EXPENSES — POR PERSONA"]?.insuredAmount).toBe(5000);
+    expect(c["MEDICAL EXPENSES — POR EVENTO"]?.insuredAmount).toBe(25000);
+  });
+
+  it("records an add-on's figure as a premium, never as a sum insured", () => {
+    // $40 is what legal aid COST. As `insuredAmount` it would read on the
+    // review screen as a $40 liability limit.
+    const c = byRisk(p);
+    expect(c["LEGAL AID"]?.premium).toBe(40);
+    expect(c["LEGAL AID"]?.insuredAmount).toBeNull();
+    expect(c["ROADSIDE ASSISTANCE"]?.premium).toBe(40);
+  });
+
+  it("unpacks section 9's parenthesised limit and deductible", () => {
+    const c = byRisk(p);
+    const theft = c["ELITE / ELITE PLUS — PARTIAL THEFT: EXCLUDED"];
+    expect(theft?.insuredAmount).toBe(0);
+    expect(theft?.deductible).toBe("0.00 DLLS. POR EVENTO");
+    expect(c["ELITE / ELITE PLUS — VANDALISM: EXCLUDED"]).toBeDefined();
+  });
+
+  it("emits each coverage once even though the PDF prints the face twice", () => {
+    // The real upload is ORIGINAL + AGENT COPY + receipt + three travel
+    // cards, all concatenated into one string before parsing.
+    const doubled = page(ANA_AUTO_AMPLIA.text + "\n\n" + ANA_AUTO_AMPLIA.text);
+    expect(parsePolicy(doubled).coverages).toHaveLength(p.coverages.length);
+    expect(parsePolicy(doubled).vehicles).toHaveLength(1);
+  });
+});
+
+describe("policy type, as a name for confirm to resolve", () => {
+  it("names ANA's two faces after the legacy tables they belong to", () => {
+    expect(parsePolicy(ANA_AUTO_AMPLIA).policyTypeName).toBe("AUTO");
+    expect(parsePolicy(ANA_AUTO_RC_DIAS).policyTypeName).toBe("AUTO");
+    expect(parsePolicy(ANA_LICENCIA).policyTypeName).toBe("LICENCIAS");
+  });
+
+  it("emits a NAME, never an id — the parser must not need a database", () => {
+    // Anything id-shaped here would mean the parser had reached for the DB.
+    for (const p of [ANA_AUTO_AMPLIA, ANA_AUTO_RC_DIAS, ANA_LICENCIA]) {
+      expect(parsePolicy(p).policyTypeName).toMatch(/^[A-Z_]+$/);
+    }
+  });
+
+  it("leaves the type unnamed when no parser claimed the page", () => {
+    expect(parsePolicy(page("a laundry receipt")).policyTypeName).toBeNull();
+  });
+});
+
+describe("parsePolicy / ANA responsabilidad civil por días", () => {
+  const p = parsePolicy(ANA_AUTO_RC_DIAS);
+
+  it("reads a by-the-day term rather than defaulting to a year", () => {
+    // Left at the schema's 365 default this weekend policy would sit in the
+    // renewals window a year out.
+    expect(p.policyFrom?.toISOString().slice(0, 10)).toBe("2026-07-23");
+    expect(p.policyTo?.toISOString().slice(0, 10)).toBe("2026-07-26");
+    expect(p.coveragePeriodDays).toBe(3);
+  });
+
+  it("reads the clave when the agent's phone occupies the left cell", () => {
+    // The by-the-day products print "(661) 612 12 55" ahead of the clave, so
+    // it is no longer the first thing on its line.
+    expect(p.notes.join(" | ")).toMatch(/clave de agente: 70175/);
+  });
+
+  it("marks the excluded sections as excluded, not as insured for zero", () => {
+    const risks = p.coverages.map((c) => c.risk);
+    expect(risks).toContain("MATERIAL DAMAGE — VEHICLE: EXCLUDED");
+    expect(risks).toContain("TOTAL THEFT — TOWING: EXCLUDED");
+    // The liability sections are what this product actually sells, and they
+    // are NOT excluded.
+    expect(risks).toContain("LIABILITY FOR PROPERTY DAMAGE TO THIRD PARTIES");
+  });
+});
+
+describe("parsePolicy / ANA driver's policy (licencia)", () => {
+  const p = parsePolicy(ANA_LICENCIA);
+
+  it("reads the holder off the numbered POLICY HOLDER list", () => {
+    expect(p.policyNumber).toBe("700489616");
+    expect(p.insuredName).toBe("PAMELA DENISE WAGONER");
+    expect(p.legalAddress).toBe("49305 HIGHWAY 74 SPC 10, PALM DESERT, CA, 92260");
+    expect(p.zip).toBe("92260");
+  });
+
+  it("lists one driver, not one per printed copy of the page", () => {
+    // The face renders three times in the real PDF; an unbounded walk
+    // returns the same person three times, which reads as a three-driver
+    // policy rather than as a parse bug.
+    const tripled = page([ANA_LICENCIA.text, ANA_LICENCIA.text, ANA_LICENCIA.text].join("\n\n"));
+    expect(p.drivers).toHaveLength(1);
+    expect(parsePolicy(tripled).drivers).toHaveLength(1);
+  });
+
+  it("splits the phone off the name even without the printed column gap", () => {
+    // The phone shares the name cell, and the only thing marking it off is
+    // white space — which the OCR seam is free to collapse. Depending on the
+    // gap surviving is what put "PAMELA DENISE WAGONER Ph.3102001538" in the
+    // insured field, where it matched no customer.
+    const collapsed = page(ANA_LICENCIA.text.replace(/ {2,}/g, " "));
+    expect(parsePolicy(collapsed).insuredName).toBe("PAMELA DENISE WAGONER");
+  });
+
+  it("drops the four empty driver slots", () => {
+    // Slots 2-5 print an empty NAME and a bare "NONE" licence.
+    expect(p.drivers.map((d) => d.fullName)).toEqual(["PAMELA DENISE WAGONER"]);
+    expect(p.drivers[0].licenseNumber).toBe("N0017668");
+    expect(p.drivers[0].phone).toBe("3102001538");
+  });
+
+  it("insures no vehicle", () => {
+    expect(p.vehicles).toEqual([]);
+    expect(p.notes.join(" | ")).toMatch(/no ampara un veh[íi]culo determinado/);
+  });
+
+  it("separates the SUM INSURED and PREMIUM columns by position", () => {
+    // Both columns print the same shape ("100,000.00 usd." / "18.70 usd.")
+    // and neither is labelled per row — only the offset tells them apart.
+    const c = byRisk(p);
+    const pd = c["LIABILITY FOR PROPERTY DAMAGE TO THIRD PARTIES"];
+    expect(pd?.insuredAmount).toBe(100000);
+    expect(pd?.premium).toBe(18.7);
+  });
+
+  it("reads the trailing Per Person / Per Accident labels on this layout", () => {
+    // They FOLLOW their amount here and PRECEDE it on the automobile face.
+    const c = byRisk(p);
+    expect(c["BODILY INJURY LIABILITY — POR PERSONA"]?.insuredAmount).toBe(100000);
+    expect(c["BODILY INJURY LIABILITY — POR EVENTO"]?.insuredAmount).toBe(200000);
+    expect(c["MEDICAL EXPENSES — POR PERSONA"]?.insuredAmount).toBe(4000);
+    expect(c["MEDICAL EXPENSES — POR EVENTO"]?.insuredAmount).toBe(20000);
+  });
+
+  it("charges a section's premium once, not once per limit", () => {
+    const c = byRisk(p);
+    expect(c["BODILY INJURY LIABILITY — POR PERSONA"]?.premium).toBe(54.27);
+    expect(c["BODILY INJURY LIABILITY — POR EVENTO"]?.premium).toBeNull();
+  });
+
+  it("handles the section order this layout uses", () => {
+    // CATASTROPHIC LIABILITY prints ABOVE MEDICAL EXPENSES here and below it
+    // on the automobile face; blocks are keyed by where the labels land.
+    const c = byRisk(p);
+    expect(c["CATASTROPHIC LIABILITY FOR DEATH OF THIRD PARTIES"]?.insuredAmount).toBe(0);
+    expect(c["LEGAL AID"]?.premium).toBe(30);
+    expect(c["ROADSIDE ASSISTANCE"]?.premium).toBe(30);
+  });
+
+  it("carries the excluded-risk sentence that defines the product", () => {
+    expect(p.notes.join(" | ")).toMatch(/riesgos excluidos: Collision, overtuning/);
+  });
+});

apps/api/src/policy-ocr/policy-matcher.service.spec.ts

@@ -0,0 +1,100 @@
+import { PolicyMatcherService } from "./policy-matcher.service";
+import type { PrismaService } from "../prisma/prisma.service";
+import type { ParsedPolicy } from "./parsers/policy-parser";
+
+function parsed(over: Partial<ParsedPolicy> = {}): ParsedPolicy {
+  return {
+    provider: "GMX",
+    policyNumber: null,
+    insuredName: null,
+    notes: [],
+    coverages: [],
+    vehicles: [],
+    drivers: [],
+    ...over,
+  } as unknown as ParsedPolicy;
+}
+
+function prismaStub(policies: unknown[], customers: { id: string; name: string }[]) {
+  const findManyPolicy = jest.fn().mockResolvedValue(policies);
+  const findManyCustomer = jest.fn().mockResolvedValue(customers);
+  return {
+    prisma: {
+      policy: { findMany: findManyPolicy },
+      customer: { findMany: findManyCustomer },
+    } as unknown as PrismaService,
+    findManyPolicy,
+    findManyCustomer,
+  };
+}
+
+const BOOK = [
+  { id: "cust-1", name: "WAGONER, PAMELA" },
+  { id: "cust-2", name: "SMITH, JOHN" },
+];
+
+describe("PolicyMatcherService name suggestions", () => {
+  it("suggests a customer when the policy number is new", async () => {
+    const { prisma } = prismaStub([], BOOK);
+    const svc = new PolicyMatcherService(prisma);
+
+    const r = await svc.match(
+      parsed({ policyNumber: "P-999", insuredName: "PAMELA DENISE WAGONER" } as never),
+    );
+
+    expect(r.customerSuggestions).toEqual([
+      expect.objectContaining({ customerId: "cust-1", tier: "PARTIAL" }),
+    ]);
+    // The suggestion is surfaced, never applied.
+    expect(r.customerId).toBeNull();
+    expect(r.confident).toBe(false);
+    expect(r.note).toContain("posibles clientes por nombre: WAGONER, PAMELA");
+  });
+
+  it("suggests when the policy number could not be read at all", async () => {
+    const { prisma } = prismaStub([], BOOK);
+    const svc = new PolicyMatcherService(prisma);
+
+    const r = await svc.match(parsed({ insuredName: "PAMELA WAGONER" } as never));
+
+    expect(r.customerSuggestions[0]).toMatchObject({ customerId: "cust-1", tier: "EXACT" });
+    expect(r.customerId).toBeNull();
+    expect(r.note).toBe(
+      "no se pudo leer el número de póliza; posible cliente por nombre: WAGONER, PAMELA",
+    );
+  });
+
+  it("does not touch the book when the policy number hits", async () => {
+    const { prisma, findManyCustomer } = prismaStub(
+      [
+        {
+          id: "pol-1",
+          policyNumber: "P-1",
+          customerId: "cust-2",
+          customer: { name: "SMITH, JOHN" },
+        },
+      ],
+      BOOK,
+    );
+    const svc = new PolicyMatcherService(prisma);
+
+    const r = await svc.match(
+      parsed({ policyNumber: "P-1", insuredName: "PAMELA WAGONER" } as never),
+    );
+
+    expect(r.confident).toBe(true);
+    expect(r.customerId).toBe("cust-2");
+    expect(r.customerSuggestions).toEqual([]);
+    expect(findManyCustomer).not.toHaveBeenCalled();
+  });
+
+  it("reads the customer book once across a batch", async () => {
+    const { prisma, findManyCustomer } = prismaStub([], BOOK);
+    const svc = new PolicyMatcherService(prisma);
+
+    await svc.match(parsed({ policyNumber: "A", insuredName: "PAMELA WAGONER" } as never));
+    await svc.match(parsed({ policyNumber: "B", insuredName: "JOHN SMITH" } as never));
+
+    expect(findManyCustomer).toHaveBeenCalledTimes(1);
+  });
+});

apps/api/src/policy-ocr/policy-matcher.service.ts

@@ -1,6 +1,12 @@
 import { Injectable } from "@nestjs/common";
 import { PrismaService } from "../prisma/prisma.service";
 import type { ParsedPolicy } from "./parsers/policy-parser";
+import {
+  suggestCustomersByName,
+  suggestionNote,
+  type CustomerNameRow,
+  type CustomerNameSuggestion,
+} from "./name-matcher";
 
 export interface MatchResult {
   policyId: string | null;
@@ -14,8 +20,24 @@ export interface MatchResult {
    * the policy number is shared across customers and a human must pick.
    */
   candidates: { policyId: string; customerId: string; customerName: string; policyNumber: string }[];
+  /**
+   * Customers whose name resembles the printed insured name. Populated only
+   * when the policy number resolved to nothing, and never used to set
+   * `customerId` or `confident` — see the class comment.
+   */
+  customerSuggestions: CustomerNameSuggestion[];
 }
 
+/**
+ * How long the customer book is reused across documents in a batch.
+ *
+ * A twenty-page batch would otherwise read all 1536 rows twenty times. The
+ * only cost of the staleness is that a customer created in the last minute
+ * is not suggested — the picker still finds them, so nothing is lost that a
+ * reviewer cannot do in one click.
+ */
+const BOOK_TTL_MS = 60_000;
+
 /**
  * Resolves a parsed policy page to an existing Policy (and its customer) the
  * office already holds.
@@ -33,14 +55,29 @@ export interface MatchResult {
  * policy numbers across customers do occur (same group policy bound by two
  * related parties), and picking one arbitrarily would silently book the
  * wrong coverage.
+ *
+ * On that zero-hit path only, the printed name is used to *rank the picker*
+ * — see `name-matcher.ts`. That is not a walk-back of the rule above: the
+ * suggestion never reaches `customerId` or `confident`, a human still picks,
+ * and the ranking exists because the office writes names surname-first
+ * ("WAGONER, PAMELA") while carriers print them given-name-first ("PAMELA
+ * DENISE WAGONER"), so the reviewer is retyping a name the machine could
+ * have offered.
  */
 @Injectable()
 export class PolicyMatcherService {
+  private book: { rows: CustomerNameRow[]; loadedAt: number } | null = null;
+
   constructor(private readonly prisma: PrismaService) {}
 
   async match(parsed: ParsedPolicy): Promise<MatchResult> {
     if (!parsed.policyNumber) {
-      return this.unmatched("no se pudo leer el número de póliza");
+      // No number to search on, so the page goes to review with a picker —
+      // the same place the name suggestions help.
+      return this.unmatched(
+        "no se pudo leer el número de póliza",
+        await this.suggestByName(parsed.insuredName),
+      );
     }
 
     const rows = await this.prisma.policy.findMany({
@@ -61,22 +98,33 @@ export class PolicyMatcherService {
     }));
 
     if (rows.length === 0) {
+      const suggestions = await this.suggestByName(parsed.insuredName);
+      const hint = suggestionNote(suggestions);
       return {
         policyId: null,
         customerId: null,
-        note: `no se encontró ninguna póliza con el número ${parsed.policyNumber}`,
+        note: [
+          `no se encontró ninguna póliza con el número ${parsed.policyNumber}`,
+          hint,
+        ]
+          .filter(Boolean)
+          .join("; "),
         confident: false,
         candidates: [],
+        customerSuggestions: suggestions,
       };
     }
 
     if (rows.length > 1) {
+      // The policy number did find rows; the reviewer picks among those, and
+      // adding name guesses on top would only add noise.
       return {
         policyId: null,
         customerId: null,
         note: `${rows.length} pólizas comparten el número ${parsed.policyNumber}`,
         confident: false,
         candidates,
+        customerSuggestions: [],
       };
     }
 
@@ -86,16 +134,47 @@ export class PolicyMatcherService {
       note: `coincidencia exacta por número de póliza ${parsed.policyNumber}`,
       confident: true,
       candidates,
+      customerSuggestions: [],
     };
   }
 
-  private unmatched(note: string): MatchResult {
+  private async suggestByName(
+    insuredName: string | null | undefined,
+  ): Promise<CustomerNameSuggestion[]> {
+    if (!insuredName) return [];
+    return suggestCustomersByName(insuredName, await this.customerBook());
+  }
+
+  /**
+   * The whole customer book, held briefly. 1536 rows of `{id, name}` is a
+   * few hundred kilobytes and the comparison is pure token-set work, so
+   * scanning it beats any SQL approximation — and a `LIKE` search would in
+   * any case have to guess which token is the surname, which is the one
+   * thing the office's own data does not agree on.
+   */
+  private async customerBook(): Promise<CustomerNameRow[]> {
+    if (this.book && Date.now() - this.book.loadedAt < BOOK_TTL_MS) {
+      return this.book.rows;
+    }
+    const rows = await this.prisma.customer.findMany({
+      select: { id: true, name: true },
+    });
+    this.book = { rows, loadedAt: Date.now() };
+    return rows;
+  }
+
+  private unmatched(
+    note: string,
+    customerSuggestions: CustomerNameSuggestion[] = [],
+  ): MatchResult {
+    const hint = suggestionNote(customerSuggestions);
     return {
       policyId: null,
       customerId: null,
-      note,
+      note: [note, hint].filter(Boolean).join("; "),
       confident: false,
       candidates: [],
+      customerSuggestions,
     };
   }
 }

apps/api/src/policy-ocr/policy-ocr.dto.ts

@@ -3,10 +3,13 @@ import {
   IsArray,
   IsDateString,
   IsEnum,
+  IsInt,
   IsNumber,
   IsObject,
   IsOptional,
   IsString,
+  Max,
+  Min,
   MinLength,
   ValidateNested,
 } from "class-validator";
@@ -19,6 +22,11 @@ export class ConfirmPolicyDocumentDto {
 
   /** Required when creating a new Policy; ignored if `policyId` is set. */
   @IsOptional() @IsString() customerId?: string;
+  /** Reviewer's explicit lookup picks. Both beat the parsed name; omitted,
+   *  the service resolves `policy_types` / `insurance_providers` by name and
+   *  leaves the FK null when there is no such row. */
+  @IsOptional() @IsString() policyTypeId?: string;
+  @IsOptional() @IsString() insuranceProviderId?: string;
   /** Set when the document matched an existing Policy. */
   @IsOptional() @IsString() policyId?: string;
 
@@ -37,6 +45,9 @@ export class ConfirmPolicyDocumentDto {
   @IsOptional() @IsNumber() brokerFee?: number;
   @IsOptional() @IsNumber() total?: number;
   @IsOptional() @IsString() premiumPayment?: string;
+  /** Printed term in days. Omitted leaves the parsed value (or the schema's
+   *  365 default) in place; ANA sells 3- and 4-day tourist policies. */
+  @IsOptional() @IsInt() @Min(1) @Max(3660) coveragePeriodDays?: number;
   /** Coverages parsed off the PDF, passed through verbatim to Policy.coveragesJson. */
   @IsOptional() @IsObject() coveragesJson?: unknown;
 
@@ -70,6 +81,7 @@ export class ReviewPolicyDocumentDto {
   @IsOptional() @IsNumber() brokerFee?: number;
   @IsOptional() @IsNumber() total?: number;
   @IsOptional() @IsString() premiumPayment?: string;
+  @IsOptional() @IsInt() @Min(1) @Max(3660) coveragePeriodDays?: number;
   @IsOptional() @IsObject() coveragesJson?: unknown;
 
   /** Set by the reviewer when the document matched an existing Policy. */

apps/api/src/policy-ocr/policy-ocr.service.ts

@@ -10,7 +10,7 @@ import { PrismaService } from "../prisma/prisma.service";
 import { StorageService } from "../storage/storage.service";
 import type { UploadedFileLike } from "../storage/upload-file";
 import { OCR_PROVIDER, type OcrPage, type OcrProvider } from "../statements/ocr/ocr.provider";
-import { parsePolicy } from "./parsers/policy-parser";
+import { parsePolicy, type ParsedDriver, type ParsedVehicle } from "./parsers/policy-parser";
 import { PolicyMatcherService } from "./policy-matcher.service";
 import type {
   ConfirmPolicyBatchDto,
@@ -69,8 +69,11 @@ export class PolicyOcrService {
       );
     }
 
+    // The provider is not asked of the uploader and not assumed: `process`
+    // sets it from what the parsers actually claimed, so the batch label can
+    // never contradict its own documents. Until then it says so.
     const batch = await this.prisma.policyOcrBatch.create({
-      data: { provider: "GMX", uploadedById, label, fileCount: files.length },
+      data: { provider: "por detectar", uploadedById, label, fileCount: files.length },
     });
 
     const copies = files.map((f) => ({ buffer: f.buffer, name: f.originalname }));
@@ -112,6 +115,7 @@ export class PolicyOcrService {
 
     let fileOrdinal = 0;
     let globalPageOrdinal = 0;
+    const providersSeen = new Set<string>();
     for (const file of files) {
       fileOrdinal += 1;
       const sourceKey = `policy-ocr/${batchId}/source-${fileOrdinal}.pdf`;
@@ -155,6 +159,7 @@ export class PolicyOcrService {
         if (parsed.provider === "") {
           throw new Error("no se reconoció el proveedor");
         }
+        providersSeen.add(parsed.provider);
         const match = await this.matcher.match(parsed);
         const notes = [...parsed.notes, match.note].filter(Boolean);
         // Confident when exactly one Policy carries the printed number —
@@ -193,12 +198,23 @@ export class PolicyOcrService {
               ? (parsed.coverages as unknown as Prisma.InputJsonValue)
               : Prisma.DbNull,
             extractedPremiumPayment: parsed.premiumPayment,
+            extractedCoveragePeriodDays: parsed.coveragePeriodDays,
+            extractedVehiclesJson: parsed.vehicles.length
+              ? (parsed.vehicles as unknown as Prisma.InputJsonValue)
+              : Prisma.DbNull,
+            extractedDriversJson: parsed.drivers.length
+              ? (parsed.drivers as unknown as Prisma.InputJsonValue)
+              : Prisma.DbNull,
+            extractedPolicyTypeName: parsed.policyTypeName,
             matchedPolicyId: match.policyId,
             matchedCustomerId: match.customerId,
             matchCandidates: match.candidates.length
               ? (match.candidates as unknown as Prisma.InputJsonValue)
               : Prisma.DbNull,
-            matchNote: notes.join("; ").slice(0, 190),
+            customerSuggestions: match.customerSuggestions.length
+              ? (match.customerSuggestions as unknown as Prisma.InputJsonValue)
+              : Prisma.DbNull,
+            matchNote: notes.join("; "),
           },
         });
       } catch (err) {
@@ -211,7 +227,7 @@ export class PolicyOcrService {
             pageNumber: fileOrdinal,
             storageKey: sourceKey,
             status: "OCR_FAILED",
-            matchNote: (err as Error).message.slice(0, 190),
+            matchNote: (err as Error).message,
           },
         });
       }
@@ -219,7 +235,13 @@ export class PolicyOcrService {
 
     await this.prisma.policyOcrBatch.update({
       where: { id: batchId },
-      data: { status: "READY_FOR_REVIEW" },
+      data: {
+        status: "READY_FOR_REVIEW",
+        // Whatever the parsers claimed. A mixed upload is labelled as mixed
+        // rather than as whichever provider happened to come first — the
+        // review header is the only place staff see what they dropped in.
+        provider: [...providersSeen].sort().join(" + ") || "desconocido",
+      },
     });
   }
 
@@ -349,6 +371,7 @@ export class PolicyOcrService {
           ? (dto.coveragesJson as Prisma.InputJsonValue)
           : undefined,
         extractedPremiumPayment: dto.premiumPayment ?? undefined,
+        extractedCoveragePeriodDays: dto.coveragePeriodDays ?? undefined,
         matchedPolicyId,
         matchedCustomerId,
         status: dto.forceConfirm ? "CONFIRMED" : "MATCHED",
@@ -447,14 +470,18 @@ export class PolicyOcrService {
         );
       }
 
-      // 1. Resolve target Policy (create or update). Field selection: every
+      // 1. Resolve the lookup rows the parser can only name. The reviewer's
+      // explicit pick always wins; the parsed name is the fallback.
+      const lookups = await this.resolveLookups(item, doc);
+
+      // 2. Resolve target Policy (create or update). Field selection: every
       // non-null `extracted*` on the doc (post-review) is written. Null is
       // preserved — never overwrite an existing Policy's `netPremium` with
       // null because the certificate page didn't carry one.
       let policyId = item.policyId ?? null;
 
       if (policyId) {
-        const updateData = buildPolicyUpdateFromDoc(item, doc);
+        const updateData = buildPolicyUpdateFromDoc(item, doc, lookups);
         await this.prisma.policy.update({
           where: { id: policyId },
           data: updateData,
@@ -467,21 +494,25 @@ export class PolicyOcrService {
             `Documento página ${doc.pageNumber}: falta número de póliza.`,
           );
         }
-        const createData = buildPolicyCreateFromDoc(item, doc, item.customerId!);
+        const createData = buildPolicyCreateFromDoc(item, doc, item.customerId!, lookups);
         const created = await this.prisma.policy.create({
           data: createData,
         });
         policyId = created.id;
       }
 
-      // 2. Attach the source PDF as a PolicyDocument. `doc.storageKey`
+      // 3. Vehicles and named drivers, for the providers whose face carries
+      // them (ANA's automobile and driver's policies; never GMX Hogar).
+      await this.applyVehiclesAndDrivers(doc, policyId);
+
+      // 4. Attach the source PDF as a PolicyDocument. `doc.storageKey`
       // already points at the exact upload (`policy-ocr/{batchId}/source-N.pdf`)
       // so the attach is just a stream copy into the policy's namespace —
       // the previous per-page "which file did this page come from" walk is
       // gone because one PDF = one doc now.
-      await this.attachSourcePdf(doc.storageKey, policyId);
+      await this.attachSourcePdf(doc.storageKey, policyId, doc.provider);
 
-      // 3. Optionally post the premium to the ledger. Only when staff
+      // 5. Optionally post the premium to the ledger. Only when staff
       // explicitly asked (`postPremium` true) and netPremium parses — without
       // that gate a missing premium would silently book $0.
       let postedTransactionId: string | null = null;
@@ -539,13 +570,149 @@ export class PolicyOcrService {
     };
   }
 
+  /**
+   * Turn the two things the parser can only NAME into foreign keys.
+   *
+   * The parser is a pure function over text and never touches the database,
+   * so it emits `policyTypeName` ("AUTO") and `provider` ("ANA"). Resolving
+   * them here keeps that boundary and means a renamed lookup row is a data
+   * change rather than a parser change.
+   *
+   * **Resolve, never create.** A missing `policy_types` row is a signal that
+   * a human deleted it (that is exactly how M_EMPR disappeared), and silently
+   * recreating it would undo that decision with no record. The field stays
+   * null and the reviewer can add the row through the lookups screen.
+   *
+   * An explicit pick from the reviewer always beats the parsed name.
+   */
+  private async resolveLookups(
+    item: ConfirmPolicyDocumentDto,
+    doc: { extractedPolicyTypeName: string | null; provider: string | null },
+  ): Promise<{ policyTypeId?: string; insuranceProviderId?: string }> {
+    const out: { policyTypeId?: string; insuranceProviderId?: string } = {};
+
+    if (item.policyTypeId) {
+      out.policyTypeId = item.policyTypeId;
+    } else if (doc.extractedPolicyTypeName) {
+      const row = await this.prisma.policyType.findUnique({
+        where: { name: doc.extractedPolicyTypeName },
+        select: { id: true },
+      });
+      if (row) out.policyTypeId = row.id;
+    }
+
+    if (item.insuranceProviderId) {
+      out.insuranceProviderId = item.insuranceProviderId;
+    } else if (doc.provider) {
+      const name = PROVIDER_ROW_NAME[doc.provider] ?? doc.provider;
+      const row = await this.prisma.insuranceProvider.findFirst({
+        where: { name },
+        select: { id: true },
+      });
+      if (row) out.insuranceProviderId = row.id;
+    }
+
+    return out;
+  }
+
+  /**
+   * Write the parsed `Vehicle` and `InsuredDriver` rows onto the policy.
+   *
+   * Both inserts are skipped when an equivalent row is already on the policy.
+   * The reason is `confirmBatch` applying to an EXISTING policy: the office
+   * uploads a renewal for a car already on file, and a blind insert would
+   * leave the customer with the same VIN listed twice with no way to tell
+   * which row the renewal belongs to. Matching is on the identifier the
+   * document actually prints — the VIN for a vehicle (falling back to the
+   * plate, since ANA's TRAILER/TOWING slots have no VIN), the licence number
+   * for a driver (falling back to the name).
+   *
+   * Nothing is ever updated or deleted here. A vehicle whose plate changed
+   * lands as a second row for a human to reconcile, which is the safe half
+   * of the mistake: an over-write would destroy the only record of what was
+   * insured last term.
+   */
+  private async applyVehiclesAndDrivers(
+    doc: { extractedVehiclesJson: Prisma.JsonValue | null; extractedDriversJson: Prisma.JsonValue | null },
+    policyId: string,
+  ): Promise<void> {
+    const vehicles = asArray<ParsedVehicle>(doc.extractedVehiclesJson);
+    const drivers = asArray<ParsedDriver>(doc.extractedDriversJson);
+    if (vehicles.length === 0 && drivers.length === 0) return;
+
+    const policy = await this.prisma.policy.findUnique({
+      where: { id: policyId },
+      select: { customerId: true },
+    });
+    if (!policy) return;
+
+    if (vehicles.length) {
+      const existing = await this.prisma.vehicle.findMany({
+        where: { policyId },
+        select: { vinNumber: true, licensePlate: true },
+      });
+      const seen = new Set(
+        existing.flatMap((v) =>
+          [v.vinNumber, v.licensePlate].filter((k): k is string => !!k).map(norm),
+        ),
+      );
+      for (const v of vehicles) {
+        const key = norm(v.vinNumber ?? v.licensePlate ?? "");
+        if (!key || seen.has(key)) continue;
+        seen.add(key);
+        await this.prisma.vehicle.create({
+          data: {
+            policyId,
+            customerId: policy.customerId,
+            make: v.make,
+            // ANA prints one BODY cell, not separate model/body columns, so
+            // it lands on `bodyType`; `model` stays null rather than being
+            // guessed out of the same string.
+            bodyType: v.bodyType,
+            modelYear: v.modelYear,
+            vinNumber: v.vinNumber,
+            licensePlate: v.licensePlate,
+            // "VEHICLE" / "TRAILER" / "TOWING" — the printed slot, which is
+            // the difference between the insured car and the trailer behind
+            // it and has no column of its own.
+            notes: v.item && v.item !== "VEHICLE" ? v.item : null,
+          },
+        });
+      }
+    }
+
+    if (drivers.length) {
+      const existing = await this.prisma.insuredDriver.findMany({
+        where: { policyId },
+        select: { licenseNumber: true, fullName: true },
+      });
+      const seen = new Set(
+        existing.flatMap((d) =>
+          [d.licenseNumber, d.fullName].filter((k): k is string => !!k).map(norm),
+        ),
+      );
+      for (const d of drivers) {
+        const key = norm(d.licenseNumber ?? d.fullName ?? "");
+        if (!key || seen.has(key)) continue;
+        seen.add(key);
+        await this.prisma.insuredDriver.create({
+          data: { policyId, fullName: d.fullName, licenseNumber: d.licenseNumber },
+        });
+      }
+    }
+  }
+
   /**
    * Stream the source PDF (`sourceKey`, set by `process` on the doc row)
    * into the policy's storage namespace and create a `PolicyDocument`
    * pointer. Trivial now that the doc row holds the exact source key —
    * the old per-page "which file did this page come from" walk is gone.
    */
-  private async attachSourcePdf(sourceKey: string, policyId: string): Promise<void> {
+  private async attachSourcePdf(
+    sourceKey: string,
+    policyId: string,
+    provider: string | null,
+  ): Promise<void> {
     const got = await this.storage.getStream(sourceKey);
     const chunks: Buffer[] = [];
     for await (const c of got.stream) chunks.push(c as Buffer);
@@ -556,7 +723,10 @@ export class PolicyOcrService {
     await this.prisma.policyDocument.create({
       data: {
         policyId,
-        documentType: "GMX_POLICY",
+        // Named after whichever parser claimed the page. Was hardcoded
+        // `GMX_POLICY`, which mislabelled every ANA upload as a GMX
+        // document in the policy's file list.
+        documentType: `${provider ?? "OCR"}_POLICY`,
         storageKey: newKey,
       },
     });
@@ -588,6 +758,27 @@ export class PolicyOcrService {
   }
 }
 
+/**
+ * The parser's provider code is not the carrier's row name in
+ * `insurance_providers`, and the two namespaces are allowed to differ.
+ *
+ * ANA is the case that forces this: the office's book is filed under
+ * "ANA SEGUROS" (738 policies). A bare "ANA" row also existed with 1 policy
+ * and is merged away by `20260815160000_policy_type_repair`, so an exact-name
+ * lookup on the parser's "ANA" would find nothing at all after that migration.
+ *
+ * Anything not listed resolves by its own name.
+ */
+const PROVIDER_ROW_NAME: Record<string, string> = {
+  ANA: "ANA SEGUROS",
+};
+
+/** The lookup FKs resolved for one document, absent when unresolvable. */
+interface ResolvedLookups {
+  policyTypeId?: string;
+  insuranceProviderId?: string;
+}
+
 /** Map a (post-review) doc + final confirmed fields onto a `Policy.update`
  *  payload. Every field that is null in both inputs is omitted so we never
  *  write null over a value the Policy already carries (the GMX certificate
@@ -611,7 +802,9 @@ function buildPolicyUpdateFromDoc(
     extractedTotal: Prisma.Decimal | null;
     extractedCoveragesJson: Prisma.JsonValue | null;
     extractedPremiumPayment: string | null;
+    extractedCoveragePeriodDays: number | null;
   },
+  lookups: ResolvedLookups,
 ): Prisma.PolicyUpdateInput {
   const numOrUndef = (a: number | undefined, b: Prisma.Decimal | null): Prisma.Decimal | undefined => {
     if (a != null) return new Prisma.Decimal(a);
@@ -631,10 +824,23 @@ function buildPolicyUpdateFromDoc(
 
   return {
     policyNumber: strOrUndef(item.policyNumber, doc.extractedPolicyNumber),
+    // `connect` rather than a raw id: this is the CHECKED update input. Left
+    // undefined when unresolved, so an existing Policy never loses a type or
+    // carrier it already had because this document could not name one.
+    policyType: lookups.policyTypeId ? { connect: { id: lookups.policyTypeId } } : undefined,
+    insuranceProvider: lookups.insuranceProviderId
+      ? { connect: { id: lookups.insuranceProviderId } }
+      : undefined,
     agentName: strOrUndef(item.agentName, doc.extractedAgentName),
     policyFrom: dateOrUndef(item.policyFrom, doc.extractedPolicyFrom),
     policyTo: dateOrUndef(item.policyTo, doc.extractedPolicyTo),
     policyDate: dateOrUndef(item.policyDate, doc.extractedPolicyDate),
+    // Left undefined when the document didn't print a term, so the schema
+    // default (365) stands for GMX. ANA's by-the-day policies DO print one,
+    // and the default would otherwise turn a 4-day tourist policy into an
+    // annual one on the renewals screen.
+    coveragePeriodDays:
+      item.coveragePeriodDays ?? doc.extractedCoveragePeriodDays ?? undefined,
     currency: strOrUndef(item.currency, doc.extractedCurrency) as Currency | undefined,
     netPremium: numOrUndef(item.netPremium, doc.extractedNetPremium),
     policyFee: numOrUndef(item.policyFee, doc.extractedPolicyFee),
@@ -683,8 +889,10 @@ function buildPolicyCreateFromDoc(
     extractedTotal: Prisma.Decimal | null;
     extractedCoveragesJson: Prisma.JsonValue | null;
     extractedPremiumPayment: string | null;
+    extractedCoveragePeriodDays: number | null;
   },
   customerId: string,
+  lookups: ResolvedLookups,
 ): Prisma.PolicyUncheckedCreateInput {
   const numOrUndef = (a: number | undefined, b: Prisma.Decimal | null): Prisma.Decimal | undefined => {
     if (a != null) return new Prisma.Decimal(a);
@@ -712,10 +920,18 @@ function buildPolicyCreateFromDoc(
   return {
     policyNumber,
     customerId,
+    policyTypeId: lookups.policyTypeId,
+    insuranceProviderId: lookups.insuranceProviderId,
     agentName: strOrUndef(item.agentName, doc.extractedAgentName),
     policyFrom: dateOrUndef(item.policyFrom, doc.extractedPolicyFrom),
     policyTo: dateOrUndef(item.policyTo, doc.extractedPolicyTo),
     policyDate: dateOrUndef(item.policyDate, doc.extractedPolicyDate),
+    // Left undefined when the document didn't print a term, so the schema
+    // default (365) stands for GMX. ANA's by-the-day policies DO print one,
+    // and the default would otherwise turn a 4-day tourist policy into an
+    // annual one on the renewals screen.
+    coveragePeriodDays:
+      item.coveragePeriodDays ?? doc.extractedCoveragePeriodDays ?? undefined,
     currency: strOrUndef(item.currency, doc.extractedCurrency) as Currency | undefined,
     netPremium: numOrUndef(item.netPremium, doc.extractedNetPremium),
     policyFee: numOrUndef(item.policyFee, doc.extractedPolicyFee),
@@ -764,4 +980,18 @@ function strOrUndefDb(a: string | undefined, b: string | null): string | undefin
   if (a != null && a !== "") return a;
   if (b != null && b !== "") return b;
   return undefined;
+}
+
+/** A JSON column the parser wrote as an array, read back as one. Anything
+ *  else (null, DbNull, a legacy object shape) is an empty list rather than a
+ *  crash — these columns are only ever populated by the parser, so a
+ *  surprise shape means old data, not a caller to reject. */
+function asArray<T>(value: Prisma.JsonValue | null): T[] {
+  return Array.isArray(value) ? (value as unknown as T[]) : [];
+}
+
+/** Compare identifiers the way a person would: case- and space-insensitive.
+ *  VINs and plates are printed inconsistently ("8BPX206" vs "8BPX 206"). */
+function norm(s: string): string {
+  return s.replace(/\s+/g, "").toUpperCase();
 }

apps/api/src/statements/ocr/tesseract.provider.spec.ts

@@ -7,8 +7,13 @@ import { parseBboxLayout } from "./tesseract.provider";
  * that grouping is what left `PERIODO FACTURADO` with no value next to it and
  * every period field empty on a batch whose text was perfectly readable.
  */
+/**
+ * Boxes are sized from the text, at 6 units a character: the reassembler now
+ * reads the space BETWEEN two boxes, so a fixed width would put a fabricated
+ * gap after every short word and every row would come back column-padded.
+ */
 function word(x: number, y: number, text: string): string {
-  return `<word xMin="${x}" yMin="${y}" xMax="${x + 20}" yMax="${y + 8}">${text}</word>`;
+  return `<word xMin="${x}" yMin="${y}" xMax="${x + text.length * 6}" yMax="${y + 8}">${text}</word>`;
 }
 
 function doc(...lines: string[]): string {
@@ -26,23 +31,57 @@ describe("parseBboxLayout", () => {
   it("rejoins a label with the value printed beside it in another flow", () => {
     const [page] = parseBboxLayout(
       doc(
-        word(20, 100, "PERIODO") + word(45, 100, "FACTURADO:"),
+        word(20, 100, "PERIODO") + word(68, 100, "FACTURADO:"),
         word(300, 100.4, "20260630-20260630"),
         padding(),
       ),
       1,
     );
     expect(page).not.toBeNull();
-    expect(page!.text).toContain("PERIODO FACTURADO: 20260630-20260630");
+    expect(page!.text).toMatch(/PERIODO FACTURADO:\s+20260630-20260630/);
   });
 
   it("keeps genuinely separate lines apart", () => {
     const [page] = parseBboxLayout(
-      doc(word(20, 100, "Cuenta:") + word(80, 100, "0900003463"), word(20, 130, "Nombre:"), padding()),
+      doc(word(20, 100, "Cuenta:") + word(68, 100, "0900003463"), word(20, 130, "Nombre:"), padding()),
       1,
     );
-    expect(page!.text.split("\n")).toContain("Cuenta: 0900003463");
-    expect(page!.text.split("\n")).toContain("Nombre:");
+    const lines = page!.text.split("\n").map((l) => l.trim());
+    expect(lines).toContain("Cuenta: 0900003463");
+    expect(lines).toContain("Nombre:");
+  });
+
+  /**
+   * The layout is data. A borderless table separates its cells with nothing
+   * but white space, so the parsers read a run of spaces as a cell boundary
+   * (`INSURED\s{2,}`) and a column offset as a column (`SUM INSURED` vs
+   * `PREMIUM`). Both regressed to nothing when this collapsed every gap to a
+   * single space, and the fixtures — taken from `pdftotext -layout`, which
+   * prints the gaps — could not see it.
+   */
+  it("preserves the gap between two cells of a borderless table", () => {
+    const [page] = parseBboxLayout(
+      doc(word(20, 100, "INSURED") + word(300, 100, "PAMELA") + word(340, 100, "WAGONER"), padding()),
+      1,
+    );
+    const line = page!.text.split("\n").find((l) => l.includes("INSURED"))!;
+    expect(line).toMatch(/INSURED\s{2,}PAMELA WAGONER/);
+  });
+
+  it("preserves the blank line between two blocks", () => {
+    const [page] = parseBboxLayout(
+      doc(word(20, 100, "Insured"), word(20, 112, "wraps"), word(20, 200, "Next"), padding()),
+      1,
+    );
+    const lines = page!.text.split("\n").map((l) => l.trim());
+    // The wrapped continuation stays attached; the next block is cut off from
+    // it, which is what stops a "join until the cell ends" walk running away.
+    expect(lines.slice(lines.indexOf("Insured"), lines.indexOf("Next") + 1)).toEqual([
+      "Insured",
+      "wraps",
+      "",
+      "Next",
+    ]);
   });
 
   it("scales point coordinates into the render's pixel space", () => {

apps/api/src/statements/ocr/tesseract.provider.ts

@@ -254,6 +254,23 @@ export function parseBboxLayout(xhtml: string, scale: number): (OcrPage | null)[
  * Rows are cut when a word's vertical centre leaves the band established by
  * the row's first word, which tolerates the sub-pixel baseline differences
  * between fonts on one line without merging two genuinely separate lines.
+ *
+ * Vertical WHITE SPACE is preserved as a blank line. Rows alone are not the
+ * whole layout: on a form, the blank between two blocks is what says where a
+ * cell's wrapped value stops, and dropping it leaves parsers that walk a
+ * block ("keep joining until the cell ends") running to the end of the page.
+ * That is not hypothetical — the GMX PVL especificación read its whole first
+ * page as the insured's name, because the fixtures were taken from
+ * `pdftotext -layout` (which prints the blanks) while the runtime fed it this
+ * function's output (which did not).
+ *
+ * Horizontal white space is preserved the same way, by padding each word out
+ * to its own column. The same fixture mismatch bit here: a run of spaces is
+ * the ONLY thing separating two cells of a borderless table, so ANA's
+ * `INSURED\s{2,}` label matches and its `SUM INSURED` / `PREMIUM` column
+ * split (taken from `head.search()` offsets) both need real offsets. Joining
+ * on one space put every driver's-policy premium in the sum-insured column
+ * and left the phone glued to the insured's name.
  */
 function toVisualRows(words: OcrWord[]): string {
   const centre = (w: OcrWord) => w.top + w.height / 2;
@@ -281,14 +298,85 @@ function toVisualRows(words: OcrWord[]): string {
   }
   if (current.length) rows.push(current);
 
-  return rows
-    .map((r) =>
-      [...r]
-        .sort((a, b) => a.left - b.left)
-        .map((w) => w.text)
-        .join(" "),
-    )
-    .join("\n");
+  const charWidth = estimateCharWidth(words);
+  const out: string[] = [];
+  rows.forEach((r, i) => {
+    if (i > 0 && isBlankBetween(rows[i - 1], r)) out.push("");
+    out.push(layoutRow(r, charWidth));
+  });
+  return out.join("\n");
+}
+
+/**
+ * One row rendered at its printed column offsets.
+ *
+ * Words that merely follow one another inside the same cell are separated by
+ * exactly one space, whatever the column arithmetic says: one `charWidth` for
+ * a page that mixes fonts leaves a rounding error on every word, and letting
+ * that accumulate sprinkles `\s{2,}` runs through ordinary prose — which is
+ * the very thing the parsers read as a cell boundary. Only a gap wide enough
+ * to be deliberate (more than one blank character) is rendered as one, and
+ * only there is the word re-anchored to its true column, so the offsets a
+ * column split depends on stay honest while values stay clean.
+ */
+function layoutRow(row: OcrWord[], charWidth: number): string {
+  let line = "";
+  let right = 0;
+
+  for (const w of [...row].sort((a, b) => a.left - b.left)) {
+    const col = Math.round(w.left / charWidth);
+    if (!line.length) {
+      line = " ".repeat(Math.max(0, col));
+    } else if (w.left - right > charWidth * 1.5) {
+      line += " ".repeat(Math.max(2, col - line.length));
+    } else {
+      line += " ";
+    }
+    line += w.text;
+    right = w.left + w.width;
+  }
+
+  return line.trimEnd();
+}
+
+/**
+ * Width of one character, in the same units the word boxes use.
+ *
+ * The median of each word's own width-per-character: robust to the handful of
+ * oversized headings and to the wide-tracked letterhead, both of which would
+ * drag a mean. Only words of 3+ characters vote, since a one-character box is
+ * mostly side bearing. Falls back to a value derived from line height when a
+ * page has nothing long enough to measure.
+ */
+function estimateCharWidth(words: OcrWord[]): number {
+  const samples = words
+    .filter((w) => w.text.length >= 3 && w.width > 0)
+    .map((w) => w.width / w.text.length)
+    .sort((a, b) => a - b);
+  if (samples.length) return samples[Math.floor(samples.length / 2)];
+  const heights = words.map((w) => w.height).filter((h) => h > 0);
+  return heights.length ? Math.max(...heights) / 2 : 1;
+}
+
+/**
+ * Does the space between two consecutive rows read as an empty line?
+ *
+ * Measured against the taller of the two rows so a heading and its body text
+ * are judged on their own scale. On the real documents the two populations do
+ * not overlap: consecutive lines of one paragraph sit at 0.3–1.1 line heights
+ * apart, and anything the reader sees as blank-separated starts at 2.1. The
+ * threshold is placed in that empty middle, biased high — a missed blank only
+ * restores today's behaviour, while a spurious one would cut a wrapped value
+ * short.
+ */
+function isBlankBetween(prev: OcrWord[], row: OcrWord[]): boolean {
+  const bottom = Math.max(...prev.map((w) => w.top + w.height));
+  const top = Math.min(...row.map((w) => w.top));
+  const unit = Math.max(
+    ...prev.map((w) => w.height),
+    ...row.map((w) => w.height),
+  );
+  return unit > 0 && top - bottom > unit * 1.6;
 }
 
 /**

apps/web/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@jorgecuadros/web",
-  "version": "1.0.18",
+  "version": "1.0.21",
   "private": true,
   "scripts": {
     "dev": "next dev -p 4500",

apps/web/src/app/polizas/[id]/page.tsx

@@ -648,10 +648,79 @@ function SiniestrosSection({ data }: { data: PolicyDetail }) {
 }
 
 /* -------------------------------------------------------- Coberturas */
-/** The legacy tables carry per-line coverage columns the target schema does
- *  not model; the migration preserved them verbatim in `coveragesJson`. */
+/**
+ * `coveragesJson` holds two unrelated shapes and the section renders each on
+ * its own terms:
+ *
+ * - **A Spanish-keyed object** — the legacy per-line coverage columns the
+ *   target schema does not model, preserved verbatim by the migration. Every
+ *   policy imported from Access carries this one.
+ * - **A `ParsedCoverage[]` array** — written by the policy OCR confirm step
+ *   (GMX's coverage table, ANA's numbered risk sections).
+ *
+ * Running the object renderer over the array is what used to happen, and it
+ * produced a row per array index labelled "0", "1", "2" with `[object
+ * Object]` as its value — not a crash, so nothing surfaced it.
+ */
+interface StoredCoverage {
+  risk?: string;
+  insuredAmount?: number | null;
+  deductible?: string | null;
+  lossParticipation?: string | null;
+  premium?: number | null;
+}
+
 function CoberturasSection({ data }: { data: PolicyDetail }) {
-  const entries = Object.entries(data.coveragesJson ?? {}).filter(
+  const raw = data.coveragesJson ?? null;
+
+  if (Array.isArray(raw)) {
+    const rows = (raw as StoredCoverage[]).filter((c) => c && c.risk);
+    if (rows.length === 0) return null;
+    return (
+      <section className="section">
+        <SectionHead rule="seguros" title="Coberturas" count={rows.length} />
+        <div className="card">
+          <div className="tx-scroll">
+            <table className="tx-table">
+              <thead>
+                <tr>
+                  <th>Riesgo</th>
+                  <th className="num">Suma asegurada</th>
+                  <th className="num">Prima</th>
+                  <th>Deducible</th>
+                  <th>Participación</th>
+                </tr>
+              </thead>
+              <tbody>
+                {rows.map((c, i) => (
+                  <tr key={i}>
+                    <td>{c.risk}</td>
+                    <td className="num">
+                      {c.insuredAmount == null
+                        ? "—"
+                        : formatMoney(c.insuredAmount.toString(), data.currency)}
+                    </td>
+                    <td className="num">
+                      {c.premium == null
+                        ? "—"
+                        : formatMoney(c.premium.toString(), data.currency)}
+                    </td>
+                    <td>{c.deductible ?? "—"}</td>
+                    <td>{c.lossParticipation ?? "—"}</td>
+                  </tr>
+                ))}
+              </tbody>
+            </table>
+          </div>
+          <div className="section-note" style={{ padding: "0 22px 18px" }}>
+            Coberturas leídas del PDF de la aseguradora.
+          </div>
+        </div>
+      </section>
+    );
+  }
+
+  const entries = Object.entries(raw ?? {}).filter(
     ([, v]) => v !== null && v !== "" && v !== 0,
   );
   if (entries.length === 0) return null;

apps/web/src/app/polizas/captura/page.tsx

@@ -4,7 +4,7 @@ import { AppShell } from "@/components/AppShell";
 import { PolicyCaptura } from "@/components/PolicyCaptura";
 
 /**
- * OCR mode of the policy intake screen. Drops the GMX PDF, walks through
+ * OCR mode of the policy intake screen. Drops the GMX or A.N.A. PDF, walks through
  * per-page review, confirms. Same wrapper as `/polizas/nuevo` (manual)
  * with `initialMode="auto"`, so the tab strip is identical and swapping
  * modes doesn't drop state.

apps/web/src/components/PolicyCaptura.tsx

@@ -12,7 +12,7 @@ import { useCan } from "@/lib/abilities";
  * two ways in:
  *
  * - **manual** — `PolicyForm` keys every field by hand.
- * - **auto** — `PolicyOcrIntake` uploads a GMX PDF, OCR proposes the
+ * - **auto** — `PolicyOcrIntake` uploads a GMX or A.N.A. PDF, OCR proposes the
  *   policy, a human still confirms.
  *
  * Both end at the same place (a `Policy` row on a customer's file) so they
@@ -28,7 +28,7 @@ export type PolicyCaptureMode = "manual" | "auto";
 const MODE_HINT: Record<PolicyCaptureMode, string> = {
   manual:
     "Captura cada campo a mano. Use esta opción cuando la póliza llega en papel, en un correo sin PDF legible, o cuando hay que revisar cada dato.",
-  auto: "Suelte el PDF descargado del portal de GMX y el sistema propondrá los campos. Nada se registra sin tu confirmación.",
+  auto: "Suelte el PDF descargado del portal de GMX o de A.N.A. y el sistema propondrá los campos. Nada se registra sin tu confirmación.",
 };
 
 export function PolicyCaptura({ initialMode = "manual" }: { initialMode?: PolicyCaptureMode }) {

apps/web/src/components/PolicyOcrIntake.tsx

@@ -13,9 +13,12 @@ import type { PolicyOcrBatch, PolicyOcrBatchStatus } from "@/lib/types";
 
 /**
  * Insurance OCR intake — mirror of StatementIntake, scoped to the insurance
- * side. Today the only provider is GMX; the parser dispatches on a brand
- * wordmark (`Grupo Mexicano de Seguros` / `gmx.com.mx` / the GMX letterhead)
- * and a new portal only needs a new BRAND entry plus a parser file.
+ * side. GMX and A.N.A. today; the parser dispatches on a brand wordmark
+ * (`Grupo Mexicano de Seguros` / `gmx.com.mx`, `A.N.A. Compañía de Seguros` /
+ * `anaseguros.com.mx`) and a new portal only needs a new BRAND entry plus a
+ * parser file. The uploader is never asked which provider a file came from —
+ * a batch may mix them, and the pipeline labels the batch from what the
+ * parsers actually claimed.
  *
  * Lives inside the `Pólizas` page rather than a top-level route because it
  * is one mode of one job (staff uploading whatever PDFs the office has on
@@ -102,8 +105,8 @@ export function PolicyOcrIntake() {
           <div className="state-box">Cargando…</div>
         ) : batches.length === 0 ? (
           <div className="state-box">
-            Todavía no hay lotes de pólizas. Descargue el certificado del portal
-            de GMX y suéltelo arriba.
+            Todavía no hay lotes de pólizas. Descargue la póliza del portal de
+            GMX o de A.N.A. y suéltela arriba.
           </div>
         ) : (
           <div className="tx-scroll">
@@ -183,14 +186,14 @@ function UploadCard({ onDone }: { onDone: () => void }) {
   return (
     <section className="card" style={{ padding: 16 }}>
       <h2 className="section-title" style={{ marginTop: 0 }}>
-        Subir PDFs de pólizas (GMX)
+        Subir PDFs de pólizas (GMX / A.N.A.)
       </h2>
       <div className="inline-form" style={{ flexWrap: "wrap", gap: 12 }}>
         <label>
           <span className="page-sub">Referencia (opcional)</span>
           <input
             className="input"
-            placeholder="ej. GMX julio 2026"
+            placeholder="ej. ANA agosto 2026"
             value={label}
             onChange={(e) => setLabel(e.target.value)}
           />

apps/web/src/components/PolicyOcrReview.tsx

@@ -21,6 +21,7 @@ import type {
   PolicyOcrBatchDetail,
   PolicyOcrConfirmDocument,
   PolicyOcrCoverage,
+  PolicyOcrCustomerSuggestion,
   PolicyOcrDocument,
   PolicyOcrReviewInput,
 } from "@/lib/types";
@@ -274,6 +275,7 @@ function DocumentRow({ doc, customerIndex, canReview, onSave, onReject }: Docume
     netPremium: doc.extractedNetPremium ?? "",
     total: doc.extractedTotal ?? "",
     premiumPayment: doc.extractedPremiumPayment ?? "",
+    coveragePeriodDays: doc.extractedCoveragePeriodDays?.toString() ?? "",
     postPremium: doc.extractedNetPremium != null && Number(doc.extractedNetPremium) > 0,
   });
   const [customerId, setCustomerId] = useState(
@@ -311,6 +313,7 @@ function DocumentRow({ doc, customerIndex, canReview, onSave, onReject }: Docume
         netPremium: numOrUndef(v.netPremium),
         total: numOrUndef(v.total),
         premiumPayment: trimOrUndef(v.premiumPayment),
+        coveragePeriodDays: numOrUndef(v.coveragePeriodDays),
         matchedPolicyId: policyId || undefined,
         matchedCustomerId: !policyId && customerId ? customerId : undefined,
         forceConfirm: true,
@@ -332,6 +335,7 @@ function DocumentRow({ doc, customerIndex, canReview, onSave, onReject }: Docume
         netPremium: reviewInput.netPremium,
         total: reviewInput.total,
         premiumPayment: reviewInput.premiumPayment,
+        coveragePeriodDays: reviewInput.coveragePeriodDays,
         coveragesJson: (doc.extractedCoveragesJson ?? undefined) as
           | PolicyOcrCoverage[]
           | undefined,
@@ -348,6 +352,7 @@ function DocumentRow({ doc, customerIndex, canReview, onSave, onReject }: Docume
   const locked = doc.status === "POSTED" || doc.status === "REJECTED";
   const matchedExisting = !!doc.matchedPolicy;
   const candidates = doc.matchCandidates ?? [];
+  const suggestions: PolicyOcrCustomerSuggestion[] = doc.customerSuggestions ?? [];
 
   return (
     <article className="card" style={{ padding: 16 }}>
@@ -360,6 +365,13 @@ function DocumentRow({ doc, customerIndex, canReview, onSave, onReject }: Docume
         {doc.extractedInsuredName && (
           <span className="page-sub">· {doc.extractedInsuredName}</span>
         )}
+        {/* Read-only: the parser names the type, the confirm step resolves it
+            to a policy_types row. Reassigning it is the policy screen's job,
+            where the full picker already lives. */}
+        {doc.extractedPolicyTypeName && (
+          <span className="tag">{doc.extractedPolicyTypeName}</span>
+        )}
+        {doc.provider && <span className="tag">{doc.provider}</span>}
       </header>
 
       <div className="doc-detail">
@@ -454,6 +466,17 @@ function DocumentRow({ doc, customerIndex, canReview, onSave, onReject }: Docume
                 onChange={(e) => set("policyDate", e.target.value)}
               />
             </Field>
+            {/* ANA sells 3- and 4-day tourist policies; left blank the
+                póliza keeps the 365-day default. */}
+            <Field label="Días de vigencia">
+              <input
+                className="input"
+                type="number"
+                min={1}
+                value={v.coveragePeriodDays}
+                onChange={(e) => set("coveragePeriodDays", e.target.value)}
+              />
+            </Field>
             <Field label="Moneda">
               <select
                 className="input select"
@@ -523,6 +546,7 @@ function DocumentRow({ doc, customerIndex, canReview, onSave, onReject }: Docume
                   <tr>
                     <th>Riesgo</th>
                     <th className="num">Suma</th>
+                    <th className="num">Prima</th>
                     <th>Deducible</th>
                     <th>Participación</th>
                   </tr>
@@ -534,6 +558,14 @@ function DocumentRow({ doc, customerIndex, canReview, onSave, onReject }: Docume
                       <td className="num">
                         {formatMoney(c.insuredAmount?.toString() ?? null, v.currency)}
                       </td>
+                      {/* ANA's add-on sections print what the coverage COST
+                          where the others print what it pays. Kept in its own
+                          column so the two are never added together. */}
+                      <td className="num">
+                        {c.premium == null
+                          ? "—"
+                          : formatMoney(c.premium.toString(), v.currency)}
+                      </td>
                       <td>{c.deductible ?? "—"}</td>
                       <td>{c.lossParticipation ?? "—"}</td>
                     </tr>
@@ -543,6 +575,66 @@ function DocumentRow({ doc, customerIndex, canReview, onSave, onReject }: Docume
             </details>
           )}
 
+          {/*
+           * Vehicles and named drivers are read-only here: they are written
+           * as their own Vehicle / InsuredDriver rows on confirm, and the
+           * policy screen is where they get edited. Showing them is what
+           * lets a reviewer catch a misread VIN before it is applied.
+           */}
+          {doc.extractedVehiclesJson && doc.extractedVehiclesJson.length > 0 && (
+            <details>
+              <summary>Unidades ({doc.extractedVehiclesJson.length})</summary>
+              <table className="tx-table" style={{ marginTop: 8 }}>
+                <thead>
+                  <tr>
+                    <th>Tipo</th>
+                    <th>Año</th>
+                    <th>Marca</th>
+                    <th>Carrocería</th>
+                    <th>Serie</th>
+                    <th>Placas</th>
+                  </tr>
+                </thead>
+                <tbody>
+                  {doc.extractedVehiclesJson.map((veh, i) => (
+                    <tr key={i}>
+                      <td>{veh.item}</td>
+                      <td>{veh.modelYear ?? "—"}</td>
+                      <td>{veh.make ?? "—"}</td>
+                      <td>{veh.bodyType ?? "—"}</td>
+                      <td>{veh.vinNumber ?? "—"}</td>
+                      <td>{veh.licensePlate ?? "—"}</td>
+                    </tr>
+                  ))}
+                </tbody>
+              </table>
+            </details>
+          )}
+
+          {doc.extractedDriversJson && doc.extractedDriversJson.length > 0 && (
+            <details>
+              <summary>Conductores ({doc.extractedDriversJson.length})</summary>
+              <table className="tx-table" style={{ marginTop: 8 }}>
+                <thead>
+                  <tr>
+                    <th>Nombre</th>
+                    <th>Licencia</th>
+                    <th>Teléfono</th>
+                  </tr>
+                </thead>
+                <tbody>
+                  {doc.extractedDriversJson.map((d, i) => (
+                    <tr key={i}>
+                      <td>{d.fullName}</td>
+                      <td>{d.licenseNumber ?? "—"}</td>
+                      <td>{d.phone ?? "—"}</td>
+                    </tr>
+                  ))}
+                </tbody>
+              </table>
+            </details>
+          )}
+
           {candidates.length > 1 && (
             <Field label="Póliza destino">
               <select
@@ -580,6 +672,34 @@ function DocumentRow({ doc, customerIndex, canReview, onSave, onReject }: Docume
             </Field>
           )}
 
+          {/*
+           * Name suggestions, never a preselection. The office writes
+           * customers surname-first and carriers print them given-name-first,
+           * so without this the reviewer retypes a name the parser already
+           * read. One click fills the picker above; nothing is chosen until
+           * they click. Kept outside the <Field> label — a label must not
+           * wrap other interactive controls.
+           */}
+          {!policyId && !customerId && !locked && canReview && suggestions.length > 0 && (
+            <div className="row" style={{ gap: 8, flexWrap: "wrap" }}>
+              <span className="page-sub">Sugerencias por nombre:</span>
+              {suggestions.map((s) => (
+                <button
+                  key={s.customerId}
+                  type="button"
+                  className="btn btn-ghost btn-sm"
+                  onClick={() => {
+                    setCustomerId(s.customerId);
+                    setCustomerName(s.customerName);
+                  }}
+                >
+                  {s.customerName}
+                  {s.tier === "PARTIAL" && <span className="page-sub"> · parcial</span>}
+                </button>
+              ))}
+            </div>
+          )}
+
           <label className="field">
             <input
               type="checkbox"

apps/web/src/lib/api.ts

@@ -1432,7 +1432,7 @@ export function statementPageUrl(documentId: string): string {
   return `${API_ORIGIN}/statements/documents/${documentId}/page`;
 }
 
-/* ----------------------------------------------------- Policy OCR (GMX) */
+/* ----------------------------------------------- Policy OCR (GMX / ANA) */
 
 export function getPolicyOcrStatus(): Promise<{
   ocrAvailable: boolean;

apps/web/src/lib/types.ts

@@ -1405,7 +1405,7 @@ export interface DiscardBatchResult {
   rejected: number;
 }
 
-/* ------------------------------------------ Policy OCR intake (GMX) */
+/* ------------------------------------- Policy OCR intake (GMX / ANA) */
 
 export type PolicyOcrBatchStatus =
   | "UPLOADED"
@@ -1447,6 +1447,27 @@ export interface PolicyOcrCoverage {
   insuredAmount: number | null;
   deductible: string | null;
   lossParticipation: string | null;
+  /** What the coverage COST, where the layout prints it separately from what
+   *  it pays out (ANA's add-on sections). GMX never prints one. */
+  premium?: number | null;
+}
+
+/** A row of ANA's `ITEM / YEAR / MAKE / BODY / SERIAL No. / PLATES` table. */
+export interface PolicyOcrVehicle {
+  item: string;
+  modelYear: string | null;
+  make: string | null;
+  bodyType: string | null;
+  vinNumber: string | null;
+  licensePlate: string | null;
+}
+
+export interface PolicyOcrDriver {
+  fullName: string;
+  licenseNumber: string | null;
+  address: string | null;
+  phone: string | null;
+  email: string | null;
 }
 
 export interface PolicyOcrMatchCandidate {
@@ -1456,6 +1477,18 @@ export interface PolicyOcrMatchCandidate {
   policyNumber: string;
 }
 
+/**
+ * A customer whose name resembles the printed insured name. A suggestion,
+ * not a match — the API never preselects one.
+ */
+export interface PolicyOcrCustomerSuggestion {
+  customerId: string;
+  customerName: string;
+  /** `EXACT` = same name tokens in any order; `PARTIAL` = one contains the other. */
+  tier: "EXACT" | "PARTIAL";
+  score: number;
+}
+
 export interface PolicyOcrDocument {
   id: string;
   pageNumber: number;
@@ -1478,6 +1511,11 @@ export interface PolicyOcrDocument {
   extractedTotal: string | null;
   extractedCoveragesJson: PolicyOcrCoverage[] | null;
   extractedPremiumPayment: string | null;
+  extractedCoveragePeriodDays: number | null;
+  extractedVehiclesJson: PolicyOcrVehicle[] | null;
+  extractedDriversJson: PolicyOcrDriver[] | null;
+  /** `PolicyType.name` the parser read, resolved to an id only at confirm. */
+  extractedPolicyTypeName: string | null;
   matchedPolicy: {
     id: string;
     policyNumber: string | null;
@@ -1486,6 +1524,7 @@ export interface PolicyOcrDocument {
   } | null;
   matchedCustomer: { id: string; name: string } | null;
   matchCandidates: PolicyOcrMatchCandidate[] | null;
+  customerSuggestions: PolicyOcrCustomerSuggestion[] | null;
   matchNote: string | null;
 }
 
@@ -1505,6 +1544,7 @@ export interface PolicyOcrReviewInput {
   brokerFee?: number;
   total?: number;
   premiumPayment?: string;
+  coveragePeriodDays?: number;
   coveragesJson?: PolicyOcrCoverage[];
   matchedPolicyId?: string;
   matchedCustomerId?: string;
@@ -1530,7 +1570,11 @@ export interface PolicyOcrConfirmDocument {
   brokerFee?: number;
   total?: number;
   premiumPayment?: string;
+  coveragePeriodDays?: number;
   coveragesJson?: PolicyOcrCoverage[];
+  /** Explicit lookup picks; both beat the name the parser read. */
+  policyTypeId?: string;
+  insuranceProviderId?: string;
   postPremium?: boolean;
 }
 

deploy/galactus/jorgecuadros-app.compose.yml

@@ -61,6 +61,11 @@ services:
       INGEST_DIR: /data/ingest
       BACKUP_DIR: /data/backups
       MIGRATION_ENV: prod
+      # The API applies pending Prisma migrations at container start, before
+      # Nest listens, and refuses to start if they fail (docker/api-entrypoint.sh).
+      # Set false ONLY when the schema is being moved by hand — the app will
+      # then boot against whatever schema it finds.
+      RUN_MIGRATIONS: ${RUN_MIGRATIONS:-true}
       # Credentials the "Operaciones" screen runs mysqldump/mysql as. NOT the
       # application user: --single-transaction needs the global RELOAD privilege
       # and the app user has only ALL ON jorgecuadros.*, so every backup, sync

deploy/jorgecuadros-app.stack.yml

@@ -41,6 +41,11 @@ services:
       INGEST_DIR: /data/ingest
       BACKUP_DIR: /data/backups
       MIGRATION_ENV: prod
+      # The API applies pending Prisma migrations at container start, before
+      # Nest listens, and refuses to start if they fail (docker/api-entrypoint.sh).
+      # Set false ONLY when the schema is being moved by hand — the app will
+      # then boot against whatever schema it finds.
+      RUN_MIGRATIONS: ${RUN_MIGRATIONS:-true}
       # Credentials the "Operaciones" screen runs mysqldump/mysql as. NOT the
       # application user: --single-transaction needs the global RELOAD privilege
       # and the app user has only ALL ON jorgecuadros.*, so every backup, sync

docker/api-entrypoint.sh

@@ -0,0 +1,102 @@
+#!/bin/sh
+# Apply pending Prisma migrations, then hand off to the API.
+#
+# WHY THE CONTAINER AND NOT THE DEPLOY WORKFLOW
+#
+# The workflow still has its own `prisma migrate deploy` step and that is not
+# redundant: it runs BEFORE the new images are pulled, i.e. while the OLD code
+# is still serving, which is the order expand/contract migrations are designed
+# around (see docs/DEPLOY_AND_MIGRATIONS.md). Doing it here as well closes the
+# gaps that step cannot:
+#
+#   - The runner has to reach MySQL directly. When it cannot, the deploy is run
+#     with `skip_migrate=true` and the schema silently does not move — the app
+#     then boots against a schema that is one release behind, which surfaces
+#     later as a column-not-found at runtime rather than as a failed deploy.
+#   - A container restarted by `restart: unless-stopped` after a host reboot,
+#     or a stack re-applied by hand in Portainer, never goes through the
+#     workflow at all.
+#
+# `migrate deploy` is idempotent, so running it in both places costs one
+# no-op query on the normal path.
+#
+# THE API DOES NOT START IF THE MIGRATION FAILS. That is deliberate: serving
+# against a schema that does not match the code is worse than being down,
+# because the failures it produces are partial and silent (a write to a column
+# that does not exist yet fails for one feature while the rest of the app looks
+# healthy). The container exits non-zero and Docker's restart policy retries.
+set -e
+
+SCHEMA=/repo/packages/database/prisma/schema.prisma
+
+log() { echo "[entrypoint] $*"; }
+
+if [ "${RUN_MIGRATIONS:-true}" != "true" ]; then
+  log "RUN_MIGRATIONS=${RUN_MIGRATIONS} — skipping migrations, starting the API"
+  exec "$@"
+fi
+
+if [ -z "${DATABASE_URL}" ]; then
+  log "DATABASE_URL is unset; cannot migrate." >&2
+  log "Set it, or set RUN_MIGRATIONS=false if you migrate out of band." >&2
+  exit 1
+fi
+
+# pnpm's hoisted linker normally puts the CLI in the root .bin, but the
+# workspace package keeps its own link too. Accept either rather than pinning
+# a layout detail of the installer — the Dockerfile asserts at build time that
+# one of these exists, so a missing CLI breaks the image build, not a deploy.
+PRISMA=""
+for candidate in /repo/node_modules/.bin/prisma \
+                 /repo/packages/database/node_modules/.bin/prisma; do
+  if [ -x "$candidate" ]; then
+    PRISMA="$candidate"
+    break
+  fi
+done
+if [ -z "$PRISMA" ]; then
+  log "prisma CLI not found in this image; cannot migrate." >&2
+  exit 1
+fi
+
+# Retry ONLY a connection failure (P1001). On a full bring-up the database
+# container can still be starting, and on galactus the API additionally has to
+# resolve the host's MagicDNS name — a lookup that is unreliable for the first
+# moments after a host reboot (see the dns block in the app compose file, and
+# docs/DEPLOY_AND_MIGRATIONS.md).
+#
+# Every other failure exits immediately. Retrying a migration that is actually
+# broken just delays the same error behind a minute of noise, and P3005 in
+# particular needs a human.
+attempt=1
+max="${MIGRATE_MAX_ATTEMPTS:-20}"
+delay="${MIGRATE_RETRY_SECONDS:-3}"
+
+while : ; do
+  log "prisma migrate deploy (attempt ${attempt}/${max})"
+  if output=$("$PRISMA" migrate deploy --schema "$SCHEMA" 2>&1); then
+    printf '%s\n' "$output"
+    log "migrations up to date"
+    break
+  fi
+  printf '%s\n' "$output" >&2
+
+  if ! printf '%s' "$output" | grep -q 'P1001'; then
+    log "migrate deploy FAILED — refusing to start the API." >&2
+    if printf '%s' "$output" | grep -q 'P3005'; then
+      log "P3005: the database has tables but no migration history. This is a" >&2
+      log "database that predates Prisma migrations. Baseline it ONCE with:" >&2
+      log "  npx prisma@5 migrate resolve --applied 0000_init --schema $SCHEMA" >&2
+    fi
+    exit 1
+  fi
+
+  if [ "$attempt" -ge "$max" ]; then
+    log "database unreachable after ${max} attempts — giving up." >&2
+    exit 1
+  fi
+  attempt=$((attempt + 1))
+  sleep "$delay"
+done
+
+exec "$@"

docker/api.Dockerfile

@@ -113,5 +113,21 @@ ENV APP_VERSION=$APP_VERSION \
     GIT_SHA=$GIT_SHA \
     BUILD_DATE=$BUILD_DATE
 
+# Pending migrations are applied at container start, before Nest listens —
+# see the header of the script for why this is done here as well as in the
+# deploy workflow. Asserted at BUILD time so a missing prisma CLI breaks the
+# image build rather than a production boot: the runtime layer copies
+# /repo/node_modules wholesale, and which of these two paths carries the bin
+# is an implementation detail of pnpm's hoisted linker.
+COPY docker/api-entrypoint.sh /usr/local/bin/api-entrypoint.sh
+RUN chmod +x /usr/local/bin/api-entrypoint.sh
+RUN for c in /repo/node_modules/.bin/prisma \
+             /repo/packages/database/node_modules/.bin/prisma; do \
+      if [ -x "$c" ]; then echo "prisma CLI found at $c"; exit 0; fi; \
+    done; \
+    echo "FATAL: prisma CLI is not in the runtime layer; api-entrypoint.sh cannot migrate" >&2; \
+    exit 1
+
 EXPOSE 3001
+ENTRYPOINT ["/usr/local/bin/api-entrypoint.sh"]
 CMD ["node", "apps/api/dist/main.js"]

docs/BACKLOG.md

@@ -12,8 +12,8 @@ carries the reasoning. Close an item *there* as well as here, or the two drift.
 **Verified against dev at compile time** (re-run before trusting the numbers):
 
 ```
-policy_types: AUTO, LICENCIAS, MULT
-policies NULL policyTypeId: 5
+policy_types: AUTO, LICENCIAS, MULT      (+ M_EMPR after 20260815160000)
+policies NULL policyTypeId: 5            (0 after 20260815160000)
 policies pending liquidación: 226
 customers: 1536
 last tag: v1.0.6 (2026-08-02 02:06 UTC) — 14 commits, 5 migrations behind HEAD
@@ -98,16 +98,31 @@ in the same phone call — (55) 5480-4000.
 
 ## 2. Live data defects — open, and confirmed open today
 
-### 2.1 `policy_types` is missing `INCENDIO` and `M_EMPR`, and 5 policies are orphaned
+### 2.1 ~~`policy_types` missing rows + 5 orphaned policies~~ — FIXED 2026-08-15
 
 `policyTypeId` is `String?` with a plain relation, so Prisma's default is
-`SetNull`. The spec's recommended `onDelete: Restrict` was **never applied**.
-Five `m_empr` policies lost their ramo; four of them are pending liquidación
-and are invisible to every ramo-filtered query — including the pending report
-§2 is supposed to produce.
+`SetNull`, and `removePolicyType()` had no in-use guard — deleting a lookup row
+returned 200 and silently blanked the ramo on every policy using it. That is
+what happened to `M_EMPR` and its 5 `m_empr` policies.
 
-Fix alongside the liquidación work (3.1), since it distorts that feature's own
-report. Source: INSURANCE "Two defects found while verifying this spec".
+Closed by `20260815160000_policy_type_repair` plus the guard in
+`policies.service.ts`:
+
+- `M_EMPR` restored and the 5 policies re-pointed at it, scoped to
+  `policyTypeId IS NULL AND legacySourceTable = 'm_empr'` so it cannot claim a
+  policy blanked for some other reason. Idempotent; verified against dev inside
+  a rolled-back transaction.
+- **`INCENDIO` deliberately not recreated.** The legacy `INCENDIO` table has
+  1 row and it never loaded, so the type has zero policies — restoring it would
+  only add a dead option to the type picker.
+- Deleting an in-use policy type, carrier or adjuster now **refuses** with the
+  name and the count. `claims.adjusterId` had the identical `SET NULL` trap and
+  is guarded too. `onDelete: Restrict` at the schema level was not applied —
+  the application guard gives a Spanish message the operator can act on, where
+  a raw FK error would not.
+- The duplicate `ANA` carrier row (1 policy) was merged into `ANA SEGUROS`
+  (738), since OCR now assigns the carrier automatically and two rows would
+  keep splitting the book.
 
 ### 2.2 ≤41 MULT second settlements were dropped in migration
 
@@ -165,12 +180,17 @@ Each of these is a known, deliberate stopping point rather than a bug.
 - No `SKIPPED_NO_EMAIL` worklist (see 1.9).
 
 **Policy OCR** — [`POLICY_OCR.md`](POLICY_OCR.md)
-- **GMX only.** The dispatcher is a `[provider, pattern]` table plus a parser
-  map, so a second carrier is one function and two entries — but no other
-  layout has been seen, and guessing produces a parser nobody can verify.
+- **GMX and A.N.A. only.** The dispatcher is a `[provider, pattern]` table plus
+  a parser map, so a third carrier is one function and two entries — but no
+  other layout has been seen, and guessing produces a parser nobody can verify.
 - **The `recibo` PDF is unread.** The GMX certificate carries no premium at
   all; reading the separate receipt and pairing it to its certificate is what
-  would let `postPremium` stop being a manual tick.
+  would let `postPremium` stop being a manual tick. A.N.A. prints its premium
+  on the face, so this is a GMX-only gap.
+- **No `insuranceProviderId` beyond the two OCR carriers.** Confirm resolves
+  the parser's provider to an `insurance_providers` row by name, so GMX and
+  A.N.A. land correctly; a policy typed in by hand still gets whatever the
+  operator picks.
 - **No versioning.** A re-issued policy arrives as a new certificate with the
   same number and confirm updates the existing row. Nothing records that this
   is the 2027 issue of that policy.

docs/DEPLOY_AND_MIGRATIONS.md

@@ -56,9 +56,11 @@ workflow's last step fails if the API does not report the tag you dispatched.
    `pre-migrate-<tag>-<timestamp>.sql.gz`, which is exactly what the
    **Operaciones** admin screen lists and can restore. A dump taken on the CI
    runner would be unreachable by the only restore path the platform has.
-3. **`prisma migrate deploy`** — as a workflow *step*, never the container
-   `CMD`. If it were the CMD, N replicas would race each other applying the
-   same migration.
+3. **`prisma migrate deploy`** — as a workflow *step*, so the schema moves
+   while the OLD code is still serving, which is the order expand/contract is
+   designed around. **The api container repeats this at start** (below); the
+   command is idempotent, so on the normal path the container's run is a
+   no-op.
 4. **app** — the new api + web images.
 5. **Verify** — `GET /version` on the running API must report the dispatched
    tag.
@@ -141,6 +143,59 @@ Commit the generated `migrations/<timestamp>_add_foo/` directory. `db push` is
 now a local-scratch tool only — using it against a database with history
 desynchronises it from `_prisma_migrations`.
 
+If you hand-write a migration instead of generating one, check it against what
+Prisma would have produced before committing — a hand-written file that drifts
+from `schema.prisma` fails on the *next* deploy, not this one:
+
+```bash
+prisma migrate diff \
+  --from-schema-datamodel <schema.prisma at the previous commit> \
+  --to-schema-datamodel packages/database/prisma/schema.prisma --script
+```
+
+## Migrations also run at container start
+
+`docker/api-entrypoint.sh` is the api image's `ENTRYPOINT`. It runs
+`prisma migrate deploy` and only then `exec`s the API. **If the migration
+fails the container exits non-zero and the API never listens.**
+
+That is the point. Serving against a schema that does not match the code is
+worse than being down, because the failures are partial and silent — a write
+to a column that does not exist yet breaks one feature while the rest of the
+app looks healthy.
+
+This does not replace the workflow step, which still runs first and against
+the old code. It covers what that step cannot:
+
+- **`skip_migrate: true`.** Previously that left the schema behind with no
+  further safety net, and the mismatch surfaced later as a runtime error. Now
+  it just moves the migration into the container, so it is a safe choice when
+  the runner cannot reach MySQL.
+- **Restarts that never touch the workflow** — `restart: unless-stopped`
+  bringing the stack back after a host reboot, or a stack re-applied by hand
+  in Portainer.
+
+Behaviour worth knowing:
+
+| | |
+|---|---|
+| `RUN_MIGRATIONS=false` | Skip and start anyway. Plumbed through both app stack files. For when the schema is being moved by hand. |
+| `DATABASE_URL` unset | Refuses to start (it would have failed at Nest boot anyway, but this says why). |
+| Cannot reach the database (**P1001**) | Retries, default 20 × 3s. Covers a cold `db` container and galactus's MagicDNS lookup right after a reboot. `MIGRATE_MAX_ATTEMPTS` / `MIGRATE_RETRY_SECONDS` tune it. |
+| Any other failure | Exits at once. Retrying a broken migration only delays the same error; **P3005** additionally prints the `migrate resolve --applied 0000_init` hint. |
+
+**On replicas.** Both stacks are `replicas: 1` and must stay that way for an
+unrelated reason (the servicios email sweep has no DB lock — see the caveats
+below). If that ever changes, concurrent `migrate deploy` runs are safe on
+their own: Prisma takes a database advisory lock, so the others block and then
+find nothing pending. They would each pay the wait at startup, not corrupt
+anything.
+
+The prisma CLI has to be present in the runtime layer for any of this. The
+image copies `/repo/node_modules` wholesale so it already is, and the
+Dockerfile **asserts it at build time** — a missing CLI breaks the image
+build rather than a production boot.
+
 ## galactus vs cubex
 
 `galactus` is standalone Docker (Portainer endpoint **3**), `cubex` is a 3-node
@@ -319,9 +374,10 @@ backup does them (see `deploy/scripts/pre-migrate-backup.mjs`):
   Portainer serves a self-signed certificate. It is scoped to that one step,
   which talks to nothing but Portainer. Replacing the certificate and dropping
   the flag is the real fix.
-- The runner lives on cubex and must reach the target host's Portainer (9443)
-  **and** MySQL (3306). If it cannot reach 3306, run the migration by hand from
-  a host that can and dispatch with `skip_migrate: true`.
+- The runner lives on cubex and must reach the target host's Portainer (9443).
+  It should also reach MySQL (3306) for the migrate step, but that is no longer
+  load-bearing — dispatch with `skip_migrate: true` and the api container
+  applies the migrations itself at start.
 - `bootstrap: true` lets the pre-migrate backup be skipped when no API container
   exists yet. Use it for a first-ever deploy only — it is the one switch that
   lets a migration run with no restore point.

docs/POLICY_OCR.md

@@ -34,7 +34,7 @@ was **reused, not copied**.
 |---|---|
 | API module | `apps/api/src/policy-ocr/` (service, controller, DTOs, matcher, parser) |
 | Shared OCR seam | `apps/api/src/ocr/ocr.module.ts` |
-| Tables | `policy_ocr_batches`, `policy_ocr_documents` (`20260801000000_policy_ocr_intake`) |
+| Tables | `policy_ocr_batches`, `policy_ocr_documents` (`20260801000000_policy_ocr_intake`, extended by `20260815120000_policy_ocr_ana` and `20260815160000_policy_type_repair`) |
 | Web | `components/PolicyCaptura.tsx` (tab shell), `PolicyOcrIntake.tsx` (upload), `PolicyOcrReview.tsx` (review queue) |
 | Abilities | `policy:ingest`, `policy:ocr-review` — both **STAFF** |
 
@@ -84,8 +84,9 @@ feature's core assumption.
 Utility statements arrive **bundled, one customer per page** — so there, one
 page is one document and the parser runs per page. A policy PDF is the
 opposite: the GMX certificate is a 2-page document where page 1 carries the
-contract header and page 2 carries the per-coverage table, and **both pages
-describe the same policy**. So the pipeline concatenates every page's text
+contract header and page 2 carries the per-coverage table (and the PVL
+especificación runs to ten), and **every page describes the same policy**. So
+the pipeline concatenates every page's text
 (`\n\n` between pages, which also keeps `ocrRawText` readable for debugging)
 and runs the parser and the matcher exactly **once per file**.
 
@@ -133,6 +134,203 @@ customers do occur (one group policy bound by two related parties), and
 picking arbitrarily would silently book the wrong coverage against the wrong
 person.
 
+### Name suggestions on the zero-hit path
+
+When the policy number finds nothing — the new-policy case, where a human has
+to pick a customer anyway — `name-matcher.ts` ranks the customer book against
+the printed insured name and the review screen offers the top three as
+one-click buttons above the picker. They are written to
+`policy_ocr_documents.customerSuggestions`, deliberately **not** to
+`matchCandidates`, so a name hint can never be read as a policy-number hit.
+Nothing sets `matchedCustomerId`; the rule above is unchanged.
+
+The problem is only ordering: the office books customers surname-first
+(`WAGONER, PAMELA`) and carriers print them given-name-first
+(`PAMELA DENISE WAGONER`), so a string compare never hits while a **token-set**
+compare does. Names are normalized (accents folded, so OCR's `MUNOZ` reaches
+the book's `MUÑOZ`; initials, `DE`/`LA`/`Y`, `JR`, `S.A. DE C.V.` and any token
+containing a digit dropped — ANA prints the phone hard against the name as
+`Ph.3102001538`). Two tiers:
+
+| Tier | Rule |
+|---|---|
+| `EXACT` | identical token sets, any order |
+| `PARTIAL` | one set contains the other, ≥2 shared tokens, **and** the surname is present |
+
+Both thresholds come from measuring the real book (1536 customers):
+
+- 1487 distinct token sets, so `EXACT` cross-person collisions are ~0
+- loosen to surname + first given name and 131 customers (8.5%) collide —
+  the book holds `MCWILLIAMS, BRIAN MICHAEL` *and* `MCWILLIAMS, BRIAN`
+- 185 surnames are shared by 524 customers, so one token is never evidence;
+  hence the ≥2 floor and the explicit surname requirement, which is what stops
+  `JERRY MARILYN` reaching `ESTRADA, JERRY & MARILYN` on given names alone
+
+Replaying every book row as a carrier would print it (given-name-first, joint
+spouse dropped): 97.9% top-ranked correct, 0.9% no suggestion, 1.2% a
+different row — and all but two of those are the same human on a duplicate or
+variant row (`MOLNAR, JANOS` vs `MOLNAR,  JANOS`, `IBARRA, ISMAEL &`). The
+two genuine wrong-person cases are `CUADROS, JORGE JR` against three
+`CUADROS, JORGE H.`, and they appear as a tie in the list rather than as a
+single answer.
+
+A blob is refused outright (>8 tokens or >80 characters): GMX's especificación
+has no field labels and the parser has been seen handing its whole first page
+over as `insuredName`, which would find a surname somewhere in the prose.
+`(SIN NOMBRE)` — 14 rows the migration left — is skipped on both sides.
+
+**Not used for utility statements.** There the registrant genuinely is not the
+customer (the `CATT, RANDY` finding above), so the same trick would be wrong,
+not merely noisy.
+
+## GMX ships two unrelated documents for the same policy
+
+The office downloads both from the same portal, and either can land in a
+batch. They share only the brand and the policy number, so `parseGmx` is a
+two-line dispatcher over two real parsers — both returning `provider: "GMX"`,
+because the matcher keys on the policy number alone and must not care which
+artifact was uploaded.
+
+| | **Caratula** (`…_Traduccion.pdf`) | **Especificación** (`…-CondicionesParticulares.pdf`) |
+|---|---|---|
+| Language | English (free translation) | Spanish |
+| Shape | boxed header table + 4-column coverage table | 10 pages of prose, no tables at all |
+| Header fields | Policy / Insured / Broker / Term / From / To / Currency | insured name, risk location, property description |
+| Dates, broker, currency field | yes | **none printed** |
+| Coverages | one row per risk | section heading + `Límite Máximo de Responsabilidad:` |
+| Parser | `parseGmxCaratula` | `parseGmxEspecificacion` |
+
+Selected by `isEspecificacion` on the PVL page header
+(`ESPECIFICACIÓN QUE SE ADHIERE`, `PVL Hogar`, `Nombre del asegurado`).
+
+Three things about the especificación are worth knowing before touching it:
+
+- **The policy number's group widths differ between the two.** The caratula
+  reads `007-037-07005947-0000-02` and the especificación
+  `07-037-07006957-00000-01` — 2 digits in the first group, 5 in the fourth.
+  The original parser pinned the widths, so it read one family and returned
+  null on the other. `POLICY_NUMBER_SHAPE` now matches the shape, and since
+  the especificación prints the number on all ten page headers, the ten
+  readings cross-check each other (disagreement is noted, not resolved — the
+  same rule the zona federal parser applies to its clave).
+- **Coverages are found by anchoring on the limit label and walking backwards
+  for the heading.** There is no row shape to match. A heading is a short line
+  *preceded by a blank line* — that last condition is the whole trick, since
+  length alone cannot tell a heading from the wrapped tail of the paragraph
+  above it (`efectuados.`, `Y CADA PÉRDIDA.`), and without it coverages get
+  named after the last word of the preceding prose.
+- **Vigencia, agente and prima are absent by design**, not unread. The parser
+  says so in a note, so a reviewer seeing three empty fields does not read it
+  as a broken parse.
+
+  **These three are keyed in by hand** — confirmed 2026-08-14 with Luz, who
+  handles GMX policies at the office. The review screen already has editable
+  inputs for all three, and `postPremium` enables off the *typed* premium, so
+  a hand-entered prima posts to the ledger exactly like a parsed one. No code
+  change was needed to support this; it is a process decision, recorded here
+  because the parser's own note now instructs the reviewer accordingly.
+
+  > **A blank vigencia is silently permanent.** `Policy.policyTo` is nullable
+  > and the renewals window query filters `policyTo: { gte, lte }`
+  > (`renewals.service.ts`), so a policy confirmed without one **never matches
+  > and never gets a renewal notice** — no error, no warning, and nothing
+  > later notices. This is why the parser's note names the consequence instead
+  > of just listing the missing fields.
+
+An excluded catastrophic risk is recorded as excluded **in the risk label**
+(`Terremoto o erupción volcánica — Sección Edificio: EXCLUIDO`) with a null
+amount, never as `0`: a coverage insured for zero and an excluded coverage are
+the same number and very different facts, and `ParsedCoverage` has no field
+for the distinction.
+
+## A.N.A. ships two unrelated faces too
+
+`A.N.A. Compañía de Seguros` is the Rosarito office's tourist auto book. Same
+split as GMX, different reason: GMX ships two *documents about one policy*,
+A.N.A. ships two *products*.
+
+| | **AUTOMOBILE** (`SPECIAL POLICY FOR TOURISTS`) | **DRIVER´S POLICY** (the office says *licencia*) |
+|---|---|---|
+| Insures | a specific car | up to five named drivers, whatever they drive |
+| Vehicle table | `ITEM / YEAR / MAKE / BODY / SERIAL No. / PLATES` | **none** |
+| Insured | one `INSURED` cell | numbered `POLICY HOLDER` list |
+| Value columns | one (`LIMIT OF LIABILITY`) | two (`SUM INSURED`, `PREMIUM`) |
+| Sections | 9, numbered | 6, unnumbered, **in a different order** |
+| Parser | `parseAnaAutomobile` | `parseAnaDriverPolicy` |
+
+Selected by `isAnaDriverPolicy` on the title band.
+
+The **four automobile products** the office sells — amplia and responsabilidad
+civil, each annual or by-the-day — are the **same layout with different
+numbers**. "Amplia" prints a vehicle value and `COVERED` on sections 1–2;
+"resp. civil" prints `0.00` and `EXCLUDED`. That is data, not a layout, so
+there is one parser rather than four.
+
+Things worth knowing before touching the ANA parsers:
+
+- **These are born-digital portal PDFs**, so `pdftotext -layout` returns exact
+  glyphs and exact columns. The driver's policy parser uses that: `SUM
+  INSURED` and `PREMIUM` print the same shape (`100,000.00 usd.` /
+  `18.70 usd.`) with no per-row label, so **horizontal position is the only
+  thing that separates them**. The split is computed from the header's own
+  column offsets rather than hardcoded, because they shift between products.
+  If a scan ever arrives without column fidelity, every amount is reported as
+  a sum insured and the reviewer is told the split failed.
+- **The money row is read positionally, not by finding six amounts.** An
+  unused `DISCOUNT` prints as a bare `-`, so an "amounts in order" reading
+  shifts every value one column left on a discounted policy. The parser
+  requires exactly six whitespace-separated cells or reports the row unread.
+- **Each PDF prints its face two or three times** (ORIGINAL, AGENT COPY, then
+  a summary receipt and three travel ID cards), and the pipeline concatenates
+  every page before parsing. The coverage walk is bounded to the first copy
+  and the driver list to the first `POLICY HOLDER` block. Unbounded, the
+  licencia returns the same person three times — which reads as a
+  three-driver policy, not as a bug, so nothing downstream would catch it.
+- **Two five-digit numbers sit in the header band** and only one is the agent
+  clave: the other is the agent's own postal code
+  (`ROSARITO, BAJA CALIFORNIA 22710`). Likewise the agent's street address
+  reads `BENITO JUAREZ 25 No.50 INT 38`, three lines above the `No.` cell that
+  holds the policy number — hence the two-space floor after `No.`.
+- **Sections 6–8 print a PREMIUM where the others print a limit.** $40 is what
+  legal aid *cost*, not a $40 liability limit, so it lands on
+  `ParsedCoverage.premium` (a field GMX never fills) and gets its own column
+  on the review screen. Adding the two together would be meaningless.
+- **`coveragePeriodDays` matters here and nowhere else.** A.N.A. sells 3- and
+  4-day policies. `Policy.coveragePeriodDays` defaults to 365, so a weekend
+  policy left at the default sits in the renewals window a year out. The term
+  is derived from the two dates and cross-checked against the printed `DAYS`
+  cell; a disagreement is noted rather than resolved.
+
+Exclusions follow the GMX rule — recorded in the risk label
+(`MATERIAL DAMAGE — VEHICLE: EXCLUDED`) with a null amount, never as `0`. It
+matters more here: a responsabilidad-civil policy prints `0.00` for material
+damage, so the two are visually identical on the page.
+
+### Vehicles and drivers
+
+A.N.A. is the first provider whose face carries either, so confirm now writes
+`Vehicle` and `InsuredDriver` rows alongside the `Policy`
+(`applyVehiclesAndDrivers`). The parsed values are stored on the document as
+`extractedVehiclesJson` / `extractedDriversJson` and shown read-only in the
+review queue, so a misread VIN is catchable before it is applied.
+
+Both inserts skip a row that already exists on the policy, matched on the
+identifier the document prints — VIN then plate for a vehicle (A.N.A.'s
+TRAILER and TOWING slots have no VIN), licence number then name for a driver.
+The case that forces this is confirming a **renewal** onto an existing policy:
+a blind insert leaves the customer with the same VIN listed twice and no way
+to tell which row the renewal belongs to.
+
+Nothing is ever updated or deleted there. A vehicle whose plate changed lands
+as a second row for a human to reconcile — the safe half of the mistake, since
+an overwrite would destroy the only record of what was insured last term.
+
+The vehicle table is parsed **by token role, not by column offset**, because
+`BODY` is the cell that wraps: `PACIFICA` is one token and `GENESIS SEDAN` is
+two, so a fixed token count reads the VIN out of the wrong slot on the second.
+The 17-character VIN is the anchor and `BODY` is whatever sits between the
+make and it.
+
 ## What the parser reads, and the field it cannot
 
 `ParsedPolicy` fields are all nullable on purpose: each carrier prints a
@@ -145,6 +343,12 @@ insured, broker (→ `Policy.agentName`), legal address, ZIP, `policyFrom` /
 per-coverage table (risk, insured amount, deductible, loss participation)
 preserved verbatim.
 
+Read from an A.N.A. face: all of the above except additional insured and
+broker parens, **plus** the premium (A.N.A. prints it — see below), the policy
+fee, the total, the term in days, the vehicle table, and the named drivers
+with their US licence numbers. The tax and the agent clave have no column in
+the schema and ride in the notes.
+
 > **The GMX certificate carries no premium.** Not "sometimes missing" — the
 > document does not have the figure. It lives on GMX's **separate `recibo`
 > PDF**. The parser leaves `netPremium` / `policyFee` / `brokerFee` / `total`
@@ -157,20 +361,71 @@ This is also why confirm never overwrites an existing `Policy.netPremium`
 with null: the certificate not carrying a premium is not evidence that the
 premium is gone.
 
+A.N.A.'s faces do print one — the `DISCOUNT / PREMIUM / POLICY FEE / TAX /
+LOCAL TAX / TOTAL` row is on the same page — so an ANA document reaches the
+review queue with `netPremium` populated and `postPremium` already ticked.
+
 Deductible and loss participation are stored as **strings** (`"5%"`, `"20%"`,
 `"USD 1,000"`) — they are printed as a mix of percentages, currency amounts
 and free text, and normalising them would lose the distinction.
 
+## Policy type and carrier
+
+Confirm sets `Policy.policyTypeId` and `Policy.insuranceProviderId` from what
+the parser read.
+
+| document | `policyTypeName` |
+|---|---|
+| ANA `AUTOMOBILE` | `AUTO` |
+| ANA `DRIVER´S POLICY` | `LICENCIAS` |
+| GMX caratula **and** especificación | `MULT` |
+
+The parser emits a **name**, never an id — it is a pure function over text and
+must not reach for the database, so `resolveLookups()` in the service turns the
+name into a foreign key. A renamed lookup row is then a data change rather than
+a parser change.
+
+**Resolve, never create.** A missing `policy_types` row means a human deleted
+it, and silently recreating it would undo that with no record. The field stays
+null and the reviewer adds the row through the lookups screen. An explicit
+`policyTypeId` / `insuranceProviderId` on the confirm payload always wins.
+
+Two judgement calls worth recording:
+
+- **GMX is `MULT`, not `INCENDIO`.** The caratula's own header reads "Multiple
+  Policy / Home" and the especificación is "PVL Hogar" — one product, two
+  artifacts. `MULT` is the live row carrying 769 of them; `INCENDIO` is
+  fire-only and no policy in the book has ever used it.
+- **The parser's provider code is not the carrier's row name.** The office's
+  book is filed under `ANA SEGUROS`, so `PROVIDER_ROW_NAME` maps `ANA` onto it.
+  A bare `ANA` row with 1 policy also existed and is merged away by
+  `20260815160000_policy_type_repair`.
+
+> **Deleting a lookup row used to be silent data loss.** `policies.policyTypeId`,
+> `policies.insuranceProviderId` and `claims.adjusterId` are all
+> `ON DELETE SET NULL`, and the lookups screen deleted unconditionally — so the
+> delete returned 200 and blanked the field on every row that used it. That is
+> how `M_EMPR` vanished and left 5 policies with no ramo, found months later by
+> querying. All three deletes now refuse while the row is in use, naming it and
+> the count. See `assertLookupUnused` and BACKLOG §2.1.
+
 ## Confirm: what actually gets written
 
 Per confirmed document, in order:
 
 1. **The `Policy` row** — updated if a policy was matched, created under the
    picked customer if not. Only non-null `extracted*` fields are written; null
-   never overwrites existing data.
-2. **A `PolicyDocument`** — the source PDF is streamed into the policy's
-   storage namespace and attached, so the paperwork stays with the policy.
-3. **Optionally a `Transaction`** — `INSURANCE` domain, negative amount
+   never overwrites existing data. `policyTypeId` and `insuranceProviderId` are
+   resolved first (above) and left untouched when unresolvable, so an existing
+   policy never loses a type or carrier it already had.
+2. **`Vehicle` and `InsuredDriver` rows** — for the providers whose face
+   carries them (A.N.A.; never GMX Hogar), skipping any that already exist on
+   the policy. See *Vehicles and drivers* above.
+3. **A `PolicyDocument`** — the source PDF is streamed into the policy's
+   storage namespace and attached, so the paperwork stays with the policy. Its
+   `documentType` is named after whichever parser claimed the page
+   (`ANA_POLICY`, `GMX_POLICY`).
+4. **Optionally a `Transaction`** — `INSURANCE` domain, negative amount
    (a charge), `captureSource: "OCR"`, `captureRef` = the document id.
 
 The ledger write is **opt-in twice over**: staff must tick `postPremium`
@@ -210,20 +465,71 @@ feature is disabled.
 
 ## Tests
 
-`apps/api/src/policy-ocr/parsers/policy-parser.spec.ts` — 8 cases, all
-against verbatim text extracted from one real document,
-`HC_Folio_000767_Traduccion.pdf`: provider detection from the wordmark and
-from the footer URL, the header fields, every coverage row off the second
-page, the deductible/loss-participation strings, the missing-premium note,
-the broker line with the agent-number parens absent, and a page with no GMX
-signal at all (which must yield no provider rather than a bad guess).
+`apps/api/src/policy-ocr/parsers/policy-parser.spec.ts` — 58 cases against
+verbatim text extracted from five real documents, indentation and blank lines
+included (the column positions are what the parser reads, so a cleaned-up
+fixture would test nothing — and on A.N.A.'s driver's policy the offsets are
+literally the only thing separating two columns).
+
+From `HC_Folio_000767_Traduccion.pdf` (caratula): provider detection from the
+wordmark and from the footer URL, the header fields, every coverage row off
+the second page, the deductible/loss-participation strings, the
+missing-premium note, the broker line with the agent-number parens absent,
+and a page with no GMX signal at all (which must yield no provider rather
+than a bad guess).
+
+From `007_LGS-HGMX_07006957_01_0-CondicionesParticulares.pdf`
+(especificación): the differently-grouped policy number, the risk location
+read across its wrapped line, the empty `Asegurado Adicional` cell that must
+not capture the next line, the absent-by-design fields, currency taken from
+the USD limits rather than the M.N. sublimits in the body prose, a limit
+split under `Edificio` / `Contenidos` sub-labels, a limit printed on the
+label's own line, a deductible stated as a sentence *above* its limit, a
+sublimit block whose amount sits after both a blank line and a page break,
+the excluded earthquake coverage, and the hydrometeorological deductible and
+coinsurance pulled from their own per-zone block.
+
+Plus four cases in `apps/api/src/policies/lookup-delete-guard.spec.ts` pinning
+the refusal that stops a lookup delete from silently blanking the rows that use
+it, and four in the parser suite on the policy-type NAME each document yields.
+
+From the three A.N.A. PDFs: brand detection (and that GMX's layout rules
+cannot claim an ANA page), the header band, DD MM YYYY read out of three
+separate column cells, the six money cells with `DISCOUNT` printed as a bare
+`-`, the vehicle row with a one-word and a two-word `BODY` cell, the empty
+TRAILER/TOWING slots, the agent street number and postal code that must *not*
+be read as the policy number and clave, the declared values labelled by item
+slot, the `$500.00` inside the deductible sentence that is not a sum insured,
+the per-person/per-accident split in both of its printed forms, an add-on's
+figure recorded as a premium, section 9's parenthesised limit, the by-the-day
+term, the excluded sections, the column-position split on the driver's policy,
+its different section order, and — for both faces — that a doubled or tripled
+input yields one set of coverages and one driver rather than one per copy.
+
+`apps/api/src/policy-ocr/name-matcher.spec.ts` — 21 cases on the customer name
+suggestions, every fixture name lifted from the real book: the reversed name,
+the printed middle name, the exact row outranking the row that merely contains
+it, the joint account reached from one spouse (and refused when only given
+names are printed), the Spanish double surname with the comma in either place,
+the 54 rows with no comma at all, the `(SIN NOMBRE)` placeholder, a bare shared
+surname, and the page-sized blob. Four more in
+`policy-matcher.service.spec.ts` pin the wiring: suggestions on the zero-hit
+and unreadable-number paths, no book read at all when the policy number hits,
+and one book read across a batch.
+
+Four of the GMX cases are regression tests for ways the parser can silently attach
+the *wrong* value rather than none — a neighbouring coverage's prose read as
+a deductible, the page-level `DEDUCIBLES:` paragraph read as one, a coverage
+named after a wrapped prose tail, and one section's per-zone deductible
+adopted by the coverage above it. Each was a real defect caught by running
+the parser against the full ten-page document.
 
 ## Not built
 
-- **Only GMX.** The dispatcher (`detectPolicyProvider`) is a table of
-  `[provider, pattern]` pairs plus a `parsers` map, so adding ANA or Qualitas
-  is a parser function and two entries — but no other carrier's layout has
-  been seen yet, and guessing at one produces a parser nobody can verify.
+- **GMX and A.N.A. only.** The dispatcher (`detectPolicyProvider`) is a table
+  of `[provider, pattern]` pairs plus a `parsers` map, so adding Qualitas is a
+  parser function and two entries — but no other carrier's layout has been
+  seen yet, and guessing at one produces a parser nobody can verify.
 - **The `recibo` PDF.** Reading the premium off GMX's separate receipt
   document, and pairing it to the certificate it belongs to, is the obvious
   next piece. It is what would let `postPremium` stop being a manual tick.

package.json

@@ -1,6 +1,6 @@
 {
   "name": "jorgecuadros-platform",
-  "version": "1.0.18",
+  "version": "1.0.21",
   "private": true,
   "workspaces": [
     "apps/*",

packages/database/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@jorgecuadros/database",
-  "version": "1.0.18",
+  "version": "1.0.21",
   "private": true,
   "main": "generated/client/index.js",
   "types": "generated/client/index.d.ts",

packages/database/prisma/migrations/20260815120000_policy_ocr_ana/migration.sql

@@ -0,0 +1,12 @@
+-- ANA Seguros policy OCR: the vehicle/driver tables and the printed term in
+-- days that ANA's tourist book carries and GMX's property book does not.
+ALTER TABLE `policy_ocr_documents`
+    ADD COLUMN `extractedCoveragePeriodDays` INTEGER NULL,
+    ADD COLUMN `extractedVehiclesJson` JSON NULL,
+    ADD COLUMN `extractedDriversJson` JSON NULL;
+
+-- The parser note trail outgrew VARCHAR(191): a nine-section ANA policy runs
+-- past it routinely, and the notes that got cut were the tail ones — the
+-- "could not read X" warnings the reviewer most needs.
+ALTER TABLE `policy_ocr_documents`
+    MODIFY COLUMN `matchNote` TEXT NULL;

packages/database/prisma/migrations/20260815160000_policy_type_repair/migration.sql

@@ -0,0 +1,55 @@
+-- The policy type the OCR parser read the product as, resolved to a
+-- `policy_types` row at confirm time.
+ALTER TABLE `policy_ocr_documents`
+    ADD COLUMN `extractedPolicyTypeName` VARCHAR(191) NULL;
+
+-- ---------------------------------------------------------------------------
+-- Repair: M_EMPR was deleted from the lookups screen and took its policies'
+-- type with it.
+--
+-- `policies.policyTypeId` is ON DELETE SET NULL, and removePolicyType() had no
+-- in-use guard, so deleting the row silently blanked the field on every policy
+-- referencing it — 5 of them, all from the legacy `m_empr` table. The guard
+-- against a repeat ships in the same change as this migration. What follows
+-- repairs what already happened.
+--
+-- Idempotent on purpose: `policy_types.name` is UNIQUE so the INSERT IGNORE is
+-- a no-op once the row exists, and the UPDATE is scoped to rows that are still
+-- null AND came from that one legacy table, so it can never claim a policy
+-- whose type was blanked for some other reason.
+INSERT IGNORE INTO `policy_types` (`id`, `name`) VALUES (UUID(), 'M_EMPR');
+
+UPDATE `policies` p
+    JOIN `policy_types` pt ON pt.`name` = 'M_EMPR'
+    SET p.`policyTypeId` = pt.`id`
+    WHERE p.`policyTypeId` IS NULL
+      AND p.`legacySourceTable` = 'm_empr';
+
+-- INCENDIO is deliberately NOT recreated. It is the other row the migration
+-- would have produced, but no policy in the book has ever carried it, so
+-- adding it back would only put a dead option in the type picker.
+
+-- ---------------------------------------------------------------------------
+-- Merge the duplicate ANA carrier.
+--
+-- `insurance_providers` holds both "ANA" (1 policy) and "ANA SEGUROS" (738).
+-- They are one carrier, and OCR is about to start assigning it automatically —
+-- picking either row while both exist would keep splitting the book.
+--
+-- "ANA SEGUROS" is the survivor because it is where the 738 already are.
+--
+-- Written as joins rather than subqueries so that BOTH statements are no-ops
+-- when either row is absent (a fresh database, or one where this was already
+-- tidied by hand). A subquery form would resolve to NULL and blank the
+-- carrier off every ANA policy.
+UPDATE `policies` p
+    JOIN `insurance_providers` dup  ON dup.`id` = p.`insuranceProviderId` AND dup.`name` = 'ANA'
+    JOIN `insurance_providers` keep ON keep.`name` = 'ANA SEGUROS'
+    SET p.`insuranceProviderId` = keep.`id`;
+
+DELETE dup FROM `insurance_providers` dup
+    JOIN `insurance_providers` keep ON keep.`name` = 'ANA SEGUROS'
+    WHERE dup.`name` = 'ANA'
+      -- Belt and braces: never drop a row that still has policies hanging off
+      -- it, whatever the UPDATE above did or did not manage to move.
+      AND NOT EXISTS (SELECT 1 FROM `policies` p WHERE p.`insuranceProviderId` = dup.`id`);

packages/database/prisma/migrations/20260815180000_policy_ocr_customer_suggestions/migration.sql

@@ -0,0 +1,7 @@
+-- Ranked customers whose name matches the printed insured name, for the
+-- documents whose policy number found nothing and therefore need a customer
+-- picked by hand. Kept in its own column rather than folded into
+-- `matchCandidates`, which the review screen reads as policy-number hits —
+-- a name is a suggestion and must never be able to masquerade as a match.
+ALTER TABLE `policy_ocr_documents`
+    ADD COLUMN `customerSuggestions` JSON NULL;

packages/database/prisma/schema.prisma

@@ -379,8 +379,11 @@ model PolicyDocument {
 /// source PDF and optionally writes a premium Transaction.
 model PolicyOcrBatch {
   id           String               @id @default(uuid())
-  /// Which insurance provider portal the batch came from. "GMX" today;
-  /// future providers (AXA, GNP, …) extend the parser, not this table.
+  /// Which insurance provider portal the batch came from — "GMX", "ANA", or
+  /// "GMX + ANA" when one upload mixed them. Set by the pipeline from what
+  /// the parsers actually claimed, not asked of the uploader, so it can
+  /// never contradict the documents. Future providers extend the parser,
+  /// not this table.
   provider     String               @default("GMX")
   status       PolicyOcrBatchStatus @default(UPLOADED)
   uploadedById String
@@ -442,10 +445,27 @@ model PolicyOcrDocument {
   extractedBrokerFee         Decimal?  @db.Decimal(12, 2)
   extractedTotal             Decimal?  @db.Decimal(12, 2)
   /// Per-coverage rows from the GMX "Material damages" / "Additional risk"
-  /// tables — preserved verbatim so a missing premium receipt still leaves
-  /// the coverages auditable.
+  /// tables and ANA's numbered risk sections — preserved verbatim so a
+  /// missing premium receipt still leaves the coverages auditable.
   extractedCoveragesJson     Json?
   extractedPremiumPayment    String?
+  /// Printed term length. ANA sells 3- and 4-day tourist policies, so
+  /// leaving `Policy.coveragePeriodDays` at its 365 default would overstate
+  /// a weekend policy by a year.
+  extractedCoveragePeriodDays Int?
+  /// `ParsedVehicle[]` off ANA's ITEM/YEAR/MAKE/BODY/SERIAL/PLATES table.
+  /// Written to `Vehicle` rows on confirm; kept here so the review screen
+  /// shows what was read before anything is applied.
+  extractedVehiclesJson      Json?
+  /// `ParsedDriver[]` — the insured on ANA's automobile face, the numbered
+  /// POLICY HOLDER list on its driver's policy. Written to `InsuredDriver`
+  /// rows on confirm.
+  extractedDriversJson       Json?
+  /// The `PolicyType.name` the parser read the product as ("AUTO",
+  /// "LICENCIAS", "MULT"). A NAME, not an id — the parser never touches the
+  /// database, so confirm resolves it against `policy_types` and leaves
+  /// `Policy.policyTypeId` null if there is no such row.
+  extractedPolicyTypeName    String?
 
   // Match by `Policy.policyNumber` → existing Policy / Customer.
   matchedPolicyId   String?
@@ -456,7 +476,18 @@ model PolicyOcrDocument {
   /// normal; >1 means the policy number is shared across customers and a
   /// human must pick.
   matchCandidates   Json?
-  matchNote         String?
+  /// `CustomerNameSuggestion[]` — customers whose name matches the printed
+  /// insured name, ranked. A SUGGESTION, never a match: it is deliberately
+  /// kept out of `matchCandidates` so the review screen cannot mistake a
+  /// name hint for a policy-number hit, and it never sets
+  /// `matchedCustomerId`. Only populated when the policy number found
+  /// nothing, which is exactly when staff have to pick a customer by hand.
+  customerSuggestions Json?
+  /// Text, not VARCHAR(191): this carries the parser's whole note trail, and
+  /// a multi-section ANA policy runs past 191 characters routinely. Silently
+  /// truncating it drops the tail notes, which are the ones that say what
+  /// could NOT be read.
+  matchNote         String?   @db.Text
 
   reviewedById String?
   reviewedBy   User?     @relation("PolicyOcrDocumentReviewer", fields: [reviewedById], references: [id])