chore(release): v1.0.6

Cut by rmancinas via the "Cut release" workflow. Pushing the tag triggers build.yml; deploy separately with tag=1.0.6.

.env.example

@@ -4,6 +4,15 @@ SESSION_SECRET=change-me-to-a-random-string
 WEB_ORIGIN=http://localhost:3000
 NEXT_PUBLIC_API_ORIGIN=http://localhost:3001
 
+# Object storage (MinIO / S3) for document blobs and scanned receipt pages.
+# Without S3_ENDPOINT + credentials the API still boots, but every document
+# upload/download and the whole recibo OCR intake are disabled. Credentials fall
+# back to MINIO_ROOT_USER / MINIO_ROOT_PASSWORD when the S3_* pair is unset.
+S3_ENDPOINT=http://localhost:9000
+S3_BUCKET=jorgecuadros-documents
+S3_ACCESS_KEY=
+S3_SECRET_KEY=
+
 # Login the "Operaciones" screen runs mysqldump/mysql as. Optional locally: when
 # unset it falls back to the DATABASE_URL credentials, which a dev MySQL usually
 # grants enough for. Required in any deployment, where the application user has

.gitea/workflows/build.yml

@@ -37,6 +37,16 @@ env:
 jobs:
   build:
     name: Build ${{ matrix.image }}
+    # release.yml pushes the release commit and its tag in a single `git push`,
+    # so Gitea creates two runs for the same commit: one for master, one for the
+    # tag. Only the tag run matters — it is the one that emits the X.Y.Z / X.Y
+    # image tags, and it publishes `latest` and `sha-<short>` too, since it is
+    # the same commit. Skip the branch run rather than racing or cancelling it.
+    # Ordinary pushes to master (any message but `chore(release):`) still build.
+    if: >-
+      github.event_name != 'push' ||
+      startsWith(github.ref, 'refs/tags/') ||
+      !startsWith(github.event.head_commit.message, 'chore(release):')
     runs-on: docker
     container:
       image: docker:27-dind

.gitea/workflows/release.yml

@@ -0,0 +1,270 @@
+# Cut a release: stamp the version across every package.json, commit, tag, push.
+#
+# This does NOT build and does NOT deploy. Pushing the `vX.Y.Z` tag is what
+# triggers build.yml, which publishes `X.Y.Z`, `X.Y`, `sha-<short>` and `latest`
+# image tags. Deploying stays a separate, deliberate act: once the build is
+# green, dispatch deploy-galactus.yml with `tag=X.Y.Z` (no leading v — the tag
+# carries the `v`, the image tag does not).
+#
+# Why a workflow instead of three local commands: the release commit is the one
+# thing that must be identical every time, and cutting it from a laptop is how
+# a manifest bump gets forgotten or a tag lands on an unpushed commit. Here the
+# only input is the number.
+#
+# Prereqs (once):
+#  - Repo secret RELEASE_TOKEN: a Gitea personal access token with
+#    write:repository on this repo. The built-in Actions token is deliberately
+#    NOT used — whether a push made with it re-triggers build.yml depends on the
+#    Gitea version, and a release that silently publishes no images is worse
+#    than one that fails. A PAT push is an ordinary push and always triggers.
+#    If build.yml somehow does not start, it has workflow_dispatch: run it
+#    against the new tag by hand.
+
+name: Cut release
+
+on:
+  workflow_dispatch:
+    inputs:
+      bump:
+        description: "Which part to bump (choose 'explicit' to type the number)"
+        type: choice
+        required: true
+        default: "minor"
+        options:
+          - patch
+          - minor
+          - major
+          - explicit
+      version:
+        description: "Exact version when bump=explicit (x.y.z, no leading v)"
+        required: false
+        default: ""
+
+jobs:
+  release:
+    name: Release
+    runs-on: docker
+    container:
+      image: node:20-alpine
+    steps:
+      - name: Install tools
+        run: apk add --no-cache git
+
+      - name: Preflight — RELEASE_TOKEN
+        env:
+          RELEASE_TOKEN: ${{ secrets.RELEASE_TOKEN }}
+        run: |
+          set -eu
+          if [ -z "${RELEASE_TOKEN:-}" ]; then
+            echo "::error::Secret RELEASE_TOKEN is not set. Create a Gitea PAT with"
+            echo "::error::write:repository and add it as a repo secret named RELEASE_TOKEN."
+            exit 1
+          fi
+
+      # Full history + tags: the duplicate-tag check below is meaningless
+      # against a shallow clone, which has none of them.
+      - uses: actions/checkout@v4
+        with:
+          fetch-depth: 0
+          ref: master
+          token: ${{ secrets.RELEASE_TOKEN }}
+
+      - name: Resolve the new version
+        id: ver
+        env:
+          BUMP: ${{ github.event.inputs.bump }}
+          EXPLICIT: ${{ github.event.inputs.version }}
+        run: |
+          set -eu
+          CURRENT=$(node -p "require('./package.json').version")
+          echo "current: $CURRENT"
+
+          if [ "$BUMP" = "explicit" ]; then
+            NEXT="$EXPLICIT"
+            if [ -z "$NEXT" ]; then
+              echo "::error::bump=explicit requires the version input."
+              exit 1
+            fi
+          else
+            NEXT=$(node -e '
+              const [cur, part] = process.argv.slice(1);
+              const m = /^(\d+)\.(\d+)\.(\d+)/.exec(cur);
+              if (!m) { console.error(`unparseable current version: ${cur}`); process.exit(1); }
+              let [maj, min, pat] = m.slice(1).map(Number);
+              if (part === "major") { maj += 1; min = 0; pat = 0; }
+              else if (part === "minor") { min += 1; pat = 0; }
+              else { pat += 1; }
+              process.stdout.write(`${maj}.${min}.${pat}`);
+            ' "$CURRENT" "$BUMP")
+          fi
+
+          # set-version.mjs validates the shape too, but failing here keeps the
+          # working tree clean when the input is a typo.
+          case "$NEXT" in
+            v*) echo "::error::Version must not carry a leading 'v' (got $NEXT)."; exit 1 ;;
+          esac
+          if ! printf '%s' "$NEXT" | grep -Eq '^[0-9]+\.[0-9]+\.[0-9]+(-[0-9A-Za-z.-]+)?$'; then
+            echo "::error::Invalid version: $NEXT (expected x.y.z)."
+            exit 1
+          fi
+          if [ "$NEXT" = "$CURRENT" ]; then
+            echo "::error::$NEXT is already the current version."
+            exit 1
+          fi
+          if git rev-parse -q --verify "refs/tags/v$NEXT" >/dev/null; then
+            echo "::error::Tag v$NEXT already exists. Releases are immutable — pick a new number."
+            exit 1
+          fi
+
+          echo "next: $NEXT"
+          echo "version=$NEXT" >> "$GITHUB_OUTPUT"
+
+      - name: Stamp the version across every manifest
+        run: node scripts/set-version.mjs "${{ steps.ver.outputs.version }}"
+
+      # A release whose only content is the version bump means the dispatch was
+      # a mistake — set-version.mjs already refused a no-op above, so an empty
+      # diff here means the manifests were somehow already at this number.
+      - name: Commit, tag, push
+        env:
+          RELEASE_TOKEN: ${{ secrets.RELEASE_TOKEN }}
+          VERSION: ${{ steps.ver.outputs.version }}
+          ACTOR: ${{ github.actor }}
+        run: |
+          set -eu
+          if git diff --quiet; then
+            echo "::error::No manifest changed. Nothing to release."
+            exit 1
+          fi
+
+          git config user.name "gitea-actions"
+          git config user.email "actions@git.mancinas.io"
+
+          git commit -a \
+            -m "chore(release): v${VERSION}" \
+            -m "Cut by ${ACTOR} via the \"Cut release\" workflow. Pushing the tag triggers build.yml; deploy separately with tag=${VERSION}."
+          git tag -a "v${VERSION}" -m "v${VERSION}"
+
+          # Re-point at an authenticated remote. The token is a secret, so Gitea
+          # masks it in the log; nothing here echoes the URL regardless.
+          git remote set-url origin \
+            "$(printf '%s' "${GITHUB_SERVER_URL}" | sed "s#://#://x-access-token:${RELEASE_TOKEN}@#")/${GITHUB_REPOSITORY}.git"
+
+          # One push for both refs: a commit that lands without its tag builds
+          # nothing and looks like a successful release.
+          #
+          # The output is captured because a failing *post-receive* hook does not
+          # fail the push: git prints `remote: error: ...`, updates both refs and
+          # exits 0. That is how v1.0.3 was cut — the hook 500'd, so Gitea never
+          # created the build run, and this step went green anyway.
+          if ! git push origin "HEAD:master" "refs/tags/v${VERSION}" 2>push.log; then
+            cat push.log
+            echo "::error::Push failed. Nothing was released."
+            exit 1
+          fi
+          cat push.log
+
+          if grep -q '^remote: error' push.log; then
+            echo "::warning::The remote's post-receive hook errored. Both refs landed,"
+            echo "::warning::but Gitea most likely created no workflow run for them."
+            echo "::warning::The next step checks and dispatches build.yml if needed."
+          fi
+
+          echo "sha=$(git rev-parse HEAD)" >> "$GITHUB_OUTPUT"
+        id: push
+
+      # Gitea creates workflow runs from the post-receive hook, so a hook error
+      # silently costs you the build: the tag exists, no image is ever published,
+      # and the failure only surfaces later as a 404 when deploy pulls the image.
+      # Confirm the run exists; dispatch it if it does not; fail loudly if that
+      # does not work either.
+      - name: Verify build.yml started
+        env:
+          RELEASE_TOKEN: ${{ secrets.RELEASE_TOKEN }}
+          VERSION: ${{ steps.ver.outputs.version }}
+          SHA: ${{ steps.push.outputs.sha }}
+        run: |
+          node -e '
+            const base = `${process.env.GITHUB_SERVER_URL}/api/v1/repos/${process.env.GITHUB_REPOSITORY}`;
+            const headers = { Authorization: `token ${process.env.RELEASE_TOKEN}` };
+            const sha = process.env.SHA;
+            const tag = `v${process.env.VERSION}`;
+            const sleep = (ms) => new Promise((r) => setTimeout(r, ms));
+
+            // The master push and the tag push carry the SAME commit, so a sha
+            // match alone is not enough: build.yml skips the master run by
+            // design, and that skipped run would satisfy a sha-only check even
+            // if the tag run were never created. When the API reports a ref for
+            // the run, require it to be the tag; when it reports none, fall back
+            // to the sha match rather than failing a release over a field name.
+            const isTagRun = (r) => {
+              const ref = r.head_branch || r.ref || "";
+              return !ref || ref === tag || ref === `refs/tags/${tag}`;
+            };
+
+            const started = async () => {
+              const res = await fetch(`${base}/actions/runs?limit=30`, { headers });
+              if (!res.ok) throw new Error(`runs query failed: HTTP ${res.status}`);
+              const body = await res.json();
+              return (body.workflow_runs || []).some(
+                (r) =>
+                  r.head_sha === sha &&
+                  String(r.path || "").includes("build.yml") &&
+                  isTagRun(r),
+              );
+            };
+
+            // The hook fires synchronously with the push, so a run that is coming
+            // is usually already there; the retries cover a busy instance.
+            const poll = async (attempts) => {
+              for (let i = 0; i < attempts; i++) {
+                if (await started()) return true;
+                await sleep(10_000);
+              }
+              return started();
+            };
+
+            (async () => {
+              if (await poll(3)) {
+                console.log(`build.yml is running for ${sha}.`);
+                return;
+              }
+
+              console.log(`No build.yml run for ${sha}. Dispatching against ${tag}.`);
+              const res = await fetch(
+                `${base}/actions/workflows/build.yml/dispatches`,
+                {
+                  method: "POST",
+                  headers: { ...headers, "Content-Type": "application/json" },
+                  // Must be the tag, not master: metadata-action only emits the
+                  // X.Y.Z and X.Y image tags when the ref is a semver tag. And
+                  // it must be the fully qualified ref — Gitea 404s on `v1.0.3`.
+                  body: JSON.stringify({ ref: `refs/tags/${tag}` }),
+                },
+              );
+              if (!res.ok) console.log(`Dispatch returned HTTP ${res.status}.`);
+
+              if (await poll(3)) {
+                console.log(`build.yml is running for ${sha}.`);
+                return;
+              }
+
+              console.log(`::error::${tag} is pushed but nothing is building it, and`);
+              console.log(`::error::the dispatch did not take. Run "Build and Push Images"`);
+              console.log(`::error::by hand with ref=${tag} (the tag, not master), then`);
+              console.log(`::error::deploy. Check the Gitea server log for the`);
+              console.log(`::error::post-receive error while you are at it.`);
+              process.exit(1);
+            })();
+          '
+
+      - name: Summary
+        env:
+          VERSION: ${{ steps.ver.outputs.version }}
+        run: |
+          set -eu
+          echo "Released v${VERSION}."
+          echo ""
+          echo "build.yml is now building git.mancinas.io/rmancinas/jorgecuadros-{api,web}:${VERSION}."
+          echo "When it is green, dispatch 'Deploy to galactus' with:"
+          echo "  tag=${VERSION}  scope=app  bootstrap=false  skip_migrate=false"

PLAN.md

@@ -134,7 +134,10 @@ Given the amount of near-duplicate/overlapping data across snapshot tables (mult
     - **Receipt capture module — DONE** (2026-07-27). The legacy "Editor" replacement, built on the single-movement capture from step 6. Wires up the previously-unused `Transaction.outstanding` (NOPAGO): capture flag on `POST /billing`, `?outstanding=` list filter, `POST /billing/:id/resolve-outstanding` (gated `ledger:create`, not `ledger:void` — resolving *completes* a capture), and exclusion from every balance aggregate exactly as the legacy `SALDOS ULTIMO 0`'s `HAVING NOPAGO = 0` did. Adds `POST /billing/batch` (one `$transaction`, check-level fields shared, per-line customer/amount) and `GET /billing/by-check`, plus the `cheque-count` report replacing `REPORTE CHEQUE COUNT` / `REPORTE POR CHEQUE` / `EDITA CHEQUE ALF|COUNT|NUM` — print/PDF/CSV/XLSX come free from the existing `/reportes/:slug` machinery. Web: `/estado-cuenta/lote` (the actual "Editor" screen, with live reconciliation against the physical check amount), plus an "Estado de pago" filter, a "sin fondos" row tag and a Resolver dialog on `/estado-cuenta`. No new abilities. Verified end-to-end against dev, API + browser.
       **Two pre-existing bugs found and fixed while building it:** (a) `statement()` filtered `legacySourceTable: { notIn: [...] }`, which compiles to SQL `NOT IN` — and `NULL NOT IN (…)` is NULL, so **every app-captured movement was invisible on the customer statement** (438 rows in the movement browser vs 392 on the statement) while still appearing everywhere else. This would have made the whole receipt-capture feature look broken to staff. Now NULL-safe. (b) The balances *count* query omitted the void filter its own page query applied, so the row count disagreed with the rows.
       **OCR seam:** `BillingService.createBatch(dto, opts)` is the single multi-row write path and carries three contract guarantees for the step-11 OCR module to post through — `items[i]` maps to `lines[i]` (so `StatementDocument.postedTransactionId` can be zipped back on), `opts.refs[i]` stamps `captureRef` with a duplicate-post guard that a *voided* row deliberately does not block, and `opts.source` is service-level only so an HTTP client cannot label hand-keyed rows as machine-captured. Backed by a new `TransactionCaptureSource` enum (MANUAL/BATCH/OCR) + `captureRef`, both nullable so the 40,136 migrated rows stay NULL rather than being mislabelled.
-    - **PDF/OCR auto-capture** — ingest→split→OCR→match→review pipeline for the 300+/month/service-provider statements staff currently key in by hand. Posts through the capture module above. Matching logic was checked field-by-field against `migration/transform_properties.py`'s actual output and found three real gaps to close first: no `TELEPHONE` service kind exists yet, `PROPERTY_TAX.accountNumber` was migrated from `PREDIAL` not `CLAVE` (needs verification against a real predial statement), and `GAS.meterNumber` was never populated by the migration at all.
+    - **PDF/OCR auto-capture — DONE** (2026-08-01). The ingest→split→OCR→match→review pipeline for the 300+/month/service-provider statements staff key in by hand, built in `apps/api/src/statements/` and posting through §1.2's `createBatch` seam with `source: "OCR"` and a per-document `captureRef`. Web: `/recibos` + `/recibos/:id`. Abilities `statement:ingest`/`statement:review` (STAFF — the review step is what makes machine capture safe at that tier). OCR is self-hosted **Tesseract** behind a swappable `OcrProvider` interface; `tesseract-ocr`, `tesseract-ocr-data-spa` and `poppler-utils` were added to the API image.
+      **Every decision was driven by 10 real scans (46 pages).** Shipped-parser results on them: provider 46/46, account ref 43/46, amount 42/46, due date 44/46 — and against the dev database **39/46 (85%) exact auto-match, 40/46 (87%) identified**, the rest genuine review cases. The scans are pure images (no text layer), so OCR is mandatory, and they arrive **bundled one customer per page**.
+      **The three gaps are closed, and two of them were mis-stated in the spec.** (a) `TELEPHONE` now exists and is backfilled from `Property.phone1` only — coverage is 534/18/1 across phone1/2/3, so phone is one billed line per property, not three. (b) **Clave catastral ≠ predial**: `DATMEX.clave` (934 rows, `KA903009`) is what CESPT and predial bills actually print, while `predial` — what `PROPERTY_TAX.accountNumber` holds — has only 663 distinct values across 1135 rows and appears on no statement; the clave now lives on `Property.cadastralKey` as the matcher's secondary key and predial is left untouched. (c) Gas was **not** a dead end: 160 of the 334 `DATMEX.gas` values are real account numbers (the rest are `ESTACIONARIO`/`CILINDRO` descriptors), all recovered into `GAS.meterNumber`.
+      **Matching is scoped per service kind and never reads the customer name** — a CESPT receipt prints `ARNAIZ ROSAS ELSA AURORA` for an account this office holds under `CATT, RANDY`, because the name on a utility bill is the registrant, not the current owner. Normalisation is per provider: CFE strips leading zeros off `NO. DE SERVICIO`, Telnor strips the 664 LADA down to the stored local 7 digits. Where a provider prints a payment barcode it is preferred over the printed label (one CFE label OCR'd a digit too many while its barcode was correct) and the two are cross-checked, with disagreement forcing review. Confirming a document whose service had no reference writes it back, so gas and any other cold start is a one-time cost.
     - **Multi-bank chequera — DONE** (2026-07-27). `Bank`/`BankAccount` models so Seguros (US bank) and Utilities (Mexican bank, currently SCOTHIA) can each have their own register. `bank_transactions` gained a **required** `bankAccountId` (plus an `(bankAccountId, transactionDate)` index, since every read is now filtered by account and ordered by date), and all 22,669 existing rows were backfilled onto a seeded "Utilities — Scotiabank (MXN)" account by `migration/backfill_bank_accounts.py` — a standalone step because `prisma db push` cannot add a required column to a populated table. It is idempotent and now runs inside `run_all.py` (both normal and `--sync`) ahead of `transform_bank.py`, which fails fast if the account is missing. Every read path in `bank.service.ts` is account-scoped, including `facets()` (which had no filter at all) and *both* raw-SQL rollups in `summary()`. API: `?bankAccountId=` is required on `list`/`stats`/`facets`/`summary` — **not** optional-with-an-all-accounts-default, since summing an MXN and a USD register repeats exactly the currency-collapsing mistake the billing module exists to prevent — plus a new `bank/accounts` + `bank/banks` sub-resource under a MANAGER `bank:manage-accounts` ability. Web: `/banco` gained an account picker (remembered per browser) and reads every figure in the selected account's currency, `/banco/cuentas` manages banks and accounts, and `/inicio`'s chequera card names the account it is showing instead of implying one register. An account's `currency` is immutable after creation by design — its booked movements are denominated in it. Verified against dev + browser: a second USD account showed full read/write isolation from the MXN register, whose totals were unchanged.
     - **Customer-number recycling** — promotes the legacy `NUM id` (currently only inside `customer_legacy_refs`) into a first-class, reusable `Customer.customerNumber`, automates *finding* candidates for reuse (cancelled / 1-year-inactive), and auto-assigns the lowest free number at creation — the search is automated, the release/reuse decision stays a human action. Backfill needs care: ~140 utilities rows and all insurance-only customers have no real legacy number (synthetic `rownum_N`/`insrow_N` placeholders in `transform_customers.py`, not real `NUM id`s).
     
@@ -157,7 +160,7 @@ Repo scaffolded at `jorgecuadros-platform/`: npm workspaces, NestJS API with a r
 
 **Portal live DB now in hand.** `utility_dbo.sql` (1.3 GB, 55 tables) and the portal codebase `my-jorgecuadros-web` (PHP/`mysqli`, Gitea repo, themed classic/modern, ~397 PHP files, core in `scripts/functions.php`) are both on disk — resolving the long-standing "`utility_dbo` schema unknown" blocker. Sync-relevant tables identified: statements/money (`utility_bills`, `accounting`, `email_alert_log`), customer/property (`home_owners`, `home_index`, `condominium`, `management`, `hoa_management`, `trust_assist`), portal-facing policy views (`fm2`/`fm3`/`fmt`, `full_coverage`, `mx_liability`, `usa_liability`), and portal write points (`peticion_gas`, PayPal payments, `notifications_settings`, `verification_codes`). A second dump, `jorgecuadros.sql` (38 MB, 11 tables — `pagos`/`pagosemail`/`PROPANO`/`TRUSTVENCE`/etc.), appears to be an older/partial export, not the portal live DB.
 
-**Step 11 spec written, not built.** `docs/RECEIPT_CAPTURE_SPEC.md` covers the receipt-capture ("Editor") completion plus the three net-new ops features (OCR auto-capture, multi-bank chequera, customer-number recycling) — see Build sequencing step 11 above for the summary. Written from the 2026-07-25/26 meeting notes and verified against the real migration scripts and current API code, not just designed from the meeting notes alone.
+**Step 11 is now three-quarters built.** Receipt capture, the multi-bank chequera and PDF/OCR auto-capture are all done and verified; only customer-number recycling remains unbuilt. `docs/RECEIPT_CAPTURE_SPEC.md` carries a BUILT note per section recording what shipped and, for §2, the four things real scanned statements proved the spec had wrong or unknown.
 
 **Step 12 spec written, not built.** `docs/INSURANCE_FEATURES_SPEC.md` covers the insurance half of the same meeting (renewal emails, liquidación batch, certificate + portal delivery, carrier APIs) — see Build sequencing step 12 above. Verified the same way, plus a live query of the dev DB for the counts it quotes (email coverage, pending liquidación, installment fill rates) and of the staged Parquet for the legacy settlement-slot usage. Two of the four features are much smaller than they sound: the renewal-notice table, its idempotency key and the letter body already exist, and the per-policy liquidación fields are already wired end to end.
 
@@ -183,8 +186,9 @@ Unlike the ops items above, these block design decisions, not just infrastructur
 
 **Step 11 — utilities/ops side:**
 
-- OCR provider/budget for the statement auto-capture pipeline (self-hosted vs. a paid per-page API, given 300+ statements/month/service provider).
+- ~~OCR provider/budget~~ — **CLOSED**: self-hosted Tesseract, chosen on measured accuracy against real scans, so there is no per-page cost to approve.
-- Whether `PROPERTY_TAX.accountNumber` (migrated from `DATMEX.PREDIAL`) is actually the same number as "Clave Catastral" (`DATMEX.CLAVE`) — blocks OCR matching for predial statements until confirmed against a real bill.
+- ~~Whether `PROPERTY_TAX.accountNumber` (from `DATMEX.PREDIAL`) is the same number as "Clave Catastral" (`DATMEX.CLAVE`)~~ — **CLOSED**: they are different numbers. Answered from real CESPT bills plus the staged data; the clave is now migrated separately and predial was left alone.
+- Whether the CFE figure to charge is the rounded headline/barcode amount (`$268` — what is actually paid at the window) or the exact breakdown `Total` (`$268.88`). The parser takes the barcode amount; one confirmation from Jorge would settle it.
 - The actual bank name/currency/details for the Seguros USD account, and whether any historical Seguros bank register exists to migrate. (Multi-bank support itself is **built** — this is now only the missing content: staff can open the account in `/banco/cuentas` the moment the answer arrives, and it starts empty unless a historical register turns up.)
 - The exact "1 year inactivity" / "cancelled" triggers for customer-number recycling eligibility.
 - Whether customer-number recycling should ever include true PII purge (matching the office's paper-world habit) or archive-and-reuse-the-number is sufficient — recommended default is archive-only, consistent with this project's existing never-hard-delete convention.

RESUME.md

@@ -442,3 +442,87 @@ for what's actually next.
   verified vs dev: Anular buttons admin-gated, voided rows struck + excluded from totals,
   clicking Anular voids end-to-end (note: it uses a blocking `window.confirm`). Customer-detail
   mini tx list now also strikes voided rows ("(anulado)" tag) — was the last void-UI gap.
+
+---
+
+## Statement OCR intake (`/recibos`) — DONE 2026-08-01
+
+Plan step 11 §2 (`docs/RECEIPT_CAPTURE_SPEC.md` §2). The last big utilities
+feature: staff scan the month's utility bills and the machine proposes customer
++ amount per page, instead of keying 300+ statements per company by hand. Built
+in `apps/api/src/statements/` and `apps/web/src/app/recibos/`, posting through
+step 11 §1.2's `BillingService.createBatch` seam (`source: "OCR"`, per-document
+`captureRef`) so machine and hand capture share one write path and one audit
+trail. Abilities `statement:ingest` / `statement:review`, both STAFF.
+
+**Verified end to end against the live dev API + MinIO**, not just built: real
+CFE and Telnor scans uploaded over HTTP, OCR'd, matched, confirmed against a
+check, and the resulting rows checked in MySQL — negative (charge) amounts,
+`captureSource = OCR`, concept auto-derived from the batch's service kind,
+`captureRef` linking each transaction back to its page. Re-confirming a posted
+batch is refused. All test data was removed afterwards.
+
+**Everything here was decided from 10 real scanned statements (46 pages), not
+from the sample-free spec.** Shipped-parser results on them: provider 46/46,
+account reference 43/46, amount 42/46, due date 44/46; matched against the dev
+database, **39/46 (85%) exact auto-match, 40/46 (87%) identified**. The rest are
+real review cases (one shared account number, three phones not on file, one
+clave not in the book, one page too poor to read).
+
+Findings that corrected the spec, each of which changed the build:
+
+- **The scans have no text layer at all** — they are camera images of paper, so
+  OCR is mandatory rather than a convenience, and they arrive **bundled, one
+  customer per page**.
+- **Clave catastral is not predial.** `DATMEX.clave` (934 rows, `KA903009`) is
+  what CESPT and predial bills print; `DATMEX.predial` — which
+  `PROPERTY_TAX.accountNumber` holds — has only 663 distinct values across 1135
+  rows and appears on no statement. The clave now lives on
+  `Property.cadastralKey` as the matcher's secondary key; predial was left
+  untouched. This is the question that had been blocking predial matching.
+- **Gas was recoverable after all.** The spec said no legacy gas number existed;
+  in fact 160 of 334 `DATMEX.gas` values are real account numbers (the rest are
+  `ESTACIONARIO`/`CILINDRO` descriptors). Recovered into `GAS.meterNumber`.
+- **Phone is one billed line per property** (534 / 18 / 1 across phone1/2/3), so
+  `TELEPHONE` — a new `ServiceKind` — backfills from `phone1` only.
+- **Never match on the printed name.** A CESPT receipt for account `5365218`
+  reads `ARNAIZ ROSAS ELSA AURORA`; the office's book, corroborated by the
+  clave, has `CATT, RANDY`. The name on a utility bill is the registrant, not
+  the current owner.
+
+`migration/backfill_statement_match_fields.py` closes those three data gaps on
+an existing database (idempotent, wired into `run_all.py` after
+`transform_properties.py`, which now produces them directly on a full rebuild).
+Applied to dev: 934 claves, 160 gas numbers, 534 TELEPHONE rows.
+
+Implementation notes worth keeping:
+
+- OCR is self-hosted **Tesseract** behind an `OcrProvider` interface — the
+  provider question is closed on measured accuracy, and a managed API stays a
+  one-line swap in `statements.module.ts`. `tesseract-ocr`,
+  `tesseract-ocr-data-spa` and `poppler-utils` were added to the API image; if
+  they are missing the module reports itself unavailable and only this feature
+  is disabled.
+- **Payment barcodes beat printed labels.** One CFE label OCR'd a digit too
+  many while its barcode was correct, so the barcode is the source and the label
+  the cross-check; disagreement forces review.
+- **Detect the provider by brand first, layout only as a fallback** — and never
+  interleave the two passes. A scanned CESPT header came back as `E BAJA ES
+  PAGO / EALIFORNIA`, which is why the layout fallback exists; a Telnor page
+  contains words a CFE layout rule would otherwise claim, which is why ordering
+  matters.
+- **Parse amounts by separator position.** A real Telnor bill OCR'd as
+  `$ 649,00`; stripping commas as thousands separators turns that into $64,900.
+- Two of the three layouts are line-oriented, but the CESPT "RECIBO" is a
+  **table** whose values sit under column headers — that one needs the word
+  boxes, which is why `OcrPage` carries geometry and not just text.
+- Confirming a document whose matched service had no reference **writes the
+  reference back** (only into an empty field, and only when exactly one blank
+  service of that kind is a candidate), so gas and any other cold start is a
+  one-time cost rather than a permanent queue.
+- Handwritten folder numbers on the bills (`9`, `405`) are **not** used for
+  matching — Tesseract read `405` as `205`.
+
+**Open:** whether the CFE charge should be the rounded barcode/headline figure
+(`$268`, what is paid at the window — what the parser uses today) or the exact
+breakdown total (`$268.88`). One question for Jorge.

apps/api/jest.config.js

@@ -0,0 +1,7 @@
+/** @type {import('jest').Config} */
+module.exports = {
+  rootDir: "src",
+  testEnvironment: "node",
+  testRegex: ".*\\.spec\\.ts$",
+  transform: { "^.+\\.ts$": "ts-jest" },
+};

apps/api/nest-cli.json

@@ -3,6 +3,7 @@
   "collection": "@nestjs/schematics",
   "sourceRoot": "src",
   "compilerOptions": {
-    "deleteOutDir": true
+    "deleteOutDir": true,
+    "tsConfigPath": "tsconfig.build.json"
   }
 }

apps/api/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@jorgecuadros/api",
-  "version": "1.0.1",
+  "version": "1.0.6",
   "private": true,
   "scripts": {
     "build": "nest build",

apps/api/src/app.module.ts

@@ -9,6 +9,8 @@ import { CustomersModule } from "./customers/customers.module";
 import { PoliciesModule } from "./policies/policies.module";
 import { PropertiesModule } from "./properties/properties.module";
 import { BillingModule } from "./billing/billing.module";
+import { StatementsModule } from "./statements/statements.module";
+import { PolicyOcrModule } from "./policy-ocr/policy-ocr.module";
 import { BankModule } from "./bank/bank.module";
 import { OpsModule } from "./ops/ops.module";
 import { ReportsModule } from "./reports/reports.module";
@@ -26,6 +28,8 @@ import { AppController } from "./app.controller";
     PoliciesModule,
     PropertiesModule,
     BillingModule,
+    StatementsModule,
+    PolicyOcrModule,
     BankModule,
     OpsModule,
     ReportsModule,

apps/api/src/auth/abilities.ts

@@ -24,6 +24,8 @@ export type Ability =
   | "policy:create"
   | "policy:update"
   | "policy:delete"
+  | "policy:ingest"
+  | "policy:ocr-review"
   | "property:create"
   | "property:update"
   | "property:delete"
@@ -32,6 +34,8 @@ export type Ability =
   | "bank:create"
   | "bank:void"
   | "bank:manage-accounts"
+  | "statement:ingest"
+  | "statement:review"
   | "lookup:manage"
   | "user:manage"
   | "db:manage";
@@ -44,6 +48,10 @@ export const ABILITY_MIN: Record<Ability, Role> = {
   "policy:create": "STAFF",
   "policy:update": "STAFF",
   "policy:delete": "MANAGER",
+  // Insurance OCR intake is the same trust tier as statement OCR: STAFF can
+  // upload + confirm, nothing reaches the books unconfirmed.
+  "policy:ingest": "STAFF",
+  "policy:ocr-review": "STAFF",
   "property:create": "STAFF",
   "property:update": "STAFF",
   "property:delete": "MANAGER",
@@ -54,6 +62,12 @@ export const ABILITY_MIN: Record<Ability, Role> = {
   // Opening or renaming a chequera is rarer and higher-stakes than posting a
   // movement into one — a wrong account silently mixes two sets of books.
   "bank:manage-accounts": "MANAGER",
+  // Uploading a stack of scans and reviewing what the OCR read are both
+  // "capturing a receipt" — the same trust tier as ledger:create, since
+  // confirming a statement *is* capturing it. The review step is what makes
+  // this safe at STAFF level: nothing reaches the ledger unconfirmed.
+  "statement:ingest": "STAFF",
+  "statement:review": "STAFF",
   "lookup:manage": "MANAGER",
   "user:manage": "ADMIN",
   "db:manage": "ADMIN",

apps/api/src/billing/billing.module.ts

@@ -5,5 +5,8 @@ import { BillingService } from "./billing.service";
 @Module({
   controllers: [BillingController],
   providers: [BillingService],
+  // The statements module posts confirmed OCR captures through
+  // BillingService.createBatch rather than writing Transaction rows itself.
+  exports: [BillingService],
 })
 export class BillingModule {}

apps/api/src/ocr/ocr.module.ts

@@ -0,0 +1,18 @@
+import { Module } from "@nestjs/common";
+import { OCR_PROVIDER } from "../statements/ocr/ocr.provider";
+import { TesseractOcrProvider } from "../statements/ocr/tesseract.provider";
+
+/**
+ * Lifts the OCR seam out of StatementsModule so other modules (today:
+ * PolicyOcrModule) can inject OCR_PROVIDER without taking on the rest of
+ * the statement intake. StatementsModule itself imports this and gets the
+ * provider the same way.
+ *
+ * The concrete engine is still bound here — Tesseract today, a managed
+ * extraction API later is a one-line change in this file.
+ */
+@Module({
+  providers: [{ provide: OCR_PROVIDER, useClass: TesseractOcrProvider }],
+  exports: [OCR_PROVIDER],
+})
+export class OcrModule {}

apps/api/src/ops/ops.service.ts

@@ -262,9 +262,26 @@ export class OpsService implements OnModuleInit {
    * deploy/scripts/pre-migrate-backup.mjs — the two write into the same volume
    * and both are listed as restore points by this same screen.
    *
-   * --set-gtid-purged=OFF: the production server is the replication SOURCE with
+   * The dumper is probed at runtime rather than assumed. This command runs
-   * GTID on, so without it every dump embeds SET @@GLOBAL.GTID_PURGED and is
+   * inside the API image, whose `mysql-client` is Alpine's — i.e. MariaDB's —
-   * unrestorable onto the very server it came from.
+   * where `mysqldump` is a deprecation-warning shim over `mariadb-dump` that
+   * rejects --set-gtid-purged outright:
+   *   mysqldump: unknown variable 'set-gtid-purged=OFF'
+   * which failed every backup, including the safety backups SYNC and REIMPORT
+   * take first. MariaDB's dumper emits no GTID state unless asked (--gtid), so
+   * there is nothing to suppress there; the flag is passed only when the dumper
+   * on PATH advertises it, and the real binary is called directly only in the
+   * MariaDB case (calling `mariadb-dump` whenever it merely exists would pick
+   * it over a MySQL `mysqldump` earlier in PATH on a host carrying both).
+   *
+   * The probe is a command substitution, not `--help | grep -q`: PIPEFAIL is in
+   * effect and grep closing the pipe early would make a supported flag look
+   * unsupported.
+   *
+   * --set-gtid-purged=OFF (MySQL only): the production server is the
+   * replication SOURCE with GTID on, so without it every dump embeds
+   * SET @@GLOBAL.GTID_PURGED and is unrestorable onto the very server it came
+   * from.
    *
    * The table-count assertion is not belt-and-braces: `gzip -t` passes on the
    * ~372-byte output of a mysqldump that died on its first statement, so a
@@ -277,8 +294,12 @@ export class OpsService implements OnModuleInit {
    */
   private dumpCommand(flags: string, db: string, out: string): string {
     return (
-      `( mysqldump ${flags} --single-transaction --routines --triggers ` +
+      `DUMP=mysqldump; GTID=; ` +
-      `--no-tablespaces --set-gtid-purged=OFF ${db} | gzip -c > ${out} && ` +
+      `case "$(mysqldump --help 2>/dev/null || true)" in ` +
+      `*set-gtid-purged*) GTID=--set-gtid-purged=OFF;; ` +
+      `*) command -v mariadb-dump >/dev/null 2>&1 && DUMP=mariadb-dump;; esac; ` +
+      `( $DUMP ${flags} --single-transaction --routines --triggers ` +
+      `--no-tablespaces $GTID ${db} | gzip -c > ${out} && ` +
       `gzip -t ${out} && ` +
       `TABLAS=$(gunzip -c ${out} | grep -c 'CREATE TABLE') && ` +
       `echo "tablas capturadas: $TABLAS" && ` +

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

@@ -0,0 +1,147 @@
+import type { OcrPage } from "../../statements/ocr/ocr.provider";
+import {
+  detectPolicyProvider,
+  parsePolicy,
+  type ParsedCoverage,
+} from "./policy-parser";
+
+/**
+ * Verbatim excerpts of what the GMX portal's translation PDF actually
+ * rendered through pdftotext — same convention as the statement parser
+ * tests, where invented-clean input would test nothing because clean input
+ * is not the failure mode.
+ */
+function page(text: string): OcrPage {
+  return { text, words: [], confidence: 0.95 };
+}
+
+describe("detectPolicyProvider", () => {
+  it("claims GMX from the brand wordmark on the letterhead", () => {
+    expect(
+      detectPolicyProvider(
+        "Grupo Mexicano de Seguros, S.A. de C.V.\nTecoyotitla 412, Edificio GMX",
+      ),
+    ).toBe("GMX");
+  });
+
+  it("claims GMX from the 'gmx.com.mx' footer URL", () => {
+    expect(detectPolicyProvider("JUNTOS EL RIESGO ES MENOR\nwww.gmx.com.mx")).toBe("GMX");
+  });
+});
+
+describe("parsePolicy / GMX", () => {
+  // Verbatim text extracted from ~/Downloads/HC_Folio_000767_Traduccion.pdf via
+  // `pdftotext -layout`. Two pages joined by "\n\n".
+  const GMX_FULL = page(
+    "Multiple Policy\nHome\n" +
+      "Policy 007-037-07005947-0000-02 in accordance with the enclosed clauses, to insurance:\n" +
+      "Insured JON ASHLEY STRABALA\n" +
+      "Additional insured VIVIAN\n" +
+      "Legal address BONAMPACK No. EXT26 No.INT 0 COL. Punta Bandera, Tijuana, Baja California, C.P. 22550\n" +
+      "ZIP 22550 Income Tax No. XEXX-010101-000\n" +
+      "Broker (1176) Jorge Humberto Cuadros\n" +
+      "Term 12 months\n" +
+      "From 19/07/2026\n" +
+      "To 19/07/2027 at twelve hours (noon) Mexico City time.\n" +
+      "Currency DOLARES Premium payment CONTADO\n" +
+      "Free translation from the Spanish Insurance contract. The English text is just copy given by courtesy. In case of a dispute, the Spanish will prevail over the English version.\n" +
+      "Agreed clauses:\n" +
+      "•The insured and GMX Hereby declared...\n" +
+      "From the above, the present contract shall not be considered under the condition mentioned within article 36-B from the Insurance Companies General Law. Therefore it shall not be required its registration before the Comision National de Seguros y Fianzas.\n" +
+      "July 23, 2026\n" +
+      "Authority sign.\n" +
+      "Grupo Mexicano de Seguros, S.A. de C.V.\n" +
+      "Tecoyotitla 412, Edificio GMX\n" +
+      "JUNTOS EL RIESGO ES MENOR\n" +
+      "www.gmx.com.mx\n\n" +
+      "Risk Insured Amount Deductible Loss Participation\n" +
+      "Building $350,000.00 Not applies Not applies\n" +
+      "Contents $60,000.00 Not applies Not applies\n" +
+      "ADDITIONAL RISK\n" +
+      "Risk Insured Amount Deductible Loss Participation\n" +
+      "Debris removal Building $35,000.00 Not applies Not applies\n" +
+      "Debris removal Contents $6,000.00 Not applies Not applies\n" +
+      "Outdoors Constructions $10,000.00 5% 10%\n" +
+      "Coverage Extention Covered Not applies Not applies\n" +
+      "All Risk Covered Not applies Not applies\n" +
+      "Earthquake and/or volcanic eruption Covered 2% of the sum insured for each damage structure 20%\n" +
+      "Extra Expenses $41,000.00 Not applies Not applies\n" +
+      "Robbery with violence $10,000.00 Not applies Not applies\n" +
+      "Jewerly $3,900.00 Not applies Not applies\n" +
+      "Electronic Equipment $10,000.00 Not applies Not applies\n" +
+      "Glasses $10,000.00 Not applies Not applies\n" +
+      "Tenant $200,000.00 Not applies Not applies\n" +
+      "Family $200,000.00 Not applies Not applies\n" +
+      "Family $200,000.00 Not applies Not applies\n" +
+      "Domestic workers $7,010.00 Not applies Not applies\n" +
+      "VALUES ADDED, HOME GMX",
+  );
+
+  it("extracts the policy number, insured name, broker, dates, and currency", () => {
+    const p = parsePolicy(GMX_FULL);
+    expect(p.provider).toBe("GMX");
+    expect(p.policyNumber).toBe("007-037-07005947-0000-02");
+    expect(p.insuredName).toBe("JON ASHLEY STRABALA");
+    expect(p.additionalInsured).toBe("VIVIAN");
+    expect(p.agentName).toBe("Jorge Humberto Cuadros");
+    expect(p.policyFrom?.toISOString().slice(0, 10)).toBe("2026-07-19");
+    expect(p.policyTo?.toISOString().slice(0, 10)).toBe("2027-07-19");
+    expect(p.policyDate?.toISOString().slice(0, 10)).toBe("2026-07-23");
+    expect(p.currency).toBe("USD");
+    expect(p.zip).toBe("22550");
+    expect(p.legalAddress).toContain("BONAMPACK");
+    expect(p.premiumPayment).toBe("CONTADO");
+  });
+
+  it("extracts every coverage row off the second page table", () => {
+    const p = parsePolicy(GMX_FULL);
+    const byName = Object.fromEntries(p.coverages.map((c) => [c.risk, c]));
+    expect(byName.Building?.insuredAmount).toBe(350000);
+    expect(byName.Contents?.insuredAmount).toBe(60000);
+    expect(byName["Debris removal Building"]?.insuredAmount).toBe(35000);
+    expect(byName["Outdoors Constructions"]?.insuredAmount).toBe(10000);
+    expect(byName["Outdoors Constructions"]?.deductible).toBe("5%");
+    expect(byName["Outdoors Constructions"]?.lossParticipation).toBe("10%");
+    // Free-text coverage cells kept verbatim (the policy form surfaces them
+    // as observations, not as numbers).
+    expect(byName["Earthquake and/or volcanic eruption"]?.insuredAmount).toBeNull();
+    expect(byName["Earthquake and/or volcanic eruption"]?.deductible).toContain("2%");
+    expect(byName["Earthquake and/or volcanic eruption"]?.lossParticipation).toBe("20%");
+    expect(byName["All Risk"]?.insuredAmount).toBeNull();
+    expect(p.coverages.length).toBeGreaterThan(10);
+  });
+
+  it("leaves premium fields null on the certificate page and notes it", () => {
+    const p = parsePolicy(GMX_FULL);
+    expect(p.netPremium).toBeNull();
+    expect(p.total).toBeNull();
+    expect(p.policyFee).toBeNull();
+    expect(p.notes.join(" ")).toMatch(/prima/i);
+  });
+
+  it("still parses when the broker parens are missing", () => {
+    const p = parsePolicy(
+      page(
+        "Insured JON ASHLEY STRABALA\nBroker Jorge Humberto Cuadros\n" +
+          "From 19/07/2026\nTo 19/07/2027\nCurrency DOLARES\n" +
+          "Grupo Mexicano de Seguros",
+      ),
+    );
+    expect(p.agentName).toBe("Jorge Humberto Cuadros");
+  });
+
+  it("rejects a page that carries no GMX signal at all", () => {
+    const p = parsePolicy(page("Random unrelated document with no policy data."));
+    expect(p.provider).toBe("");
+    expect(p.notes.join(" ")).toContain("no se reconoció el proveedor");
+  });
+
+  it("captures the deductible / loss-participation columns verbatim as strings", () => {
+    const p = parsePolicy(GMX_FULL);
+    const eq = p.coverages.find((c) => c.risk === "Earthquake and/or volcanic eruption");
+    expect(eq).toBeDefined();
+    const eqTyped = eq as ParsedCoverage;
+    expect(eqTyped.deductible).toContain("sum insured");
+    expect(eqTyped.lossParticipation).toBe("20%");
+  });
+});

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

@@ -0,0 +1,422 @@
+import type { OcrPage } from "../../statements/ocr/ocr.provider";
+
+/**
+ * What one parsed policy page yields. All fields are nullable because each
+ * provider prints a different subset (GMX's certificate has no premium
+ * breakdown, only insured amounts; GMX's receipt page would carry the
+ * premium), and the matcher + the review queue both work better with
+ * "field was read" vs "field was not" rather than guessing.
+ */
+export interface ParsedPolicy {
+  /** "GMX" today; the dispatcher lives on `detectProvider`. */
+  provider: string;
+  policyNumber: string | null;
+  insuredName: string | null;
+  additionalInsured: string | null;
+  /** The "Broker" line on GMX — mapped onto `Policy.agentName`. */
+  agentName: string | null;
+  legalAddress: string | null;
+  zip: string | null;
+  policyFrom: Date | null;
+  policyTo: Date | null;
+  /** Signature/issue date — `Policy.policyDate`. */
+  policyDate: Date | null;
+  /** "MXN" | "USD" | …, derived from the printed currency word. */
+  currency: string | null;
+  netPremium: number | null;
+  policyFee: number | null;
+  brokerFee: number | null;
+  total: number | null;
+  /** "CONTADO" / "MENSUAL" / … — premium-payment cadence text. */
+  premiumPayment: string | null;
+  /**
+   * GMX prints per-coverage rows in a table: Building / Contents /
+   * Earthquake / … with insured amount, deductible, loss participation.
+   * Preserved verbatim so a missing premium receipt still leaves the
+   * coverages auditable on the Policy row.
+   */
+  coverages: ParsedCoverage[];
+  /** Human-readable trail of what was read, surfaced in the review queue. */
+  notes: string[];
+}
+
+export interface ParsedCoverage {
+  /** "Building", "Contents", "Debris removal Building", "Earthquake…". */
+  risk: string;
+  insuredAmount: number | null;
+  deductible: string | null;
+  lossParticipation: string | null;
+}
+
+// --- shared helpers ---------------------------------------------------------
+
+const DIGIT_CONFUSIONS: Record<string, string> = {
+  O: "0", o: "0", D: "0", I: "1", l: "1", "|": "1", S: "5", B: "8",
+};
+
+/**
+ * Tesseract confuses these glyphs inside numeric runs with some regularity.
+ * Same map and same caveat as the statement parser: ONLY apply to fields
+ * known to be digits, never to free text.
+ */
+function toDigits(s: string | null | undefined): string {
+  if (!s) return "";
+  return s
+    .split("")
+    .map((c) => DIGIT_CONFUSIONS[c] ?? c)
+    .join("")
+    .replace(/\D/g, "");
+}
+
+/**
+ * Parse a printed amount, treating `,` and `.` by position rather than by
+ * assumption. Same algorithm as the statement parser — kept here so the
+ * policy module is self-contained, since importing from `../../statements`
+ * would couple two unrelated domains through a helper.
+ */
+function money(s: string | null | undefined): number | null {
+  if (!s) return null;
+  const cleaned = s.replace(/[\s$]/g, "");
+
+  let m = cleaned.match(/^(\d{1,3}(?:[.,]\d{3})+)([.,]\d{1,2})?$/);
+  if (m) {
+    const whole = m[1].replace(/[.,]/g, "");
+    const cents = m[2] ? m[2].slice(1) : "";
+    return Number(cents ? `${whole}.${cents.padEnd(2, "0")}` : whole);
+  }
+
+  m = cleaned.match(/^(\d+)[.,](\d{2})$/);
+  if (m) return Number(`${m[1]}.${m[2]}`);
+
+  const n = Number(cleaned.replace(/[,.]/g, ""));
+  return Number.isFinite(n) ? n : null;
+}
+
+function firstMatch(text: string, patterns: RegExp[]): string | null {
+  for (const p of patterns) {
+    const m = text.match(p);
+    if (m?.[1]) return m[1].trim();
+  }
+  return null;
+}
+
+const MONTHS: Record<string, number> = {
+  ENE: 0, FEB: 1, MAR: 2, ABR: 3, MAY: 4, JUN: 5,
+  JUL: 6, AGO: 7, SEP: 8, OCT: 9, NOV: 10, DIC: 11,
+};
+
+/**
+ * DD/MM/YYYY (GMX) and the dash-separated ISO variants. Two-digit years are
+ * windowed: < 50 → 20YY, ≥ 50 → 19YY, matching what a 1950-2049 window
+ * expects from a paper document.
+ */
+function parseDate(raw: string | null | undefined): Date | null {
+  if (!raw) return null;
+  const s = raw.trim();
+
+  let m = s.match(/^(\d{1,2})\/(\d{1,2})\/(\d{4})$/);
+  if (m) return utc(+m[3], +m[2] - 1, +m[1]);
+
+  m = s.match(/^(\d{1,2})[-\s/]([A-Z]{3})[-\s/](\d{2,4})$/i);
+  if (m && MONTHS[m[2].toUpperCase()] !== undefined) {
+    const yr = +m[3];
+    const y = m[3].length === 2 ? (yr < 50 ? 2000 + yr : 1900 + yr) : yr;
+    return utc(y, MONTHS[m[2].toUpperCase()], +m[1]);
+  }
+
+  m = s.match(/^(\d{4})[-/](\d{1,2})[-/](\d{1,2})$/);
+  if (m) return utc(+m[1], +m[2] - 1, +m[3]);
+
+  // "July 23, 2026" — the signature date on the GMX certificate.
+  m = s.match(/^([A-Za-z]+)\s+(\d{1,2}),\s*(\d{4})$/);
+  if (m) {
+    const MONTH_NAMES: Record<string, number> = {
+      january: 0, february: 1, march: 2, april: 3, may: 4, june: 5,
+      july: 6, august: 7, september: 8, october: 9, november: 10, december: 11,
+    };
+    const mo = MONTH_NAMES[m[1].toLowerCase()];
+    if (mo !== undefined) return utc(+m[3], mo, +m[2]);
+  }
+
+  return null;
+}
+
+function utc(y: number, mo: number, d: number): Date | null {
+  const dt = new Date(Date.UTC(y, mo, d));
+  return Number.isNaN(dt.getTime()) ? null : dt;
+}
+
+/** Map the printed currency word onto an ISO code. */
+function currencyCode(raw: string | null | undefined): string | null {
+  if (!raw) return null;
+  const s = raw.trim().toUpperCase();
+  if (s.startsWith("PESO") || s === "MXN" || s.includes("NACIONAL")) return "MXN";
+  if (s.startsWith("DOLAR") || s === "USD" || s.includes("DOLLAR")) return "USD";
+  if (s === "EUR" || s.includes("EURO")) return "EUR";
+  return null;
+}
+
+// --- provider detection -----------------------------------------------------
+
+/**
+ * Brand first, layout as a fallback. Same ordering rule as the statement
+ * parser: a brand wordmark is the cheapest, most reliable discriminator, and
+ * a layout rule that runs first can wrongly claim a page that happens to
+ * carry the same shape string (the statement parser's lesson with CFE vs
+ * GAS on "PERIODO FACTURADO").
+ */
+const BRAND: [string, RegExp][] = [
+  ["GMX", /\bGMX\b|Grupo\s*Mexicano\s*de\s*Seguros|gmx\.com\.mx|JUNTOS\s*EL\s*RIESGO\s*ES\s*MENOR/i],
+];
+
+const LAYOUT: [string, RegExp][] = [
+  ["GMX", /Multiple\s*Policy|IMPUESTO\s*PREDIAL[\s\S]{0,80}EN\s*FECHA|Material\s*damages\s*Section/i],
+];
+
+export function detectPolicyProvider(text: string): string | null {
+  for (const group of [BRAND, LAYOUT]) {
+    for (const [name, pattern] of group) {
+      if (pattern.test(text)) return name;
+    }
+  }
+  return null;
+}
+
+// --- parsers ----------------------------------------------------------------
+
+const PARSERS: Record<string, (page: OcrPage) => ParsedPolicy> = {
+  GMX: parseGmx,
+};
+
+const EMPTY_COVERAGE: ParsedCoverage = {
+  risk: "",
+  insuredAmount: null,
+  deductible: null,
+  lossParticipation: null,
+};
+
+export function parsePolicy(page: OcrPage): ParsedPolicy {
+  const provider = detectPolicyProvider(page.text);
+  if (!provider) {
+    return {
+      provider: "",
+      policyNumber: null,
+      insuredName: null,
+      additionalInsured: null,
+      agentName: null,
+      legalAddress: null,
+      zip: null,
+      policyFrom: null,
+      policyTo: null,
+      policyDate: null,
+      currency: null,
+      netPremium: null,
+      policyFee: null,
+      brokerFee: null,
+      total: null,
+      premiumPayment: null,
+      coverages: [],
+      notes: ["no se reconoció el proveedor"],
+    };
+  }
+  return PARSERS[provider](page);
+}
+
+// --- GMX --------------------------------------------------------------------
+
+/**
+ * GMX policy certificate layout (this is the translation PDF — the Spanish
+ * version is the canonical source, but every GMX portal download is a
+ * translation so the parser can rely on these English labels).
+ *
+ * Page 1 carries the contract header in a single boxed table:
+ *   Policy | Insured | Additional insured | Legal address | ZIP | Income Tax No.
+ *   Broker | Term | From | To | Currency | Premium payment
+ * followed by an "Agreed clauses" block, the signature date, and the GMX
+ * letterhead.
+ *
+ * Page 2 carries the per-coverage table (Risk / Insured Amount / Deductible /
+ * Loss Participation) under "Material damages Section" and "ADDITIONAL RISK".
+ *
+ * Premium / total / fees are NOT on the certificate page — they live on
+ * GMX's separate "recibo" PDF. The parser leaves them null and flags the
+ * gap in `notes`; the matcher still proposes a Policy update from the
+ * certificate alone, and the staff confirm step fills premium in by hand
+ * or after a follow-up receipt upload.
+ */
+function parseGmx(page: OcrPage): ParsedPolicy {
+  const text = page.text;
+  const notes: string[] = [];
+
+  // ----- header table (page 1) --------------------------------------------
+  // The Policy row repeats the number in a long run:
+  //   "Policy 007-037-07005947-0000-02 in accordance with the enclosed clauses…"
+  // so taking the first token-shaped number is correct; the trailing prose
+  // never looks like one. The dashes are part of the printed number — keep
+  // them (don't run toDigits, which would flatten them).
+  const policyNumber = firstMatch(text, [
+    /\bPolicy\s+([0-9OIlSBD]{3,4}[-\s][0-9OIlSBD]{3}[-\s][0-9OIlSBD]{8}[-\s][0-9OIlSBD]{4}[-\s][0-9OIlSBD]{2})/i,
+    /\bPolicy\s+([0-9OIlSBD][0-9OIlSBD\s-]{9,30})/,
+  ]);
+
+  // "Insured   JON ASHLEY STRABALA" — label, then 1+ whitespace, then the name.
+  // Names can carry accents (ÁVILA) or apostrophes (O'NEILL); the label is
+  // always upper-case English on this layout, so case is reliable.
+  const insuredName = labelValue(text, /^Insured\s+([A-ZÁÉÍÓÚÑ'][A-ZÁÉÍÓÚÑ '\-.]+)$/m);
+  const additionalInsured = labelValue(text, /^Additional\s+insured\s+([A-ZÁÉÍÓÚÑ '\-.]+)$/m);
+
+  // Legal address is a single long line; the parser keeps it whole.
+  const legalAddress = labelValue(text, /^Legal\s+address\s+(.+)$/m);
+  const zip = labelValue(text, /^ZIP\s+(\d{4,6})\b/m);
+  if (!zip && legalAddress) {
+    // Last resort: zip often appears at the tail of the address run too
+    // ("…C.P. 22550"). Cheap regex, no false-positive cost on this layout.
+    const m = legalAddress.match(/\b(\d{5})\b/);
+    if (m) notes.push(`ZIP leído de la dirección (${m[1]})`);
+  }
+
+  // Broker line on GMX: "(1176) Jorge Humberto Cuadros" — the number is the
+  // agent code, the name is what lands on `Policy.agentName`. The parens
+  // are optional: a future layout or scan drop them.
+  const brokerRaw = labelValue(text, /^Broker\s+(?:\(\d+\)\s*)?(.+)$/m);
+  const agentName = brokerRaw?.trim() ?? null;
+
+  // Term: "12 months" — informational, not a free-standing date. Stored in
+  // notes; the UI can derive `coveragePeriodDays` from From/To anyway.
+  const term = firstMatch(text, [/^Term\s+(\d+\s+months?)$/m]);
+  if (term) notes.push(`vigencia: ${term}`);
+
+  const policyFrom = parseDate(
+    labelValue(text, /^From\s+(\d{1,2}\/\d{1,2}\/\d{4})\b/m),
+  );
+  const policyTo = parseDate(
+    firstMatch(text, [/^To\s+(\d{1,2}\/\d{1,2}\/\d{4})\b/m]),
+  );
+
+  // "at twelve hours (noon) Mexico City time." — kept in notes only.
+  if (/twelve\s*hours|noon/i.test(text)) notes.push("vencimiento a las 12:00 hora del centro");
+
+  // The Currency / Premium payment cells sit next to each other on one
+  // line; pull them with bounded matches so the trailing label of the
+  // adjacent cell doesn't swallow the wrong value.
+  const currency = currencyCode(labelValue(text, /^Currency\s+(\S+?)(?:\s+Premium\s+payment|$)/m));
+  const premiumPayment = labelValue(text, /Premium\s+payment\s+(\S+)$/m);
+
+  // ----- signature date (page 1) -----------------------------------------
+  // Appears above the signature line on its own: "July 23, 2026".
+  const dateMatch = text.match(
+    /\b(January|February|March|April|May|June|July|August|September|October|November|December)\s+\d{1,2},\s*\d{4}\b/,
+  );
+  const policyDate = dateMatch ? parseDate(dateMatch[0]) : null;
+  if (!policyDate) notes.push("no se pudo leer la fecha de firma");
+
+  // ----- coverages table (page 2) -----------------------------------------
+  const coverages = parseGmxCoverages(text, notes);
+
+  if (!policyNumber) notes.push("no se pudo leer el número de póliza");
+  if (!policyFrom || !policyTo) notes.push("no se pudo leer el período de vigencia");
+  // Premium fields are expected to be missing on the certificate page; flag
+  // it explicitly so the reviewer knows to look for a separate receipt.
+  if (!text.match(/Prima\s*neta|net\s*premium/i)) {
+    notes.push("esta página no trae prima; revisar el recibo de GMX por separado");
+  }
+
+  return {
+    provider: "GMX",
+    policyNumber: policyNumber ? policyNumber.replace(/\s+/g, "") : null,
+    insuredName,
+    additionalInsured,
+    agentName,
+    legalAddress,
+    zip,
+    policyFrom,
+    policyTo,
+    policyDate,
+    currency,
+    netPremium: null,
+    policyFee: null,
+    brokerFee: null,
+    total: null,
+    premiumPayment,
+    coverages,
+    notes,
+  };
+}
+
+/**
+ * Read the value that follows a `LABEL` on the same line. Used by every
+ * "Label    Value" cell on the GMX header table — matches on the line
+ * itself rather than across the page, so a label that also appears in body
+ * text can't accidentally claim a different cell.
+ */
+function labelValue(text: string, pattern: RegExp): string | null {
+  const m = text.match(pattern);
+  if (!m?.[1]) return null;
+  return m[1].replace(/\s+/g, " ").trim();
+}
+
+/**
+ * Walk the GMX per-coverage table on page 2.
+ *
+ * Real sample row (single-line representation of the table after pdftotext
+ * flattens it; the real layout uses fixed columns):
+ *   "Building                  $350,000.00   Not applies   Not applies"
+ *
+ * The four columns are:
+ *   Risk (left), Insured Amount ($ figure OR the word "Covered"),
+ *   Deductible (free text — "Not applies", "5%", "2% of the sum insured…"),
+ *   Loss Participation (same).
+ *
+ * "Covered" means the coverage is included with no dollar cap. We record
+ * the word so the review queue surfaces it instead of inventing a number.
+ *
+ * Deductible / Loss Participation are kept as printed strings, not
+ * converted to numbers — a "20%" loss participation is a different field
+ * shape from a "$5,000" deductible and the JSON column lets the UI render
+ * either verbatim.
+ *
+ * Multi-line cells (the "Earthquake" row's deductible wraps to three lines
+ * because the column is narrow) are collapsed by joining consecutive
+ * non-table-body lines onto the previous row's deductible cell before
+ * applying the column regex.
+ */
+function parseGmxCoverages(text: string, notes: string[]): ParsedCoverage[] {
+  const out: ParsedCoverage[] = [];
+
+  // Stop at "VALUES ADDED" — the trailing prose section (homeowner
+  // services, legal text) is not a coverage table. Re-enter at
+  // "ADDITIONAL RISK" for the second coverage block on page 2.
+  const segments = text.split(/VALUES\s*ADDED/i)[0].split(/ADDITIONAL\s*RISK/i);
+
+  // `[ \t]` (not `\s`) inside a cell: the deductible/loss-participation
+  // columns may wrap onto several lines in the raw `pdftotext` output, and
+  // matching across newlines silently swallows the next row.
+  const re = /^([A-Za-zÁÉÍÓÚÑ][A-Za-zÁÉÍÓÚÑ /\-.]+?)[ \t]+(\$[\d,.]+|Covered|Not[ \t]+applies)[ \t]+(\S+(?:[ \t]\S+){0,8})[ \t]+(\S+(?:[ \t]\S+){0,8})[ \t]*$/gim;
+  let m: RegExpExecArray | null;
+  for (const seg of segments) {
+    re.lastIndex = 0;
+    while ((m = re.exec(seg)) !== null) {
+      const risk = m[1].trim();
+      const amountCell = m[2].trim();
+      const deductible = m[3].trim();
+      const lossParticipation = m[4].trim();
+
+      // Skip the "Risk / Insured Amount / Deductible / Loss Participation"
+      // header row itself, which matches the same regex.
+      if (/^Risk$/i.test(risk) && /Insured\s*Amount/i.test(amountCell)) continue;
+
+      out.push({
+        risk,
+        insuredAmount:
+          amountCell === "Covered" || amountCell === "Not applies"
+            ? null
+            : money(amountCell),
+        deductible,
+        lossParticipation,
+      });
+    }
+  }
+
+  if (out.length === 0) notes.push("no se encontraron coberturas en la tabla");
+  return out;
+}

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

@@ -0,0 +1,101 @@
+import { Injectable } from "@nestjs/common";
+import { PrismaService } from "../prisma/prisma.service";
+import type { ParsedPolicy } from "./parsers/policy-parser";
+
+export interface MatchResult {
+  policyId: string | null;
+  customerId: string | null;
+  /** Why it landed here — shown in the review queue verbatim. */
+  note: string;
+  /** True only for an unambiguous hit on `Policy.policyNumber`. */
+  confident: boolean;
+  /**
+   * Every policy that carries the parsed number, with its customer. >1 means
+   * the policy number is shared across customers and a human must pick.
+   */
+  candidates: { policyId: string; customerId: string; customerName: string; policyNumber: string }[];
+}
+
+/**
+ * Resolves a parsed policy page to an existing Policy (and its customer) the
+ * office already holds.
+ *
+ * **Match on `Policy.policyNumber` alone, never on the printed insured name.**
+ * The certificate's "Insured" line is the account's registrant, which drifts
+ * from the current owner — the same problem the statement matcher cites for
+ * utility bills ("ARNAIZ ROSAS ELSA AURORA" on a CESPT receipt for a
+ * customer this office holds as "CATT, RANDY"). Names are surfaced for the
+ * reviewer to sanity-check and never feed matching.
+ *
+ * A policy number that matches zero rows means the policy is new: the
+ * review screen then offers a customer picker and the confirm step creates
+ * the row. Multiple hits are surfaced rather than auto-picked — duplicate
+ * policy numbers across customers do occur (same group policy bound by two
+ * related parties), and picking one arbitrarily would silently book the
+ * wrong coverage.
+ */
+@Injectable()
+export class PolicyMatcherService {
+  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");
+    }
+
+    const rows = await this.prisma.policy.findMany({
+      where: { policyNumber: parsed.policyNumber },
+      select: {
+        id: true,
+        policyNumber: true,
+        customerId: true,
+        customer: { select: { name: true } },
+      },
+    });
+
+    const candidates = rows.map((r) => ({
+      policyId: r.id,
+      customerId: r.customerId,
+      customerName: r.customer.name,
+      policyNumber: r.policyNumber,
+    }));
+
+    if (rows.length === 0) {
+      return {
+        policyId: null,
+        customerId: null,
+        note: `no se encontró ninguna póliza con el número ${parsed.policyNumber}`,
+        confident: false,
+        candidates: [],
+      };
+    }
+
+    if (rows.length > 1) {
+      return {
+        policyId: null,
+        customerId: null,
+        note: `${rows.length} pólizas comparten el número ${parsed.policyNumber}`,
+        confident: false,
+        candidates,
+      };
+    }
+
+    return {
+      policyId: candidates[0].policyId,
+      customerId: candidates[0].customerId,
+      note: `coincidencia exacta por número de póliza ${parsed.policyNumber}`,
+      confident: true,
+      candidates,
+    };
+  }
+
+  private unmatched(note: string): MatchResult {
+    return {
+      policyId: null,
+      customerId: null,
+      note,
+      confident: false,
+      candidates: [],
+    };
+  }
+}

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

@@ -0,0 +1,161 @@
+import {
+  Body,
+  Controller,
+  Get,
+  Param,
+  Patch,
+  Post,
+  Query,
+  Req,
+  Res,
+  StreamableFile,
+  UploadedFiles,
+  UseGuards,
+  UseInterceptors,
+} from "@nestjs/common";
+import { FilesInterceptor } from "@nestjs/platform-express";
+import type { Request, Response } from "express";
+import { AuthenticatedGuard } from "../auth/authenticated.guard";
+import { AbilityGuard } from "../auth/ability.guard";
+import { RequireAbility } from "../auth/require-ability.decorator";
+import { AuditService } from "../common/audit.service";
+import type { UploadedFileLike } from "../storage/upload-file";
+import { PolicyOcrService } from "./policy-ocr.service";
+import {
+  ConfirmPolicyBatchDto,
+  CreatePolicyOcrBatchDto,
+  ReviewPolicyDocumentDto,
+} from "./policy-ocr.dto";
+
+/**
+ * Insurance OCR intake (policy_ocr_intake).
+ *
+ * Mirrors StatementsController shape: one batch = one upload session of
+ * policy PDFs from a provider portal (GMX today), one document per page.
+ * Confirming a batch delegates nothing to a separate billing path —
+ * everything goes through `Policy` (and optionally a Transaction for the
+ * premium), the same tables the manual `PolicyForm` writes.
+ */
+@Controller("policy-ocr")
+@UseGuards(AuthenticatedGuard, AbilityGuard)
+export class PolicyOcrController {
+  constructor(
+    private readonly policyOcr: PolicyOcrService,
+    private readonly audit: AuditService,
+  ) {}
+
+  private actingId(req: Request): string {
+    return (req.user as { id: string } | undefined)?.id ?? "";
+  }
+
+  @Get("status")
+  async status() {
+    return {
+      ocrAvailable: await this.policyOcr.ocrAvailable(),
+      storageAvailable: this.policyOcr.storageAvailable(),
+    };
+  }
+
+  @Get("batches")
+  listBatches(@Query("page") page?: string, @Query("pageSize") pageSize?: string) {
+    return this.policyOcr.listBatches(
+      Math.max(1, Number(page) || 1),
+      Math.min(100, Math.max(1, Number(pageSize) || 25)),
+    );
+  }
+
+  @Get("batches/:id")
+  getBatch(@Param("id") id: string) {
+    return this.policyOcr.getBatch(id);
+  }
+
+  @Get("batches/:id/documents")
+  listDocuments(@Param("id") id: string) {
+    return this.policyOcr.listDocuments(id);
+  }
+
+  /**
+ * The source PDF for a parsed policy document. One PDF = one parsed policy,
+ * so this returns the entire upload (typically multi-page for insurance
+ * certificates). The review screen embeds it in an iframe.
+ */
+@Get("documents/:id/page")
+async pageImage(
+  @Param("id") id: string,
+  @Res({ passthrough: true }) res: Response,
+) {
+  const { stream, contentType, contentLength } = await this.policyOcr.pageImage(id);
+  res.set({
+    // The doc row stores the source PDF, not a rendered page image.
+    "Content-Type": contentType ?? "application/pdf",
+    ...(contentLength ? { "Content-Length": String(contentLength) } : {}),
+  });
+  return new StreamableFile(stream);
+}
+
+  // --- writes ---------------------------------------------------------------
+
+  @Post("batches")
+  @RequireAbility("policy:ingest")
+  @UseInterceptors(
+    FilesInterceptor("files", 25, { limits: { fileSize: 50 * 1024 * 1024 } }),
+  )
+  async createBatch(
+    @UploadedFiles() files: UploadedFileLike[] | undefined,
+    @Body() _dto: CreatePolicyOcrBatchDto,
+    @Query("label") label: string | undefined,
+    @Req() req: Request,
+  ) {
+    const batch = await this.policyOcr.createBatch(
+      files ?? [],
+      this.actingId(req),
+      label ?? _dto.label,
+    );
+    void this.audit.log(this.actingId(req), "policyOcr.batch.create", {
+      batchId: batch.id,
+      fileCount: batch.fileCount,
+    });
+    return batch;
+  }
+
+  @Patch("documents/:id")
+  @RequireAbility("policy:ocr-review")
+  async review(
+    @Param("id") id: string,
+    @Body() dto: ReviewPolicyDocumentDto,
+    @Req() req: Request,
+  ) {
+    const doc = await this.policyOcr.review(id, dto, this.actingId(req));
+    void this.audit.log(this.actingId(req), "policyOcr.document.review", {
+      documentId: id,
+      status: doc.status,
+    });
+    return doc;
+  }
+
+  @Post("documents/:id/reject")
+  @RequireAbility("policy:ocr-review")
+  async reject(@Param("id") id: string, @Req() req: Request) {
+    const doc = await this.policyOcr.reject(id, this.actingId(req));
+    void this.audit.log(this.actingId(req), "policyOcr.document.reject", {
+      documentId: id,
+    });
+    return doc;
+  }
+
+  @Post("batches/:id/confirm")
+  @RequireAbility("policy:ocr-review")
+  async confirm(
+    @Param("id") id: string,
+    @Body() dto: ConfirmPolicyBatchDto,
+    @Req() req: Request,
+  ) {
+    const result = await this.policyOcr.confirmBatch(id, dto, this.actingId(req));
+    void this.audit.log(this.actingId(req), "policyOcr.batch.confirm", {
+      batchId: id,
+      applied: result.applied,
+      postedTransactions: result.postedTransactions,
+    });
+    return result;
+  }
+}

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

@@ -0,0 +1,85 @@
+import { Type } from "class-transformer";
+import {
+  IsArray,
+  IsDateString,
+  IsEnum,
+  IsNumber,
+  IsObject,
+  IsOptional,
+  IsString,
+  MinLength,
+  ValidateNested,
+} from "class-validator";
+
+/** One document's confirmed-after-review state. The service reads these
+ *  fields and writes them onto either a matched Policy or a freshly created
+ *  one. Anything null here is not written. */
+export class ConfirmPolicyDocumentDto {
+  @IsString() documentId!: string;
+
+  /** Required when creating a new Policy; ignored if `policyId` is set. */
+  @IsOptional() @IsString() customerId?: string;
+  /** Set when the document matched an existing Policy. */
+  @IsOptional() @IsString() policyId?: string;
+
+  @IsOptional() @IsString() policyNumber?: string;
+  @IsOptional() @IsString() insuredName?: string;
+  @IsOptional() @IsString() additionalInsured?: string;
+  @IsOptional() @IsString() agentName?: string;
+  @IsOptional() @IsString() legalAddress?: string;
+  @IsOptional() @IsString() zip?: string;
+  @IsOptional() @IsDateString() policyFrom?: string;
+  @IsOptional() @IsDateString() policyTo?: string;
+  @IsOptional() @IsDateString() policyDate?: string;
+  @IsOptional() @IsEnum(["MXN", "USD", "EUR"]) currency?: "MXN" | "USD" | "EUR";
+  @IsOptional() @IsNumber() netPremium?: number;
+  @IsOptional() @IsNumber() policyFee?: number;
+  @IsOptional() @IsNumber() brokerFee?: number;
+  @IsOptional() @IsNumber() total?: number;
+  @IsOptional() @IsString() premiumPayment?: string;
+  /** Coverages parsed off the PDF, passed through verbatim to Policy.coveragesJson. */
+  @IsOptional() @IsObject() coveragesJson?: unknown;
+
+  /** When true, write a Transaction(domain=INSURANCE, amount=-netPremium)
+   *  in addition to creating/updating the Policy. Skipped if netPremium is
+   *  null or zero. */
+  @IsOptional() postPremium?: boolean;
+}
+
+export class ConfirmPolicyBatchDto {
+  @IsArray()
+  @ValidateNested({ each: true })
+  @Type(() => ConfirmPolicyDocumentDto)
+  documents!: ConfirmPolicyDocumentDto[];
+}
+
+/** Staff correction of one document's extracted fields or its match. */
+export class ReviewPolicyDocumentDto {
+  @IsOptional() @IsString() policyNumber?: string;
+  @IsOptional() @IsString() insuredName?: string;
+  @IsOptional() @IsString() additionalInsured?: string;
+  @IsOptional() @IsString() agentName?: string;
+  @IsOptional() @IsString() legalAddress?: string;
+  @IsOptional() @IsString() zip?: string;
+  @IsOptional() @IsDateString() policyFrom?: string;
+  @IsOptional() @IsDateString() policyTo?: string;
+  @IsOptional() @IsDateString() policyDate?: string;
+  @IsOptional() @IsString() currency?: string;
+  @IsOptional() @IsNumber() netPremium?: number;
+  @IsOptional() @IsNumber() policyFee?: number;
+  @IsOptional() @IsNumber() brokerFee?: number;
+  @IsOptional() @IsNumber() total?: number;
+  @IsOptional() @IsString() premiumPayment?: string;
+  @IsOptional() @IsObject() coveragesJson?: unknown;
+
+  /** Set by the reviewer when the document matched an existing Policy. */
+  @IsOptional() @IsString() matchedPolicyId?: string;
+  /** Set by the reviewer when creating a new Policy. */
+  @IsOptional() @IsString() matchedCustomerId?: string;
+  /** Force-confirm a doc even when the matcher left it ambiguous. */
+  @IsOptional() forceConfirm?: boolean;
+}
+
+export class CreatePolicyOcrBatchDto {
+  @IsOptional() @IsString() @MinLength(1) label?: string;
+}

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

@@ -0,0 +1,18 @@
+import { Module } from "@nestjs/common";
+import { OcrModule } from "../ocr/ocr.module";
+import { PolicyOcrController } from "./policy-ocr.controller";
+import { PolicyOcrService } from "./policy-ocr.service";
+import { PolicyMatcherService } from "./policy-matcher.service";
+
+/**
+ * Reuses the OCR seam from OcrModule unchanged: the Tesseract provider is
+ * bound there and `OcrProvider` is the only thing the parsers touch. This
+ * module registers its own controller + service + matcher; nothing about
+ * utility ingestion needs to know about it.
+ */
+@Module({
+  imports: [OcrModule],
+  controllers: [PolicyOcrController],
+  providers: [PolicyOcrService, PolicyMatcherService],
+})
+export class PolicyOcrModule {}

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

@@ -0,0 +1,720 @@
+import {
+  BadRequestException,
+  Inject,
+  Injectable,
+  Logger,
+  NotFoundException,
+} from "@nestjs/common";
+import { Currency, Prisma } from "@jorgecuadros/database";
+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 { PolicyMatcherService } from "./policy-matcher.service";
+import type {
+  ConfirmPolicyBatchDto,
+  ConfirmPolicyDocumentDto,
+  ReviewPolicyDocumentDto,
+} from "./policy-ocr.dto";
+
+/**
+ * Insurance OCR intake — mirrors the statement pipeline at
+ * `apps/api/src/statements/statements.service.ts`. Reuses the OCR seam and
+ * Tesseract binding unchanged; the parsers and matcher are policy-specific.
+ *
+ * Why a parallel pipeline rather than a column on StatementDocument: the
+ * matcher keys on `Policy.policyNumber`, the confirm step writes to a
+ * different table (`Policy`, not `Transaction`), and the review UI shows
+ * different fields. Sharing one queue would either bloat the row with null
+ * columns or force the review screen to branch on a discriminator — both
+ * worse than a thin second table.
+ */
+@Injectable()
+export class PolicyOcrService {
+  private readonly logger = new Logger(PolicyOcrService.name);
+
+  constructor(
+    private readonly prisma: PrismaService,
+    private readonly storage: StorageService,
+    private readonly matcher: PolicyMatcherService,
+    @Inject(OCR_PROVIDER) private readonly ocr: OcrProvider,
+  ) {}
+
+  ocrAvailable(): Promise<boolean> {
+    return this.ocr.available();
+  }
+
+  storageAvailable(): boolean {
+    return this.storage.available;
+  }
+
+  // --- ingest ---------------------------------------------------------------
+
+  async createBatch(
+    files: UploadedFileLike[],
+    uploadedById: string,
+    label?: string,
+  ) {
+    if (!files?.length) throw new BadRequestException("No se recibió ningún archivo.");
+    if (!(await this.ocr.available())) {
+      throw new BadRequestException(
+        "El servidor no tiene OCR instalado; no se pueden leer PDFs de pólizas.",
+      );
+    }
+    if (!this.storage.available) {
+      throw new BadRequestException(
+        "El almacenamiento de documentos no está configurado; no se pueden " +
+          "guardar los PDFs escaneados.",
+      );
+    }
+
+    const batch = await this.prisma.policyOcrBatch.create({
+      data: { provider: "GMX", uploadedById, label, fileCount: files.length },
+    });
+
+    const copies = files.map((f) => ({ buffer: f.buffer, name: f.originalname }));
+    void this.process(batch.id, copies).catch(async (err) => {
+      this.logger.error(`Policy OCR batch ${batch.id} failed: ${(err as Error).message}`);
+      await this.prisma.policyOcrBatch.update({
+        where: { id: batch.id },
+        data: { status: "FAILED", error: (err as Error).message },
+      });
+    });
+
+    return batch;
+  }
+
+  /**
+   * Render → text → parse → match, **one PolicyOcrDocument row per uploaded
+   * file**. The GMX certificate is a 2-page PDF where page 1 carries the
+   * contract header and page 2 carries the per-coverage table — both pages
+   * describe the SAME policy, so the parser concatenates them and the
+   * matcher runs once. `pageNumber` on the row is repurposed as the file
+   * ordinal within the batch (1, 2, 3…) — the unique constraint
+   * `(batchId, pageNumber)` still holds and lets a single batch carry many
+   * policies.
+   *
+   * The doc's `storageKey` is the SOURCE PDF (`policy-ocr/{batchId}/source-N.pdf`)
+   * rather than a rendered page image, so the review screen can embed the
+   * exact artifact the office received. The rendered page PNGs are still
+   * stored under `policy-ocr/{batchId}/page-M.png` for any future re-OCR or
+   * image-based audit, but they aren't used as `storageKey` for the document.
+   */
+  private async process(
+    batchId: string,
+    files: { buffer: Buffer; name?: string }[],
+  ) {
+    await this.prisma.policyOcrBatch.update({
+      where: { id: batchId },
+      data: { status: "PROCESSING" },
+    });
+
+    let fileOrdinal = 0;
+    let globalPageOrdinal = 0;
+    for (const file of files) {
+      fileOrdinal += 1;
+      const sourceKey = `policy-ocr/${batchId}/source-${fileOrdinal}.pdf`;
+      await this.storage.put(sourceKey, file.buffer, "application/pdf");
+
+      const pages = await this.ocr.renderPages(file.buffer);
+      const textLayer = await this.ocr.textPages(file.buffer).catch(() => []);
+
+      // One OcrPage per rendered page: text-layer wins when present (cheap,
+      // exact), OCR the rendered image when it isn't. Same precedence rule
+      // as the statement OCR pipeline.
+      const perPageOcr: OcrPage[] = [];
+      for (const [index, image] of pages.entries()) {
+        globalPageOrdinal += 1;
+        const pageStorageKey = `policy-ocr/${batchId}/page-${globalPageOrdinal}.png`;
+        await this.storage.put(pageStorageKey, image, "image/png");
+
+        const embedded = textLayer[index] ?? null;
+        const pageOcr = embedded ?? (await this.ocr.recognize(image));
+        perPageOcr.push(pageOcr);
+      }
+
+      // Concatenate every page's text with a blank line between pages so the
+      // parser's anchored regexes (^From$, ^Currency\s+...) still work
+      // across page boundaries — pdftotext -bbox-layout produces newline-
+      // separated text per page already, the `\n\n` just preserves a clear
+      // boundary in ocrRawText for debugging.
+      const mergedText = perPageOcr.map((p) => p.text).join("\n\n");
+      const avgConfidence =
+        perPageOcr.length === 0
+          ? 0
+          : perPageOcr.reduce((s, p) => s + p.confidence, 0) / perPageOcr.length;
+      const synthetic: OcrPage = {
+        text: mergedText,
+        words: [],
+        confidence: avgConfidence,
+      };
+
+      try {
+        const parsed = parsePolicy(synthetic);
+        if (parsed.provider === "") {
+          throw new Error("no se reconoció el proveedor");
+        }
+        const match = await this.matcher.match(parsed);
+        const notes = [...parsed.notes, match.note].filter(Boolean);
+        // Confident when exactly one Policy carries the printed number —
+        // the only unambiguous hit we trust. A new policy (no match) still
+        // needs a customer pick, so it stays in review.
+        const trusted = match.confident && parsed.policyNumber != null;
+
+        await this.prisma.policyOcrDocument.create({
+          data: {
+            batchId,
+            pageNumber: fileOrdinal,
+            storageKey: sourceKey,
+            status: trusted ? "MATCHED" : "NEEDS_REVIEW",
+            ocrRawText: mergedText,
+            ocrConfidence: new Prisma.Decimal(avgConfidence.toFixed(3)),
+            provider: parsed.provider,
+            extractedPolicyNumber: parsed.policyNumber,
+            extractedInsuredName: parsed.insuredName,
+            extractedAdditionalInsured: parsed.additionalInsured,
+            extractedAgentName: parsed.agentName,
+            extractedLegalAddress: parsed.legalAddress,
+            extractedZip: parsed.zip,
+            extractedPolicyFrom: parsed.policyFrom,
+            extractedPolicyTo: parsed.policyTo,
+            extractedPolicyDate: parsed.policyDate,
+            extractedCurrency: parsed.currency,
+            extractedNetPremium:
+              parsed.netPremium != null ? new Prisma.Decimal(parsed.netPremium) : null,
+            extractedPolicyFee:
+              parsed.policyFee != null ? new Prisma.Decimal(parsed.policyFee) : null,
+            extractedBrokerFee:
+              parsed.brokerFee != null ? new Prisma.Decimal(parsed.brokerFee) : null,
+            extractedTotal:
+              parsed.total != null ? new Prisma.Decimal(parsed.total) : null,
+            extractedCoveragesJson: parsed.coverages.length
+              ? (parsed.coverages as unknown as Prisma.InputJsonValue)
+              : Prisma.DbNull,
+            extractedPremiumPayment: parsed.premiumPayment,
+            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),
+          },
+        });
+      } catch (err) {
+        // The file as a whole failed to parse (no provider, parse exception).
+        // One OCR_FAILED row per file is the right granularity — the page
+        // images are still on disk for a re-run after a parser fix.
+        await this.prisma.policyOcrDocument.create({
+          data: {
+            batchId,
+            pageNumber: fileOrdinal,
+            storageKey: sourceKey,
+            status: "OCR_FAILED",
+            matchNote: (err as Error).message.slice(0, 190),
+          },
+        });
+      }
+    }
+
+    await this.prisma.policyOcrBatch.update({
+      where: { id: batchId },
+      data: { status: "READY_FOR_REVIEW" },
+    });
+  }
+
+  // --- reads ----------------------------------------------------------------
+
+  async listBatches(page: number, pageSize: number) {
+    const [total, items] = await this.prisma.$transaction([
+      this.prisma.policyOcrBatch.count(),
+      this.prisma.policyOcrBatch.findMany({
+        orderBy: { createdAt: "desc" },
+        skip: (page - 1) * pageSize,
+        take: pageSize,
+        include: {
+          uploadedBy: { select: { name: true } },
+          _count: { select: { documents: true } },
+        },
+      }),
+    ]);
+    return { items, total, page, pageSize, pageCount: Math.ceil(total / pageSize) };
+  }
+
+  async getBatch(id: string) {
+    const batch = await this.prisma.policyOcrBatch.findUnique({
+      where: { id },
+      include: { uploadedBy: { select: { name: true } } },
+    });
+    if (!batch) throw new NotFoundException("Lote no encontrado.");
+
+    const counts = await this.prisma.policyOcrDocument.groupBy({
+      by: ["status"],
+      where: { batchId: id },
+      _count: { _all: true },
+    });
+    return {
+      ...batch,
+      byStatus: Object.fromEntries(counts.map((c) => [c.status, c._count._all])),
+    };
+  }
+
+  async listDocuments(batchId: string) {
+    return this.prisma.policyOcrDocument.findMany({
+      where: { batchId },
+      orderBy: { pageNumber: "asc" },
+      include: {
+        matchedCustomer: { select: { id: true, name: true } },
+        matchedPolicy: {
+          select: {
+            id: true,
+            policyNumber: true,
+            customerId: true,
+            customer: { select: { name: true } },
+          },
+        },
+      },
+    });
+  }
+
+  /**
+   * The source PDF for the document, so the review screen can show the
+   * exact artifact the office uploaded (the browser's PDF viewer handles
+   * scrolling, zoom, and selection natively). The rendered page PNGs
+   * remain on disk under `policy-ocr/{batchId}/page-N.png` for any
+   * future re-OCR, but the doc row points here at the source.
+   */
+  async pageImage(documentId: string) {
+    const doc = await this.prisma.policyOcrDocument.findUnique({
+      where: { id: documentId },
+      select: { storageKey: true },
+    });
+    if (!doc) throw new NotFoundException("Documento no encontrado.");
+    return this.storage.getStream(doc.storageKey);
+  }
+
+  // --- review ---------------------------------------------------------------
+
+  async review(id: string, dto: ReviewPolicyDocumentDto, reviewedById: string) {
+    const doc = await this.prisma.policyOcrDocument.findUnique({ where: { id } });
+    if (!doc) throw new NotFoundException("Documento no encontrado.");
+    if (doc.status === "POSTED") {
+      throw new BadRequestException("Este documento ya fue aplicado.");
+    }
+
+    // Trusting a customer-supplied pair (policyId, customerId) without
+    // cross-check is how a document lands on the wrong customer's ledger;
+    // pin them here from the DB.
+    let matchedPolicyId = dto.matchedPolicyId ?? doc.matchedPolicyId;
+    let matchedCustomerId = doc.matchedCustomerId;
+
+    if (matchedPolicyId) {
+      const p = await this.prisma.policy.findUnique({
+        where: { id: matchedPolicyId },
+        select: { customerId: true },
+      });
+      if (!p) throw new BadRequestException("Póliza no encontrada.");
+      matchedCustomerId = p.customerId;
+    } else if (dto.matchedCustomerId) {
+      const c = await this.prisma.customer.findUnique({
+        where: { id: dto.matchedCustomerId },
+        select: { id: true },
+      });
+      if (!c) throw new BadRequestException("Cliente no encontrado.");
+      matchedCustomerId = c.id;
+    }
+
+    return this.prisma.policyOcrDocument.update({
+      where: { id },
+      data: {
+        extractedPolicyNumber: dto.policyNumber ?? undefined,
+        extractedInsuredName: dto.insuredName ?? undefined,
+        extractedAdditionalInsured: dto.additionalInsured ?? undefined,
+        extractedAgentName: dto.agentName ?? undefined,
+        extractedLegalAddress: dto.legalAddress ?? undefined,
+        extractedZip: dto.zip ?? undefined,
+        extractedPolicyFrom: dto.policyFrom ? new Date(dto.policyFrom) : undefined,
+        extractedPolicyTo: dto.policyTo ? new Date(dto.policyTo) : undefined,
+        extractedPolicyDate: dto.policyDate ? new Date(dto.policyDate) : undefined,
+        extractedCurrency: dto.currency ?? undefined,
+        extractedNetPremium:
+          dto.netPremium != null ? new Prisma.Decimal(dto.netPremium) : undefined,
+        extractedPolicyFee:
+          dto.policyFee != null ? new Prisma.Decimal(dto.policyFee) : undefined,
+        extractedBrokerFee:
+          dto.brokerFee != null ? new Prisma.Decimal(dto.brokerFee) : undefined,
+        extractedTotal:
+          dto.total != null ? new Prisma.Decimal(dto.total) : undefined,
+        extractedCoveragesJson: dto.coveragesJson
+          ? (dto.coveragesJson as Prisma.InputJsonValue)
+          : undefined,
+        extractedPremiumPayment: dto.premiumPayment ?? undefined,
+        matchedPolicyId,
+        matchedCustomerId,
+        status: dto.forceConfirm ? "CONFIRMED" : "MATCHED",
+        reviewedById,
+        reviewedAt: new Date(),
+      },
+    });
+  }
+
+  async reject(id: string, reviewedById: string) {
+    const doc = await this.prisma.policyOcrDocument.findUnique({ where: { id } });
+    if (!doc) throw new NotFoundException("Documento no encontrado.");
+    if (doc.status === "POSTED") {
+      throw new BadRequestException("Este documento ya fue aplicado.");
+    }
+    return this.prisma.policyOcrDocument.update({
+      where: { id },
+      data: { status: "REJECTED", reviewedById, reviewedAt: new Date() },
+    });
+  }
+
+  // --- confirm --------------------------------------------------------------
+
+  /**
+   * Apply every confirmed document: create or update the Policy, attach the
+   * source PDF as a PolicyDocument, and (when staff asked + premium parses)
+   * write a Transaction row. Each step is guarded by status checks so a
+   * double-confirm cannot re-apply a document.
+   */
+  async confirmBatch(batchId: string, dto: ConfirmPolicyBatchDto, reviewedById: string) {
+    const batch = await this.prisma.policyOcrBatch.findUnique({ where: { id: batchId } });
+    if (!batch) throw new NotFoundException("Lote no encontrado.");
+
+    const results: { documentId: string; policyId: string; postedTransactionId: string | null }[] = [];
+
+    for (const item of dto.documents) {
+      const doc = await this.prisma.policyOcrDocument.findUnique({
+        where: { id: item.documentId },
+      });
+      if (!doc) {
+        throw new BadRequestException(`Documento ${item.documentId} no encontrado.`);
+      }
+      if (doc.status === "POSTED") {
+        throw new BadRequestException(
+          `El documento página ${doc.pageNumber} ya fue aplicado.`,
+        );
+      }
+      if (!item.policyId && !item.customerId) {
+        throw new BadRequestException(
+          `Documento página ${doc.pageNumber}: falta póliza destino o cliente.`,
+        );
+      }
+
+      // 1. 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);
+        await this.prisma.policy.update({
+          where: { id: policyId },
+          data: updateData,
+        });
+      } else {
+        // Create under the picked customer. `policyNumber` is the only field
+        // that must be present.
+        if (!item.policyNumber && !doc.extractedPolicyNumber) {
+          throw new BadRequestException(
+            `Documento página ${doc.pageNumber}: falta número de póliza.`,
+          );
+        }
+        const createData = buildPolicyCreateFromDoc(item, doc, item.customerId!);
+        const created = await this.prisma.policy.create({
+          data: createData,
+        });
+        policyId = created.id;
+      }
+
+      // 2. 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);
+
+      // 3. 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;
+      const premium =
+        item.netPremium != null
+          ? item.netPremium
+          : doc.extractedNetPremium != null
+          ? Number(doc.extractedNetPremium)
+          : null;
+      if (item.postPremium && premium && premium > 0) {
+        const tx = await this.prisma.transaction.create({
+          data: {
+            customerId: (await this.policyCustomerId(policyId))!,
+            domain: "INSURANCE",
+            amount: new Prisma.Decimal(-Math.abs(premium)),
+            transactionDate: doc.extractedPolicyDate ?? doc.extractedPolicyFrom ?? new Date(),
+            currency: (item.currency ??
+              doc.extractedCurrency ??
+              "MXN") as Currency,
+            reference: item.policyNumber ?? doc.extractedPolicyNumber ?? null,
+            period: null,
+            captureSource: "OCR",
+            captureRef: doc.id,
+            message: `Prima de póliza ${item.policyNumber ?? doc.extractedPolicyNumber ?? ""}`,
+          },
+        });
+        postedTransactionId = tx.id;
+      }
+
+      await this.prisma.policyOcrDocument.update({
+        where: { id: doc.id },
+        data: {
+          status: "POSTED",
+          matchedPolicyId: policyId,
+          reviewedById,
+          reviewedAt: new Date(),
+          createdPolicyId: item.policyId ? null : policyId,
+          postedTransactionId,
+        },
+      });
+
+      results.push({
+        documentId: doc.id,
+        policyId,
+        postedTransactionId,
+      });
+    }
+
+    await this.closeIfDone(batchId);
+
+    return {
+      applied: results.length,
+      policies: results.map((r) => r.policyId),
+      postedTransactions: results.filter((r) => r.postedTransactionId).length,
+    };
+  }
+
+  /**
+   * Stream the source PDF (`sourceKey`, set by `process` on the doc row)
+   * into the policy's storage namespace and create a `PolicyDocument`
+   * pointer. Trivial now that the doc row holds the exact source key —
+   * the old per-page "which file did this page come from" walk is gone.
+   */
+  private async attachSourcePdf(sourceKey: string, policyId: string): Promise<void> {
+    const got = await this.storage.getStream(sourceKey);
+    const chunks: Buffer[] = [];
+    for await (const c of got.stream) chunks.push(c as Buffer);
+    const buf = Buffer.concat(chunks);
+
+    const newKey = `policy/${policyId}/${Date.now()}-${crypto.randomUUID()}.pdf`;
+    await this.storage.put(newKey, buf, "application/pdf");
+    await this.prisma.policyDocument.create({
+      data: {
+        policyId,
+        documentType: "GMX_POLICY",
+        storageKey: newKey,
+      },
+    });
+  }
+
+  private async policyCustomerId(policyId: string): Promise<string | null> {
+    const p = await this.prisma.policy.findUnique({
+      where: { id: policyId },
+      select: { customerId: true },
+    });
+    return p?.customerId ?? null;
+  }
+
+  private async closeIfDone(batchId: string) {
+    const open = await this.prisma.policyOcrDocument.count({
+      where: {
+        batchId,
+        status: { in: ["PENDING_OCR", "NEEDS_REVIEW", "MATCHED", "CONFIRMED"] },
+      },
+    });
+    if (open === 0) {
+      await this.prisma.policyOcrBatch.update({
+        where: { id: batchId },
+        data: { status: "COMPLETED", completedAt: new Date() },
+      });
+    }
+  }
+}
+
+/** 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
+ *  has no premium — we must not blank the existing Policy.netPremium). */
+function buildPolicyUpdateFromDoc(
+  item: ConfirmPolicyDocumentDto,
+  doc: {
+    extractedPolicyNumber: string | null;
+    extractedInsuredName: string | null;
+    extractedAdditionalInsured: string | null;
+    extractedAgentName: string | null;
+    extractedLegalAddress: string | null;
+    extractedZip: string | null;
+    extractedPolicyFrom: Date | null;
+    extractedPolicyTo: Date | null;
+    extractedPolicyDate: Date | null;
+    extractedCurrency: string | null;
+    extractedNetPremium: Prisma.Decimal | null;
+    extractedPolicyFee: Prisma.Decimal | null;
+    extractedBrokerFee: Prisma.Decimal | null;
+    extractedTotal: Prisma.Decimal | null;
+    extractedCoveragesJson: Prisma.JsonValue | null;
+    extractedPremiumPayment: string | null;
+  },
+): Prisma.PolicyUpdateInput {
+  const numOrUndef = (a: number | undefined, b: Prisma.Decimal | null): Prisma.Decimal | undefined => {
+    if (a != null) return new Prisma.Decimal(a);
+    if (b != null) return b;
+    return undefined;
+  };
+  const dateOrUndef = (a: string | undefined, b: Date | null): Date | undefined => {
+    if (a) return new Date(a);
+    if (b) return b;
+    return undefined;
+  };
+  const strOrUndef = (a: string | undefined, b: string | null): string | undefined => {
+    if (a != null && a !== "") return a;
+    if (b != null && b !== "") return b;
+    return undefined;
+  };
+
+  return {
+    policyNumber: strOrUndef(item.policyNumber, doc.extractedPolicyNumber),
+    agentName: strOrUndef(item.agentName, doc.extractedAgentName),
+    policyFrom: dateOrUndef(item.policyFrom, doc.extractedPolicyFrom),
+    policyTo: dateOrUndef(item.policyTo, doc.extractedPolicyTo),
+    policyDate: dateOrUndef(item.policyDate, doc.extractedPolicyDate),
+    currency: strOrUndef(item.currency, doc.extractedCurrency) as Currency | undefined,
+    netPremium: numOrUndef(item.netPremium, doc.extractedNetPremium),
+    policyFee: numOrUndef(item.policyFee, doc.extractedPolicyFee),
+    brokerFee: numOrUndef(item.brokerFee, doc.extractedBrokerFee),
+    total: numOrUndef(item.total, doc.extractedTotal),
+    // coveragesJson / observations: freeform, keep the GMX data when present.
+    coveragesJson:
+      item.coveragesJson !== undefined
+        ? (item.coveragesJson as Prisma.InputJsonValue)
+        : doc.extractedCoveragesJson != null
+        ? (doc.extractedCoveragesJson as Prisma.InputJsonValue)
+        : undefined,
+    // Premium payment cadence ("CONTADO") and insured-name fields land in
+    // `observations` so the PolicyForm's edits stay the source of truth for
+    // structured fields. The reviewer can move them by hand if needed.
+    observations: joinObservations(
+      doc.extractedInsuredName,
+      doc.extractedAdditionalInsured,
+      doc.extractedLegalAddress,
+      doc.extractedZip,
+      doc.extractedPremiumPayment,
+      item,
+    ),
+  };
+}
+
+/** Same shape as `buildPolicyUpdateFromDoc`, but for `Policy.create`. The
+ *  `customerId` is supplied separately and `policyNumber` is required (a
+ *  Policy without a number can't be re-matched by the OCR pipeline). */
+function buildPolicyCreateFromDoc(
+  item: ConfirmPolicyDocumentDto,
+  doc: {
+    extractedPolicyNumber: string | null;
+    extractedInsuredName: string | null;
+    extractedAdditionalInsured: string | null;
+    extractedAgentName: string | null;
+    extractedLegalAddress: string | null;
+    extractedZip: string | null;
+    extractedPolicyFrom: Date | null;
+    extractedPolicyTo: Date | null;
+    extractedPolicyDate: Date | null;
+    extractedCurrency: string | null;
+    extractedNetPremium: Prisma.Decimal | null;
+    extractedPolicyFee: Prisma.Decimal | null;
+    extractedBrokerFee: Prisma.Decimal | null;
+    extractedTotal: Prisma.Decimal | null;
+    extractedCoveragesJson: Prisma.JsonValue | null;
+    extractedPremiumPayment: string | null;
+  },
+  customerId: string,
+): Prisma.PolicyUncheckedCreateInput {
+  const numOrUndef = (a: number | undefined, b: Prisma.Decimal | null): Prisma.Decimal | undefined => {
+    if (a != null) return new Prisma.Decimal(a);
+    if (b != null) return b;
+    return undefined;
+  };
+  const dateOrUndef = (a: string | undefined, b: Date | null): Date | undefined => {
+    if (a) return new Date(a);
+    if (b) return b;
+    return undefined;
+  };
+  const strOrUndef = (a: string | undefined, b: string | null): string | undefined => {
+    if (a != null && a !== "") return a;
+    if (b != null && b !== "") return b;
+    return undefined;
+  };
+
+  const policyNumber =
+    strOrUndef(item.policyNumber, doc.extractedPolicyNumber);
+  if (!policyNumber) {
+    // Caller already guards this; the throw is a type-narrowing aid.
+    throw new Error("policyNumber required for create");
+  }
+
+  return {
+    policyNumber,
+    customerId,
+    agentName: strOrUndef(item.agentName, doc.extractedAgentName),
+    policyFrom: dateOrUndef(item.policyFrom, doc.extractedPolicyFrom),
+    policyTo: dateOrUndef(item.policyTo, doc.extractedPolicyTo),
+    policyDate: dateOrUndef(item.policyDate, doc.extractedPolicyDate),
+    currency: strOrUndef(item.currency, doc.extractedCurrency) as Currency | undefined,
+    netPremium: numOrUndef(item.netPremium, doc.extractedNetPremium),
+    policyFee: numOrUndef(item.policyFee, doc.extractedPolicyFee),
+    brokerFee: numOrUndef(item.brokerFee, doc.extractedBrokerFee),
+    total: numOrUndef(item.total, doc.extractedTotal),
+    coveragesJson:
+      item.coveragesJson !== undefined
+        ? (item.coveragesJson as Prisma.InputJsonValue)
+        : doc.extractedCoveragesJson != null
+        ? (doc.extractedCoveragesJson as Prisma.InputJsonValue)
+        : undefined,
+    observations: joinObservations(
+      doc.extractedInsuredName,
+      doc.extractedAdditionalInsured,
+      doc.extractedLegalAddress,
+      doc.extractedZip,
+      doc.extractedPremiumPayment,
+      item,
+    ),
+  };
+}
+
+function joinObservations(
+  insured: string | null,
+  additional: string | null,
+  address: string | null,
+  zip: string | null,
+  premiumPayment: string | null,
+  item: ConfirmPolicyDocumentDto,
+): string | undefined {
+  const lines: string[] = [];
+  const insuredName = strOrUndefDb(item.insuredName, insured);
+  if (insuredName) lines.push(`Asegurado: ${insuredName}`);
+  const additionalInsured = strOrUndefDb(item.additionalInsured, additional);
+  if (additionalInsured) lines.push(`Asegurado adicional: ${additionalInsured}`);
+  const legalAddress = strOrUndefDb(item.legalAddress, address);
+  if (legalAddress) lines.push(`Dirección: ${legalAddress}`);
+  const zipVal = strOrUndefDb(item.zip, zip);
+  if (zipVal) lines.push(`C.P.: ${zipVal}`);
+  const cadence = strOrUndefDb(item.premiumPayment, premiumPayment);
+  if (cadence) lines.push(`Pago de prima: ${cadence}`);
+  return lines.length ? lines.join("\n") : undefined;
+}
+
+function strOrUndefDb(a: string | undefined, b: string | null): string | undefined {
+  if (a != null && a !== "") return a;
+  if (b != null && b !== "") return b;
+  return undefined;
+}

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

@@ -0,0 +1,71 @@
+/**
+ * The OCR seam. Everything above this interface works in terms of page text and
+ * word boxes, so the concrete engine is swappable without touching the parsers,
+ * the matcher, or the schema.
+ *
+ * The shipped implementation is self-hosted Tesseract (see tesseract.provider).
+ * That choice is evidence-based rather than assumed: run against 46 pages of
+ * real scanned CFE, CESPT and Telnor statements, it identified the provider on
+ * 46/46 and extracted a usable account reference on 43/46, and on a later
+ * corpus of 19 scanned municipal predial receipts it read the provider on
+ * 19/19 and an identifier on 18/19 — well past the bar for a queue whose whole
+ * point is that a human confirms every row. A
+ * managed document-extraction API (Textract, Document Intelligence, Document
+ * AI) fits behind this same interface if per-page accuracy ever proves
+ * insufficient, with no schema change — but at 300+ pages/month/company it
+ * would carry a real recurring cost for accuracy that is not currently the
+ * bottleneck.
+ */
+
+/** One OCR'd word, with where it sits on the page. */
+export interface OcrWord {
+  text: string;
+  /** Pixel box in the rendered page image. */
+  left: number;
+  top: number;
+  width: number;
+  height: number;
+  /** Engine confidence for this word, 0..1. */
+  confidence: number;
+}
+
+export interface OcrPage {
+  /** Full page text, reading order, newline-separated. */
+  text: string;
+  /**
+   * Word boxes. Needed because several of the real layouts are *tables* — the
+   * CESPT "RECIBO" prints `No. DE CUENTA` as a column header with the value in
+   * the row beneath it, which line-oriented text cannot associate. Parsers fall
+   * back to geometry for exactly those fields.
+   */
+  words: OcrWord[];
+  /** Mean word confidence across the page, 0..1. */
+  confidence: number;
+}
+
+export interface OcrProvider {
+  /** True when the engine is actually usable in this deployment. */
+  available(): Promise<boolean>;
+  /** Split a PDF into one rendered page image per page. */
+  renderPages(pdf: Buffer): Promise<Buffer[]>;
+  /** OCR a single rendered page image. */
+  recognize(pageImage: Buffer): Promise<OcrPage>;
+  /**
+   * Read a PDF's own text layer, one entry per page, `null` where the page has
+   * none worth using.
+   *
+   * Not every statement is a scan. The gas company e-mails born-digital CFDI
+   * invoices whose text is already exact and already positioned — running those
+   * through a rasteriser and a character recogniser can only lose information
+   * (one sample turned `MEDIDOR: VM01014426` into `ar (LTR): 014420`) while
+   * costing about a minute of CPU per page for the privilege. Where the layer
+   * exists it is strictly better input for the same parsers, so it is tried
+   * first and OCR remains the fallback for genuine scans.
+   *
+   * Positions are reported in the same pixel space `recognize` uses, so the
+   * geometric helpers in the parsers work unchanged on either source.
+   */
+  textPages(pdf: Buffer): Promise<(OcrPage | null)[]>;
+}
+
+export const OCR_PROVIDER = Symbol("OCR_PROVIDER");

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

@@ -0,0 +1,67 @@
+import { parseBboxLayout } from "./tesseract.provider";
+
+/**
+ * Shaped like real `pdftotext -bbox-layout` output: the gas invoice lays its
+ * header out as two columns of independent text flows, so poppler puts a label
+ * and the value printed beside it in *different* `<line>` elements. Trusting
+ * 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.
+ */
+function word(x: number, y: number, text: string): string {
+  return `<word xMin="${x}" yMin="${y}" xMax="${x + 20}" yMax="${y + 8}">${text}</word>`;
+}
+
+function doc(...lines: string[]): string {
+  return `<doc><page width="612" height="792">${lines
+    .map((l) => `<flow><block><line>${l}</line></block></flow>`)
+    .join("")}</page></doc>`;
+}
+
+/** Enough words on the page to clear the "is this a real text layer" floor. */
+function padding(): string {
+  return Array.from({ length: 50 }, (_, i) => word(10, 400 + i * 10, `w${i}`)).join("");
+}
+
+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(300, 100.4, "20260630-20260630"),
+        padding(),
+      ),
+      1,
+    );
+    expect(page).not.toBeNull();
+    expect(page!.text).toContain("PERIODO FACTURADO: 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()),
+      1,
+    );
+    expect(page!.text.split("\n")).toContain("Cuenta: 0900003463");
+    expect(page!.text.split("\n")).toContain("Nombre:");
+  });
+
+  it("scales point coordinates into the render's pixel space", () => {
+    // Word boxes have to land in the same coordinate space tesseract reports,
+    // or the geometric helpers the parsers share silently stop finding values.
+    const [page] = parseBboxLayout(doc(word(72, 144, "X") + padding()), 300 / 72);
+    const x = page!.words.find((w) => w.text === "X")!;
+    expect(x.left).toBeCloseTo(300);
+    expect(x.top).toBeCloseTo(600);
+  });
+
+  it("reports no text layer for a scan carrying a few stray glyphs", () => {
+    expect(parseBboxLayout(doc(word(10, 10, "3") + word(40, 10, "of") + word(60, 10, "5")), 1)).toEqual([
+      null,
+    ]);
+  });
+
+  it("decodes the entities poppler escapes", () => {
+    const [page] = parseBboxLayout(doc(word(10, 10, "A&amp;B") + padding()), 1);
+    expect(page!.text).toContain("A&B");
+  });
+});

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

@@ -0,0 +1,350 @@
+import { Injectable, Logger, ServiceUnavailableException } from "@nestjs/common";
+import { ConfigService } from "@nestjs/config";
+import { execFile } from "node:child_process";
+import { mkdtemp, readFile, readdir, rm, writeFile } from "node:fs/promises";
+import { tmpdir } from "node:os";
+import { join } from "node:path";
+import { promisify } from "node:util";
+import type { OcrPage, OcrProvider, OcrWord } from "./ocr.provider";
+
+const run = promisify(execFile);
+
+/**
+ * Self-hosted OCR: `pdftoppm` (poppler) to rasterise, `tesseract` to read.
+ *
+ * Both are external binaries rather than a native npm addon, which keeps the
+ * pnpm workspace free of a compiled dependency and makes the alpine runtime
+ * image a two-package change (see docker/api.Dockerfile). Like StorageService,
+ * a missing binary degrades rather than crashes the API: the module reports
+ * itself unavailable and statement ingest returns 503, while every other
+ * feature keeps working.
+ *
+ * The settings below are not arbitrary — they were measured against the real
+ * scanned samples:
+ *   - 300 DPI grayscale. The source scans are phone photos of paper at ~5MB a
+ *     page; below 300 the small print (RMU, clave catastral) stops resolving,
+ *     above it costs time for no additional fields.
+ *   - `--psm 6` ("assume a single uniform block of text"). The default page
+ *     segmentation splits these dense forms into columns and interleaves them,
+ *     which destroys the label-then-value adjacency every parser depends on.
+ *   - Spanish traineddata, with a graceful fall back to English if the language
+ *     pack is absent — an accented label reads worse but the digits, which are
+ *     what actually gets matched, are unaffected.
+ */
+@Injectable()
+export class TesseractOcrProvider implements OcrProvider {
+  private readonly logger = new Logger(TesseractOcrProvider.name);
+  private readonly dpi: number;
+  private readonly lang: string;
+  private probe: Promise<boolean> | null = null;
+
+  constructor(config: ConfigService) {
+    this.dpi = Number(config.get("OCR_DPI") ?? 300);
+    this.lang = config.get<string>("OCR_LANG") ?? "spa";
+  }
+
+  /** Cached — the binaries do not appear or vanish while the process runs. */
+  available(): Promise<boolean> {
+    if (!this.probe) {
+      this.probe = (async () => {
+        try {
+          await Promise.all([
+            run("tesseract", ["--version"]),
+            run("pdftoppm", ["-v"]),
+          ]);
+          return true;
+        } catch {
+          this.logger.warn(
+            "OCR unavailable: `tesseract` and/or `pdftoppm` not found on PATH. " +
+              "Statement ingest is disabled; every other feature is unaffected.",
+          );
+          return false;
+        }
+      })();
+    }
+    return this.probe;
+  }
+
+  private async require(): Promise<void> {
+    if (!(await this.available())) {
+      throw new ServiceUnavailableException(
+        "El servicio de OCR no está disponible en este servidor.",
+      );
+    }
+  }
+
+  private async scratch<T>(fn: (dir: string) => Promise<T>): Promise<T> {
+    const dir = await mkdtemp(join(tmpdir(), "stmt-ocr-"));
+    try {
+      return await fn(dir);
+    } finally {
+      await rm(dir, { recursive: true, force: true });
+    }
+  }
+
+  async renderPages(pdf: Buffer): Promise<Buffer[]> {
+    await this.require();
+    return this.scratch(async (dir) => {
+      const src = join(dir, "in.pdf");
+      await writeFile(src, pdf);
+      // -gray: these are grayscale scans already; colour triples the bytes
+      // handed to tesseract for no gain in character recognition.
+      await run("pdftoppm", [
+        "-r",
+        String(this.dpi),
+        "-gray",
+        "-png",
+        src,
+        join(dir, "page"),
+      ]);
+      const files = (await readdir(dir))
+        .filter((f) => f.startsWith("page") && f.endsWith(".png"))
+        // pdftoppm zero-pads its page numbers, so lexical order is page order.
+        .sort();
+      return Promise.all(files.map((f) => readFile(join(dir, f))));
+    });
+  }
+
+  /**
+   * `pdftotext -bbox-layout` — the same poppler package `pdftoppm` comes from,
+   * so this costs no extra dependency in the runtime image.
+   *
+   * A page is only accepted when it carries a real text layer. Scanned PDFs
+   * frequently contain a handful of stray glyphs (a scanner watermark, a page
+   * number stamped by the MFP), and treating those as the page's text would
+   * hand every parser an almost-empty string and silently take OCR out of the
+   * loop — so a floor of MIN_TEXT_WORDS words has to be present before the
+   * layer is believed.
+   */
+  async textPages(pdf: Buffer): Promise<(OcrPage | null)[]> {
+    await this.require();
+    return this.scratch(async (dir) => {
+      const src = join(dir, "in.pdf");
+      await writeFile(src, pdf);
+      const out = join(dir, "out.html");
+      try {
+        await run("pdftotext", ["-bbox-layout", src, out]);
+      } catch (err) {
+        this.logger.warn(
+          `pdftotext failed; falling back to OCR for this file: ${(err as Error).message}`,
+        );
+        return [];
+      }
+      // Points to pixels at the render DPI, so word boxes from either source
+      // land in one coordinate space and `valueUnder`'s thresholds hold.
+      return parseBboxLayout(await readFile(out, "utf8"), this.dpi / 72);
+    });
+  }
+
+  async recognize(pageImage: Buffer): Promise<OcrPage> {
+    await this.require();
+    return this.scratch(async (dir) => {
+      const img = join(dir, "page.png");
+      await writeFile(img, pageImage);
+
+      // One tesseract invocation produces both outputs; TSV carries the word
+      // boxes and per-word confidence, and its text can be reassembled into
+      // reading order, so there is no need to run the engine twice.
+      const out = join(dir, "out");
+      try {
+        await run("tesseract", [img, out, "-l", this.lang, "--psm", "6", "tsv"]);
+      } catch (err) {
+        if (this.lang !== "eng") {
+          this.logger.warn(
+            `Tesseract failed with lang "${this.lang}", retrying with "eng": ${
+              (err as Error).message
+            }`,
+          );
+          await run("tesseract", [img, out, "-l", "eng", "--psm", "6", "tsv"]);
+        } else {
+          throw err;
+        }
+      }
+
+      const tsv = await readFile(`${out}.tsv`, "utf8");
+      return parseTsv(tsv);
+    });
+  }
+}
+
+/**
+ * Below this many words a "text layer" is scanner debris, not a document.
+ * The real born-digital samples carry 400+ words a page; the scanned ones
+ * carry none at all, so the exact threshold is not delicate.
+ */
+const MIN_TEXT_WORDS = 40;
+
+const ENTITIES: Record<string, string> = {
+  amp: "&",
+  lt: "<",
+  gt: ">",
+  quot: '"',
+  apos: "'",
+};
+
+function decodeEntities(s: string): string {
+  return s.replace(/&(#x?[0-9a-fA-F]+|[a-z]+);/g, (whole, body: string) => {
+    if (body[0] === "#") {
+      const code =
+        body[1] === "x" || body[1] === "X"
+          ? parseInt(body.slice(2), 16)
+          : parseInt(body.slice(1), 10);
+      return Number.isFinite(code) ? String.fromCodePoint(code) : whole;
+    }
+    return ENTITIES[body] ?? whole;
+  });
+}
+
+/**
+ * Turn `pdftotext -bbox-layout`'s XHTML into one OcrPage per PDF page.
+ *
+ * Parsed with regexes rather than an XML library on purpose: the output is
+ * machine-generated by poppler with a fixed element shape (`page` > `flow` >
+ * `block` > `line` > `word`), and the alternative is a parser dependency in
+ * the API for one file format read in one place. Only `page` and `word` are
+ * consulted — see below for why poppler's own `line` grouping is discarded.
+ *
+ * `confidence` is 1 for every word: these are the document's own characters,
+ * not a recognition guess.
+ */
+export function parseBboxLayout(xhtml: string, scale: number): (OcrPage | null)[] {
+  const pages: (OcrPage | null)[] = [];
+
+  for (const pageMatch of xhtml.matchAll(/<page\b[^>]*>([\s\S]*?)<\/page>/g)) {
+    const words: OcrWord[] = [];
+
+    for (const w of pageMatch[1].matchAll(
+      /<word\s+xMin="([\d.eE+-]+)"\s+yMin="([\d.eE+-]+)"\s+xMax="([\d.eE+-]+)"\s+yMax="([\d.eE+-]+)"\s*>([\s\S]*?)<\/word>/g,
+    )) {
+      const text = decodeEntities(w[5]).trim();
+      if (!text) continue;
+      const left = Number(w[1]) * scale;
+      const top = Number(w[2]) * scale;
+      words.push({
+        text,
+        left,
+        top,
+        width: Number(w[3]) * scale - left,
+        height: Number(w[4]) * scale - top,
+        confidence: 1,
+      });
+    }
+
+    pages.push(
+      words.length >= MIN_TEXT_WORDS
+        ? { text: toVisualRows(words), words, confidence: 1 }
+        : null,
+    );
+  }
+
+  return pages;
+}
+
+/**
+ * Reassemble words into the rows a reader sees, left to right.
+ *
+ * Poppler's own `<line>` grouping cannot be used for this. It groups by text
+ * flow, and these invoices lay their fields out as two columns of independent
+ * flows — so `PERIODO FACTURADO:` and the `20260630-20260630` printed beside
+ * it end up in different `<line>` elements, and every label-then-value pattern
+ * in the parsers misses a value that is plainly there on the page. Regrouping
+ * by vertical position restores the adjacency, and matches what tesseract
+ * hands back for the scanned version of the same layout.
+ *
+ * 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.
+ */
+function toVisualRows(words: OcrWord[]): string {
+  const centre = (w: OcrWord) => w.top + w.height / 2;
+  const sorted = [...words].sort((a, b) => centre(a) - centre(b) || a.left - b.left);
+
+  const rows: OcrWord[][] = [];
+  let current: OcrWord[] = [];
+  let band = 0;
+
+  for (const w of sorted) {
+    if (!current.length) {
+      current = [w];
+      band = centre(w);
+      continue;
+    }
+    // Half the word's own height: tall headings and body text both sit within
+    // their own line's band, and neither reaches into the next one.
+    if (Math.abs(centre(w) - band) <= Math.max(w.height, current[0].height) / 2) {
+      current.push(w);
+    } else {
+      rows.push(current);
+      current = [w];
+      band = centre(w);
+    }
+  }
+  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");
+}
+
+/**
+ * Turn tesseract's TSV into words plus reassembled text.
+ *
+ * Columns are: level, page_num, block_num, par_num, line_num, word_num, left,
+ * top, width, height, conf, text. Rows with level < 5 are structural (page,
+ * block, paragraph, line) and carry no text; only level 5 is a word. A conf of
+ * -1 marks a structural row, so those are dropped rather than averaged in —
+ * including them would drag every page's confidence toward zero.
+ */
+export function parseTsv(tsv: string): OcrPage {
+  const lines = tsv.split("\n");
+  const header = lines[0]?.split("\t") ?? [];
+  const col = (name: string) => header.indexOf(name);
+  const iLeft = col("left");
+  const iTop = col("top");
+  const iWidth = col("width");
+  const iHeight = col("height");
+  const iConf = col("conf");
+  const iText = col("text");
+  const iLine = col("line_num");
+  const iBlock = col("block_num");
+
+  const words: OcrWord[] = [];
+  // Keyed by block+line so the reassembled text preserves the engine's own
+  // reading order instead of sorting words by raw y, which interleaves columns.
+  const byLine = new Map<string, string[]>();
+
+  for (let i = 1; i < lines.length; i++) {
+    const f = lines[i].split("\t");
+    if (f.length <= iText) continue;
+    const text = f[iText]?.trim();
+    if (!text) continue;
+    const confidence = Number(f[iConf]);
+    if (!Number.isFinite(confidence) || confidence < 0) continue;
+
+    words.push({
+      text,
+      left: Number(f[iLeft]) || 0,
+      top: Number(f[iTop]) || 0,
+      width: Number(f[iWidth]) || 0,
+      height: Number(f[iHeight]) || 0,
+      confidence: confidence / 100,
+    });
+
+    const key = `${f[iBlock]}:${f[iLine]}`;
+    const bucket = byLine.get(key);
+    if (bucket) bucket.push(text);
+    else byLine.set(key, [text]);
+  }
+
+  const text = [...byLine.values()].map((w) => w.join(" ")).join("\n");
+  const confidence = words.length
+    ? words.reduce((sum, w) => sum + w.confidence, 0) / words.length
+    : 0;
+
+  return { text, words, confidence };
+}

apps/api/src/statements/parsers/statement-parser.spec.ts

@@ -0,0 +1,270 @@
+import type { OcrPage } from "../ocr/ocr.provider";
+import {
+  detectProvider,
+  normalizeCadastralKey,
+  normalizeZofematKey,
+  parseStatement,
+} from "./statement-parser";
+
+/**
+ * Every string in this file is a verbatim excerpt of what the OCR engine
+ * actually returned for a real receipt — misreads, dropped spaces, mangled
+ * accents and all. That is the point: these are the specific ways these five
+ * layouts have been observed to fail, and the assertions pin down what the
+ * parser is supposed to do about each one. Inventing clean input here would
+ * test nothing, because clean input was never the problem.
+ */
+function page(text: string): OcrPage {
+  return { text, words: [], confidence: 0.9 };
+}
+
+describe("detectProvider", () => {
+  it("reads a Rosarito predial receipt as predial, not as a water bill", () => {
+    // "Clave Catastral" is also a CESPT structural marker, so a predial page
+    // whose header OCR'd badly must still not be claimed by the CESPT rule.
+    expect(
+      detectProvider(
+        "e | Clave Catastral. KP-128-105 IMPUESTO PREDIAL ea rita\n" +
+          "TASA | VALOR FISCAL | BIMESTRES | INCISO. | IMPUESTO",
+      ),
+    ).toBe("PREDIAL ROSARITO");
+  });
+
+  it("keeps telling the three municipalities apart by their RFC", () => {
+    expect(detectProvider("R.F.C. ATB-541201-KK2")).toBe("PREDIAL TIJUANA");
+    expect(detectProvider("R.F.C. AMP-981201-HJ4")).toBe("PREDIAL ROSARITO");
+    expect(detectProvider("MEN-540301-9J5")).toBe("PREDIAL ENSENADA");
+  });
+
+  it("does not let the CFE rule claim a gas bill over 'PERIODO FACTURADO'", () => {
+    expect(
+      detectProvider("Orden de Facturación: 000009801640\nPERIODO FACTURADO: 20260630-20260630"),
+    ).toBe("GAS TIJUANA");
+  });
+});
+
+describe("normalizeCadastralKey", () => {
+  it("keeps a letter in the third position instead of digitising it", () => {
+    // `MMB01041` is a real key on file; mapping its B to 8 produced a key that
+    // matches no property at all.
+    expect(normalizeCadastralKey("MM-B01-041", [])).toBe("MMB01041");
+  });
+
+  it("repairs the spurious I tesseract inserts into the prefix", () => {
+    expect(normalizeCadastralKey("MIM-200-010", [])).toBe("MM200010");
+  });
+
+  it("digitises confusable glyphs from position four onward", () => {
+    expect(normalizeCadastralKey("KP-1O8-O45", [])).toBe("KP108045");
+  });
+
+  it("flags a prefix it had to truncate", () => {
+    const notes: string[] = [];
+    expect(normalizeCadastralKey("KPX-128-106", notes)).toBe("KP128106");
+    expect(notes).toHaveLength(1);
+  });
+});
+
+describe("parsePredialTijuana", () => {
+  const TIJUANA = page(
+    "Hats | AYUNTAMIENTO DE TIJUANA, BC $2,613.00 23/01/2026\n" +
+      "y) TELEFONO: 973-7000 R.F.C. ATB-541201-KK2\n" +
+      "ER AÑO VALOR FISCAL TASA IMPUESTO |CONCEPTO IMPORTE\n" +
+      "ED ca 2026 1,207,15778 246 2,969.61 1102 - IMPUESTO PREDIAL 2,969.61\n" +
+      "55164964310126000002613000054192\n" +
+      "se 0 O (54427 [a] | TOTALAPAGAR: 2,613.00\n" +
+      "Dc 1097 : FECHA VENCE : 31/ENE/2026",
+  );
+
+  it("splits the payment barcode into account, deadline and amount", () => {
+    const p = parseStatement(TIJUANA);
+    expect(p.provider).toBe("PREDIAL TIJUANA");
+    expect(p.serviceKind).toBe("PROPERTY_TAX");
+    expect(p.accountRef).toBe("55164964");
+    expect(p.amount).toBe(2613);
+    expect(p.dueDate?.toISOString().slice(0, 10)).toBe("2026-01-31");
+    expect(p.period).toBe("2026");
+  });
+
+  it("reads the printed total even when the space in the label is lost", () => {
+    // The real page OCR'd the label as "TOTALAPAGAR:", and it is that reading
+    // that cross-checks the barcode's amount.
+    expect(parseStatement(TIJUANA).crossChecked).toBe(true);
+  });
+
+  it("refuses to trust a barcode the printed total contradicts", () => {
+    const p = parseStatement(
+      page(
+        "R.F.C. ATB-541201-KK2\n" +
+          "55164964310126000002613000054192\n" +
+          "TOTAL A PAGAR: 9,613.00\nFECHA VENCE : 31/ENE/2026",
+      ),
+    );
+    expect(p.crossChecked).toBe(false);
+    expect(p.notes.join(" ")).toContain("no coincide");
+  });
+});
+
+describe("parsePredialRosarito", () => {
+  it("takes the rounded Total, not the Sub Total printed above it", () => {
+    const p = parseStatement(
+      page(
+        "AYUNTAMIENTO MUNICIPAL DE PLAYAS DE ROSARITO, B.C.\n" +
+          "Ce Clave Catastral: + JR-400-008 7 | IMPUESTO PREDIAL\n" +
+          "SUPERFICIE: 228.31 ZONA 30025 “Redondeo IT049 -$0.39 Sub Total $5,409.39\n" +
+          "¿XTEMPORANEO DESPUES DE: 31/01/2026 Elaboro: MGLG\n" +
+          "Total | $5,409.00\n" +
+          "| Periodo por Pagar: 2026/1 2026/6",
+      ),
+    );
+    expect(p.cadastralKey).toBe("JR400008");
+    expect(p.amount).toBe(5409);
+    expect(p.dueDate?.toISOString().slice(0, 10)).toBe("2026-01-31");
+    expect(p.period).toBe("2026");
+  });
+
+  it("is not fooled by the unspaced 'SubTotal' spelling", () => {
+    // This exact page read $9,624.85 off a receipt for $9,625.00 while the
+    // lookbehind still assumed a space.
+    const p = parseStatement(
+      page(
+        "AMP-981201-HJ4 IMPUESTO PREDIAL\n" +
+          "SUPERFICIE. 367.62 ZONA:30151 | Redondco 17049 $0.15 SubTotal $9,624.85\n" +
+          ": Total | $9,625.00",
+      ),
+    );
+    expect(p.amount).toBe(9625);
+  });
+});
+
+describe("parsePredialEnsenada", () => {
+  const totals = (tail: string) =>
+    page(
+      "IMPRESION MAQUINA REGISTRADORA ez | MUNICIPIO DE ENSENADA\n" +
+        "+7] DATOS. DEL.CAUSANTE alta A pe CLAVE MM-200-010 2 CUENTA\n" +
+        `ES g € S| TOTALES 12,744.47 0.00 0.00 324.56 0.00 13,069.03 ${tail} |`,
+    );
+
+  it("reads the paid total off the TOTALES row however the label OCR'd", () => {
+    expect(parseStatement(totals("TOTA LA A $5,797.00")).amount).toBe(5797);
+    expect(parseStatement(totals("orAL: M7 z] $14,414.00")).amount).toBe(14414);
+    expect(parseStatement(totals("| TOTAL: = $6 246.00")).amount).toBe(6246);
+  });
+
+  it("reports no amount rather than one whose $ was misread as an 8", () => {
+    // `TOTAL: A 82,203.00` is a $2,203.00 receipt. Posting $82,203 would look
+    // entirely ordinary in the ledger, so this page must go to review instead.
+    const p = parseStatement(totals("TOTAL: A 82,203.00"));
+    expect(p.amount).toBeNull();
+    expect(p.notes.join(" ")).toContain("capturarlo a mano");
+  });
+
+  it("never falls back to the assessed total on the same row", () => {
+    expect(parseStatement(totals("yo: se TE= 58/4690]")).amount).toBeNull();
+  });
+});
+
+describe("parseGas", () => {
+  const gas = (...cuentas: string[]) =>
+    page(
+      "GTI4608032K2 COMPAÑIA DE GAS DE TIJUANA\n" +
+        "Fecha de Vencimiento: 2026/08/08\n" +
+        cuentas.map((c) => `Cuenta: ${c}`).join("\n") +
+        "\nPERIODO FACTURADO: 20260630-20260630\nTOTAL A PAGAR: $275.82",
+    );
+
+  it("strips the printed leading zero to the stored account number", () => {
+    const p = parseStatement(gas("0900003463", "0900003463", "0900003463"));
+    expect(p.serviceKind).toBe("GAS");
+    expect(p.accountRef).toBe("900003463");
+    expect(p.amount).toBe(275.82);
+    expect(p.dueDate?.toISOString().slice(0, 10)).toBe("2026-08-08");
+    expect(p.period).toBe("2026-06");
+    expect(p.crossChecked).toBe(true);
+  });
+
+  it("takes the majority reading but still sends a disagreement to review", () => {
+    const p = parseStatement(gas("0900003463", "0900003463", "0900003468"));
+    expect(p.accountRef).toBe("900003463");
+    expect(p.crossChecked).toBe(false);
+  });
+
+  it("claims no cross-check from a single printing", () => {
+    expect(parseStatement(gas("0900003463")).crossChecked).toBeNull();
+  });
+});
+
+describe("parseZonaFederal", () => {
+  /**
+   * The Tijuana zona federal receipt, trimmed to the rows the parser reads.
+   * Verbatim from page 7 of the August 2026 batch, including the two ways the
+   * heading OCR'd: the clave line is struck through by the office's own
+   * highlighter, which is what cost two of eight pages their concession clave.
+   */
+  const zf = (clave: string, body = "") =>
+    page(
+      "ESIZ <pYl Av. Independencia y Esq. Paseo del CentenaxiaiiArlhnto de Tijuana, B.C.\n" +
+        "Teléfono: 9737000 R.F.C. ATB-541201-BK2 0070000146 12:54 PM\n" +
+        "Zona Federal Marítimo Terrestre\n" +
+        `${clave} Nombre: DENNIS JOHN SEIN Concesión:\n` +
+        "Periodo Construcción Tasa  Ornato Tasa Impuesto Actualiza. Recargo Multa Importe\n" +
+        "2026-2 / 2026-2 316.40 35.00 0.00 12.11 1,845.66 0.00 27.13 1,000.00 2,872.79\n" +
+        "SubTotal 1,845.66 0.00 27.13 1,000.00 2,872.79\n" +
+        "Concepto: Derechos de ocupación de Zona Federal Marítimo Terrestre\n" +
+        body,
+    );
+
+  it("is not claimed by the predial parser that shares its RFC and header", () => {
+    // Tijuana bills predial and zona federal from the same treasury, so
+    // "Ayuntamiento de Tijuana" and ATB-541201 identify neither on their own.
+    expect(detectProvider("R.F.C. ATB-541201-BK2\nZona Federal Marítimo Terrestre")).toBe(
+      "ZONA FEDERAL TIJUANA",
+    );
+    expect(parseStatement(zf("Clave: 14-D -014")).serviceKind).toBe("FEDERAL_ZONE");
+  });
+
+  it("still recognises the layout when the heading itself did not survive OCR", () => {
+    // Real: page 1 came back as "Zona Ledera) Maritimo Terrestre".
+    expect(
+      detectProvider("Zona Ledera) Maritimo Terrestre\nClave EJ -012% Nombre: STEFAN"),
+    ).toBe("ZONA FEDERAL TIJUANA");
+  });
+
+  it("reads the clave through the loose spacing the receipt prints", () => {
+    expect(parseStatement(zf("Clave: 14-D -014")).accountRef).toBe("14D014");
+    expect(parseStatement(zf("Clave: 14-A-119")).accountRef).toBe("14A119");
+  });
+
+  it("keeps the letter instead of digitising it", () => {
+    // toDigits maps D to 0 and B to 8; a real 14-D -014 must not become 140014.
+    expect(normalizeZofematKey("14-D -014")).toBe("14D014");
+    expect(normalizeZofematKey("12-B -013")).toBe("12B013");
+  });
+
+  it("takes the payable amount from the SubTotal row, rounded to whole pesos", () => {
+    // The municipality rounds and prints the difference as "Ajuste Ley Hacienda
+    // Mpal"; 2,872.79 is charged as $2,873.00.
+    expect(parseStatement(zf("Clave: 14-D -014")).amount).toBe(2873);
+  });
+
+  it("prefers the printed total and cross-checks it against the subtotal", () => {
+    const p = parseStatement(zf("Clave: 14-D -014", "Total a pagar $2,873.00"));
+    expect(p.amount).toBe(2873);
+    expect(p.crossChecked).toBe(true);
+  });
+
+  it("sends a printed total that contradicts the subtotal to review", () => {
+    const p = parseStatement(zf("Clave: 14-D -014", "Total a pagar $2,973.00"));
+    expect(p.crossChecked).toBe(false);
+  });
+
+  it("translates the printed bimester into the ledger's own vocabulary", () => {
+    expect(parseStatement(zf("Clave: 14-D -014")).period).toBe("MAR/APR");
+  });
+
+  it("leaves the clave blank rather than guessing when the marker ate it", () => {
+    const p = parseStatement(zf("Clave EJ -012%"));
+    expect(p.accountRef).toBeNull();
+    expect(p.notes.join(" ")).toContain("clave");
+  });
+});

apps/api/src/statements/parsers/statement-parser.ts

@@ -0,0 +1,872 @@
+import type { ServiceKind } from "@jorgecuadros/database";
+import type { OcrPage, OcrWord } from "../ocr/ocr.provider";
+
+/**
+ * What one parsed statement page yields. `accountRef` is already normalised to
+ * the form the migrated `PropertyService` columns hold, so the matcher compares
+ * like with like and never has to know about provider-specific formatting.
+ */
+export interface ParsedStatement {
+  /**
+   * "CFE" | "CESPT" | "TELNOR" | "GAS TIJUANA" | "PREDIAL TIJUANA" |
+   * "PREDIAL ROSARITO" | "PREDIAL ENSENADA" | "ZONA FEDERAL TIJUANA", or null
+   * when no parser claimed the page.
+   */
+  provider: string | null;
+  serviceKind: ServiceKind | null;
+  accountRef: string | null;
+  /** Clave catastral, when printed — a second key to match on. */
+  cadastralKey: string | null;
+  amount: number | null;
+  dueDate: Date | null;
+  period: string | null;
+  /**
+   * Independent corroboration of `accountRef`. CFE and Telnor both print a
+   * payment barcode that repeats the account number (and the amount), so when
+   * the barcode and the label agree the extraction is near-certainly right;
+   * when they disagree, or only one is present, the page is worth a human
+   * glance. Null when the layout has no second source.
+   */
+  crossChecked: boolean | null;
+  /** Human-readable trail of what was read, surfaced in the review queue. */
+  notes: string[];
+}
+
+// --- shared helpers ---------------------------------------------------------
+
+/**
+ * Tesseract confuses these glyphs inside numeric runs with some regularity —
+ * a real clave catastral `KB078025` came back as `KBO78025`. Applied ONLY to
+ * fields known to be digits, never to free text, where it would corrupt words.
+ */
+const DIGIT_CONFUSIONS: Record<string, string> = {
+  O: "0",
+  o: "0",
+  D: "0",
+  I: "1",
+  l: "1",
+  "|": "1",
+  S: "5",
+  B: "8",
+};
+
+export function toDigits(s: string | null | undefined): string {
+  if (!s) return "";
+  return s
+    .split("")
+    .map((c) => DIGIT_CONFUSIONS[c] ?? c)
+    .join("")
+    .replace(/\D/g, "");
+}
+
+/**
+ * Parse a printed amount, treating `,` and `.` by position rather than by
+ * assumption. A real Telnor bill OCR'd as "$ 649,00" — blindly stripping commas
+ * as thousands separators turned $649.00 into $64,900, a hundredfold error that
+ * would post silently. Two trailing digits after a single separator are always
+ * cents here; a separator followed by three digits is a thousands group.
+ */
+function money(s: string | null | undefined): number | null {
+  if (!s) return null;
+  const cleaned = s.replace(/[\s$]/g, "");
+
+  // 1.234,56 or 1,234.56 — grouped thousands plus optional cents.
+  let m = cleaned.match(/^(\d{1,3}(?:[.,]\d{3})+)([.,]\d{1,2})?$/);
+  if (m) {
+    const whole = m[1].replace(/[.,]/g, "");
+    const cents = m[2] ? m[2].slice(1) : "";
+    return Number(cents ? `${whole}.${cents.padEnd(2, "0")}` : whole);
+  }
+
+  // 649,00 / 649.00 — a single separator with exactly two digits after it.
+  m = cleaned.match(/^(\d+)[.,](\d{2})$/);
+  if (m) return Number(`${m[1]}.${m[2]}`);
+
+  const n = Number(cleaned.replace(/[,.]/g, ""));
+  return Number.isFinite(n) ? n : null;
+}
+
+function firstMatch(text: string, patterns: RegExp[]): string | null {
+  for (const p of patterns) {
+    const m = text.match(p);
+    if (m?.[1]) return m[1].trim();
+  }
+  return null;
+}
+
+/** Every capture of `pattern` across the page, in order. */
+function allMatches(text: string, pattern: RegExp): string[] {
+  const out: string[] = [];
+  const re = new RegExp(pattern.source, pattern.flags.includes("g") ? pattern.flags : `${pattern.flags}g`);
+  for (const m of text.matchAll(re)) {
+    if (m[1]) out.push(m[1].trim());
+  }
+  return out;
+}
+
+const MONTHS: Record<string, number> = {
+  ENE: 0, FEB: 1, MAR: 2, ABR: 3, MAY: 4, JUN: 5,
+  JUL: 6, AGO: 7, SEP: 8, OCT: 9, NOV: 10, DIC: 11,
+};
+
+/** Parses the three date shapes these statements actually print. */
+export function parseDate(raw: string | null | undefined): Date | null {
+  if (!raw) return null;
+  const s = raw.trim().toUpperCase();
+
+  // 16/07/2026
+  let m = s.match(/^(\d{1,2})\/(\d{1,2})\/(\d{4})$/);
+  if (m) return utc(+m[3], +m[2] - 1, +m[1]);
+
+  // 22-JUL-2026  /  22 JUN 26  /  31/ENE/2026 (Tijuana predial)
+  m = s.match(/^(\d{1,2})[-\s/]([A-Z]{3})[A-Z]*[-\s/](\d{2,4})$/);
+  if (m && MONTHS[m[2]] !== undefined) {
+    const y = m[3].length === 2 ? 2000 + +m[3] : +m[3];
+    return utc(y, MONTHS[m[2]], +m[1]);
+  }
+
+  // 2026-07-22 (already normalised, e.g. decoded from a barcode) and the
+  // 2026/08/08 the gas bill prints — same field order, different separator.
+  m = s.match(/^(\d{4})[-/](\d{2})[-/](\d{2})$/);
+  if (m) return utc(+m[1], +m[2] - 1, +m[3]);
+
+  return null;
+}
+
+function utc(y: number, mo: number, d: number): Date | null {
+  const dt = new Date(Date.UTC(y, mo, d));
+  return Number.isNaN(dt.getTime()) ? null : dt;
+}
+
+/**
+ * Read the value printed *underneath* a column header.
+ *
+ * The CESPT "RECIBO" is a table: `No. DE CUENTA` is a header cell and its value
+ * sits in the row below it, so no amount of label-adjacent regex on line text
+ * can associate the two. This walks the word boxes instead — find the header
+ * word, then take the nearest word below it whose horizontal centre falls
+ * within the column.
+ */
+export function valueUnder(
+  page: OcrPage,
+  header: RegExp,
+  opts: { maxDy?: number; tolerance?: number; match?: RegExp } = {},
+): string | null {
+  const { maxDy = 300, tolerance = 200, match } = opts;
+  const centre = (w: OcrWord) => ({
+    x: w.left + w.width / 2,
+    y: w.top + w.height / 2,
+  });
+
+  for (const h of page.words.filter((w) => header.test(w.text))) {
+    const hc = centre(h);
+    const below = page.words
+      .filter((w) => {
+        const c = centre(w);
+        return c.y > hc.y && c.y <= hc.y + maxDy && Math.abs(c.x - hc.x) <= tolerance;
+      })
+      .sort((a, b) => centre(a).y - centre(b).y);
+
+    for (const w of below) {
+      if (!match || match.test(w.text)) return w.text;
+    }
+  }
+  return null;
+}
+
+// --- provider detection -----------------------------------------------------
+
+/**
+ * Brand wordmarks first, page structure only as a fallback — and the two passes
+ * must not be interleaved. Scanned logos OCR badly (one CESPT header came back
+ * as "E BAJA ES PAGO / EALIFORNIA", with neither "CESPT" nor "COMISIÓN ESTATAL"
+ * readable), so the structural pass is what rescues those pages. But a Telnor
+ * bill contains the words "Pagar antes de", which a CFE structural rule
+ * evaluated first will happily claim — running all brand checks before any
+ * structural check is what keeps that from happening.
+ */
+const BRAND: [string, RegExp][] = [
+  ["CFE", /comisi[oó]n federal de electricidad|CFE.?contigo|Suministrador de Servicios/i],
+  ["CESPT", /CESPT|COMISI[OÓ]N ESTATAL DE SERVICIOS/i],
+  ["TELNOR", /TELNOR|TELEFONOS DEL NOROESTE/i],
+  ["GAS TIJUANA", /COMPA[ÑN][IÍ]?A\s*DE\s*GAS\s*DE\s*TIJUANA|bajagas/i],
+  // Ahead of the predial rules on purpose. Tijuana's zona federal receipt is
+  // issued by the same treasury and carries the same header — "Ayuntamiento de
+  // Tijuana", the same address, the same `ATB-541201` RFC — so every predial
+  // discriminator matches it too, and whichever rule is asked first wins the
+  // page. What only the zona federal layout says is "Marítimo Terrestre", which
+  // survived OCR on all eight sample pages even where the heading above it came
+  // back as "Zona Ledera) Maritimo Terrestre" and the printed concession clave
+  // was lost under a highlighter mark.
+  ["ZONA FEDERAL TIJUANA", /ZOFEMAT|Mar[ií]timo\s*Terrestre|ocupaci[oó]n\s*de\s*Zona\s*Federal/i],
+  // The municipal RFCs are the single most reliable discriminator on a predial
+  // receipt: they are printed in a clean monospaced run on every layout, they
+  // never change, and they say which of the three city treasuries issued the
+  // page — which the wordmarks alone do not, since a Tijuana receipt also
+  // carries "PLAYAS DE TIJUANA" and a Rosarito one "TIJUANA ENSENADA".
+  ["PREDIAL TIJUANA", /AYUNTAMIENTO\s*DE\s*TIJUANA|ATB.?541201/i],
+  ["PREDIAL ROSARITO", /AYUNTAMIENTO\s*MUNICIPAL\s*DE\s*PLAYAS\s*DE\s*ROSARITO|AMP.?981201|rosarito\.gob/i],
+  ["PREDIAL ENSENADA", /MUNICIPIO\s*DE\s*ENSENADA|MEN.?540301/i],
+];
+
+/**
+ * The predial rules come first because a Rosarito receipt prints "Clave
+ * Catastral" as a boxed label — the very string the CESPT structural rule
+ * looks for — so a page whose municipal header failed to OCR would otherwise
+ * be claimed as a water bill and matched against the wrong column entirely.
+ * "IMPUESTO PREDIAL" appears on all three municipal layouts and on none of the
+ * utility ones, so it is the safe first question to ask.
+ */
+const LAYOUT: [string, RegExp][] = [
+  // Same reasoning as the brand pass, one rule earlier: the concept line
+  // "Derechos de ocupación de Zona Federal Marítimo Terrestre" is printed on
+  // the stub of every zona federal page and on no other layout, and it read
+  // cleanly on 8 of 8 samples — including the two whose heading did not.
+  ["ZONA FEDERAL TIJUANA", /Derechos\s*de\s*ocupaci[oó]n/i],
+  ["PREDIAL TIJUANA", /IMPUESTO\s*PREDIAL[\s\S]*?(?:CERTIFICACION\s*DE\s*CAJA|PASEO\s*DEL\s*CENTENARIO|PAGA\s*TU\s*PREDIAL)/i],
+  ["PREDIAL ENSENADA", /(?:IMPUESTO\s*PREDIAL[\s\S]*?TRANSPENINSULAR)|(?:IMPRESION\s*MAQUINA\s*REGISTRADORA)/i],
+  ["PREDIAL ROSARITO", /IMPUESTO\s*PREDIAL/i],
+  ["GAS TIJUANA", /Orden\s*de\s*Facturaci[oó]n|FACTOR\s*DE\s*PRESI[OÓ]N|GAS\s*LP/i],
+  ["CFE", /NO\.?\s*DE\s*SERVICIO|L[IÍ]MITE\s*DE\s*PAGO|PERIODO\s*FACTURADO/i],
+  ["CESPT", /SALDO\s+CORRIENTE|CLAVE\s*CATASTRAL|No\.?\s*DE\s*CUENTA/i],
+  ["TELNOR", /Mes\s*de\s*Facturaci[oó]n|Pagar\s*antes\s*de/i],
+];
+
+export function detectProvider(text: string): string | null {
+  for (const group of [BRAND, LAYOUT]) {
+    for (const [name, pattern] of group) {
+      if (pattern.test(text)) return name;
+    }
+  }
+  return null;
+}
+
+// --- CFE (electric) ---------------------------------------------------------
+
+function parseCfe(page: OcrPage): ParsedStatement {
+  const text = page.text;
+  const notes: string[] = [];
+
+  // The payment barcode line repeats the service number, the due date (YYMMDD)
+  // and the amount in one fixed-width run, and reads far more reliably than the
+  // label: on one sample the label came back as "0059603001917" (a digit too
+  // many) while its barcode gave the correct "005960300191". So the barcode
+  // wins, and the label becomes the cross-check rather than the source.
+  const barcode = text.match(/\b01\s+([0-9OIlSBD]{12})\s+([0-9OIlSBD]{6})\s+([0-9OIlSBD]{9})\b/);
+  const label = firstMatch(text, [/NO\.?\s*DE\s*SERVICIO\s*[:;.]?\s*([0-9OIlSBD]{10,14})/i]);
+
+  let accountRef: string | null = null;
+  let amount: number | null = null;
+  let dueDate: Date | null = null;
+  let crossChecked: boolean | null = null;
+
+  if (barcode) {
+    // Leading zeros are print padding: DATMEX.rpu holds the bare 10 digits.
+    accountRef = toDigits(barcode[1]).replace(/^0+/, "");
+    amount = Number(toDigits(barcode[3]));
+    const d = toDigits(barcode[2]);
+    dueDate = parseDate(`20${d.slice(0, 2)}-${d.slice(2, 4)}-${d.slice(4, 6)}`);
+    notes.push("importe y vencimiento leídos del código de barras");
+    if (label) {
+      crossChecked = toDigits(label).replace(/^0+/, "") === accountRef;
+      if (!crossChecked) {
+        notes.push(
+          `el número impreso (${toDigits(label).replace(/^0+/, "")}) no coincide con el código de barras`,
+        );
+      }
+    }
+  } else if (label) {
+    accountRef = toDigits(label).replace(/^0+/, "");
+    notes.push("sin código de barras legible; número tomado de la etiqueta");
+  }
+
+  if (amount == null) {
+    amount = money(firstMatch(text, [/TOTAL\s*A\s*PAGAR\s*[:;.]?\s*\$?\s*([\d,]+\.?\d*)/i]));
+  }
+  if (!dueDate) {
+    dueDate = parseDate(
+      firstMatch(text, [/L[IÍ]MITE\s*DE\s*PAGO\s*[:;.]?\s*(\d{1,2}\s+\w{3}\s+\d{2,4})/i]),
+    );
+  }
+
+  return {
+    provider: "CFE",
+    serviceKind: "ELECTRIC",
+    accountRef: accountRef || null,
+    cadastralKey: null,
+    amount,
+    dueDate,
+    period: firstMatch(text, [
+      /PERIODO\s*FACTURADO\s*[:;.]?\s*(\d{1,2}\s+\w{3}\s+\d{2}\s*-\s*\d{1,2}\s+\w{3}\s+\d{2})/i,
+    ]),
+    crossChecked,
+    notes,
+  };
+}
+
+// --- CESPT (water) ----------------------------------------------------------
+
+/**
+ * Two different layouts arrive under the same brand:
+ *   - the line-oriented "COMPROBANTE DE PAGO" (`Cuenta : 7604192`), and
+ *   - the tabular "RECIBO", where `No. DE CUENTA` is a column header.
+ * Line patterns are tried first; anything they miss falls through to the
+ * geometric read, which is what the tabular layout needs.
+ */
+function parseCespt(page: OcrPage): ParsedStatement {
+  const text = page.text;
+  const notes: string[] = [];
+
+  let account = firstMatch(text, [/Cuenta\s*[:;.]?\s*([0-9OIlSBD]{5,9})/i]);
+  if (!account) {
+    account = valueUnder(page, /^CUENTA$/i, { match: /^[0-9OIlSBD]{5,9}$/ });
+    if (account) notes.push("número de cuenta leído de la columna del recibo");
+  }
+
+  let clave = firstMatch(text, [/Cve\.?\s*Cat\.?\s*[:;.]?\s*([A-Z]{2}\s?[0-9OIlSBD]{6})/i]);
+  if (!clave) {
+    clave = valueUnder(page, /^CATASTRAL$/i, { match: /^[A-Z]{2}[0-9OIlSBD]{6}$/i });
+    if (clave) notes.push("clave catastral leída de la columna del recibo");
+  }
+
+  let due = firstMatch(text, [/Fecha\s*Venc\s*[:;.]?\s*(\d{2}\/\d{2}\/\d{4})/i]);
+  if (!due) due = valueUnder(page, /^VENCIMIENTO$/i, { match: /^\d{2}\/\d{2}\/\d{4}$/ });
+
+  const amount = money(
+    firstMatch(text, [
+      /TOTAL\s*[:;.]?\s*\$?\s*([\d,]+\.\d{2})/i,
+      /SALDO\s+CORRIENTE[^\n]*?([\d,]+\.\d{2})/i,
+    ]),
+  );
+
+  // Leading zeros are print padding here too: the RECIBO prints `0457341` for
+  // what DATMEX.agua holds as `457341`.
+  const accountRef = account ? toDigits(account).replace(/^0+/, "") : null;
+  const cadastralKey = clave
+    ? clave.replace(/\s/g, "").slice(0, 2).toUpperCase() +
+      toDigits(clave.replace(/\s/g, "").slice(2))
+    : null;
+
+  return {
+    provider: "CESPT",
+    serviceKind: "WATER",
+    accountRef: accountRef || null,
+    cadastralKey: cadastralKey || null,
+    amount,
+    dueDate: parseDate(due),
+    period: null,
+    crossChecked: null,
+    notes,
+  };
+}
+
+// --- TELNOR (telephone) -----------------------------------------------------
+
+function parseTelnor(page: OcrPage): ParsedStatement {
+  const text = page.text;
+  const notes: string[] = [];
+
+  const label = firstMatch(text, [
+    /Tel[eé]fono\s*[:;.]?\s*([0-9OIlSBD]{3}\s?[0-9OIlSBD]{3}\s?[0-9OIlSBD]{4})/i,
+  ]);
+  // The payment stub prints phone (10 digits) + amount in cents (9) + a check
+  // digit: `6646093444 000099900 7` for a $999.00 bill. Reading the amount as
+  // 10 digits swallows the check digit and inflates the figure 100-fold.
+  const barcode = text.match(/\b(\d{10})(\d{9})\d\b/);
+
+  let accountRef: string | null = null;
+  let crossChecked: boolean | null = null;
+
+  // The bill prints the number with its 664 Tijuana LADA; DATMEX stores the
+  // bare local 7 digits, so the LADA is dropped rather than the stored value
+  // being padded — padding would guess at an area code for the 500+ existing
+  // rows that never recorded one.
+  if (label) accountRef = toDigits(label).slice(-7);
+  if (barcode) {
+    const fromBarcode = barcode[1].slice(-7);
+    if (accountRef) {
+      crossChecked = fromBarcode === accountRef;
+      if (!crossChecked) notes.push("el teléfono impreso no coincide con el código de barras");
+    } else {
+      accountRef = fromBarcode;
+      notes.push("teléfono leído del código de barras");
+    }
+  }
+
+  let amount = money(firstMatch(text, [/Total\s*a\s*Pagar\s*[:;.]?\s*\$?\s*([\d,]+\.?\d{0,2})/i]));
+  if (amount == null && barcode) {
+    amount = Number(barcode[2]) / 100;
+    notes.push("importe leído del código de barras");
+  }
+
+  return {
+    provider: "TELNOR",
+    serviceKind: "TELEPHONE",
+    accountRef: accountRef || null,
+    cadastralKey: null,
+    amount,
+    dueDate: parseDate(
+      firstMatch(text, [/Pagar\s*antes\s*de\s*[:;.]?\s*(\d{2}-\w{3}-\d{4})/i]),
+    ),
+    period: firstMatch(text, [/Mes\s*de\s*Facturaci[oó]n\s*[:;.]?\s*(\w+)/i]),
+    crossChecked,
+    notes,
+  };
+}
+
+// --- GAS (Compañía de Gas de Tijuana / bajagas) ------------------------------
+
+/**
+ * These arrive as born-digital CFDI PDFs rather than scans, so the text layer
+ * (see `TesseractOcrProvider.textPages`) usually reads them exactly and the
+ * patterns below only have to be tolerant enough for the scanned case.
+ *
+ * The account number is printed three times — supply address, fiscal data, and
+ * the payment stub at the foot — which is a free cross-check: three readings
+ * that agree are near-certainly right, and any disagreement means one of them
+ * was misread and the page deserves a human glance.
+ *
+ * `Cuenta` is what the matcher compares, not `Contrato`. The migration
+ * recovered gas references out of `PropertyService.notes` into `meterNumber`
+ * and what sat there is the 9-digit account (`900003463`), printed here with a
+ * leading zero as `0900003463`.
+ */
+function parseGas(page: OcrPage): ParsedStatement {
+  const text = page.text;
+  const notes: string[] = [];
+
+  const seen = allMatches(text, /Cuenta\s*[:;.]?\s*([0-9OIlSBD]{6,12})/i).map((s) =>
+    toDigits(s).replace(/^0+/, ""),
+  );
+  const distinct = [...new Set(seen.filter(Boolean))];
+
+  let accountRef: string | null = null;
+  let crossChecked: boolean | null = null;
+  if (distinct.length === 1) {
+    accountRef = distinct[0];
+    if (seen.length > 1) crossChecked = true;
+  } else if (distinct.length > 1) {
+    // Majority wins — the stub and the two address blocks print the same
+    // number, so a single divergent reading is the misread one. It still goes
+    // to review: `crossChecked: false` is what keeps the batch from
+    // auto-matching a number one of three readings disagreed with.
+    const tally = new Map<string, number>();
+    for (const s of seen) tally.set(s, (tally.get(s) ?? 0) + 1);
+    accountRef = [...tally.entries()].sort((a, b) => b[1] - a[1])[0][0];
+    crossChecked = false;
+    notes.push(`el número de cuenta se leyó de ${distinct.length} formas distintas (${distinct.join(", ")})`);
+  }
+
+  const amount = money(
+    firstMatch(text, [
+      /TOTAL\s*A\s*PAGAR\s*[:;.]?\s*\$\s*([\d,]+\.\d{2})/i,
+      /Total\s*a\s*pagar\s*[:;.]?\s*\$\s*([\d,]+\.\d{2})/i,
+    ]),
+  );
+
+  // `20260630-20260630` — the range the bill was cut for. Both ends are the
+  // same reading date on every sample, so the period is reported as the ISO
+  // month rather than a range no ledger row would ever be searched by.
+  const facturado = firstMatch(text, [/PERIODO\s*FACTURADO\s*[:;.]?\s*(\d{8})\s*-\s*\d{8}/i]);
+  const period = facturado ? `${facturado.slice(0, 4)}-${facturado.slice(4, 6)}` : null;
+
+  return {
+    provider: "GAS TIJUANA",
+    serviceKind: "GAS",
+    accountRef: accountRef || null,
+    cadastralKey: null,
+    amount,
+    dueDate: parseDate(
+      firstMatch(text, [/Fecha\s*de\s*Vencimiento\s*[:;.]?\s*(\d{4}\s*\/\s*\d{2}\s*\/\s*\d{2})/i])?.replace(
+        /\s/g,
+        "",
+      ),
+    ),
+    period,
+    crossChecked,
+    notes,
+  };
+}
+
+// --- PREDIAL (municipal property tax) ---------------------------------------
+
+/**
+ * Normalise a printed clave catastral to the eight-character form
+ * `Property.cadastralKey` holds. The municipalities print it grouped
+ * (`KP-128-106`, `MM-B01-041`); the stored value drops the separators
+ * (`KP128106`, `MMB01041`).
+ *
+ * The shape is *not* two letters and six digits, which is the assumption that
+ * has to be resisted here. Across the 932 distinct claves on file, characters
+ * four through eight are digits without exception, but the third is a digit in
+ * 917 of them and one of `A`, `B`, `H`, `T` in the other fifteen. Running the
+ * whole tail through `toDigits` — which maps `B` to `8` — is what turned a real
+ * `MMB01041` into a nonexistent `MM801041`, so only positions four onward get
+ * that treatment and a letter in the third position is kept as printed.
+ *
+ * That leaves a genuine ambiguity at that one position: a `B` there might be a
+ * misread `8`, and 34 stored claves do carry an `8` there against six with a
+ * `B`. It is left as read rather than guessed, because a page that fails to
+ * match lands in the review queue where a human fixes it in seconds, while a
+ * page that matches the wrong property posts a charge to the wrong customer.
+ *
+ * The two-letter prefix is the other fragile part. Tesseract inserts a spurious
+ * `I` into letter pairs with some regularity — a real `MM-200-010` came back as
+ * `MIM-200-010` — so a run longer than two letters has its `I`/`L` dropped
+ * first, which recovers exactly that case. Anything still not two letters is
+ * truncated and flagged, because a wrong prefix silently matches the wrong
+ * property or, more often, nothing at all.
+ */
+export function normalizeCadastralKey(
+  raw: string,
+  notes: string[],
+): string | null {
+  const m = raw.match(/^([A-Za-z|]{2,5})[-\s]?([A-Za-z0-9|]{3})[-\s]?([0-9OIlSBD]{3})$/);
+  if (!m) return null;
+
+  let letters = m[1].toUpperCase().replace(/[^A-Z]/g, "");
+  if (letters.length > 2) {
+    const stripped = letters.replace(/[IL]/g, "");
+    if (stripped.length === 2) {
+      letters = stripped;
+    } else {
+      letters = letters.slice(0, 2);
+      notes.push(`la clave catastral se leyó como "${m[1]}"; se tomó "${letters}"`);
+    }
+  }
+  if (letters.length !== 2) return null;
+
+  const third = m[2][0].toUpperCase();
+  const tail =
+    (/[A-Z]/.test(third) ? third : toDigits(third)) +
+    toDigits(m[2].slice(1)) +
+    toDigits(m[3]);
+
+  return tail.length === 6 ? letters + tail : null;
+}
+
+/** The grouped clave as printed, anchored to its label when one survived OCR. */
+const GROUPED_CLAVE = "[A-Z|]{2,5}-[A-Z0-9OIlSBD]{3}-[0-9OIlSBD]{3}";
+
+function findCadastralKey(text: string, notes: string[]): string | null {
+  const labelled = firstMatch(text, [
+    new RegExp(`Clave\\s*Catastral\\s*[^A-Z0-9]{0,8}(${GROUPED_CLAVE})`, "i"),
+    new RegExp(`CLAVE\\s*[^A-Z0-9]{0,8}(${GROUPED_CLAVE})`, "i"),
+  ]);
+  if (labelled) return normalizeCadastralKey(labelled, notes);
+
+  // Ensenada's label ("CLAVE") lands inside a table header that OCRs into
+  // noise more often than not, so the bare grouped shape is accepted as a
+  // fallback. It is distinctive enough — two letters and two three-character
+  // groups joined by hyphens appears nowhere else on these pages.
+  const bare = firstMatch(text, [new RegExp(`\\b(${GROUPED_CLAVE})\\b`)]);
+  return bare ? normalizeCadastralKey(bare, notes) : null;
+}
+
+/**
+ * Tijuana: a "CERTIFICACIÓN DE CAJA" whose payment barcode is one 32-digit run
+ * of `account(8) + due date(DDMMYY) + amount(9) + folio(9)`, verified against
+ * all five sample pages. Municipal totals are whole pesos (the receipt itself
+ * carries a "Redondeo" line), so the barcode amount needs no decimal point.
+ *
+ * No clave catastral is printed anywhere on this layout — the 8-digit
+ * municipal account is the only identifier, and it is not a number the legacy
+ * database ever held. Until a reviewer confirms one, every Tijuana page lands
+ * in review; confirming teaches the matcher (see `learnAccountRefs`) so the
+ * same property matches itself next year.
+ */
+function parsePredialTijuana(page: OcrPage): ParsedStatement {
+  const text = page.text;
+  const notes: string[] = [];
+
+  const barcode = text.match(/(?<![0-9OIlSBD])([0-9OIlSBD]{32})(?![0-9OIlSBD])/);
+  const printedTotal = money(
+    firstMatch(text, [/TOTAL\s*A?\s*PAGAR\s*[:;.]?\s*\$?\s*([\d,]+\.?\d{0,2})/i]),
+  );
+
+  let accountRef: string | null = null;
+  let amount: number | null = printedTotal;
+  let dueDate: Date | null = null;
+  let crossChecked: boolean | null = null;
+
+  if (barcode) {
+    const run = toDigits(barcode[1]);
+    const d = run.slice(8, 14);
+    const fromBarcode = Number(run.slice(14, 23));
+    accountRef = run.slice(0, 8);
+    dueDate = parseDate(`20${d.slice(4, 6)}-${d.slice(2, 4)}-${d.slice(0, 2)}`);
+    notes.push("cuenta, importe y vencimiento leídos del código de barras");
+
+    if (printedTotal != null) {
+      // Guarding the money, not the account number: the printed total is the
+      // figure a human would key, so when the two disagree one of them is a
+      // misread peso amount and nothing should post unreviewed.
+      crossChecked = Math.abs(printedTotal - fromBarcode) < 0.5;
+      if (!crossChecked) {
+        notes.push(
+          `el total impreso (${printedTotal}) no coincide con el código de barras (${fromBarcode})`,
+        );
+      }
+    }
+    if (amount == null) amount = fromBarcode;
+  }
+
+  if (!dueDate) {
+    dueDate = parseDate(
+      firstMatch(text, [/FECHA\s*VENCE\s*[:;.]?\s*(\d{1,2}\/\w{3}\/\d{4})/i]),
+    );
+  }
+
+  return {
+    provider: "PREDIAL TIJUANA",
+    serviceKind: "PROPERTY_TAX",
+    accountRef: accountRef || null,
+    cadastralKey: null,
+    amount,
+    dueDate,
+    // The fiscal year, which is what the legacy ledger's `period` holds for
+    // predial ("2026" is its single most common value). It is read from the
+    // assessment table's year column, and failing that from the deadline: a
+    // predial bill for year N falls due on 31 January of year N.
+    period:
+      firstMatch(text, [/VALOR\s*FISCAL[\s\S]{0,160}?\b(20\d{2})\b/i]) ??
+      (dueDate ? String(dueDate.getUTCFullYear()) : null),
+    crossChecked,
+    notes,
+  };
+}
+
+/**
+ * Rosarito: a wide "CERTIFICACIÓN DE CAJA" keyed by clave catastral, with no
+ * account number of its own — the clave is the identifier, which is exactly
+ * what `Property.cadastralKey` holds, so these match on the first pass.
+ *
+ * The total is read with a negative lookbehind on "Sub": the receipt prints
+ * `Sub Total $5,409.39` (before the peso rounding) directly above
+ * `Total $5,409.00`, and taking the first "Total" on the page books 39 cents
+ * that the municipality did not charge. The lookbehind allows zero spaces
+ * because the label prints both ways — `Sub Total` on one sample and
+ * `SubTotal` on the next, and the tight one is what slipped past a fixed
+ * `Sub\s` and read $9,624.85 off a receipt for $9,625.00.
+ */
+function parsePredialRosarito(page: OcrPage): ParsedStatement {
+  const notes: string[] = [];
+  const text = page.text;
+
+  return {
+    provider: "PREDIAL ROSARITO",
+    serviceKind: "PROPERTY_TAX",
+    accountRef: null,
+    cadastralKey: findCadastralKey(text, notes),
+    amount: money(firstMatch(text, [/(?<!Sub\s{0,3})Total\s*[|:;.]?\s*\$\s*([\d,]+\.\d{2})/i])),
+    // "EXTEMPORANEO DESPUES DE: 31/01/2026" — the leading E is regularly eaten
+    // by the box rule printed over it, so the anchor starts at "XTEMPORANEO".
+    dueDate: parseDate(
+      firstMatch(text, [/XTEMPOR[AÁ]NEO\s*DESPU[EÉ]S\s*DE\s*[:;.]?\s*(\d{2}\/\d{2}\/\d{4})/i]),
+    ),
+    period: firstMatch(text, [/Periodo\s*por\s*Pagar\s*[:;.]?\s*(20\d{2})/i]),
+    crossChecked: null,
+    notes,
+  };
+}
+
+/**
+ * Ensenada: a dot-matrix "IMPRESION MAQUINA REGISTRADORA" statement, by some
+ * distance the worst-scanning of the three. Matching is by clave catastral.
+ *
+ * The amount is read positionally rather than by label, because the label does
+ * not survive: across five real pages the same word came back as `TOTAL:`,
+ * `TOTA LA A` and `orAL:`. What is stable is the row — the summary line that
+ * starts `TOTALES` carries the assessed figures across it and the amount
+ * actually paid last, at the right margin.
+ *
+ * That last figure must carry a literal `$`. On a real sample the paid total
+ * printed as `TOTAL: A $2,203.00` and OCR'd as `TOTAL: A 82,203.00` — the
+ * dollar sign read as an 8, a mistake that would post a $2,203 charge as
+ * $82,203 and look entirely ordinary in the ledger. Requiring the `$` costs
+ * that page its amount and sends it to review, which is the only acceptable
+ * failure here. The unprefixed figures earlier on the row are deliberately not
+ * a fallback: they are the tax assessed before the early-payment discount, not
+ * what was paid.
+ */
+function parsePredialEnsenada(page: OcrPage): ParsedStatement {
+  const notes: string[] = [];
+  const text = page.text;
+
+  const totalsRow = text.split("\n").find((l) => /TOTALES/i.test(l)) ?? "";
+  const figures = allMatches(totalsRow, /\$\s*(\d[\d,.\s]*\.\d{2})/);
+  const amount = figures.length ? money(figures[figures.length - 1]) : null;
+  if (amount == null) {
+    notes.push("no se pudo leer el importe con certeza; capturarlo a mano");
+  }
+
+  return {
+    provider: "PREDIAL ENSENADA",
+    serviceKind: "PROPERTY_TAX",
+    accountRef: null,
+    cadastralKey: findCadastralKey(text, notes),
+    amount,
+    // This layout prints no payment deadline at all — it is a receipt for a
+    // payment already made at the municipal window.
+    dueDate: null,
+    period: firstMatch(text, [/A[ÑN]O\s*[\s\S]{0,60}?\b(20\d{2})\b/i]),
+    crossChecked: null,
+    notes,
+  };
+}
+
+// --- ZONA FEDERAL (ZOFEMAT, Tijuana) ----------------------------------------
+
+/**
+ * Normalise the concession clave the zona federal receipt is keyed by.
+ *
+ * It is printed grouped and loosely spaced — `12-T -012`, `14-A-119`,
+ * `14-K -031` — and is a different shape from the cadastral key entirely: two
+ * digits, one letter, three digits. The letter is kept as printed rather than
+ * digitised, for the same reason `normalizeCadastralKey` keeps its third
+ * character: `toDigits` maps `B` to `8` and `D` to `0`, and a real `14-D -014`
+ * run through it becomes `140014`, which is not a clave at all.
+ *
+ * Stored without separators, because nothing on file holds this value yet (see
+ * `parseZonaFederal`) so the canonical form is ours to pick, and a bare run
+ * cannot be broken by the hyphen the scan renders as a dash, a minus or
+ * nothing.
+ */
+export function normalizeZofematKey(raw: string): string | null {
+  const m = raw.match(/^([0-9OIlSBD]{2})\s*-\s*([A-Za-z])\s*-?\s*([0-9OIlSBD]{3})$/);
+  if (!m) return null;
+  const zone = toDigits(m[1]);
+  const lot = toDigits(m[3]);
+  if (zone.length !== 2 || lot.length !== 3) return null;
+  return `${zone}${m[2].toUpperCase()}${lot}`;
+}
+
+/**
+ * The bimester the receipt prints as `2026-2 / 2026-2`, rendered in the
+ * vocabulary the ledger already speaks.
+ *
+ * All 258 legacy FEDERAL ZONE transactions carry a period of `JAN/FEB`,
+ * `MAR/APR`, `MAY/JUN` or `NOV/DEC`, and their payment dates confirm the
+ * ordering — JAN/FEB was paid in March, MAR/APR in May, MAY/JUN in July,
+ * NOV/DEC in January, i.e. always the month after the bimester closes. The
+ * receipts agree: the two `2026-3` samples fall due 17/07/2026 with no
+ * surcharge, which is bimester three, May and June. Writing `2026-3` instead
+ * would leave the OCR-posted rows unsearchable alongside every hand-keyed one.
+ */
+const BIMESTERS = ["JAN/FEB", "MAR/APR", "MAY/JUN", "JUL/AUG", "SEP/OCT", "NOV/DEC"];
+
+/**
+ * Tijuana's "Zona Federal Marítimo Terrestre" — the federal maritime-zone
+ * occupancy fee, billed by the municipality for beachfront lots.
+ *
+ * Nothing on file identifies these. `PropertyService.accountNumber` for
+ * FEDERAL_ZONE holds DATMEX.zfed, which is not a reference at all but an
+ * amount: its 77 values include `246.06`, `2369.09`, `22653.94` and a negative
+ * `-1679`, and the concession claves these receipts are keyed by appear nowhere
+ * in the database. So the clave goes to `meterNumber` (see `scopedRefField`),
+ * every page starts cold, and the first confirm teaches the match — the same
+ * arrangement Tijuana predial needed, for the same reason.
+ *
+ * The amount is taken from the SubTotal row rather than the "Total a pagar"
+ * box, which is printed on a grey fill and OCR'd on only 1 of 8 sample pages
+ * while the SubTotal row read on 8 of 8. The two differ by design: the
+ * municipality rounds to whole pesos and prints the difference on its own
+ * "Ajuste Ley Hacienda Mpal" line — `-$0.05` against a 591.05 subtotal, `$0.21`
+ * against 2,872.79 — so the payable figure is the rounded subtotal, and where
+ * the printed box did read, it agreed.
+ */
+function parseZonaFederal(page: OcrPage): ParsedStatement {
+  const text = page.text;
+  const notes: string[] = [];
+
+  // Printed twice, once on the receipt and once on the stub below it, which is
+  // a free second reading: on one sample the heading was struck through by the
+  // office's own highlighter and only the stub survived.
+  const claves = [
+    ...new Set(
+      allMatches(text, /Clave\s*[:;.]?\s*([0-9OIlSBD]{2}\s*-\s*[A-Za-z]\s*-?\s*[0-9OIlSBD]{3})/i)
+        .map(normalizeZofematKey)
+        .filter((k): k is string => k != null),
+    ),
+  ];
+
+  const accountRef = claves[0] ?? null;
+  let crossChecked: boolean | null = null;
+
+  const subtotalRow = text.split("\n").find((l) => /SubTotal/i.test(l)) ?? "";
+  const figures = allMatches(subtotalRow, /(\d[\d,]*\.\d{2})/);
+  // Impuesto, Actualización, Recargo, Multa, Importe — the payable one is last.
+  const importe = figures.length ? money(figures[figures.length - 1]) : null;
+  const rounded = importe != null ? Math.round(importe) : null;
+  const printed = money(
+    firstMatch(text, [/Total\s*a\s*pagar\s*[:;.]?\s*\$?\s*([\d,]+\.\d{2})/i]),
+  );
+
+  if (printed != null && rounded != null) {
+    crossChecked = Math.abs(printed - rounded) < 0.5;
+    if (!crossChecked) {
+      notes.push(
+        `el total impreso (${printed}) no coincide con el subtotal redondeado (${rounded})`,
+      );
+    }
+  } else if (rounded != null) {
+    notes.push("importe tomado del subtotal, redondeado al peso");
+  } else if (printed == null) {
+    notes.push("no se pudo leer el importe con certeza; capturarlo a mano");
+  }
+
+  // A clave read two different ways means one of the two readings is wrong and
+  // there is no third to break the tie, so the page goes to a human even if the
+  // money cross-checked.
+  if (claves.length > 1) {
+    crossChecked = false;
+    notes.push(`la clave se leyó de ${claves.length} formas distintas (${claves.join(", ")})`);
+  }
+  if (!accountRef) notes.push("no se pudo leer la clave de la concesión");
+
+  const bimester = text.match(/\b(20\d{2})\s*-\s*([1-6])\s*\/\s*20\d{2}\s*-\s*[1-6]/);
+
+  return {
+    provider: "ZONA FEDERAL TIJUANA",
+    serviceKind: "FEDERAL_ZONE",
+    accountRef,
+    cadastralKey: null,
+    amount: printed ?? rounded,
+    dueDate: parseDate(
+      firstMatch(text, [/Vencimiento\s*[:;.]?\s*(\d{2}\/\d{2}\/\d{4})/i]),
+    ),
+    period: bimester ? BIMESTERS[+bimester[2] - 1] : null,
+    crossChecked,
+    notes,
+  };
+}
+
+const PARSERS: Record<string, (page: OcrPage) => ParsedStatement> = {
+  CFE: parseCfe,
+  CESPT: parseCespt,
+  TELNOR: parseTelnor,
+  "GAS TIJUANA": parseGas,
+  "PREDIAL TIJUANA": parsePredialTijuana,
+  "PREDIAL ROSARITO": parsePredialRosarito,
+  "PREDIAL ENSENADA": parsePredialEnsenada,
+  "ZONA FEDERAL TIJUANA": parseZonaFederal,
+};
+
+const EMPTY: ParsedStatement = {
+  provider: null,
+  serviceKind: null,
+  accountRef: null,
+  cadastralKey: null,
+  amount: null,
+  dueDate: null,
+  period: null,
+  crossChecked: null,
+  notes: [],
+};
+
+/** Detect the provider and run its parser. */
+export function parseStatement(page: OcrPage): ParsedStatement {
+  const provider = detectProvider(page.text);
+  if (!provider) return { ...EMPTY, notes: ["no se reconoció el proveedor"] };
+  return PARSERS[provider](page);
+}

apps/api/src/statements/statement-matcher.service.ts

@@ -0,0 +1,236 @@
+import { Injectable } from "@nestjs/common";
+import type { ServiceKind } from "@jorgecuadros/database";
+import { PrismaService } from "../prisma/prisma.service";
+import type { ParsedStatement } from "./parsers/statement-parser";
+
+export interface MatchResult {
+  propertyServiceId: string | null;
+  customerId: string | null;
+  /** Why it landed here — shown in the review queue verbatim. */
+  note: string;
+  /** True only for an unambiguous hit on the scoped field. */
+  confident: boolean;
+  /** Populated when more than one service claims the same number. */
+  candidates: { propertyServiceId: string; customerId: string; customerName: string }[];
+}
+
+/**
+ * Resolves a parsed statement to the customer who should be billed for it.
+ *
+ * Two rules govern everything here.
+ *
+ * **Match on one scoped field, never fuzzily across all identifiers.** Each
+ * service kind has exactly one column its statements print, and only that
+ * column is consulted. A blanket search over accountNumber/meterNumber/route
+ * would let a water account number collide with an unrelated phone number, and
+ * the resulting mis-post would look perfectly ordinary in the ledger.
+ *
+ * **Never match on the customer name.** The name on a utility bill is the
+ * account's registrant, which drifts from the current owner and is often years
+ * stale — one sample CESPT receipt is printed to "ARNAIZ ROSAS ELSA AURORA"
+ * for an account this office holds under "CATT, RANDY", who is not the same
+ * person. Names are displayed for the reviewer to sanity-check, and are never
+ * an input to matching.
+ */
+/**
+ * Which `PropertyService` column a given kind's statements actually print.
+ *
+ * Exported because the same answer governs three places that must agree: the
+ * lookup here, the blank-service fill on review, and the write-back on confirm.
+ * When they disagree, a reference gets learned into a column nothing searches,
+ * and the same page returns to the review queue every month forever.
+ *
+ * `meterNumber` is doing double duty for the three kinds whose printed
+ * reference DATMEX never held in `accountNumber`:
+ *   - GAS, where the number lived in free-text notes,
+ *   - PROPERTY_TAX, where `accountNumber` holds DATMEX.predial — a 3-4 digit
+ *     office file number that is neither unique nor printed on any statement.
+ *     The Tijuana municipal receipt prints an 8-digit account and no clave
+ *     catastral at all, so it needs a column of its own; overwriting the legacy
+ *     predial numbers to make room would destroy the only link back to the
+ *     original records, and
+ *   - FEDERAL_ZONE, where `accountNumber` holds DATMEX.zfed, which is not a
+ *     reference of any kind but a peso amount: 3 of its 77 values carry cents
+ *     (`246.06`, `2369.09`, `22653.94`) and one is negative. Searching it for
+ *     the concession clave the receipt prints would never hit, and — worse —
+ *     because every row already has a value, the `[field]: null` guards in
+ *     `learnAccountRefs` and the blank-service fill would never fire either, so
+ *     the same page would return to the review queue every bimester forever.
+ */
+export function scopedRefField(
+  kind: ServiceKind,
+): "accountNumber" | "meterNumber" | null {
+  switch (kind) {
+    case "ELECTRIC": // CFE "NO. DE SERVICIO" -> DATMEX.rpu
+    case "WATER": // CESPT "Cuenta" / "No. DE CUENTA" -> DATMEX.agua
+    case "TELEPHONE": // Telnor "Teléfono" (LADA stripped) -> DATMEX.telefono
+    case "CABLE":
+      return "accountNumber";
+    case "GAS": // bajagas "Cuenta" -> recovered from notes into meterNumber
+    case "PROPERTY_TAX": // Tijuana's 8-digit municipal account
+    case "FEDERAL_ZONE": // ZOFEMAT concession clave, e.g. `12T012`
+      return "meterNumber";
+    default:
+      return null;
+  }
+}
+
+@Injectable()
+export class StatementMatcherService {
+  constructor(private readonly prisma: PrismaService) {}
+
+  async match(parsed: ParsedStatement, expectedKind: ServiceKind): Promise<MatchResult> {
+    const kind = parsed.serviceKind ?? expectedKind;
+
+    // The uploader labels a batch with one service kind. If the parser reads a
+    // page as a different provider, that is a mis-sorted page, not a match —
+    // posting it would book a phone bill as a water charge.
+    if (parsed.serviceKind && parsed.serviceKind !== expectedKind) {
+      return this.unmatched(
+        `la página parece de ${parsed.provider} (${parsed.serviceKind}) pero el lote es de ${expectedKind}`,
+      );
+    }
+
+    const field = scopedRefField(kind);
+
+    if (field && parsed.accountRef) {
+      const hit = await this.byServiceField(kind, field, parsed.accountRef);
+      if (hit) return hit;
+    }
+
+    // The clave catastral is printed on CESPT bills as well as predial ones, so
+    // it rescues a page whose account number did not OCR — which happened on
+    // real samples, where the clave read cleanly and the account number did
+    // not. On the Rosarito and Ensenada predial layouts it is not a rescue at
+    // all but the only identifier the receipt carries, so a unique hit there is
+    // as good as any account-number match and is treated as one.
+    if (parsed.cadastralKey) {
+      const primary = kind === "PROPERTY_TAX" && !parsed.accountRef;
+      const hit = await this.byCadastralKey(kind, parsed.cadastralKey, primary);
+      if (hit) return hit;
+    }
+
+    if (!field && !parsed.cadastralKey) {
+      return this.unmatched(`no hay campo de búsqueda definido para ${kind}`);
+    }
+    if (!parsed.accountRef && !parsed.cadastralKey) {
+      return this.unmatched(
+        kind === "PROPERTY_TAX"
+          ? "no se leyó ni la clave catastral ni la cuenta municipal"
+          : "no se pudo leer la referencia de la cuenta",
+      );
+    }
+    return this.unmatched(
+      parsed.accountRef
+        ? `no se encontró ningún servicio de ${kind} con la referencia ${parsed.accountRef}`
+        : `no se encontró ninguna propiedad con la clave catastral ${parsed.cadastralKey}`,
+    );
+  }
+
+  private async byServiceField(
+    kind: ServiceKind,
+    field: "accountNumber" | "meterNumber",
+    ref: string,
+  ): Promise<MatchResult | null> {
+    const rows = await this.prisma.propertyService.findMany({
+      where: { kind, [field]: ref },
+      select: {
+        id: true,
+        property: {
+          select: { customerId: true, customer: { select: { name: true } } },
+        },
+      },
+    });
+    if (rows.length === 0) return null;
+
+    const candidates = rows.map((r) => ({
+      propertyServiceId: r.id,
+      customerId: r.property.customerId,
+      customerName: r.property.customer.name,
+    }));
+
+    // Duplicate account numbers do occur in the legacy data (the office's own
+    // DUPLICADOS report existed for a reason), so every candidate is surfaced
+    // for the reviewer to choose rather than one being picked arbitrarily.
+    if (rows.length > 1) {
+      return {
+        propertyServiceId: null,
+        customerId: null,
+        note: `${rows.length} servicios comparten la referencia ${ref}`,
+        confident: false,
+        candidates,
+      };
+    }
+
+    return {
+      propertyServiceId: candidates[0].propertyServiceId,
+      customerId: candidates[0].customerId,
+      note: `coincidencia exacta por ${field === "accountNumber" ? "número de cuenta" : "medidor"} ${ref}`,
+      confident: true,
+      candidates,
+    };
+  }
+
+  private async byCadastralKey(
+    kind: ServiceKind,
+    key: string,
+    /** True when the clave is the identifier the statement was issued against. */
+    primary: boolean,
+  ): Promise<MatchResult | null> {
+    const props = await this.prisma.property.findMany({
+      where: { cadastralKey: key },
+      select: {
+        customerId: true,
+        customer: { select: { name: true } },
+        services: { where: { kind }, select: { id: true } },
+      },
+    });
+    if (props.length === 0) return null;
+
+    const candidates = props.flatMap((p) =>
+      (p.services.length ? p.services.map((s) => s.id) : [null]).map((sid) => ({
+        propertyServiceId: sid as string,
+        customerId: p.customerId,
+        customerName: p.customer.name,
+      })),
+    );
+
+    if (candidates.length > 1) {
+      return {
+        propertyServiceId: null,
+        customerId: null,
+        note: `${candidates.length} propiedades comparten la clave catastral ${key}`,
+        confident: false,
+        candidates,
+      };
+    }
+
+    // When the clave is the *secondary* key — a utility bill that also happens
+    // to print it — the page is left for review, because the clave was not the
+    // number the statement was issued against and confirming is what teaches
+    // the matcher the account number for next month. When it is the primary key
+    // (Rosarito and Ensenada predial, which print nothing else), a unique hit
+    // is a real match and there is no second number to learn.
+    return {
+      propertyServiceId: candidates[0].propertyServiceId ?? null,
+      customerId: candidates[0].customerId,
+      note: primary
+        ? `coincidencia exacta por clave catastral ${key}`
+        : `identificado por clave catastral ${key}; confirme para registrar también el número de cuenta`,
+      // A clave with no service row of the right kind behind it still needs a
+      // human: there is nothing to attach the posting to.
+      confident: primary && candidates[0].propertyServiceId != null,
+      candidates,
+    };
+  }
+
+  private unmatched(note: string): MatchResult {
+    return {
+      propertyServiceId: null,
+      customerId: null,
+      note,
+      confident: false,
+      candidates: [],
+    };
+  }
+}

apps/api/src/statements/statement.dto.ts

@@ -0,0 +1,52 @@
+import {
+  IsBoolean,
+  IsEnum,
+  IsInt,
+  IsNumber,
+  IsOptional,
+  IsString,
+  MinLength,
+} from "class-validator";
+import { Currency, ServiceKind, StatementDocumentStatus } from "@jorgecuadros/database";
+
+export class CreateStatementBatchDto {
+  @IsEnum(ServiceKind) serviceKind!: ServiceKind;
+  @IsOptional() @IsString() label?: string;
+}
+
+/** Staff correction of one document's extracted fields or its match. */
+export class ReviewDocumentDto {
+  @IsOptional() @IsString() accountRef?: string;
+  @IsOptional() @IsNumber() amount?: number;
+  @IsOptional() @IsString() period?: string;
+  @IsOptional() @IsString() dueDate?: string;
+  @IsOptional() @IsString() matchedPropertyServiceId?: string;
+  @IsOptional() @IsString() matchedCustomerId?: string;
+  // Restricted to the review-reachable states: a client cannot declare a
+  // document POSTED, because only a successful ledger write may do that.
+  @IsOptional()
+  @IsEnum(StatementDocumentStatus)
+  status?: Extract<StatementDocumentStatus, "MATCHED" | "NEEDS_REVIEW" | "CONFIRMED">;
+}
+
+/**
+ * Post a batch's confirmed documents. The check-level fields are shared by
+ * every line, exactly as on the manual batch-capture screen — an OCR batch is
+ * still "these receipts, paid by this check".
+ */
+export class ConfirmBatchDto {
+  @IsString() @MinLength(1) checkNumber!: string;
+  @IsString() @MinLength(1) transactionDate!: string;
+  @IsOptional() @IsEnum(Currency) currency?: Currency;
+  /** Overrides the concept derived from the batch's service kind. */
+  @IsOptional() @IsString() typeId?: string;
+  /** Post as outstanding (sin fondos) — captured but not yet funded. */
+  @IsOptional() @IsBoolean() outstanding?: boolean;
+  /** Also post documents a reviewer explicitly marked CONFIRMED. */
+  @IsOptional() @IsBoolean() includeReviewed?: boolean;
+}
+
+export class ListBatchesQuery {
+  @IsOptional() @IsInt() page?: number;
+  @IsOptional() @IsInt() pageSize?: number;
+}

apps/api/src/statements/statements.controller.ts

@@ -0,0 +1,162 @@
+import {
+  Body,
+  Controller,
+  Get,
+  Param,
+  Patch,
+  Post,
+  Query,
+  Req,
+  Res,
+  StreamableFile,
+  UploadedFiles,
+  UseGuards,
+  UseInterceptors,
+} from "@nestjs/common";
+import { FilesInterceptor } from "@nestjs/platform-express";
+import type { ServiceKind, StatementDocumentStatus } from "@jorgecuadros/database";
+import type { Request, Response } from "express";
+import { AuthenticatedGuard } from "../auth/authenticated.guard";
+import { AbilityGuard } from "../auth/ability.guard";
+import { RequireAbility } from "../auth/require-ability.decorator";
+import { AuditService } from "../common/audit.service";
+import type { UploadedFileLike } from "../storage/upload-file";
+import { StatementsService } from "./statements.service";
+import { ConfirmBatchDto, ReviewDocumentDto } from "./statement.dto";
+
+/**
+ * Statement OCR intake (RECEIPT_CAPTURE_SPEC §2).
+ *
+ * Nothing here writes to the ledger directly — confirming a batch delegates to
+ * BillingService, so an OCR-captured charge is indistinguishable from a
+ * hand-keyed one except for its `captureSource`.
+ */
+@Controller("statements")
+@UseGuards(AuthenticatedGuard, AbilityGuard)
+export class StatementsController {
+  constructor(
+    private readonly statements: StatementsService,
+    private readonly audit: AuditService,
+  ) {}
+
+  private actingId(req: Request): string {
+    return (req.user as { id: string } | undefined)?.id ?? "";
+  }
+
+  /**
+   * Whether this deployment can ingest scans at all — the UI hides automatic
+   * capture without it. Both halves are needed: OCR to read the page, object
+   * storage to keep it.
+   */
+  @Get("status")
+  async status() {
+    return {
+      ocrAvailable: await this.statements.ocrAvailable(),
+      storageAvailable: this.statements.storageAvailable(),
+    };
+  }
+
+  @Get("batches")
+  listBatches(@Query("page") page?: string, @Query("pageSize") pageSize?: string) {
+    return this.statements.listBatches(
+      Math.max(1, Number(page) || 1),
+      Math.min(100, Math.max(1, Number(pageSize) || 25)),
+    );
+  }
+
+  @Get("batches/:id")
+  getBatch(@Param("id") id: string) {
+    return this.statements.getBatch(id);
+  }
+
+  @Get("batches/:id/documents")
+  listDocuments(@Param("id") id: string, @Query("status") status?: string) {
+    return this.statements.listDocuments(
+      id,
+      (status || undefined) as StatementDocumentStatus | undefined,
+    );
+  }
+
+  /** The rendered page, so a reviewer can compare it against what was read. */
+  @Get("documents/:id/page")
+  async pageImage(@Param("id") id: string, @Res({ passthrough: true }) res: Response) {
+    const { stream, contentType, contentLength } = await this.statements.pageImage(id);
+    res.set({
+      "Content-Type": contentType ?? "image/png",
+      ...(contentLength ? { "Content-Length": String(contentLength) } : {}),
+    });
+    return new StreamableFile(stream);
+  }
+
+  // --- writes ---------------------------------------------------------------
+
+  @Post("batches")
+  @RequireAbility("statement:ingest")
+  @UseInterceptors(
+    // A month of one company's statements is a handful of multi-page scans;
+    // 25 files at 50MB covers that with room to spare.
+    FilesInterceptor("files", 25, { limits: { fileSize: 50 * 1024 * 1024 } }),
+  )
+  async createBatch(
+    @UploadedFiles() files: UploadedFileLike[] | undefined,
+    @Query("serviceKind") serviceKind: ServiceKind,
+    @Query("label") label: string | undefined,
+    @Req() req: Request,
+  ) {
+    const batch = await this.statements.createBatch(
+      files ?? [],
+      serviceKind,
+      this.actingId(req),
+      label,
+    );
+    void this.audit.log(this.actingId(req), "statement.batch.create", {
+      batchId: batch.id,
+      serviceKind,
+      fileCount: batch.fileCount,
+    });
+    return batch;
+  }
+
+  @Patch("documents/:id")
+  @RequireAbility("statement:review")
+  async review(
+    @Param("id") id: string,
+    @Body() dto: ReviewDocumentDto,
+    @Req() req: Request,
+  ) {
+    const doc = await this.statements.review(id, dto, this.actingId(req));
+    void this.audit.log(this.actingId(req), "statement.document.review", {
+      documentId: id,
+      status: doc.status,
+    });
+    return doc;
+  }
+
+  @Post("documents/:id/reject")
+  @RequireAbility("statement:review")
+  async reject(@Param("id") id: string, @Req() req: Request) {
+    const doc = await this.statements.reject(id, this.actingId(req));
+    void this.audit.log(this.actingId(req), "statement.document.reject", {
+      documentId: id,
+    });
+    return doc;
+  }
+
+  /** Post every matched document in the batch, against one check. */
+  @Post("batches/:id/confirm")
+  @RequireAbility("statement:review")
+  async confirm(
+    @Param("id") id: string,
+    @Body() dto: ConfirmBatchDto,
+    @Req() req: Request,
+  ) {
+    const result = await this.statements.confirmBatch(id, dto, this.actingId(req));
+    void this.audit.log(this.actingId(req), "statement.batch.confirm", {
+      batchId: id,
+      posted: result.posted,
+      total: result.total,
+      checkNumber: dto.checkNumber,
+    });
+    return result;
+  }
+}

apps/api/src/statements/statements.module.ts

@@ -0,0 +1,18 @@
+import { Module } from "@nestjs/common";
+import { BillingModule } from "../billing/billing.module";
+import { OcrModule } from "../ocr/ocr.module";
+import { StatementsController } from "./statements.controller";
+import { StatementsService } from "./statements.service";
+import { StatementMatcherService } from "./statement-matcher.service";
+
+/**
+ * The concrete OCR engine is bound in OcrModule (see apps/api/src/ocr/) —
+ * everything downstream depends on the OcrProvider interface, so swapping
+ * Tesseract for a managed extraction API is a one-line change there.
+ */
+@Module({
+  imports: [BillingModule, OcrModule],
+  controllers: [StatementsController],
+  providers: [StatementsService, StatementMatcherService],
+})
+export class StatementsModule {}

apps/api/src/statements/statements.service.ts

@@ -0,0 +1,480 @@
+import {
+  BadRequestException,
+  Inject,
+  Injectable,
+  Logger,
+  NotFoundException,
+} from "@nestjs/common";
+import {
+  Prisma,
+  type ServiceKind,
+  type StatementDocumentStatus,
+} from "@jorgecuadros/database";
+import { PrismaService } from "../prisma/prisma.service";
+import { StorageService } from "../storage/storage.service";
+import { BillingService } from "../billing/billing.service";
+import type { UploadedFileLike } from "../storage/upload-file";
+import { OCR_PROVIDER, type OcrProvider } from "./ocr/ocr.provider";
+import { parseStatement } from "./parsers/statement-parser";
+import { StatementMatcherService, scopedRefField } from "./statement-matcher.service";
+import type { ConfirmBatchDto, ReviewDocumentDto } from "./statement.dto";
+
+/**
+ * Default ledger concept per service kind. The names are the legacy
+ * `TYPE OF TRX` values already in `type_transactions`, resolved by name once
+ * per confirm rather than hard-coded as ids, which differ per environment.
+ */
+const CONCEPT_BY_KIND: Partial<Record<ServiceKind, string>> = {
+  ELECTRIC: "ELECTRIC",
+  WATER: "WATER",
+  TELEPHONE: "TELEPHONE",
+  GAS: "GAS BUTANO",
+  PROPERTY_TAX: "PROPERTY TAXES",
+  FEDERAL_ZONE: "FEDERAL ZONE",
+  CABLE: "CABLE",
+};
+
+/** Statuses a document can still be worked on from. */
+const OPEN: StatementDocumentStatus[] = ["NEEDS_REVIEW", "MATCHED", "CONFIRMED"];
+
+@Injectable()
+export class StatementsService {
+  private readonly logger = new Logger(StatementsService.name);
+
+  constructor(
+    private readonly prisma: PrismaService,
+    private readonly storage: StorageService,
+    private readonly billing: BillingService,
+    private readonly matcher: StatementMatcherService,
+    @Inject(OCR_PROVIDER) private readonly ocr: OcrProvider,
+  ) {}
+
+  ocrAvailable(): Promise<boolean> {
+    return this.ocr.available();
+  }
+
+  /** Scans are stored as blobs, so no object storage means no intake. */
+  storageAvailable(): boolean {
+    return this.storage.available;
+  }
+
+  // --- ingest ---------------------------------------------------------------
+
+  /**
+   * Accept a batch of scanned PDFs and start processing.
+   *
+   * Processing is kicked off but deliberately not awaited: 300 pages of OCR is
+   * minutes of CPU, far past any sane HTTP timeout. The caller gets the batch
+   * id immediately and polls its status, which is also what lets the review
+   * queue show partial progress.
+   */
+  async createBatch(
+    files: UploadedFileLike[],
+    serviceKind: ServiceKind,
+    uploadedById: string,
+    label?: string,
+  ) {
+    if (!files?.length) throw new BadRequestException("No se recibió ningún archivo.");
+    if (!(await this.ocr.available())) {
+      throw new BadRequestException(
+        "El servidor no tiene OCR instalado; no se pueden procesar recibos.",
+      );
+    }
+    // Checked here rather than at the first `put`, which would only surface as
+    // a FAILED batch minutes later.
+    if (!this.storage.available) {
+      throw new BadRequestException(
+        "El almacenamiento de documentos no está configurado; no se pueden " +
+          "guardar los recibos escaneados.",
+      );
+    }
+
+    const batch = await this.prisma.statementBatch.create({
+      data: { serviceKind, uploadedById, label, fileCount: files.length },
+    });
+
+    // Buffers are held for the background pass; the request's own copies would
+    // otherwise be garbage once the response is sent.
+    const copies = files.map((f) => ({ buffer: f.buffer, name: f.originalname }));
+    void this.process(batch.id, copies, serviceKind).catch(async (err) => {
+      this.logger.error(`Batch ${batch.id} failed: ${(err as Error).message}`);
+      await this.prisma.statementBatch.update({
+        where: { id: batch.id },
+        data: { status: "FAILED", error: (err as Error).message },
+      });
+    });
+
+    return batch;
+  }
+
+  /** Render → OCR → parse → match, one document row per page. */
+  private async process(
+    batchId: string,
+    files: { buffer: Buffer; name?: string }[],
+    serviceKind: ServiceKind,
+  ) {
+    await this.prisma.statementBatch.update({
+      where: { id: batchId },
+      data: { status: "PROCESSING" },
+    });
+
+    let pageNumber = 0;
+    for (const file of files) {
+      // The source PDF is kept as well as the page images: it is the artifact
+      // the office actually received, and the only way to re-run a corrected
+      // parser over the original later.
+      const sourceKey = `statement/${batchId}/source-${pageNumber + 1}.pdf`;
+      await this.storage.put(sourceKey, file.buffer, "application/pdf");
+
+      const pages = await this.ocr.renderPages(file.buffer);
+      // Page images are still rendered and stored for every file, text layer or
+      // not: the review screen shows the reviewer the page, and "what the
+      // parser read" is only checkable against a picture of the paper.
+      const textLayer = await this.ocr.textPages(file.buffer).catch(() => []);
+
+      for (const [index, image] of pages.entries()) {
+        pageNumber += 1;
+        const storageKey = `statement/${batchId}/page-${pageNumber}.png`;
+        await this.storage.put(storageKey, image, "image/png");
+
+        try {
+          const embedded = textLayer[index] ?? null;
+          const ocr = embedded ?? (await this.ocr.recognize(image));
+          const parsed = parseStatement(ocr);
+          if (embedded) {
+            parsed.notes.unshift("texto leído del PDF original, sin OCR");
+          }
+          const match = await this.matcher.match(parsed, serviceKind);
+
+          const notes = [...parsed.notes, match.note].filter(Boolean);
+          // A confident field match is only trusted when nothing contradicts
+          // it: a barcode that disagrees with the printed number means one of
+          // the two was misread, and which one is a judgement call.
+          const trusted = match.confident && parsed.crossChecked !== false;
+
+          await this.prisma.statementDocument.create({
+            data: {
+              batchId,
+              pageNumber,
+              storageKey,
+              status: trusted ? "MATCHED" : "NEEDS_REVIEW",
+              ocrRawText: ocr.text,
+              ocrConfidence: new Prisma.Decimal(ocr.confidence.toFixed(3)),
+              provider: parsed.provider,
+              extractedAccountRef: parsed.accountRef,
+              extractedAmount:
+                parsed.amount != null ? new Prisma.Decimal(parsed.amount) : null,
+              extractedPeriod: parsed.period,
+              extractedDueDate: parsed.dueDate,
+              extractedCadastralKey: parsed.cadastralKey,
+              matchedPropertyServiceId: match.propertyServiceId,
+              matchedCustomerId: match.customerId,
+              matchNote: notes.join("; ").slice(0, 190),
+            },
+          });
+        } catch (err) {
+          // One unreadable page must not abandon the other 299.
+          await this.prisma.statementDocument.create({
+            data: {
+              batchId,
+              pageNumber,
+              storageKey,
+              status: "OCR_FAILED",
+              matchNote: (err as Error).message.slice(0, 190),
+            },
+          });
+        }
+      }
+    }
+
+    await this.prisma.statementBatch.update({
+      where: { id: batchId },
+      data: { status: "READY_FOR_REVIEW" },
+    });
+  }
+
+  // --- reads ----------------------------------------------------------------
+
+  async listBatches(page: number, pageSize: number) {
+    const [total, items] = await this.prisma.$transaction([
+      this.prisma.statementBatch.count(),
+      this.prisma.statementBatch.findMany({
+        orderBy: { createdAt: "desc" },
+        skip: (page - 1) * pageSize,
+        take: pageSize,
+        include: {
+          uploadedBy: { select: { name: true } },
+          _count: { select: { documents: true } },
+        },
+      }),
+    ]);
+    return { items, total, page, pageSize, pageCount: Math.ceil(total / pageSize) };
+  }
+
+  async getBatch(id: string) {
+    const batch = await this.prisma.statementBatch.findUnique({
+      where: { id },
+      include: { uploadedBy: { select: { name: true } } },
+    });
+    if (!batch) throw new NotFoundException("Lote no encontrado.");
+
+    const counts = await this.prisma.statementDocument.groupBy({
+      by: ["status"],
+      where: { batchId: id },
+      _count: { _all: true },
+    });
+    const totals = await this.prisma.statementDocument.aggregate({
+      where: { batchId: id, status: { in: OPEN } },
+      _sum: { extractedAmount: true },
+    });
+
+    return {
+      ...batch,
+      byStatus: Object.fromEntries(counts.map((c) => [c.status, c._count._all])),
+      pendingTotal: totals._sum.extractedAmount?.toFixed(2) ?? "0.00",
+    };
+  }
+
+  async listDocuments(batchId: string, status?: StatementDocumentStatus) {
+    return this.prisma.statementDocument.findMany({
+      where: { batchId, ...(status ? { status } : {}) },
+      orderBy: { pageNumber: "asc" },
+      include: {
+        matchedCustomer: { select: { id: true, name: true } },
+        matchedPropertyService: {
+          select: {
+            id: true,
+            kind: true,
+            accountNumber: true,
+            meterNumber: true,
+            property: { select: { id: true, addressLine1: true } },
+          },
+        },
+      },
+    });
+  }
+
+  /** The rendered page image, so a reviewer can read what the parser read. */
+  async pageImage(documentId: string) {
+    const doc = await this.prisma.statementDocument.findUnique({
+      where: { id: documentId },
+      select: { storageKey: true },
+    });
+    if (!doc) throw new NotFoundException("Documento no encontrado.");
+    return this.storage.getStream(doc.storageKey);
+  }
+
+  // --- review ---------------------------------------------------------------
+
+  /** Staff correction of an extracted field or of the match itself. */
+  async review(id: string, dto: ReviewDocumentDto, reviewedById: string) {
+    const doc = await this.prisma.statementDocument.findUnique({ where: { id } });
+    if (!doc) throw new NotFoundException("Documento no encontrado.");
+    if (doc.status === "POSTED") {
+      throw new BadRequestException("Este documento ya fue registrado.");
+    }
+
+    // Changing the service implies its owner; deriving the customer here rather
+    // than trusting a client-supplied pair is what stops a page being posted to
+    // one customer's ledger against another customer's service.
+    let matchedCustomerId = doc.matchedCustomerId;
+    let matchedPropertyServiceId = dto.matchedPropertyServiceId ?? undefined;
+
+    if (dto.matchedPropertyServiceId) {
+      const svc = await this.prisma.propertyService.findUnique({
+        where: { id: dto.matchedPropertyServiceId },
+        select: { property: { select: { customerId: true } } },
+      });
+      if (!svc) throw new BadRequestException("Servicio no encontrado.");
+      matchedCustomerId = svc.property.customerId;
+    } else if (dto.matchedCustomerId) {
+      matchedCustomerId = dto.matchedCustomerId;
+
+      // A reviewer picks a *customer*, not one of their service rows. Without
+      // a service the posting still works, but the confirmed reference has
+      // nowhere to be written back, so the same account would land in review
+      // again next month — which is exactly the behaviour that is supposed to
+      // make gas (whose numbers the migration never populated) a one-time cost.
+      // So: if the batch's service kind resolves to exactly one of that
+      // customer's services that has no reference yet, attach it. Exactly one
+      // — with two candidates there is no way to tell which meter or line the
+      // bill belongs to, and guessing would write a real number onto the wrong
+      // service.
+      const batch = await this.prisma.statementBatch.findUnique({
+        where: { id: doc.batchId },
+        select: { serviceKind: true },
+      });
+      const field = batch && scopedRefField(batch.serviceKind);
+      if (batch && field) {
+        const blank = await this.prisma.propertyService.findMany({
+          where: {
+            kind: batch.serviceKind,
+            [field]: null,
+            property: { customerId: matchedCustomerId },
+          },
+          select: { id: true },
+          take: 2,
+        });
+        if (blank.length === 1) matchedPropertyServiceId = blank[0].id;
+      }
+    }
+
+    return this.prisma.statementDocument.update({
+      where: { id },
+      data: {
+        extractedAccountRef: dto.accountRef ?? undefined,
+        extractedAmount:
+          dto.amount != null ? new Prisma.Decimal(dto.amount) : undefined,
+        extractedPeriod: dto.period ?? undefined,
+        extractedDueDate: dto.dueDate ? new Date(dto.dueDate) : undefined,
+        matchedPropertyServiceId,
+        matchedCustomerId,
+        status: dto.status ?? "MATCHED",
+        reviewedById,
+        reviewedAt: new Date(),
+      },
+    });
+  }
+
+  async reject(id: string, reviewedById: string) {
+    const doc = await this.prisma.statementDocument.findUnique({ where: { id } });
+    if (!doc) throw new NotFoundException("Documento no encontrado.");
+    if (doc.status === "POSTED") {
+      throw new BadRequestException("Este documento ya fue registrado.");
+    }
+    return this.prisma.statementDocument.update({
+      where: { id },
+      data: { status: "REJECTED", reviewedById, reviewedAt: new Date() },
+    });
+  }
+
+  // --- posting --------------------------------------------------------------
+
+  /**
+   * Post every confirmable document in a batch to the ledger.
+   *
+   * This goes through `BillingService.createBatch` — the same method the manual
+   * "Editor" screen uses — rather than writing `Transaction` rows directly, so
+   * OCR-sourced and hand-keyed receipts share one write path, one validation
+   * path and one audit trail. `source: "OCR"` and a per-line `captureRef` of
+   * the document id give the duplicate-post guard something to key on, so a
+   * batch confirmed twice cannot double-charge anyone.
+   */
+  async confirmBatch(batchId: string, dto: ConfirmBatchDto, reviewedById: string) {
+    const batch = await this.prisma.statementBatch.findUnique({
+      where: { id: batchId },
+    });
+    if (!batch) throw new NotFoundException("Lote no encontrado.");
+
+    const docs = await this.prisma.statementDocument.findMany({
+      where: {
+        batchId,
+        status: { in: dto.includeReviewed ? ["MATCHED", "CONFIRMED"] : ["MATCHED"] },
+        matchedCustomerId: { not: null },
+      },
+      orderBy: { pageNumber: "asc" },
+    });
+    if (!docs.length) {
+      throw new BadRequestException("No hay documentos listos para registrar.");
+    }
+
+    const missing = docs.filter((d) => d.extractedAmount == null);
+    if (missing.length) {
+      throw new BadRequestException(
+        `Falta el importe en ${missing.length} documento(s): página(s) ` +
+          missing.map((d) => d.pageNumber).join(", "),
+      );
+    }
+
+    const typeId = dto.typeId ?? (await this.conceptFor(batch.serviceKind));
+
+    const result = await this.billing.createBatch(
+      {
+        domain: "UTILITY",
+        transactionDate: dto.transactionDate,
+        checkNumber: dto.checkNumber,
+        currency: dto.currency ?? "MXN",
+        typeId,
+        lines: docs.map((d) => ({
+          customerId: d.matchedCustomerId!,
+          // Charges are negative in this ledger: a negative amount is what the
+          // customer owes. The parser reads the printed (positive) figure, so
+          // the sign is applied here, at the single point where a statement
+          // becomes a ledger row.
+          amount: -Math.abs(Number(d.extractedAmount)),
+          reference: d.extractedAccountRef ?? undefined,
+          period: d.extractedPeriod ?? undefined,
+          outstanding: dto.outstanding ?? false,
+        })),
+      },
+      { source: "OCR", refs: docs.map((d) => d.id) },
+    );
+
+    // `items[i]` is positionally parallel to `lines[i]` (seam guarantee 1), so
+    // the created rows zip straight back onto the documents that produced them.
+    await this.prisma.$transaction(
+      docs.map((d, i) =>
+        this.prisma.statementDocument.update({
+          where: { id: d.id },
+          data: {
+            status: "POSTED",
+            postedTransactionId: result.items[i].id,
+            reviewedById,
+            reviewedAt: new Date(),
+          },
+        }),
+      ),
+    );
+
+    // Teach the matcher. When a document was matched by clave catastral or by
+    // hand because the scoped field was blank, writing the reference back means
+    // next month's statement for the same account matches on its own — this is
+    // what turns gas (whose numbers the migration never populated) from a
+    // permanent review queue into a one-time cost.
+    await this.learnAccountRefs(docs, batch.serviceKind);
+
+    await this.closeIfDone(batchId);
+
+    return { posted: result.count, total: result.total, checkNumber: dto.checkNumber };
+  }
+
+  /** Write a confirmed reference onto a service that had none. */
+  private async learnAccountRefs(
+    docs: { matchedPropertyServiceId: string | null; extractedAccountRef: string | null }[],
+    kind: ServiceKind,
+  ) {
+    const field = scopedRefField(kind);
+    if (!field) return;
+    for (const d of docs) {
+      if (!d.matchedPropertyServiceId || !d.extractedAccountRef) continue;
+      await this.prisma.propertyService.updateMany({
+        // Only fills a hole — never overwrites a number already on file, which
+        // would let one misread page rewrite good reference data.
+        where: { id: d.matchedPropertyServiceId, [field]: null },
+        data: { [field]: d.extractedAccountRef },
+      });
+    }
+  }
+
+  private async closeIfDone(batchId: string) {
+    const open = await this.prisma.statementDocument.count({
+      where: { batchId, status: { in: OPEN } },
+    });
+    if (open === 0) {
+      await this.prisma.statementBatch.update({
+        where: { id: batchId },
+        data: { status: "COMPLETED", completedAt: new Date() },
+      });
+    }
+  }
+
+  private async conceptFor(kind: ServiceKind): Promise<string | undefined> {
+    const name = CONCEPT_BY_KIND[kind];
+    if (!name) return undefined;
+    const row = await this.prisma.typeTransaction.findFirst({
+      where: { nameEn: name },
+      select: { id: true },
+    });
+    return row?.id;
+  }
+}

apps/api/src/storage/storage.service.ts

@@ -73,6 +73,16 @@ export class StorageService implements OnModuleInit {
     }
   }
 
+  /**
+   * Whether the deployment has object storage at all. Callers use this to
+   * refuse work up front instead of failing halfway through — a recibo batch
+   * that dies on its first `put` leaves a FAILED batch and no explanation the
+   * office can act on.
+   */
+  get available(): boolean {
+    return this.client !== null;
+  }
+
   private require(): S3Client {
     if (!this.client) {
       throw new ServiceUnavailableException(

apps/api/tsconfig.build.json

@@ -0,0 +1,4 @@
+{
+  "extends": "./tsconfig.json",
+  "exclude": ["node_modules", "dist", "**/*.spec.ts"]
+}

apps/web/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@jorgecuadros/web",
-  "version": "1.0.1",
+  "version": "1.0.6",
   "private": true,
   "scripts": {
     "dev": "next dev -p 4500",

apps/web/src/app/estado-cuenta/lote/page.tsx

@@ -1,557 +1,11 @@
-"use client";
-
-import { useEffect, useMemo, useState } from "react";
-import Link from "next/link";
 import { AppShell } from "@/components/AppShell";
-import { CustomerPicker } from "@/components/CustomerPicker";
+import { Captura } from "@/components/Captura";
-import { createMovementBatch, getBillingFacets, getByCheck } from "@/lib/api";
-import { useCan } from "@/lib/abilities";
-import { formatMoney, formatNumber, txTypeLabel } from "@/lib/labels";
-import type {
-  BatchCreateInput,
-  BillingFacets,
-  ByCheckResponse,
-  Currency,
-  LedgerCurrency,
-  TransactionDomain,
-} from "@/lib/types";
-
-/**
- * Batch capture by check — the "Editor" screen from the legacy system
- * (docs/RECEIPT_CAPTURE_SPEC.md §1.2).
- *
- * Staff key many customers' receipts against ONE physical check before cutting
- * it, then check that the captured total matches the check's amount. That
- * reconciliation is the whole point, so the running total is the most prominent
- * thing on the page and an optional "importe del cheque" field turns it into a
- * live difference.
- *
- * No batch entity is persisted: `checkNumber` is a plain column, and grouping
- * by it answers every by-check question (see the "Reporte por cheque" report).
- */
-
-const DOMAINS: { key: TransactionDomain; label: string }[] = [
-  { key: "UTILITY", label: "Servicios" },
-  { key: "INSURANCE", label: "Seguros" },
-  { key: "TRUST", label: "Fideicomiso" },
-];
-
-interface Line {
-  /** Local row key — lines have no server identity until the batch posts. */
-  key: number;
-  customerId: string;
-  customerName: string;
-  amount: string;
-  reference: string;
-  period: string;
-  outstanding: boolean;
-}
-
-function blankLine(key: number): Line {
-  return {
-    key,
-    customerId: "",
-    customerName: "",
-    amount: "",
-    reference: "",
-    period: "",
-    outstanding: false,
-  };
-}
 
+/** Daily capture, opened on the manual (key-by-hand) mode. */
 export default function BatchCapturePage() {
   return (
     <AppShell>
-      <BatchCapture />
+      <Captura initialMode="manual" />
     </AppShell>
   );
 }
-
-function BatchCapture() {
-  const canCapture = useCan("ledger:create");
-  const [facets, setFacets] = useState<BillingFacets | null>(null);
-
-  // Check-level fields — shared by every line.
-  const [domain, setDomain] = useState<TransactionDomain>("UTILITY");
-  const [currency, setCurrency] = useState<LedgerCurrency>("MXN");
-  const [typeId, setTypeId] = useState("");
-  const [checkNumber, setCheckNumber] = useState("");
-  const [transactionDate, setTransactionDate] = useState(
-    new Date().toISOString().slice(0, 10),
-  );
-  /** The physical check's amount, for reconciliation only — never submitted. */
-  const [checkAmount, setCheckAmount] = useState("");
-
-  const [lines, setLines] = useState<Line[]>([blankLine(1), blankLine(2), blankLine(3)]);
-  const [nextKey, setNextKey] = useState(4);
-
-  const [saving, setSaving] = useState(false);
-  const [error, setError] = useState<string | null>(null);
-  const [posted, setPosted] = useState<ByCheckResponse | null>(null);
-
-  useEffect(() => {
-    getBillingFacets().then(setFacets).catch(() => setFacets(null));
-  }, []);
-
-  const filled = lines.filter(
-    (l) => l.customerId && l.amount.trim() !== "" && Number.isFinite(Number(l.amount)),
-  );
-
-  // Charges are captured as positive numbers and signed on submit, matching
-  // MovementForm — staff type what's on the bill, not a negative.
-  const total = useMemo(
-    () =>
-      filled
-        .filter((l) => !l.outstanding)
-        .reduce((sum, l) => sum + Math.abs(Number(l.amount)), 0),
-    [filled],
-  );
-  const outstandingTotal = useMemo(
-    () =>
-      filled
-        .filter((l) => l.outstanding)
-        .reduce((sum, l) => sum + Math.abs(Number(l.amount)), 0),
-    [filled],
-  );
-
-  const checkAmt = Number(checkAmount);
-  const hasCheckAmt = checkAmount.trim() !== "" && Number.isFinite(checkAmt);
-  const diff = hasCheckAmt ? checkAmt - total : 0;
-  const reconciled = hasCheckAmt && Math.abs(diff) < 0.005;
-
-  function update(key: number, patch: Partial<Line>) {
-    setLines((ls) => ls.map((l) => (l.key === key ? { ...l, ...patch } : l)));
-  }
-
-  function addLine() {
-    setLines((ls) => [...ls, blankLine(nextKey)]);
-    setNextKey((k) => k + 1);
-  }
-
-  function removeLine(key: number) {
-    setLines((ls) => (ls.length === 1 ? ls : ls.filter((l) => l.key !== key)));
-  }
-
-  async function submit(e: React.FormEvent) {
-    e.preventDefault();
-    if (!checkNumber.trim()) {
-      setError("Indica el número de cheque.");
-      return;
-    }
-    if (filled.length === 0) {
-      setError("Captura al menos una línea con cliente y monto.");
-      return;
-    }
-    const dupes = filled
-      .map((l) => l.customerId)
-      .filter((id, i, arr) => arr.indexOf(id) !== i);
-    if (dupes.length) {
-      const names = filled
-        .filter((l) => dupes.includes(l.customerId))
-        .map((l) => l.customerName);
-      if (
-        !window.confirm(
-          `Hay más de una línea para el mismo cliente (${[...new Set(names)].join(
-            ", ",
-          )}). ¿Continuar?`,
-        )
-      )
-        return;
-    }
-
-    const payload: BatchCreateInput = {
-      domain,
-      transactionDate,
-      checkNumber: checkNumber.trim(),
-      currency: currency as Currency,
-      typeId: typeId || undefined,
-      lines: filled.map((l) => ({
-        customerId: l.customerId,
-        // Every line of a check batch is a charge the office paid out.
-        amount: -Math.abs(Number(l.amount)),
-        reference: l.reference.trim() || undefined,
-        period: l.period.trim() || undefined,
-        outstanding: l.outstanding || undefined,
-      })),
-    };
-
-    setSaving(true);
-    setError(null);
-    try {
-      await createMovementBatch(payload);
-      // Re-read through the by-check view so the confirmation shows what's
-      // actually stored (including anything captured against this check
-      // earlier), not just what this request sent.
-      setPosted(await getByCheck(payload.checkNumber));
-    } catch (e2) {
-      setError((e2 as Error)?.message ?? "No se pudo guardar el lote.");
-    } finally {
-      setSaving(false);
-    }
-  }
-
-  function reset() {
-    setPosted(null);
-    setLines([blankLine(nextKey), blankLine(nextKey + 1), blankLine(nextKey + 2)]);
-    setNextKey((k) => k + 3);
-    setCheckNumber("");
-    setCheckAmount("");
-  }
-
-  if (!canCapture) {
-    return (
-      <div className="state-box state-error">
-        No tienes permiso para capturar movimientos.
-      </div>
-    );
-  }
-
-  if (posted) {
-    return (
-      <>
-        <div className="page-head">
-          <div>
-            <h1 className="page-title">Lote capturado</h1>
-            <p className="eyebrow">
-              Cheque {posted.checkNumber} · {formatNumber(posted.count)}{" "}
-              {posted.count === 1 ? "movimiento" : "movimientos"}
-            </p>
-          </div>
-          <div style={{ display: "flex", gap: 10 }}>
-            <button type="button" className="btn btn-primary" onClick={reset}>
-              Capturar otro cheque
-            </button>
-            <Link href="/estado-cuenta" className="btn btn-outline">
-              Volver a estado de cuenta
-            </Link>
-          </div>
-        </div>
-
-        <div className="filtered-totals" style={{ marginBottom: 16 }}>
-          {posted.totals.map((t) => (
-            <div className="filtered-total" key={t.currency}>
-              <span className="filtered-total-cur">{t.currency}</span>
-              <span className="filtered-total-net">
-                Total del cheque <strong>{formatMoney(t.total, t.currency)}</strong>
-              </span>
-              <span>{formatNumber(t.count)} movimientos</span>
-            </div>
-          ))}
-          {posted.outstandingCount > 0 && (
-            <div className="filtered-total">
-              <span>
-                {formatNumber(posted.outstandingCount)} sin fondos (no suman al
-                total)
-              </span>
-            </div>
-          )}
-        </div>
-
-        <div className="tx-scroll">
-          <table className="tx-table">
-            <thead>
-              <tr>
-                <th>Cliente</th>
-                <th>Referencia</th>
-                <th>Periodo</th>
-                <th>Estado</th>
-                <th className="num">Monto</th>
-              </tr>
-            </thead>
-            <tbody>
-              {posted.items.map((i) => (
-                <tr key={i.id}>
-                  <td>
-                    <Link
-                      href={`/estado-cuenta/${i.customerId}`}
-                      className="inline-link"
-                    >
-                      {i.customerName}
-                    </Link>
-                  </td>
-                  <td>{i.reference || "—"}</td>
-                  <td>{i.period || "—"}</td>
-                  <td>{i.outstanding ? "Sin fondos" : "Pagado"}</td>
-                  <td className="num">
-                    <span className="tx-amount neg">
-                      {formatMoney(i.amount, i.currency)}
-                    </span>
-                  </td>
-                </tr>
-              ))}
-            </tbody>
-          </table>
-        </div>
-
-        <p className="muted" style={{ marginTop: 14 }}>
-          Para imprimir la conciliación, usa el reporte{" "}
-          <Link
-            href={`/reportes/cheque-count?checkNumber=${encodeURIComponent(
-              posted.checkNumber,
-            )}`}
-            className="inline-link"
-          >
-            Reporte por cheque
-          </Link>
-          .
-        </p>
-      </>
-    );
-  }
-
-  return (
-    <>
-      <div className="page-head">
-        <div>
-          <h1 className="page-title">Captura por cheque</h1>
-          <p className="eyebrow">
-            Captura los recibos de varios clientes contra un mismo cheque y
-            concilia el total antes de guardar.
-          </p>
-        </div>
-        <Link href="/estado-cuenta" className="btn btn-outline">
-          Cancelar
-        </Link>
-      </div>
-
-      {error && <div className="state-box state-error">{error}</div>}
-
-      <form onSubmit={submit}>
-        <div className="card" style={{ padding: 20, marginBottom: 16 }}>
-          <h2 className="section-title" style={{ marginBottom: 14 }}>
-            Datos del cheque
-          </h2>
-          <div className="form-grid">
-            <label className="field">
-              <span className="field-label">Número de cheque *</span>
-              <input
-                className="input"
-                value={checkNumber}
-                onChange={(e) => setCheckNumber(e.target.value)}
-                required
-              />
-            </label>
-            <label className="field">
-              <span className="field-label">Fecha *</span>
-              <input
-                className="input"
-                type="date"
-                required
-                value={transactionDate}
-                onChange={(e) => setTransactionDate(e.target.value)}
-              />
-            </label>
-            <label className="field">
-              <span className="field-label">Línea de negocio *</span>
-              <select
-                className="select"
-                value={domain}
-                onChange={(e) => setDomain(e.target.value as TransactionDomain)}
-              >
-                {DOMAINS.map((d) => (
-                  <option key={d.key} value={d.key}>
-                    {d.label}
-                  </option>
-                ))}
-              </select>
-            </label>
-            <label className="field">
-              <span className="field-label">Moneda *</span>
-              <select
-                className="select"
-                value={currency}
-                onChange={(e) => setCurrency(e.target.value as LedgerCurrency)}
-              >
-                <option value="MXN">Pesos (MXN)</option>
-                <option value="USD">Dólares (USD)</option>
-              </select>
-            </label>
-            <label className="field">
-              <span className="field-label">Concepto</span>
-              <select
-                className="select"
-                value={typeId}
-                onChange={(e) => setTypeId(e.target.value)}
-              >
-                <option value="">(sin concepto)</option>
-                {facets?.types.map((t) => (
-                  <option key={t.id} value={t.id}>
-                    {txTypeLabel({ nameEn: t.name })}
-                  </option>
-                ))}
-              </select>
-            </label>
-            <label className="field">
-              <span className="field-label">Importe del cheque</span>
-              <input
-                className="input"
-                type="number"
-                step="0.01"
-                min="0"
-                value={checkAmount}
-                onChange={(e) => setCheckAmount(e.target.value)}
-                placeholder="Para conciliar"
-              />
-            </label>
-          </div>
-        </div>
-
-        <div className="card" style={{ padding: 20, marginBottom: 16 }}>
-          <div
-            style={{
-              display: "flex",
-              justifyContent: "space-between",
-              alignItems: "center",
-              marginBottom: 14,
-            }}
-          >
-            <h2 className="section-title" style={{ margin: 0 }}>
-              Recibos ({formatNumber(filled.length)})
-            </h2>
-            <button type="button" className="btn btn-outline" onClick={addLine}>
-              Agregar línea
-            </button>
-          </div>
-
-          <div className="tx-scroll">
-            <table className="tx-table">
-              <thead>
-                <tr>
-                  <th style={{ minWidth: 240 }}>Cliente *</th>
-                  <th style={{ minWidth: 120 }}>Referencia</th>
-                  <th style={{ minWidth: 100 }}>Periodo</th>
-                  <th style={{ minWidth: 110 }} className="num">
-                    Monto *
-                  </th>
-                  <th style={{ whiteSpace: "nowrap" }}>Sin fondos</th>
-                  <th style={{ width: 1 }} />
-                </tr>
-              </thead>
-              <tbody>
-                {lines.map((l) => (
-                  <tr key={l.key}>
-                    <td>
-                      <CustomerPicker
-                        value={l.customerId}
-                        valueName={l.customerId ? l.customerName : undefined}
-                        onPick={(id, name) =>
-                          update(l.key, { customerId: id, customerName: name })
-                        }
-                      />
-                    </td>
-                    <td>
-                      <input
-                        className="input"
-                        value={l.reference}
-                        onChange={(e) =>
-                          update(l.key, { reference: e.target.value })
-                        }
-                      />
-                    </td>
-                    <td>
-                      <input
-                        className="input"
-                        value={l.period}
-                        onChange={(e) => update(l.key, { period: e.target.value })}
-                        placeholder="2026-07"
-                      />
-                    </td>
-                    <td>
-                      <input
-                        className="input num"
-                        type="number"
-                        step="0.01"
-                        min="0"
-                        value={l.amount}
-                        onChange={(e) => update(l.key, { amount: e.target.value })}
-                        placeholder="0.00"
-                      />
-                    </td>
-                    <td style={{ textAlign: "center" }}>
-                      <input
-                        type="checkbox"
-                        checked={l.outstanding}
-                        onChange={(e) =>
-                          update(l.key, { outstanding: e.target.checked })
-                        }
-                        aria-label="Sin fondos"
-                      />
-                    </td>
-                    <td>
-                      <button
-                        type="button"
-                        className="btn btn-ghost"
-                        style={{ padding: "4px 10px", fontSize: 12 }}
-                        onClick={() => removeLine(l.key)}
-                        disabled={lines.length === 1}
-                      >
-                        Quitar
-                      </button>
-                    </td>
-                  </tr>
-                ))}
-              </tbody>
-            </table>
-          </div>
-        </div>
-
-        <div className="card" style={{ padding: 20, marginBottom: 16 }}>
-          <h2 className="section-title" style={{ marginBottom: 14 }}>
-            Conciliación
-          </h2>
-          <div className="filtered-totals">
-            <div className="filtered-total">
-              <span className="filtered-total-cur">{currency}</span>
-              <span className="filtered-total-net">
-                Capturado <strong>{formatMoney(String(-total), currency)}</strong>
-              </span>
-              <span>{formatNumber(filled.filter((l) => !l.outstanding).length)} recibos</span>
-            </div>
-            {outstandingTotal > 0 && (
-              <div className="filtered-total">
-                <span>
-                  Sin fondos{" "}
-                  <strong>{formatMoney(String(-outstandingTotal), currency)}</strong>{" "}
-                  (no suma al cheque)
-                </span>
-              </div>
-            )}
-            {hasCheckAmt && (
-              <div className="filtered-total">
-                <span className="filtered-total-net">
-                  {reconciled ? (
-                    <strong className="tx-amount pos">Cuadra con el cheque</strong>
-                  ) : (
-                    <>
-                      Diferencia{" "}
-                      <strong className="tx-amount neg">
-                        {formatMoney(String(diff), currency)}
-                      </strong>
-                    </>
-                  )}
-                </span>
-              </div>
-            )}
-          </div>
-        </div>
-
-        <div className="form-actions">
-          <button
-            type="submit"
-            className="btn btn-primary"
-            disabled={saving || filled.length === 0}
-          >
-            {saving
-              ? "Guardando…"
-              : `Capturar ${formatNumber(filled.length)} ${
-                  filled.length === 1 ? "recibo" : "recibos"
-                }`}
-          </button>
-          <Link href="/estado-cuenta" className="btn btn-outline">
-            Cancelar
-          </Link>
-        </div>
-      </form>
-    </>
-  );
-}

apps/web/src/app/globals.css

@@ -837,6 +837,46 @@ button {
   display: inline-block;
 }
 
+/* Upload progress (Operaciones ingest) */
+.upload-progress {
+  display: flex;
+  flex-direction: column;
+  gap: 6px;
+  padding: 4px 0 8px;
+}
+.progress-track {
+  position: relative;
+  overflow: hidden;
+  height: 8px;
+  border-radius: 999px;
+  background: var(--paper-2);
+}
+.progress-fill {
+  height: 100%;
+  border-radius: 999px;
+  background: var(--brand-500);
+  transition: width 0.2s linear;
+}
+.progress-indeterminate .progress-fill {
+  width: 40% !important;
+  animation: progress-slide 1.2s var(--ease-out-quart) infinite;
+}
+@keyframes progress-slide {
+  0% {
+    transform: translateX(-100%);
+  }
+  100% {
+    transform: translateX(250%);
+  }
+}
+.upload-progress-stats {
+  display: flex;
+  flex-wrap: wrap;
+  gap: 12px;
+  font-size: 12px;
+  color: var(--muted);
+}
+
 @keyframes shimmer {
   0% {
     background-position: -420px 0;

apps/web/src/app/operaciones/page.tsx

@@ -1,6 +1,6 @@
 "use client";
 
-import { useCallback, useEffect, useRef, useState } from "react";
+import { Fragment, useCallback, useEffect, useRef, useState } from "react";
 import { AppShell } from "@/components/AppShell";
 import { useCan } from "@/lib/abilities";
 import {
@@ -20,6 +20,7 @@ import {
   startOpsJob,
   uploadIngest,
 } from "@/lib/api";
+import type { UploadProgress } from "@/lib/api";
 import type {
   BackupFile,
   IngestFile,
@@ -56,6 +57,7 @@ function Operaciones() {
   const [confirm, setConfirm] = useState<ConfirmState>(null);
   const [confirmText, setConfirmText] = useState("");
   const [uploading, setUploading] = useState<string | null>(null);
+  const [progress, setProgress] = useState<UploadProgress | null>(null);
   const [starting, setStarting] = useState(false);
 
   const fileInputs = useRef<Record<string, HTMLInputElement | null>>({});
@@ -119,14 +121,16 @@ function Operaciones() {
     setError(null);
     setNotice(null);
     setUploading(name);
+    setProgress(null);
     try {
-      await uploadIngest(name, file);
+      await uploadIngest(name, file, setProgress);
       setNotice(`${name} cargado.`);
       refreshLists();
     } catch (e) {
       setError((e as Error)?.message ?? "No se pudo cargar el archivo.");
     } finally {
       setUploading(null);
+      setProgress(null);
       const input = fileInputs.current[name];
       if (input) input.value = "";
     }
@@ -243,7 +247,8 @@ function Operaciones() {
             </thead>
             <tbody>
               {(ingest ?? []).map((f) => (
-                <tr key={f.name}>
+                <Fragment key={f.name}>
+                  <tr>
                     <td className="mono">{f.name}</td>
                     <td>
                       <span className={`badge ${f.present ? "badge-positive" : "badge-negative"}`}>
@@ -274,6 +279,7 @@ function Operaciones() {
                           <button
                             className="btn btn-ghost"
                             type="button"
+                            disabled={uploading === f.name}
                             onClick={() => handleDeleteIngest(f.name)}
                           >
                             Eliminar
@@ -282,6 +288,14 @@ function Operaciones() {
                       </div>
                     </td>
                   </tr>
+                  {uploading === f.name && (
+                    <tr>
+                      <td colSpan={5}>
+                        <UploadProgressBar progress={progress} />
+                      </td>
+                    </tr>
+                  )}
+                </Fragment>
               ))}
             </tbody>
           </table>
@@ -496,6 +510,61 @@ function Operaciones() {
   );
 }
 
+/** "1:05" / "0:09" — remaining time, coarse on purpose. */
+function formatEta(seconds: number): string {
+  const s = Math.max(0, Math.round(seconds));
+  if (s >= 3600) {
+    const h = Math.floor(s / 3600);
+    const m = Math.round((s % 3600) / 60);
+    return `${h} h ${m} min`;
+  }
+  return `${Math.floor(s / 60)}:${String(s % 60).padStart(2, "0")}`;
+}
+
+/**
+ * Live upload readout. The bar tracks bytes handed to the network; once those
+ * are all sent the server still has to write the file, so the tail of the
+ * upload reads "Procesando…" instead of sitting at 100%.
+ */
+function UploadProgressBar({ progress }: { progress: UploadProgress | null }) {
+  const pct = progress?.fraction != null ? Math.round(progress.fraction * 100) : null;
+  return (
+    <div className="upload-progress">
+      <div
+        className={`progress-track${pct === null ? " progress-indeterminate" : ""}`}
+        role="progressbar"
+        aria-valuemin={0}
+        aria-valuemax={100}
+        aria-valuenow={pct ?? undefined}
+      >
+        <div className="progress-fill" style={{ width: `${pct ?? 100}%` }} />
+      </div>
+      <div className="upload-progress-stats mono">
+        {progress === null ? (
+          "Preparando…"
+        ) : progress.finishing ? (
+          `Procesando en el servidor… (${formatBytes(progress.total)} enviados)`
+        ) : (
+          <>
+            {pct !== null && <strong>{pct}%</strong>}
+            {progress.total > 0 && (
+              <span>
+                {formatBytes(progress.loaded)} / {formatBytes(progress.total)}
+              </span>
+            )}
+            {progress.bytesPerSecond > 0 && (
+              <span>{formatBytes(progress.bytesPerSecond)}/s</span>
+            )}
+            {progress.secondsRemaining !== null && progress.bytesPerSecond > 0 && (
+              <span>faltan {formatEta(progress.secondsRemaining)}</span>
+            )}
+          </>
+        )}
+      </div>
+    </div>
+  );
+}
+
 function OpTile({
   title,
   desc,

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

@@ -0,0 +1,16 @@
+"use client";
+
+import { AppShell } from "@/components/AppShell";
+import { PolicyOcrReview } from "@/components/PolicyOcrReview";
+
+export default function PolicyOcrBatchPage({
+  params,
+}: {
+  params: { id: string };
+}) {
+  return (
+    <AppShell>
+      <PolicyOcrReview id={params.id} />
+    </AppShell>
+  );
+}

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

@@ -0,0 +1,21 @@
+"use client";
+
+import { AppShell } from "@/components/AppShell";
+import { PolicyCaptura } from "@/components/PolicyCaptura";
+
+/**
+ * OCR mode of the policy intake screen. Drops the GMX 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.
+ *
+ * Sister route `/polizas/captura/[id]` is the batch review screen once a
+ * batch is uploaded.
+ */
+export default function CapturaOcrPage() {
+  return (
+    <AppShell>
+      <PolicyCaptura initialMode="auto" />
+    </AppShell>
+  );
+}

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

@@ -1,41 +1,22 @@
 "use client";
 
 import { Suspense } from "react";
-import Link from "next/link";
-import { useSearchParams } from "next/navigation";
 import { AppShell } from "@/components/AppShell";
-import { PolicyForm } from "@/components/PolicyForm";
+import { PolicyCaptura } from "@/components/PolicyCaptura";
-import { useCan } from "@/lib/abilities";
 
+/**
+ * Manual mode of the policy intake screen. Shares the tab wrapper with
+ * `/polizas/captura` (OCR mode) so staff can swap between the two without
+ * losing their place. Customer picker comes from the `?customerId=`
+ * / `?customerName=` query string — used by `/clientes/[id]` when staff
+ * creates a policy from a customer detail page.
+ */
 export default function NuevaPolizaPage() {
   return (
     <AppShell>
       <Suspense fallback={null}>
-        <NuevaPoliza />
+        <PolicyCaptura initialMode="manual" />
       </Suspense>
     </AppShell>
   );
 }
-
-function NuevaPoliza() {
-  const allowed = useCan("policy:create");
-  const params = useSearchParams();
-  const customerId = params.get("customerId") ?? undefined;
-  const customerName = params.get("customerName") ?? undefined;
-
-  return (
-    <>
-      <div className="page-head">
-        <Link href="/polizas" className="back-link">← Pólizas</Link>
-        <h1 className="page-title">Nueva póliza</h1>
-      </div>
-      {allowed ? (
-        <PolicyForm fixedCustomerId={customerId} fixedCustomerName={customerName} />
-      ) : (
-        <div className="state-box state-error">
-          No tiene permisos para crear pólizas.
-        </div>
-      )}
-    </>
-  );
-}

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

@@ -57,6 +57,7 @@ export default function PolizasPage() {
 
 function PolizasBrowser() {
   const canCreate = useCan("policy:create");
+  const canIngest = useCan("policy:ingest");
   const [stats, setStats] = useState<PolicyStats | null>(null);
   const [facets, setFacets] = useState<PolicyFacets | null>(null);
 
@@ -135,6 +136,11 @@ function PolizasBrowser() {
               { slug: "vigente", label: "Por vencer (Incen.)", params: { typeName: "INCEN" } },
             ]}
           />
+          {canIngest && (
+            <Link href="/polizas/captura" className="btn btn-outline">
+              + Captura OCR
+            </Link>
+          )}
           {canCreate && (
             <Link href="/polizas/nuevo" className="btn btn-primary">+ Nueva póliza</Link>
           )}

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

@@ -0,0 +1,525 @@
+"use client";
+
+import { useCallback, useEffect, useMemo, useState } from "react";
+import Link from "next/link";
+import { AppShell } from "@/components/AppShell";
+import { CustomerPicker } from "@/components/CustomerPicker";
+import {
+  confirmStatementBatch,
+  getStatementBatch,
+  listStatementDocuments,
+  rejectStatementDocument,
+  reviewStatementDocument,
+  statementPageUrl,
+} from "@/lib/api";
+import { useCan } from "@/lib/abilities";
+import { formatDate, formatMoney, serviceKindLabel } from "@/lib/labels";
+import type {
+  ConfirmBatchInput,
+  StatementBatchDetail,
+  StatementDocument,
+  StatementDocumentStatus,
+} from "@/lib/types";
+
+/**
+ * Review queue for one batch of scanned bills.
+ *
+ * The reviewer's job is to answer one question per page — "is this the right
+ * customer for this amount?" — so the page image sits next to the extracted
+ * fields and every row can be corrected in place. Rows the matcher is sure
+ * about are pre-approved and can be posted in bulk; everything else is listed
+ * first, because that is the work.
+ */
+
+const STATUS_LABEL: Record<StatementDocumentStatus, string> = {
+  PENDING_OCR: "En proceso",
+  OCR_FAILED: "No se pudo leer",
+  NEEDS_REVIEW: "Requiere revisión",
+  MATCHED: "Identificado",
+  CONFIRMED: "Confirmado",
+  POSTED: "Registrado",
+  REJECTED: "Descartado",
+};
+
+/** Rows still needing a decision, listed before the settled ones. */
+const OPEN_FIRST: StatementDocumentStatus[] = [
+  "NEEDS_REVIEW",
+  "OCR_FAILED",
+  "MATCHED",
+  "CONFIRMED",
+  "POSTED",
+  "REJECTED",
+  "PENDING_OCR",
+];
+
+export default function RecibosBatchPage({ params }: { params: { id: string } }) {
+  return (
+    <AppShell>
+      <BatchReview id={params.id} />
+    </AppShell>
+  );
+}
+
+function BatchReview({ id }: { id: string }) {
+  const canReview = useCan("statement:review");
+  const [batch, setBatch] = useState<StatementBatchDetail | null>(null);
+  const [docs, setDocs] = useState<StatementDocument[]>([]);
+  const [error, setError] = useState<string | null>(null);
+  const [loading, setLoading] = useState(true);
+
+  const load = useCallback(async () => {
+    try {
+      const [b, d] = await Promise.all([
+        getStatementBatch(id),
+        listStatementDocuments(id),
+      ]);
+      setBatch(b);
+      setDocs(d);
+      setError(null);
+    } catch (e) {
+      setError((e as Error)?.message ?? "No se pudo cargar el lote.");
+    } finally {
+      setLoading(false);
+    }
+  }, [id]);
+
+  useEffect(() => {
+    void load();
+  }, [load]);
+
+  const processing = batch?.status === "PROCESSING" || batch?.status === "UPLOADED";
+  useEffect(() => {
+    if (!processing) return;
+    const t = setInterval(() => void load(), 4000);
+    return () => clearInterval(t);
+  }, [processing, load]);
+
+  const sorted = useMemo(
+    () =>
+      [...docs].sort(
+        (a, b) =>
+          OPEN_FIRST.indexOf(a.status) - OPEN_FIRST.indexOf(b.status) ||
+          a.pageNumber - b.pageNumber,
+      ),
+    [docs],
+  );
+
+  const readyCount = docs.filter(
+    (d) => d.status === "MATCHED" && d.matchedCustomer,
+  ).length;
+
+  if (loading) return <div className="state-box">Cargando…</div>;
+  if (!batch) return <div className="state-box state-error">{error ?? "No encontrado."}</div>;
+
+  return (
+    <div className="stack">
+      <header className="page-head">
+        <div>
+          <h1 className="page-title">
+            Recibos — {serviceKindLabel(batch.serviceKind)}
+            {batch.label ? ` · ${batch.label}` : ""}
+          </h1>
+          <p className="page-sub">
+            {formatDate(batch.createdAt)} · {docs.length} página(s) ·{" "}
+            {STATUS_LABEL_BATCH[batch.status] ?? batch.status}
+          </p>
+        </div>
+        <Link className="btn btn-ghost" href="/recibos">
+          Volver a captura
+        </Link>
+      </header>
+
+      {error && <div className="state-box state-error">{error}</div>}
+
+      {processing && (
+        <ProcessingBanner docsLength={docs.length} pendingOcr={batch.byStatus.PENDING_OCR ?? 0} />
+      )}
+
+      <SummaryCard batch={batch} readyCount={readyCount} />
+
+      {canReview && readyCount > 0 && (
+        <ConfirmCard
+          batchId={id}
+          readyCount={readyCount}
+          onDone={load}
+          setError={setError}
+        />
+      )}
+
+      <section className="stack">
+        {sorted.map((doc) => (
+          <DocumentRow
+            key={doc.id}
+            doc={doc}
+            canReview={canReview}
+            onChange={load}
+          />
+        ))}
+      </section>
+    </div>
+  );
+}
+
+const STATUS_LABEL_BATCH: Record<string, string> = {
+  UPLOADED: "Recibido",
+  PROCESSING: "Procesando",
+  READY_FOR_REVIEW: "Listo para revisar",
+  COMPLETED: "Registrado",
+  FAILED: "Falló",
+};
+
+/**
+ * Live readout while OCR is running. The backend tells us how many pages are
+ * still PENDING_OCR, so we can show real progress instead of "loading…". When
+ * the docs list hasn't caught up to the upload yet (total === 0) we fall back
+ * to the indeterminate bar.
+ */
+function ProcessingBanner({
+  docsLength,
+  pendingOcr,
+}: {
+  docsLength: number;
+  pendingOcr: number;
+}) {
+  const done = Math.max(docsLength - pendingOcr, 0);
+  const pct =
+    docsLength > 0 ? Math.min(100, Math.round((done / docsLength) * 100)) : null;
+  return (
+    <div className="card" style={{ padding: 16 }}>
+      <div className="upload-progress" style={{ padding: 0 }}>
+        <div
+          className={`progress-track${pct === null ? " progress-indeterminate" : ""}`}
+          role="progressbar"
+          aria-valuemin={0}
+          aria-valuemax={100}
+          aria-valuenow={pct ?? undefined}
+        >
+          <div className="progress-fill" style={{ width: `${pct ?? 100}%` }} />
+        </div>
+        <div className="upload-progress-stats">
+          {pct === null ? (
+            <span>Leyendo los recibos…</span>
+          ) : (
+            <>
+              <strong>{pct}%</strong>
+              <span>
+                {done} de {docsLength} página(s) leídas
+              </span>
+              {pendingOcr > 0 && <span>{pendingOcr} en cola</span>}
+            </>
+          )}
+          <span style={{ marginLeft: "auto" }}>
+            Esta pantalla se actualiza sola.
+          </span>
+        </div>
+      </div>
+    </div>
+  );
+}
+
+function SummaryCard({
+  batch,
+  readyCount,
+}: {
+  batch: StatementBatchDetail;
+  readyCount: number;
+}) {
+  const entries = Object.entries(batch.byStatus) as [StatementDocumentStatus, number][];
+  return (
+    <section className="card" style={{ padding: 16 }}>
+      <div className="inline-form" style={{ flexWrap: "wrap", gap: 20 }}>
+        {entries.map(([status, count]) => (
+          <div key={status}>
+            <div className="page-sub">{STATUS_LABEL[status] ?? status}</div>
+            <div style={{ fontSize: "1.4rem", fontWeight: 600 }}>{count}</div>
+          </div>
+        ))}
+        <div>
+          <div className="page-sub">Importe pendiente</div>
+          <div style={{ fontSize: "1.4rem", fontWeight: 600 }}>
+            {formatMoney(batch.pendingTotal, "MXN")}
+          </div>
+        </div>
+        <div>
+          <div className="page-sub">Listos para registrar</div>
+          <div style={{ fontSize: "1.4rem", fontWeight: 600 }}>{readyCount}</div>
+        </div>
+      </div>
+    </section>
+  );
+}
+
+/**
+ * Posting is by check, exactly as on the manual capture screen — an OCR batch
+ * is still "these bills, paid with this check", so the same fields are asked
+ * for and the same ledger path is used.
+ */
+function ConfirmCard({
+  batchId,
+  readyCount,
+  onDone,
+  setError,
+}: {
+  batchId: string;
+  readyCount: number;
+  onDone: () => void;
+  setError: (m: string | null) => void;
+}) {
+  const [checkNumber, setCheckNumber] = useState("");
+  const [transactionDate, setTransactionDate] = useState(
+    new Date().toISOString().slice(0, 10),
+  );
+  const [outstanding, setOutstanding] = useState(false);
+  const [includeReviewed, setIncludeReviewed] = useState(true);
+  const [busy, setBusy] = useState(false);
+  const [result, setResult] = useState<string | null>(null);
+
+  async function submit() {
+    if (!checkNumber.trim()) return;
+    setBusy(true);
+    setError(null);
+    try {
+      const input: ConfirmBatchInput = {
+        checkNumber: checkNumber.trim(),
+        transactionDate,
+        outstanding,
+        includeReviewed,
+      };
+      const r = await confirmStatementBatch(batchId, input);
+      setResult(
+        `Se registraron ${r.posted} movimiento(s) por ${formatMoney(r.total, "MXN")} con el cheque ${r.checkNumber}.`,
+      );
+      setCheckNumber("");
+      onDone();
+    } catch (e) {
+      setError((e as Error)?.message ?? "No se pudo registrar el lote.");
+    } finally {
+      setBusy(false);
+    }
+  }
+
+  return (
+    <section className="card" style={{ padding: 16 }}>
+      <h2 className="section-title" style={{ marginTop: 0 }}>
+        Registrar {readyCount} recibo(s)
+      </h2>
+      <div className="inline-form" style={{ flexWrap: "wrap", gap: 12 }}>
+        <label>
+          <span className="page-sub">Cheque</span>
+          <input
+            className="input"
+            value={checkNumber}
+            onChange={(e) => setCheckNumber(e.target.value)}
+            placeholder="Número de cheque"
+          />
+        </label>
+        <label>
+          <span className="page-sub">Fecha</span>
+          <input
+            type="date"
+            className="input"
+            value={transactionDate}
+            onChange={(e) => setTransactionDate(e.target.value)}
+          />
+        </label>
+        <label className="check">
+          <input
+            type="checkbox"
+            checked={outstanding}
+            onChange={(e) => setOutstanding(e.target.checked)}
+          />{" "}
+          Sin fondos (queda pendiente)
+        </label>
+        <label className="check">
+          <input
+            type="checkbox"
+            checked={includeReviewed}
+            onChange={(e) => setIncludeReviewed(e.target.checked)}
+          />{" "}
+          Incluir los confirmados a mano
+        </label>
+        <button
+          type="button"
+          className="btn btn-primary"
+          disabled={!checkNumber.trim() || busy}
+          onClick={submit}
+        >
+          {busy ? "Registrando…" : "Registrar"}
+        </button>
+      </div>
+      {result && (
+        <div className="state-box" style={{ marginTop: 12 }}>
+          {result}
+        </div>
+      )}
+      <p className="page-sub" style={{ marginTop: 12 }}>
+        Se registran como cargos del cliente, por la misma vía que la captura
+        manual. Un lote registrado dos veces no duplica cobros.
+      </p>
+    </section>
+  );
+}
+
+function DocumentRow({
+  doc,
+  canReview,
+  onChange,
+}: {
+  doc: StatementDocument;
+  canReview: boolean;
+  onChange: () => void;
+}) {
+  const [open, setOpen] = useState(
+    doc.status === "NEEDS_REVIEW" || doc.status === "OCR_FAILED",
+  );
+  const [amount, setAmount] = useState(doc.extractedAmount ?? "");
+  const [accountRef, setAccountRef] = useState(doc.extractedAccountRef ?? "");
+  const [customerId, setCustomerId] = useState(doc.matchedCustomer?.id ?? "");
+  const [customerName, setCustomerName] = useState(doc.matchedCustomer?.name ?? "");
+  const [busy, setBusy] = useState(false);
+  const [err, setErr] = useState<string | null>(null);
+
+  const settled = doc.status === "POSTED" || doc.status === "REJECTED";
+
+  async function save(status: "MATCHED" | "CONFIRMED") {
+    setBusy(true);
+    setErr(null);
+    try {
+      await reviewStatementDocument(doc.id, {
+        accountRef: accountRef.trim() || undefined,
+        amount: amount ? Number(amount) : undefined,
+        matchedCustomerId: customerId || undefined,
+        status,
+      });
+      onChange();
+    } catch (e) {
+      setErr((e as Error)?.message ?? "No se pudo guardar.");
+    } finally {
+      setBusy(false);
+    }
+  }
+
+  async function reject() {
+    setBusy(true);
+    setErr(null);
+    try {
+      await rejectStatementDocument(doc.id);
+      onChange();
+    } catch (e) {
+      setErr((e as Error)?.message ?? "No se pudo descartar.");
+    } finally {
+      setBusy(false);
+    }
+  }
+
+  return (
+    <div className="card" style={{ padding: 16 }}>
+      <div
+        className="inline-form"
+        style={{ justifyContent: "space-between", flexWrap: "wrap", gap: 12 }}
+      >
+        <div>
+          <strong>Página {doc.pageNumber}</strong>{" "}
+          <span className="tag">{STATUS_LABEL[doc.status] ?? doc.status}</span>{" "}
+          {doc.provider && <span className="page-sub">{doc.provider}</span>}
+          <div className="page-sub" style={{ marginTop: 4 }}>
+            {doc.matchedCustomer ? (
+              <Link href={`/clientes/${doc.matchedCustomer.id}`}>
+                {doc.matchedCustomer.name}
+              </Link>
+            ) : (
+              "Sin cliente asignado"
+            )}
+            {doc.extractedAccountRef && ` · cuenta ${doc.extractedAccountRef}`}
+            {doc.extractedCadastralKey && ` · clave ${doc.extractedCadastralKey}`}
+          </div>
+          {doc.matchNote && (
+            <div className="page-sub" style={{ marginTop: 4 }}>
+              {doc.matchNote}
+            </div>
+          )}
+        </div>
+        <div className="inline-form" style={{ gap: 8 }}>
+          <strong>
+            {doc.extractedAmount
+              ? formatMoney(doc.extractedAmount, "MXN")
+              : "sin importe"}
+          </strong>
+          <button
+            type="button"
+            className="btn btn-ghost"
+            onClick={() => setOpen((v) => !v)}
+          >
+            {open ? "Ocultar" : "Ver recibo"}
+          </button>
+        </div>
+      </div>
+
+      {open && (
+        <div style={{ marginTop: 12, display: "grid", gap: 16 }}>
+          {/* The scan itself — the reviewer's source of truth, not the OCR. */}
+          <img
+            src={statementPageUrl(doc.id)}
+            alt={`Recibo página ${doc.pageNumber}`}
+            style={{
+              maxWidth: "100%",
+              border: "1px solid var(--border, #ddd)",
+              borderRadius: 6,
+            }}
+          />
+
+          {canReview && !settled && (
+            <div className="inline-form" style={{ flexWrap: "wrap", gap: 12 }}>
+              <label>
+                <span className="page-sub">Cuenta</span>
+                <input
+                  className="input"
+                  value={accountRef}
+                  onChange={(e) => setAccountRef(e.target.value)}
+                />
+              </label>
+              <label>
+                <span className="page-sub">Importe</span>
+                <input
+                  className="input"
+                  inputMode="decimal"
+                  value={amount}
+                  onChange={(e) => setAmount(e.target.value)}
+                />
+              </label>
+              <div style={{ minWidth: 260 }}>
+                <span className="page-sub">Cliente</span>
+                <CustomerPicker
+                  value={customerId}
+                  valueName={customerName}
+                  onPick={(cid, name) => {
+                    setCustomerId(cid);
+                    setCustomerName(name);
+                  }}
+                />
+              </div>
+              <button
+                type="button"
+                className="btn btn-primary"
+                disabled={busy || !customerId}
+                onClick={() => save("MATCHED")}
+              >
+                Guardar
+              </button>
+              <button
+                type="button"
+                className="btn btn-ghost"
+                disabled={busy}
+                onClick={reject}
+              >
+                Descartar
+              </button>
+            </div>
+          )}
+
+          {err && <div className="state-box state-error">{err}</div>}
+        </div>
+      )}
+    </div>
+  );
+}

apps/web/src/app/recibos/page.tsx

@@ -0,0 +1,15 @@
+import { AppShell } from "@/components/AppShell";
+import { Captura } from "@/components/Captura";
+
+/**
+ * Same capture screen as `/estado-cuenta/lote`, opened on the automatic
+ * (scanned recibos + OCR) mode. Kept as its own route so links from a batch
+ * review page and older bookmarks land on the right tab.
+ */
+export default function RecibosPage() {
+  return (
+    <AppShell>
+      <Captura initialMode="auto" />
+    </AppShell>
+  );
+}

apps/web/src/components/AppShell.tsx

@@ -24,7 +24,15 @@ import type { AuthUser, Ability } from "@/lib/types";
  * abilities. Used by every authenticated page.
  */
 
-type NavLink = { href: string; label: string; ability?: Ability; exact?: boolean };
+type NavLink = {
+  href: string;
+  label: string;
+  ability?: Ability;
+  exact?: boolean;
+  /** Extra path prefixes that belong to this entry (e.g. a second route into
+   *  the same screen), so they highlight it instead of nothing. */
+  aliases?: string[];
+};
 type NavEntry =
   | ({ kind: "link" } & NavLink)
   | { kind: "group"; label: string; items: NavLink[] };
@@ -45,8 +53,16 @@ const NAV: NavEntry[] = [
     label: "Cobranza",
     items: [
       // Daily data-entry screen (the legacy "Editor"). Hidden from VIEWER, who
-      // can't capture anyway — the page itself also refuses.
+      // can't capture anyway — the page itself also refuses. Both capture modes
-      { href: "/estado-cuenta/lote", label: "Captura", ability: "ledger:create" },
+      // live behind this one entry: keying receipts by hand, and scanning a
+      // stack of bills for OCR (the `/recibos` route opens the same screen on
+      // its automatic tab).
+      {
+        href: "/estado-cuenta/lote",
+        label: "Captura",
+        ability: "ledger:create",
+        aliases: ["/recibos"],
+      },
       { href: "/estado-cuenta", label: "Estado de cuenta" },
       { href: "/banco", label: "Chequera" },
     ],
@@ -98,9 +114,11 @@ function activeHref(pathname: string | null): string | null {
   if (!pathname) return null;
   let best: string | null = null;
   for (const item of NAV_LINKS) {
+    const under = (href: string) =>
+      pathname === href || pathname.startsWith(`${href}/`);
     const match = item.exact
       ? pathname === item.href
-      : pathname === item.href || pathname.startsWith(`${item.href}/`);
+      : under(item.href) || (item.aliases?.some(under) ?? false);
     if (match && (best === null || item.href.length > best.length)) {
       best = item.href;
     }

apps/web/src/components/Captura.tsx

@@ -0,0 +1,90 @@
+"use client";
+
+import { useState } from "react";
+import Link from "next/link";
+import { ManualCheckCapture } from "@/components/ManualCheckCapture";
+import { StatementIntake } from "@/components/StatementIntake";
+import { useCan } from "@/lib/abilities";
+
+/**
+ * The daily capture screen (the legacy "Editor"), with two ways in:
+ *
+ * - **manual** — key each customer's receipt against one check by hand.
+ * - **auto** — scan the stack of paper bills and let OCR propose customer and
+ *   amount for every page, which a human still confirms.
+ *
+ * Both end in the same place: charges on the customer's ledger, posted against
+ * one check. They are modes of one screen rather than two menu entries because
+ * it is one job — staff pick the mode by what's on the desk that morning, a
+ * stack of bills or a keyboard.
+ *
+ * `/estado-cuenta/lote` opens on manual, `/recibos` opens on auto; both render
+ * this component, so an old bookmark still lands on the right tab.
+ */
+
+export type CaptureMode = "manual" | "auto";
+
+const MODE_HINT: Record<CaptureMode, string> = {
+  manual:
+    "Captura los recibos de varios clientes contra un mismo cheque y concilia el total antes de guardar.",
+  auto: "Escanea los recibos del mes y el sistema propone cliente e importe para cada página. Nada se registra sin tu confirmación.",
+};
+
+export function Captura({ initialMode = "manual" }: { initialMode?: CaptureMode }) {
+  const canCapture = useCan("ledger:create");
+  const canIngest = useCan("statement:ingest");
+
+  // Gating is cosmetic (the API enforces every write), but a user who only has
+  // one of the two abilities should land on the mode they can actually use.
+  const modes: { key: CaptureMode; label: string }[] = [
+    ...(canCapture ? [{ key: "manual" as const, label: "Captura manual" }] : []),
+    ...(canIngest
+      ? [{ key: "auto" as const, label: "Captura automática (OCR)" }]
+      : []),
+  ];
+
+  const [mode, setMode] = useState<CaptureMode>(
+    modes.some((m) => m.key === initialMode) ? initialMode : (modes[0]?.key ?? "manual"),
+  );
+
+  if (modes.length === 0) {
+    return (
+      <div className="state-box state-error">
+        No tienes permiso para capturar movimientos.
+      </div>
+    );
+  }
+
+  return (
+    <>
+      <div className="page-head">
+        <div>
+          <h1 className="page-title">Captura</h1>
+          <p className="eyebrow">{MODE_HINT[mode]}</p>
+        </div>
+        <Link href="/estado-cuenta" className="btn btn-outline">
+          Volver a estado de cuenta
+        </Link>
+      </div>
+
+      {modes.length > 1 && (
+        <div className="seg" role="tablist" style={{ marginBottom: 16 }}>
+          {modes.map((m) => (
+            <button
+              key={m.key}
+              type="button"
+              role="tab"
+              aria-selected={mode === m.key}
+              className={`seg-btn ${mode === m.key ? "active" : ""}`}
+              onClick={() => setMode(m.key)}
+            >
+              {m.label}
+            </button>
+          ))}
+        </div>
+      )}
+
+      {mode === "manual" ? <ManualCheckCapture /> : <StatementIntake />}
+    </>
+  );
+}

apps/web/src/components/ManualCheckCapture.tsx

@@ -0,0 +1,526 @@
+"use client";
+
+import { useEffect, useMemo, useState } from "react";
+import Link from "next/link";
+import { CustomerPicker } from "@/components/CustomerPicker";
+import { createMovementBatch, getBillingFacets, getByCheck } from "@/lib/api";
+import { formatMoney, formatNumber, txTypeLabel } from "@/lib/labels";
+import type {
+  BatchCreateInput,
+  BillingFacets,
+  ByCheckResponse,
+  Currency,
+  LedgerCurrency,
+  TransactionDomain,
+} from "@/lib/types";
+
+/**
+ * Batch capture by check — the "Editor" screen from the legacy system
+ * (docs/RECEIPT_CAPTURE_SPEC.md §1.2), and the manual half of the Captura
+ * screen (see `Captura`).
+ *
+ * Staff key many customers' receipts against ONE physical check before cutting
+ * it, then check that the captured total matches the check's amount. That
+ * reconciliation is the whole point, so the running total is the most prominent
+ * thing on the page and an optional "importe del cheque" field turns it into a
+ * live difference.
+ *
+ * No batch entity is persisted: `checkNumber` is a plain column, and grouping
+ * by it answers every by-check question (see the "Reporte por cheque" report).
+ */
+
+const DOMAINS: { key: TransactionDomain; label: string }[] = [
+  { key: "UTILITY", label: "Servicios" },
+  { key: "INSURANCE", label: "Seguros" },
+  { key: "TRUST", label: "Fideicomiso" },
+];
+
+interface Line {
+  /** Local row key — lines have no server identity until the batch posts. */
+  key: number;
+  customerId: string;
+  customerName: string;
+  amount: string;
+  reference: string;
+  period: string;
+  outstanding: boolean;
+}
+
+function blankLine(key: number): Line {
+  return {
+    key,
+    customerId: "",
+    customerName: "",
+    amount: "",
+    reference: "",
+    period: "",
+    outstanding: false,
+  };
+}
+
+export function ManualCheckCapture() {
+  const [facets, setFacets] = useState<BillingFacets | null>(null);
+
+  // Check-level fields — shared by every line.
+  const [domain, setDomain] = useState<TransactionDomain>("UTILITY");
+  const [currency, setCurrency] = useState<LedgerCurrency>("MXN");
+  const [typeId, setTypeId] = useState("");
+  const [checkNumber, setCheckNumber] = useState("");
+  const [transactionDate, setTransactionDate] = useState(
+    new Date().toISOString().slice(0, 10),
+  );
+  /** The physical check's amount, for reconciliation only — never submitted. */
+  const [checkAmount, setCheckAmount] = useState("");
+
+  const [lines, setLines] = useState<Line[]>([blankLine(1), blankLine(2), blankLine(3)]);
+  const [nextKey, setNextKey] = useState(4);
+
+  const [saving, setSaving] = useState(false);
+  const [error, setError] = useState<string | null>(null);
+  const [posted, setPosted] = useState<ByCheckResponse | null>(null);
+
+  useEffect(() => {
+    getBillingFacets().then(setFacets).catch(() => setFacets(null));
+  }, []);
+
+  const filled = lines.filter(
+    (l) => l.customerId && l.amount.trim() !== "" && Number.isFinite(Number(l.amount)),
+  );
+
+  // Charges are captured as positive numbers and signed on submit, matching
+  // MovementForm — staff type what's on the bill, not a negative.
+  const total = useMemo(
+    () =>
+      filled
+        .filter((l) => !l.outstanding)
+        .reduce((sum, l) => sum + Math.abs(Number(l.amount)), 0),
+    [filled],
+  );
+  const outstandingTotal = useMemo(
+    () =>
+      filled
+        .filter((l) => l.outstanding)
+        .reduce((sum, l) => sum + Math.abs(Number(l.amount)), 0),
+    [filled],
+  );
+
+  const checkAmt = Number(checkAmount);
+  const hasCheckAmt = checkAmount.trim() !== "" && Number.isFinite(checkAmt);
+  const diff = hasCheckAmt ? checkAmt - total : 0;
+  const reconciled = hasCheckAmt && Math.abs(diff) < 0.005;
+
+  function update(key: number, patch: Partial<Line>) {
+    setLines((ls) => ls.map((l) => (l.key === key ? { ...l, ...patch } : l)));
+  }
+
+  function addLine() {
+    setLines((ls) => [...ls, blankLine(nextKey)]);
+    setNextKey((k) => k + 1);
+  }
+
+  function removeLine(key: number) {
+    setLines((ls) => (ls.length === 1 ? ls : ls.filter((l) => l.key !== key)));
+  }
+
+  async function submit(e: React.FormEvent) {
+    e.preventDefault();
+    if (!checkNumber.trim()) {
+      setError("Indica el número de cheque.");
+      return;
+    }
+    if (filled.length === 0) {
+      setError("Captura al menos una línea con cliente y monto.");
+      return;
+    }
+    const dupes = filled
+      .map((l) => l.customerId)
+      .filter((id, i, arr) => arr.indexOf(id) !== i);
+    if (dupes.length) {
+      const names = filled
+        .filter((l) => dupes.includes(l.customerId))
+        .map((l) => l.customerName);
+      if (
+        !window.confirm(
+          `Hay más de una línea para el mismo cliente (${[...new Set(names)].join(
+            ", ",
+          )}). ¿Continuar?`,
+        )
+      )
+        return;
+    }
+
+    const payload: BatchCreateInput = {
+      domain,
+      transactionDate,
+      checkNumber: checkNumber.trim(),
+      currency: currency as Currency,
+      typeId: typeId || undefined,
+      lines: filled.map((l) => ({
+        customerId: l.customerId,
+        // Every line of a check batch is a charge the office paid out.
+        amount: -Math.abs(Number(l.amount)),
+        reference: l.reference.trim() || undefined,
+        period: l.period.trim() || undefined,
+        outstanding: l.outstanding || undefined,
+      })),
+    };
+
+    setSaving(true);
+    setError(null);
+    try {
+      await createMovementBatch(payload);
+      // Re-read through the by-check view so the confirmation shows what's
+      // actually stored (including anything captured against this check
+      // earlier), not just what this request sent.
+      setPosted(await getByCheck(payload.checkNumber));
+    } catch (e2) {
+      setError((e2 as Error)?.message ?? "No se pudo guardar el lote.");
+    } finally {
+      setSaving(false);
+    }
+  }
+
+  function reset() {
+    setPosted(null);
+    setLines([blankLine(nextKey), blankLine(nextKey + 1), blankLine(nextKey + 2)]);
+    setNextKey((k) => k + 3);
+    setCheckNumber("");
+    setCheckAmount("");
+  }
+
+  if (posted) {
+    return (
+      <>
+        <div className="page-head">
+          <div>
+            <h2 className="page-title">Lote capturado</h2>
+            <p className="eyebrow">
+              Cheque {posted.checkNumber} · {formatNumber(posted.count)}{" "}
+              {posted.count === 1 ? "movimiento" : "movimientos"}
+            </p>
+          </div>
+          <div style={{ display: "flex", gap: 10 }}>
+            <button type="button" className="btn btn-primary" onClick={reset}>
+              Capturar otro cheque
+            </button>
+            <Link href="/estado-cuenta" className="btn btn-outline">
+              Volver a estado de cuenta
+            </Link>
+          </div>
+        </div>
+
+        <div className="filtered-totals" style={{ marginBottom: 16 }}>
+          {posted.totals.map((t) => (
+            <div className="filtered-total" key={t.currency}>
+              <span className="filtered-total-cur">{t.currency}</span>
+              <span className="filtered-total-net">
+                Total del cheque <strong>{formatMoney(t.total, t.currency)}</strong>
+              </span>
+              <span>{formatNumber(t.count)} movimientos</span>
+            </div>
+          ))}
+          {posted.outstandingCount > 0 && (
+            <div className="filtered-total">
+              <span>
+                {formatNumber(posted.outstandingCount)} sin fondos (no suman al
+                total)
+              </span>
+            </div>
+          )}
+        </div>
+
+        <div className="tx-scroll">
+          <table className="tx-table">
+            <thead>
+              <tr>
+                <th>Cliente</th>
+                <th>Referencia</th>
+                <th>Periodo</th>
+                <th>Estado</th>
+                <th className="num">Monto</th>
+              </tr>
+            </thead>
+            <tbody>
+              {posted.items.map((i) => (
+                <tr key={i.id}>
+                  <td>
+                    <Link
+                      href={`/estado-cuenta/${i.customerId}`}
+                      className="inline-link"
+                    >
+                      {i.customerName}
+                    </Link>
+                  </td>
+                  <td>{i.reference || "—"}</td>
+                  <td>{i.period || "—"}</td>
+                  <td>{i.outstanding ? "Sin fondos" : "Pagado"}</td>
+                  <td className="num">
+                    <span className="tx-amount neg">
+                      {formatMoney(i.amount, i.currency)}
+                    </span>
+                  </td>
+                </tr>
+              ))}
+            </tbody>
+          </table>
+        </div>
+
+        <p className="muted" style={{ marginTop: 14 }}>
+          Para imprimir la conciliación, usa el reporte{" "}
+          <Link
+            href={`/reportes/cheque-count?checkNumber=${encodeURIComponent(
+              posted.checkNumber,
+            )}`}
+            className="inline-link"
+          >
+            Reporte por cheque
+          </Link>
+          .
+        </p>
+      </>
+    );
+  }
+
+  return (
+    <>
+      {error && <div className="state-box state-error">{error}</div>}
+
+      <form onSubmit={submit}>
+        <div className="card" style={{ padding: 20, marginBottom: 16 }}>
+          <h2 className="section-title" style={{ marginBottom: 14 }}>
+            Datos del cheque
+          </h2>
+          <div className="form-grid">
+            <label className="field">
+              <span className="field-label">Número de cheque *</span>
+              <input
+                className="input"
+                value={checkNumber}
+                onChange={(e) => setCheckNumber(e.target.value)}
+                required
+              />
+            </label>
+            <label className="field">
+              <span className="field-label">Fecha *</span>
+              <input
+                className="input"
+                type="date"
+                required
+                value={transactionDate}
+                onChange={(e) => setTransactionDate(e.target.value)}
+              />
+            </label>
+            <label className="field">
+              <span className="field-label">Línea de negocio *</span>
+              <select
+                className="select"
+                value={domain}
+                onChange={(e) => setDomain(e.target.value as TransactionDomain)}
+              >
+                {DOMAINS.map((d) => (
+                  <option key={d.key} value={d.key}>
+                    {d.label}
+                  </option>
+                ))}
+              </select>
+            </label>
+            <label className="field">
+              <span className="field-label">Moneda *</span>
+              <select
+                className="select"
+                value={currency}
+                onChange={(e) => setCurrency(e.target.value as LedgerCurrency)}
+              >
+                <option value="MXN">Pesos (MXN)</option>
+                <option value="USD">Dólares (USD)</option>
+              </select>
+            </label>
+            <label className="field">
+              <span className="field-label">Concepto</span>
+              <select
+                className="select"
+                value={typeId}
+                onChange={(e) => setTypeId(e.target.value)}
+              >
+                <option value="">(sin concepto)</option>
+                {facets?.types.map((t) => (
+                  <option key={t.id} value={t.id}>
+                    {txTypeLabel({ nameEn: t.name })}
+                  </option>
+                ))}
+              </select>
+            </label>
+            <label className="field">
+              <span className="field-label">Importe del cheque</span>
+              <input
+                className="input"
+                type="number"
+                step="0.01"
+                min="0"
+                value={checkAmount}
+                onChange={(e) => setCheckAmount(e.target.value)}
+                placeholder="Para conciliar"
+              />
+            </label>
+          </div>
+        </div>
+
+        <div className="card" style={{ padding: 20, marginBottom: 16 }}>
+          <div
+            style={{
+              display: "flex",
+              justifyContent: "space-between",
+              alignItems: "center",
+              marginBottom: 14,
+            }}
+          >
+            <h2 className="section-title" style={{ margin: 0 }}>
+              Recibos ({formatNumber(filled.length)})
+            </h2>
+            <button type="button" className="btn btn-outline" onClick={addLine}>
+              Agregar línea
+            </button>
+          </div>
+
+          <div className="tx-scroll">
+            <table className="tx-table">
+              <thead>
+                <tr>
+                  <th style={{ minWidth: 240 }}>Cliente *</th>
+                  <th style={{ minWidth: 120 }}>Referencia</th>
+                  <th style={{ minWidth: 100 }}>Periodo</th>
+                  <th style={{ minWidth: 110 }} className="num">
+                    Monto *
+                  </th>
+                  <th style={{ whiteSpace: "nowrap" }}>Sin fondos</th>
+                  <th style={{ width: 1 }} />
+                </tr>
+              </thead>
+              <tbody>
+                {lines.map((l) => (
+                  <tr key={l.key}>
+                    <td>
+                      <CustomerPicker
+                        value={l.customerId}
+                        valueName={l.customerId ? l.customerName : undefined}
+                        onPick={(id, name) =>
+                          update(l.key, { customerId: id, customerName: name })
+                        }
+                      />
+                    </td>
+                    <td>
+                      <input
+                        className="input"
+                        value={l.reference}
+                        onChange={(e) =>
+                          update(l.key, { reference: e.target.value })
+                        }
+                      />
+                    </td>
+                    <td>
+                      <input
+                        className="input"
+                        value={l.period}
+                        onChange={(e) => update(l.key, { period: e.target.value })}
+                        placeholder="2026-07"
+                      />
+                    </td>
+                    <td>
+                      <input
+                        className="input num"
+                        type="number"
+                        step="0.01"
+                        min="0"
+                        value={l.amount}
+                        onChange={(e) => update(l.key, { amount: e.target.value })}
+                        placeholder="0.00"
+                      />
+                    </td>
+                    <td style={{ textAlign: "center" }}>
+                      <input
+                        type="checkbox"
+                        checked={l.outstanding}
+                        onChange={(e) =>
+                          update(l.key, { outstanding: e.target.checked })
+                        }
+                        aria-label="Sin fondos"
+                      />
+                    </td>
+                    <td>
+                      <button
+                        type="button"
+                        className="btn btn-ghost"
+                        style={{ padding: "4px 10px", fontSize: 12 }}
+                        onClick={() => removeLine(l.key)}
+                        disabled={lines.length === 1}
+                      >
+                        Quitar
+                      </button>
+                    </td>
+                  </tr>
+                ))}
+              </tbody>
+            </table>
+          </div>
+        </div>
+
+        <div className="card" style={{ padding: 20, marginBottom: 16 }}>
+          <h2 className="section-title" style={{ marginBottom: 14 }}>
+            Conciliación
+          </h2>
+          <div className="filtered-totals">
+            <div className="filtered-total">
+              <span className="filtered-total-cur">{currency}</span>
+              <span className="filtered-total-net">
+                Capturado <strong>{formatMoney(String(-total), currency)}</strong>
+              </span>
+              <span>{formatNumber(filled.filter((l) => !l.outstanding).length)} recibos</span>
+            </div>
+            {outstandingTotal > 0 && (
+              <div className="filtered-total">
+                <span>
+                  Sin fondos{" "}
+                  <strong>{formatMoney(String(-outstandingTotal), currency)}</strong>{" "}
+                  (no suma al cheque)
+                </span>
+              </div>
+            )}
+            {hasCheckAmt && (
+              <div className="filtered-total">
+                <span className="filtered-total-net">
+                  {reconciled ? (
+                    <strong className="tx-amount pos">Cuadra con el cheque</strong>
+                  ) : (
+                    <>
+                      Diferencia{" "}
+                      <strong className="tx-amount neg">
+                        {formatMoney(String(diff), currency)}
+                      </strong>
+                    </>
+                  )}
+                </span>
+              </div>
+            )}
+          </div>
+        </div>
+
+        <div className="form-actions">
+          <button
+            type="submit"
+            className="btn btn-primary"
+            disabled={saving || filled.length === 0}
+          >
+            {saving
+              ? "Guardando…"
+              : `Capturar ${formatNumber(filled.length)} ${
+                  filled.length === 1 ? "recibo" : "recibos"
+                }`}
+          </button>
+          <Link href="/estado-cuenta" className="btn btn-outline">
+            Cancelar
+          </Link>
+        </div>
+      </form>
+    </>
+  );
+}

apps/web/src/components/PolicyCaptura.tsx

@@ -0,0 +1,124 @@
+"use client";
+
+import { useState } from "react";
+import Link from "next/link";
+import { useSearchParams } from "next/navigation";
+import { PolicyForm } from "@/components/PolicyForm";
+import { PolicyOcrIntake } from "@/components/PolicyOcrIntake";
+import { useCan } from "@/lib/abilities";
+
+/**
+ * Policy intake — mirror of `Captura.tsx` (statement OCR side): one screen,
+ * two ways in:
+ *
+ * - **manual** — `PolicyForm` keys every field by hand.
+ * - **auto** — `PolicyOcrIntake` uploads a GMX 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
+ * live as two modes of one screen rather than two menu entries — exactly the
+ * same shape Captura uses for `ManualCheckCapture` vs `StatementIntake`.
+ *
+ * `/polizas/nuevo` opens manual, `/polizas/captura` opens auto; both render
+ * this component so the tab toggle works either way and an old bookmark
+ * still lands on the right tab.
+ */
+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.",
+};
+
+export function PolicyCaptura({ initialMode = "manual" }: { initialMode?: PolicyCaptureMode }) {
+  const canCreate = useCan("policy:create");
+  const canIngest = useCan("policy:ingest");
+
+  // `/clientes/[id]` deep-links into /polizas/nuevo with the customer
+  // pre-picked so staff can fill the rest without retyping. The OCR pane
+  // ignores these — there's no customer to lock in until the batch is
+  // confirmed.
+  const params = useSearchParams();
+  const fixedCustomerId = params.get("customerId") ?? undefined;
+  const fixedCustomerName = params.get("customerName") ?? undefined;
+
+  // One user can land on either mode. The tab strip only renders when both
+  // abilities are held — a STAFF with only policy:ingest (no create) still
+  // sees the screen but only the OCR tab is offered.
+  const modes: { key: PolicyCaptureMode; label: string }[] = [
+    ...(canCreate ? [{ key: "manual" as const, label: "Captura manual" }] : []),
+    ...(canIngest ? [{ key: "auto" as const, label: "Captura automática (OCR)" }] : []),
+  ];
+
+  const [mode, setMode] = useState<PolicyCaptureMode>(
+    modes.some((m) => m.key === initialMode) ? initialMode : (modes[0]?.key ?? "manual"),
+  );
+
+  if (modes.length === 0) {
+    return (
+      <div className="state-box state-error">
+        No tienes permiso para crear ni capturar pólizas.
+      </div>
+    );
+  }
+
+  return (
+    <>
+      <div className="page-head">
+        <Link href="/polizas" className="back-link">← Pólizas</Link>
+        <h1 className="page-title">Nueva póliza</h1>
+        <p className="eyebrow">{MODE_HINT[mode]}</p>
+      </div>
+
+      {modes.length > 1 && (
+        <div className="seg" role="tablist" style={{ marginBottom: 16 }}>
+          {modes.map((m) => (
+            <button
+              key={m.key}
+              type="button"
+              role="tab"
+              aria-selected={mode === m.key}
+              className={`seg-btn ${mode === m.key ? "active" : ""}`}
+              onClick={() => setMode(m.key)}
+            >
+              {m.label}
+            </button>
+          ))}
+        </div>
+      )}
+
+      {mode === "manual" ? (
+        <ManualPane
+          fixedCustomerId={fixedCustomerId}
+          fixedCustomerName={fixedCustomerName}
+        />
+      ) : (
+        <PolicyOcrIntake />
+      )}
+    </>
+  );
+}
+
+function ManualPane({
+  fixedCustomerId,
+  fixedCustomerName,
+}: {
+  fixedCustomerId?: string;
+  fixedCustomerName?: string;
+}) {
+  const allowed = useCan("policy:create");
+  if (!allowed) {
+    return (
+      <div className="state-box state-error">
+        No tiene permisos para crear pólizas.
+      </div>
+    );
+  }
+  return (
+    <PolicyForm
+      fixedCustomerId={fixedCustomerId}
+      fixedCustomerName={fixedCustomerName}
+    />
+  );
+}

apps/web/src/components/PolicyOcrIntake.tsx

@@ -0,0 +1,232 @@
+"use client";
+
+import { useCallback, useEffect, useState } from "react";
+import Link from "next/link";
+import {
+  getPolicyOcrStatus,
+  listPolicyOcrBatches,
+  uploadPolicyOcrBatch,
+} from "@/lib/api";
+import { useCan } from "@/lib/abilities";
+import { formatDate } from "@/lib/labels";
+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.
+ *
+ * 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
+ * hand that day, mixed service vs insurance), and the matching/review queue
+ * already keys on the policyNumber → existing Policy transition that the
+ * rest of /polizas owns.
+ */
+
+const STATUS_LABEL: Record<PolicyOcrBatchStatus, string> = {
+  UPLOADED: "Recibido",
+  PROCESSING: "Procesando…",
+  READY_FOR_REVIEW: "Listo para revisar",
+  COMPLETED: "Aplicado",
+  FAILED: "Falló",
+};
+
+export function PolicyOcrIntake() {
+  const canIngest = useCan("policy:ingest");
+  const [batches, setBatches] = useState<PolicyOcrBatch[]>([]);
+  const [ocrAvailable, setOcrAvailable] = useState<boolean | null>(null);
+  const [storageAvailable, setStorageAvailable] = useState<boolean | null>(null);
+  const [loading, setLoading] = useState(true);
+  const [error, setError] = useState<string | null>(null);
+
+  const load = useCallback(async () => {
+    try {
+      const [list, status] = await Promise.all([
+        listPolicyOcrBatches(),
+        getPolicyOcrStatus(),
+      ]);
+      setBatches(list.items);
+      setOcrAvailable(status.ocrAvailable);
+      setStorageAvailable(status.storageAvailable);
+      setError(null);
+    } catch (e) {
+      setError((e as Error)?.message ?? "No se pudieron cargar los lotes.");
+    } finally {
+      setLoading(false);
+    }
+  }, []);
+
+  useEffect(() => {
+    void load();
+  }, [load]);
+
+  const working = batches.some(
+    (b) => b.status === "PROCESSING" || b.status === "UPLOADED",
+  );
+  useEffect(() => {
+    if (!working) return;
+    const t = setInterval(() => void load(), 4000);
+    return () => clearInterval(t);
+  }, [working, load]);
+
+  const ready = ocrAvailable === true && storageAvailable === true;
+
+  return (
+    <div className="stack">
+      {ocrAvailable === false && (
+        <div className="state-box state-error">
+          Este servidor no tiene OCR instalado, así que no se pueden leer PDFs
+          de pólizas escaneados. La captura manual sigue funcionando.
+        </div>
+      )}
+
+      {storageAvailable === false && (
+        <div className="state-box state-error">
+          Este servidor no tiene configurado el almacenamiento de documentos, así
+          que no hay dónde guardar los PDFs. Mientras tanto, capture las
+          pólizas a mano.
+        </div>
+      )}
+
+      {canIngest && ready && <UploadCard onDone={load} />}
+
+      {error && <div className="state-box state-error">{error}</div>}
+
+      <section className="card" style={{ padding: 16 }}>
+        <h2 className="section-title" style={{ marginTop: 0 }}>
+          Lotes
+        </h2>
+        {loading ? (
+          <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.
+          </div>
+        ) : (
+          <div className="tx-scroll">
+            <table className="tx-table">
+              <thead>
+                <tr>
+                  <th>Fecha</th>
+                  <th>Aseguradora</th>
+                  <th>Referencia</th>
+                  <th>Estado</th>
+                  <th className="num">Páginas</th>
+                  <th>Subido por</th>
+                  <th />
+                </tr>
+              </thead>
+              <tbody>
+                {batches.map((b) => (
+                  <tr key={b.id}>
+                    <td style={{ whiteSpace: "nowrap" }}>{formatDate(b.createdAt)}</td>
+                    <td>{b.provider}</td>
+                    <td>{b.label || "—"}</td>
+                    <td>
+                      <StatusTag status={b.status} />
+                      {b.error && (
+                        <div className="page-sub" style={{ marginTop: 4 }}>
+                          {b.error}
+                        </div>
+                      )}
+                    </td>
+                    <td className="num">{b._count?.documents ?? 0}</td>
+                    <td>{b.uploadedBy?.name ?? "—"}</td>
+                    <td>
+                      <Link
+                        className="btn btn-ghost"
+                        href={`/polizas/captura/${b.id}`}
+                      >
+                        Revisar
+                      </Link>
+                    </td>
+                  </tr>
+                ))}
+              </tbody>
+            </table>
+          </div>
+        )}
+      </section>
+    </div>
+  );
+}
+
+function StatusTag({ status }: { status: PolicyOcrBatchStatus }) {
+  return <span className="tag">{STATUS_LABEL[status] ?? status}</span>;
+}
+
+function UploadCard({ onDone }: { onDone: () => void }) {
+  const [files, setFiles] = useState<File[]>([]);
+  const [label, setLabel] = useState("");
+  const [busy, setBusy] = useState(false);
+  const [error, setError] = useState<string | null>(null);
+
+  async function submit() {
+    if (!files.length) return;
+    setBusy(true);
+    setError(null);
+    try {
+      await uploadPolicyOcrBatch(files, label.trim() || undefined);
+      setFiles([]);
+      setLabel("");
+      onDone();
+    } catch (e) {
+      setError((e as Error)?.message ?? "No se pudo subir el lote.");
+    } finally {
+      setBusy(false);
+    }
+  }
+
+  return (
+    <section className="card" style={{ padding: 16 }}>
+      <h2 className="section-title" style={{ marginTop: 0 }}>
+        Subir PDFs de pólizas (GMX)
+      </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"
+            value={label}
+            onChange={(e) => setLabel(e.target.value)}
+          />
+        </label>
+
+        <label>
+          <span className="page-sub">Archivos PDF</span>
+          <input
+            type="file"
+            className="input"
+            accept="application/pdf"
+            multiple
+            onChange={(e) => setFiles(Array.from(e.target.files ?? []))}
+          />
+        </label>
+
+        <button
+          type="button"
+          className="btn btn-primary"
+          disabled={!files.length || busy}
+          onClick={submit}
+        >
+          {busy ? "Subiendo…" : `Procesar ${files.length || ""}`.trim()}
+        </button>
+      </div>
+
+      {error && (
+        <div className="state-box state-error" style={{ marginTop: 12 }}>
+          {error}
+        </div>
+      )}
+
+      <p className="page-sub" style={{ marginTop: 12 }}>
+        Un lote puede traer varios PDFs. Cada página se procesa por separado; el
+        sistema busca una póliza existente por número y, si no la encuentra,
+        propone crear una nueva bajo el cliente que se elija en la revisión.
+      </p>
+    </section>
+  );
+}

apps/web/src/components/PolicyOcrReview.tsx

@@ -0,0 +1,578 @@
+"use client";
+
+import { useCallback, useEffect, useMemo, useState } from "react";
+import Link from "next/link";
+import { CustomerPicker } from "@/components/CustomerPicker";
+import {
+  confirmPolicyOcrBatch,
+  getPolicyOcrBatch,
+  listCustomers,
+  listPolicyOcrDocuments,
+  policyOcrDocumentUrl,
+  rejectPolicyOcrDocument,
+  reviewPolicyOcrDocument,
+} from "@/lib/api";
+import { useCan } from "@/lib/abilities";
+import { formatDate, formatMoney } from "@/lib/labels";
+import type {
+  CustomerListItem,
+  PolicyOcrBatchDetail,
+  PolicyOcrConfirmDocument,
+  PolicyOcrCoverage,
+  PolicyOcrDocument,
+  PolicyOcrReviewInput,
+} from "@/lib/types";
+
+const STATUS_LABEL: Record<string, string> = {
+  PENDING_OCR: "Pendiente",
+  OCR_FAILED: "Falló OCR",
+  NEEDS_REVIEW: "Para revisar",
+  MATCHED: "Listo",
+  CONFIRMED: "Confirmado",
+  POSTED: "Aplicado",
+  REJECTED: "Rechazado",
+};
+
+const OPEN_FIRST = [
+  "NEEDS_REVIEW",
+  "MATCHED",
+  "CONFIRMED",
+  "PENDING_OCR",
+  "OCR_FAILED",
+  "REJECTED",
+  "POSTED",
+];
+
+type EditMap = Record<string, PolicyOcrConfirmDocument | undefined>;
+
+export function PolicyOcrReview({ id }: { id: string }) {
+  const canReview = useCan("policy:ocr-review");
+  const [batch, setBatch] = useState<PolicyOcrBatchDetail | null>(null);
+  const [docs, setDocs] = useState<PolicyOcrDocument[]>([]);
+  const [edits, setEdits] = useState<EditMap>({});
+  const [customerIndex, setCustomerIndex] = useState<Record<string, CustomerListItem>>({});
+  const [error, setError] = useState<string | null>(null);
+  const [loading, setLoading] = useState(true);
+  const [submitting, setSubmitting] = useState(false);
+
+  const load = useCallback(async () => {
+    try {
+      const [b, d, c] = await Promise.all([
+        getPolicyOcrBatch(id),
+        listPolicyOcrDocuments(id),
+        canReview ? listCustomers({ pageSize: 200 }).then((r) => r.items) : Promise.resolve([]),
+      ]);
+      setBatch(b);
+      setDocs(d);
+      setCustomerIndex(Object.fromEntries(c.map((x) => [x.id, x])));
+      setError(null);
+    } catch (e) {
+      setError((e as Error)?.message ?? "No se pudo cargar el lote.");
+    } finally {
+      setLoading(false);
+    }
+  }, [id, canReview]);
+
+  useEffect(() => {
+    void load();
+  }, [load]);
+
+  const processing = batch?.status === "PROCESSING" || batch?.status === "UPLOADED";
+  useEffect(() => {
+    if (!processing) return;
+    const t = setInterval(() => void load(), 4000);
+    return () => clearInterval(t);
+  }, [processing, load]);
+
+  const sorted = useMemo(
+    () =>
+      [...docs].sort(
+        (a, b) =>
+          OPEN_FIRST.indexOf(a.status) - OPEN_FIRST.indexOf(b.status) ||
+          a.pageNumber - b.pageNumber,
+      ),
+    [docs],
+  );
+
+  const readyCount = Object.values(edits).filter(Boolean).length;
+
+  function setEdit(docId: string, edit: PolicyOcrConfirmDocument) {
+    setEdits((prev) => ({ ...prev, [docId]: edit }));
+  }
+
+  async function onConfirm() {
+    if (!batch) return;
+    const payload: PolicyOcrConfirmDocument[] = [];
+    for (const d of docs) {
+      const edit = edits[d.id];
+      if (!edit) continue;
+      if (!edit.policyId && !edit.customerId) {
+        setError(`Página ${d.pageNumber}: falta cliente o póliza destino.`);
+        return;
+      }
+      payload.push(edit);
+    }
+    if (!payload.length) {
+      setError("No hay documentos revisados. Guarde cada página antes de aplicar.");
+      return;
+    }
+    setSubmitting(true);
+    setError(null);
+    try {
+      await confirmPolicyOcrBatch(batch.id, { documents: payload });
+      setEdits({});
+      await load();
+    } catch (e) {
+      setError((e as Error)?.message ?? "No se pudo aplicar el lote.");
+    } finally {
+      setSubmitting(false);
+    }
+  }
+
+  if (loading) return <div className="state-box">Cargando…</div>;
+  if (!batch) return <div className="state-box state-error">{error ?? "No encontrado."}</div>;
+
+  return (
+    <div className="stack">
+      <header className="page-head">
+        <div>
+          <h1 className="page-title">
+            Pólizas — {batch.provider}
+            {batch.label ? ` · ${batch.label}` : ""}
+          </h1>
+          <p className="page-sub">
+            {formatDate(batch.createdAt)} · {docs.length} página(s) ·{" "}
+            {STATUS_LABEL[batch.status] ?? batch.status}
+          </p>
+        </div>
+        <Link className="btn btn-ghost" href="/polizas">
+          Volver a pólizas
+        </Link>
+      </header>
+
+      {processing && <div className="state-box">Procesando…</div>}
+
+      {error && <div className="state-box state-error">{error}</div>}
+
+      {canReview && readyCount > 0 && (
+        <section className="card" style={{ padding: 16 }}>
+          <h2 className="section-title" style={{ marginTop: 0 }}>
+            Aplicar lote
+          </h2>
+          <p className="page-sub" style={{ marginBottom: 12 }}>
+            {readyCount} página(s) revisada(s). Se creará o actualizará la póliza
+            y, si marcó la casilla, se registrará la prima en el estado de
+            cuenta.
+          </p>
+          <button
+            type="button"
+            className="btn btn-primary"
+            disabled={submitting}
+            onClick={onConfirm}
+          >
+            {submitting ? "Aplicando…" : "Aplicar"}
+          </button>
+        </section>
+      )}
+
+      <section className="stack">
+        {sorted.map((doc) => (
+          <DocumentRow
+            key={doc.id}
+            doc={doc}
+            customerIndex={customerIndex}
+            canReview={canReview}
+            onSave={async (edit) => {
+              await reviewPolicyOcrDocument(doc.id, edit.reviewInput);
+              setEdit(doc.id, edit.confirmInput);
+              await load();
+            }}
+            onReject={async () => {
+              await rejectPolicyOcrDocument(doc.id);
+              setEdits((prev) => {
+                const { [doc.id]: _, ...rest } = prev;
+                return rest;
+              });
+              await load();
+            }}
+          />
+        ))}
+      </section>
+    </div>
+  );
+}
+
+interface RowSaved {
+  reviewInput: PolicyOcrReviewInput;
+  confirmInput: PolicyOcrConfirmDocument;
+}
+
+interface DocumentRowProps {
+  doc: PolicyOcrDocument;
+  customerIndex: Record<string, CustomerListItem>;
+  canReview: boolean;
+  onSave: (saved: RowSaved) => Promise<void>;
+  onReject: () => Promise<void>;
+}
+
+function DocumentRow({ doc, customerIndex, canReview, onSave, onReject }: DocumentRowProps) {
+  const [v, setV] = useState({
+    policyNumber: doc.extractedPolicyNumber ?? "",
+    insuredName: doc.extractedInsuredName ?? "",
+    additionalInsured: doc.extractedAdditionalInsured ?? "",
+    agentName: doc.extractedAgentName ?? "",
+    legalAddress: doc.extractedLegalAddress ?? "",
+    zip: doc.extractedZip ?? "",
+    policyFrom: doc.extractedPolicyFrom?.slice(0, 10) ?? "",
+    policyTo: doc.extractedPolicyTo?.slice(0, 10) ?? "",
+    policyDate: doc.extractedPolicyDate?.slice(0, 10) ?? "",
+    currency: doc.extractedCurrency ?? "USD",
+    netPremium: doc.extractedNetPremium ?? "",
+    total: doc.extractedTotal ?? "",
+    premiumPayment: doc.extractedPremiumPayment ?? "",
+    postPremium: doc.extractedNetPremium != null && Number(doc.extractedNetPremium) > 0,
+  });
+  const [customerId, setCustomerId] = useState(
+    doc.matchedCustomer?.id ?? doc.matchedPolicy?.customerId ?? "",
+  );
+  const [customerName, setCustomerName] = useState(
+    doc.matchedCustomer?.name ?? doc.matchedPolicy?.customer.name ?? "",
+  );
+  const [policyId, setPolicyId] = useState(doc.matchedPolicy?.id ?? "");
+  const [busy, setBusy] = useState(false);
+  const [err, setErr] = useState<string | null>(null);
+
+  function set<K extends keyof typeof v>(k: K, val: (typeof v)[K]) {
+    setV((p) => ({ ...p, [k]: val }));
+  }
+
+  async function save() {
+    setBusy(true);
+    setErr(null);
+    try {
+      const numOrUndef = (s: string) => (s.trim() === "" ? undefined : Number(s));
+      const trimOrUndef = (s: string) => (s.trim() === "" ? undefined : s.trim());
+      const currency = v.currency || undefined;
+      const reviewInput: PolicyOcrReviewInput = {
+        policyNumber: trimOrUndef(v.policyNumber),
+        insuredName: trimOrUndef(v.insuredName),
+        additionalInsured: trimOrUndef(v.additionalInsured),
+        agentName: trimOrUndef(v.agentName),
+        legalAddress: trimOrUndef(v.legalAddress),
+        zip: trimOrUndef(v.zip),
+        policyFrom: v.policyFrom || undefined,
+        policyTo: v.policyTo || undefined,
+        policyDate: v.policyDate || undefined,
+        currency,
+        netPremium: numOrUndef(v.netPremium),
+        total: numOrUndef(v.total),
+        premiumPayment: trimOrUndef(v.premiumPayment),
+        matchedPolicyId: policyId || undefined,
+        matchedCustomerId: !policyId && customerId ? customerId : undefined,
+        forceConfirm: true,
+      };
+      const confirmInput: PolicyOcrConfirmDocument = {
+        documentId: doc.id,
+        policyId: policyId || undefined,
+        customerId: !policyId && customerId ? customerId : undefined,
+        policyNumber: reviewInput.policyNumber,
+        insuredName: reviewInput.insuredName,
+        additionalInsured: reviewInput.additionalInsured,
+        agentName: reviewInput.agentName,
+        legalAddress: reviewInput.legalAddress,
+        zip: reviewInput.zip,
+        policyFrom: reviewInput.policyFrom,
+        policyTo: reviewInput.policyTo,
+        policyDate: reviewInput.policyDate,
+        currency: (currency as "MXN" | "USD" | "EUR" | undefined) ?? undefined,
+        netPremium: reviewInput.netPremium,
+        total: reviewInput.total,
+        premiumPayment: reviewInput.premiumPayment,
+        coveragesJson: (doc.extractedCoveragesJson ?? undefined) as
+          | PolicyOcrCoverage[]
+          | undefined,
+        postPremium: v.postPremium,
+      };
+      await onSave({ reviewInput, confirmInput });
+    } catch (e) {
+      setErr((e as Error)?.message ?? "No se pudo guardar.");
+    } finally {
+      setBusy(false);
+    }
+  }
+
+  const locked = doc.status === "POSTED" || doc.status === "REJECTED";
+  const matchedExisting = !!doc.matchedPolicy;
+  const candidates = doc.matchCandidates ?? [];
+
+  return (
+    <article className="card" style={{ padding: 16 }}>
+      <header className="row" style={{ gap: 12, alignItems: "center" }}>
+        <span className="tag">{STATUS_LABEL[doc.status] ?? doc.status}</span>
+        <span className="page-sub">Página {doc.pageNumber}</span>
+        {doc.extractedPolicyNumber && (
+          <strong style={{ marginLeft: 8 }}>{doc.extractedPolicyNumber}</strong>
+        )}
+        {doc.extractedInsuredName && (
+          <span className="page-sub">· {doc.extractedInsuredName}</span>
+        )}
+      </header>
+
+      <div className="doc-detail">
+        {/*
+         * Embed the source PDF the office uploaded. One PDF = one parsed
+         * policy, so the browser's PDF viewer handles multi-page navigation
+         * natively; we don't need to render individual pages on the server.
+         */}
+        <iframe
+          src={policyOcrDocumentUrl(doc.id)}
+          title={`Póliza ${doc.extractedPolicyNumber ?? doc.pageNumber}`}
+          style={{
+            width: "100%",
+            height: 720,
+            border: "1px solid var(--border, #ddd)",
+            borderRadius: 6,
+            background: "#fff",
+          }}
+        />
+
+        <div className="stack" style={{ flex: 1, minWidth: 0 }}>
+          {doc.matchNote && <p className="page-sub">{doc.matchNote}</p>}
+
+          {matchedExisting ? (
+            <div className="state-box">
+              Coincide con la póliza{" "}
+              <strong>{doc.matchedPolicy?.policyNumber}</strong> del cliente{" "}
+              <strong>{doc.matchedPolicy?.customer.name}</strong>.
+            </div>
+          ) : candidates.length > 1 ? (
+            <div className="state-box state-warn">
+              {candidates.length} pólizas comparten este número. Elija
+              manualmente abajo.
+            </div>
+          ) : (
+            <div className="state-box">
+              No se encontró una póliza con este número. Se creará una nueva
+              bajo el cliente que elija abajo.
+            </div>
+          )}
+
+          <fieldset className="form-grid" disabled={locked || !canReview}>
+            <Field label="Número de póliza">
+              <input
+                className="input"
+                value={v.policyNumber}
+                onChange={(e) => set("policyNumber", e.target.value)}
+              />
+            </Field>
+            <Field label="Asegurado">
+              <input
+                className="input"
+                value={v.insuredName}
+                onChange={(e) => set("insuredName", e.target.value)}
+              />
+            </Field>
+            <Field label="Asegurado adicional">
+              <input
+                className="input"
+                value={v.additionalInsured}
+                onChange={(e) => set("additionalInsured", e.target.value)}
+              />
+            </Field>
+            <Field label="Agente">
+              <input
+                className="input"
+                value={v.agentName}
+                onChange={(e) => set("agentName", e.target.value)}
+              />
+            </Field>
+            <Field label="Desde">
+              <input
+                className="input"
+                type="date"
+                value={v.policyFrom}
+                onChange={(e) => set("policyFrom", e.target.value)}
+              />
+            </Field>
+            <Field label="Hasta">
+              <input
+                className="input"
+                type="date"
+                value={v.policyTo}
+                onChange={(e) => set("policyTo", e.target.value)}
+              />
+            </Field>
+            <Field label="Fecha de firma">
+              <input
+                className="input"
+                type="date"
+                value={v.policyDate}
+                onChange={(e) => set("policyDate", e.target.value)}
+              />
+            </Field>
+            <Field label="Moneda">
+              <select
+                className="input select"
+                value={v.currency}
+                onChange={(e) => set("currency", e.target.value)}
+              >
+                <option value="MXN">MXN</option>
+                <option value="USD">USD</option>
+                <option value="EUR">EUR</option>
+              </select>
+            </Field>
+            <Field label="Prima neta">
+              <input
+                className="input"
+                type="number"
+                step="0.01"
+                value={v.netPremium}
+                onChange={(e) => set("netPremium", e.target.value)}
+              />
+            </Field>
+            <Field label="Total">
+              <input
+                className="input"
+                type="number"
+                step="0.01"
+                value={v.total}
+                onChange={(e) => set("total", e.target.value)}
+              />
+            </Field>
+            <Field label="Pago de prima">
+              <input
+                className="input"
+                value={v.premiumPayment}
+                onChange={(e) => set("premiumPayment", e.target.value)}
+              />
+            </Field>
+            <Field label="Dirección">
+              <input
+                className="input"
+                value={v.legalAddress}
+                onChange={(e) => set("legalAddress", e.target.value)}
+              />
+            </Field>
+            <Field label="C.P.">
+              <input
+                className="input"
+                value={v.zip}
+                onChange={(e) => set("zip", e.target.value)}
+              />
+            </Field>
+          </fieldset>
+
+          {doc.extractedCoveragesJson && doc.extractedCoveragesJson.length > 0 && (
+            <details>
+              <summary>
+                Coberturas ({doc.extractedCoveragesJson.length}) ·{" "}
+                {formatMoney(
+                  doc.extractedCoveragesJson
+                    .map((c) => Number(c.insuredAmount ?? 0))
+                    .reduce((a, b) => a + b, 0)
+                    .toString(),
+                  v.currency,
+                )}
+              </summary>
+              <table className="tx-table" style={{ marginTop: 8 }}>
+                <thead>
+                  <tr>
+                    <th>Riesgo</th>
+                    <th className="num">Suma</th>
+                    <th>Deducible</th>
+                    <th>Participación</th>
+                  </tr>
+                </thead>
+                <tbody>
+                  {doc.extractedCoveragesJson.map((c, i) => (
+                    <tr key={i}>
+                      <td>{c.risk}</td>
+                      <td className="num">
+                        {formatMoney(c.insuredAmount?.toString() ?? null, v.currency)}
+                      </td>
+                      <td>{c.deductible ?? "—"}</td>
+                      <td>{c.lossParticipation ?? "—"}</td>
+                    </tr>
+                  ))}
+                </tbody>
+              </table>
+            </details>
+          )}
+
+          {candidates.length > 1 && (
+            <Field label="Póliza destino">
+              <select
+                className="input select"
+                value={policyId}
+                onChange={(e) => {
+                  setPolicyId(e.target.value);
+                  const found = candidates.find((c) => c.policyId === e.target.value);
+                  if (found) {
+                    setCustomerId(found.customerId);
+                    setCustomerName(customerIndex[found.customerId]?.name ?? found.customerName);
+                  }
+                }}
+              >
+                <option value="">— elegir póliza —</option>
+                {candidates.map((c) => (
+                  <option key={c.policyId} value={c.policyId}>
+                    {c.policyNumber} · {c.customerName}
+                  </option>
+                ))}
+              </select>
+            </Field>
+          )}
+
+          {!policyId && (
+            <Field label={matchedExisting ? "Cliente" : "Cliente (póliza nueva)"}>
+              <CustomerPicker
+                value={customerId}
+                valueName={customerName}
+                onPick={(id, name) => {
+                  setCustomerId(id);
+                  setCustomerName(name);
+                }}
+              />
+            </Field>
+          )}
+
+          <label className="field">
+            <input
+              type="checkbox"
+              checked={v.postPremium}
+              onChange={(e) => set("postPremium", e.target.checked)}
+              disabled={!v.netPremium || Number(v.netPremium) <= 0}
+            />{" "}
+            Registrar prima en el estado de cuenta
+          </label>
+
+          {err && <div className="state-box state-error">{err}</div>}
+
+          {!locked && canReview && (
+            <div className="row" style={{ gap: 8 }}>
+              <button type="button" className="btn btn-primary" disabled={busy} onClick={save}>
+                {busy ? "Guardando…" : "Guardar revisión"}
+              </button>
+              <button
+                type="button"
+                className="btn btn-ghost"
+                onClick={() => void onReject()}
+              >
+                Rechazar
+              </button>
+            </div>
+          )}
+        </div>
+      </div>
+    </article>
+  );
+}
+
+function Field({ label, children }: { label: string; children: React.ReactNode }) {
+  return (
+    <label className="field">
+      <span className="field-label">{label}</span>
+      {children}
+    </label>
+  );
+}

apps/web/src/components/StatementIntake.tsx

@@ -0,0 +1,279 @@
+"use client";
+
+import { useCallback, useEffect, useState } from "react";
+import Link from "next/link";
+import {
+  getStatementStatus,
+  listStatementBatches,
+  uploadStatementBatch,
+} from "@/lib/api";
+import { useCan } from "@/lib/abilities";
+import { formatDate, SERVICE_KIND_LABELS, serviceKindLabel } from "@/lib/labels";
+import type { ServiceKind, StatementBatch, StatementBatchStatus } from "@/lib/types";
+
+/**
+ * Automatic capture — statement OCR intake (docs/RECEIPT_CAPTURE_SPEC.md §2).
+ *
+ * Each utility company mails 300+ paper bills a month, one per customer, which
+ * staff otherwise key in by hand on the manual tab of the same "Captura"
+ * screen. Here they scan the stack instead, and the machine proposes customer +
+ * amount for every page; a human still confirms before anything reaches the
+ * ledger. Same daily job, same ledger path — only the input differs, which is
+ * why it lives as a mode of Captura rather than a screen of its own.
+ *
+ * One batch = one service kind, because the matcher is scoped per kind: a
+ * water account number and a phone number are compared against different
+ * columns, and mixing them in one upload is how a bill gets posted to the
+ * wrong customer.
+ */
+
+/** The kinds the parsers actually recognise today. */
+const SUPPORTED: ServiceKind[] = [
+  "ELECTRIC",
+  "WATER",
+  "TELEPHONE",
+  "GAS",
+  "PROPERTY_TAX",
+  "FEDERAL_ZONE",
+];
+/** Uploadable, but every page will land in review until a parser learns it. */
+const OTHER_KINDS: ServiceKind[] = ["CABLE"];
+
+const STATUS_LABEL: Record<StatementBatchStatus, string> = {
+  UPLOADED: "Recibido",
+  PROCESSING: "Procesando…",
+  READY_FOR_REVIEW: "Listo para revisar",
+  COMPLETED: "Registrado",
+  FAILED: "Falló",
+};
+
+export function StatementIntake() {
+  const canIngest = useCan("statement:ingest");
+  const [batches, setBatches] = useState<StatementBatch[]>([]);
+  const [ocrAvailable, setOcrAvailable] = useState<boolean | null>(null);
+  const [storageAvailable, setStorageAvailable] = useState<boolean | null>(null);
+  const [loading, setLoading] = useState(true);
+  const [error, setError] = useState<string | null>(null);
+
+  const load = useCallback(async () => {
+    try {
+      const [list, status] = await Promise.all([
+        listStatementBatches(),
+        getStatementStatus(),
+      ]);
+      setBatches(list.items);
+      setOcrAvailable(status.ocrAvailable);
+      setStorageAvailable(status.storageAvailable);
+      setError(null);
+    } catch (e) {
+      setError((e as Error)?.message ?? "No se pudieron cargar los lotes.");
+    } finally {
+      setLoading(false);
+    }
+  }, []);
+
+  useEffect(() => {
+    void load();
+  }, [load]);
+
+  // A batch of 300 pages takes minutes to OCR, so the list refreshes itself
+  // while anything is still working rather than making staff reload.
+  const working = batches.some(
+    (b) => b.status === "PROCESSING" || b.status === "UPLOADED",
+  );
+  useEffect(() => {
+    if (!working) return;
+    const t = setInterval(() => void load(), 4000);
+    return () => clearInterval(t);
+  }, [working, load]);
+
+  const ready = ocrAvailable === true && storageAvailable === true;
+
+  return (
+    <div className="stack">
+      {ocrAvailable === false && (
+        <div className="state-box state-error">
+          Este servidor no tiene OCR instalado, así que no se pueden leer recibos
+          escaneados. Usa la captura manual; el resto del sistema funciona con
+          normalidad.
+        </div>
+      )}
+
+      {storageAvailable === false && (
+        <div className="state-box state-error">
+          Este servidor no tiene configurado el almacenamiento de documentos, así
+          que no hay dónde guardar los recibos escaneados. Usa la captura manual
+          mientras se configura.
+        </div>
+      )}
+
+      {canIngest && ready && <UploadCard onDone={load} />}
+
+      {error && <div className="state-box state-error">{error}</div>}
+
+      <section className="card" style={{ padding: 16 }}>
+        <h2 className="section-title" style={{ marginTop: 0 }}>
+          Lotes
+        </h2>
+        {loading ? (
+          <div className="state-box">Cargando…</div>
+        ) : batches.length === 0 ? (
+          <div className="state-box">Todavía no hay lotes de recibos.</div>
+        ) : (
+          <div className="tx-scroll">
+            <table className="tx-table">
+              <thead>
+                <tr>
+                  <th>Fecha</th>
+                  <th>Servicio</th>
+                  <th>Referencia</th>
+                  <th>Estado</th>
+                  <th className="num">Páginas</th>
+                  <th>Subido por</th>
+                  <th />
+                </tr>
+              </thead>
+              <tbody>
+                {batches.map((b) => (
+                  <tr key={b.id}>
+                    <td style={{ whiteSpace: "nowrap" }}>{formatDate(b.createdAt)}</td>
+                    <td>{serviceKindLabel(b.serviceKind)}</td>
+                    <td>{b.label || "—"}</td>
+                    <td>
+                      <StatusTag status={b.status} />
+                      {b.error && (
+                        <div className="page-sub" style={{ marginTop: 4 }}>
+                          {b.error}
+                        </div>
+                      )}
+                    </td>
+                    <td className="num">{b._count?.documents ?? 0}</td>
+                    <td>{b.uploadedBy?.name ?? "—"}</td>
+                    <td>
+                      <Link className="btn btn-ghost" href={`/recibos/${b.id}`}>
+                        Revisar
+                      </Link>
+                    </td>
+                  </tr>
+                ))}
+              </tbody>
+            </table>
+          </div>
+        )}
+      </section>
+    </div>
+  );
+}
+
+function StatusTag({ status }: { status: StatementBatchStatus }) {
+  return <span className="tag">{STATUS_LABEL[status] ?? status}</span>;
+}
+
+function UploadCard({ onDone }: { onDone: () => void }) {
+  const [files, setFiles] = useState<File[]>([]);
+  const [serviceKind, setServiceKind] = useState<ServiceKind>("ELECTRIC");
+  const [label, setLabel] = useState("");
+  const [busy, setBusy] = useState(false);
+  const [error, setError] = useState<string | null>(null);
+
+  async function submit() {
+    if (!files.length) return;
+    setBusy(true);
+    setError(null);
+    try {
+      await uploadStatementBatch(files, serviceKind, label.trim() || undefined);
+      setFiles([]);
+      setLabel("");
+      onDone();
+    } catch (e) {
+      setError((e as Error)?.message ?? "No se pudo subir el lote.");
+    } finally {
+      setBusy(false);
+    }
+  }
+
+  const unsupported = !SUPPORTED.includes(serviceKind);
+
+  return (
+    <section className="card" style={{ padding: 16 }}>
+      <h2 className="section-title" style={{ marginTop: 0 }}>
+        Subir recibos escaneados
+      </h2>
+      <div className="inline-form" style={{ flexWrap: "wrap", gap: 12 }}>
+        <label>
+          <span className="page-sub">Servicio</span>
+          <select
+            className="input"
+            value={serviceKind}
+            onChange={(e) => setServiceKind(e.target.value as ServiceKind)}
+          >
+            <optgroup label="Con lectura automática">
+              {SUPPORTED.map((k) => (
+                <option key={k} value={k}>
+                  {SERVICE_KIND_LABELS[k] ?? k}
+                </option>
+              ))}
+            </optgroup>
+            <optgroup label="Sin lectura automática (revisión manual)">
+              {OTHER_KINDS.map((k) => (
+                <option key={k} value={k}>
+                  {SERVICE_KIND_LABELS[k] ?? k}
+                </option>
+              ))}
+            </optgroup>
+          </select>
+        </label>
+
+        <label>
+          <span className="page-sub">Referencia (opcional)</span>
+          <input
+            className="input"
+            placeholder="ej. CFE julio 2026"
+            value={label}
+            onChange={(e) => setLabel(e.target.value)}
+          />
+        </label>
+
+        <label>
+          <span className="page-sub">Archivos PDF</span>
+          <input
+            type="file"
+            className="input"
+            accept="application/pdf"
+            multiple
+            onChange={(e) => setFiles(Array.from(e.target.files ?? []))}
+          />
+        </label>
+
+        <button
+          type="button"
+          className="btn btn-primary"
+          disabled={!files.length || busy}
+          onClick={submit}
+        >
+          {busy ? "Subiendo…" : `Procesar ${files.length || ""}`.trim()}
+        </button>
+      </div>
+
+      {error && (
+        <div className="state-box state-error" style={{ marginTop: 12 }}>
+          {error}
+        </div>
+      )}
+
+      {unsupported && (
+        <div className="state-box" style={{ marginTop: 12 }}>
+          Todavía no hay lectura automática para{" "}
+          {SERVICE_KIND_LABELS[serviceKind] ?? serviceKind}: cada página quedará
+          para revisión manual. Al confirmarlas se guarda el número de cuenta,
+          así que los recibos del mes siguiente sí se reconocerán solos.
+        </div>
+      )}
+
+      <p className="page-sub" style={{ marginTop: 12 }}>
+        Un lote es de un solo servicio. Cada página del PDF se trata como un
+        recibo distinto, salvo que el proveedor imprima varias hojas por cliente.
+      </p>
+    </section>
+  );
+}

apps/web/src/lib/api.ts

@@ -16,6 +16,8 @@ import type {
   BankStats,
   BankSummary,
   BatchCreateInput,
+  ConfirmBatchInput,
+  ConfirmBatchResult,
   BatchCreateResponse,
   BillingFacets,
   BillingStats,
@@ -27,6 +29,11 @@ import type {
   CreateMovementInput,
   UpdateBankAccountInput,
   ResolveOutstandingInput,
+  ReviewDocumentInput,
+  StatementBatch,
+  StatementBatchDetail,
+  StatementDocument,
+  StatementDocumentStatus,
   CustomerDetail,
   CustomerInput,
   CustomerListResponse,
@@ -42,6 +49,12 @@ import type {
   PolicySort,
   PolicyStats,
   PolicyStatus,
+  PolicyOcrBatch,
+  PolicyOcrBatchDetail,
+  PolicyOcrDocument,
+  PolicyOcrReviewInput,
+  PolicyOcrConfirmInput,
+  PolicyOcrConfirmResult,
   LookupsResponse,
   OpsJob,
   OpsJobKind,
@@ -796,38 +809,117 @@ export function listIngest(): Promise<IngestFile[]> {
   return apiFetch<IngestFile[]>("/ops/ingest");
 }
 
+/** Live upload stats reported to `uploadFile`'s `onProgress` callback. */
+export type UploadProgress = {
+  loaded: number;
+  /** 0 when the browser can't compute the request size. */
+  total: number;
+  /** 0..1, or null when `total` is unknown. */
+  fraction: number | null;
+  /** Smoothed transfer rate. */
+  bytesPerSecond: number;
+  /** null until a rate and a total are both known. */
+  secondsRemaining: number | null;
+  /** True once the bytes are sent and we're waiting on the server's reply. */
+  finishing: boolean;
+};
+
 /**
  * Multipart upload — not JSON, so it bypasses apiFetch's Content-Type. `path`
  * is API-relative (may include a query string); `filename` overrides the part
- * name sent to the server.
+ * name sent to the server. Uses XHR rather than fetch because fetch has no way
+ * to report request-body progress.
  */
-export async function uploadFile(
+export function uploadFile(
   path: string,
   file: File,
   filename?: string,
+  onProgress?: (p: UploadProgress) => void,
 ): Promise<unknown> {
   const body = new FormData();
   body.append("file", file, filename ?? file.name);
-  const res = await fetch(`${API_ORIGIN}${path}`, {
+
-    method: "POST",
+  return new Promise((resolve, reject) => {
-    credentials: "include",
+    const xhr = new XMLHttpRequest();
-    body,
+    xhr.open("POST", `${API_ORIGIN}${path}`);
+    xhr.withCredentials = true;
+
+    if (onProgress) {
+      // Exponentially smoothed rate — raw per-chunk deltas jump around too
+      // much to read.
+      let lastAt = performance.now();
+      let lastLoaded = 0;
+      let rate = 0;
+
+      xhr.upload.onprogress = (e) => {
+        const now = performance.now();
+        const dt = (now - lastAt) / 1000;
+        if (dt >= 0.15) {
+          const sample = (e.loaded - lastLoaded) / dt;
+          rate = rate === 0 ? sample : rate * 0.7 + sample * 0.3;
+          lastAt = now;
+          lastLoaded = e.loaded;
+        }
+        const total = e.lengthComputable ? e.total : 0;
+        onProgress({
+          loaded: e.loaded,
+          total,
+          fraction: total ? e.loaded / total : null,
+          bytesPerSecond: rate,
+          secondsRemaining:
+            total && rate > 0 ? (total - e.loaded) / rate : null,
+          finishing: false,
         });
-  if (!res.ok) {
+      };
-    let message = `Error ${res.status}`;
+      // Bytes are out the door; the server still has to write the file.
-    try {
+      xhr.upload.onload = () => {
-      const b = await res.json();
+        onProgress({
-      if (b?.message) message = b.message;
+          loaded: file.size,
-    } catch {
+          total: file.size,
-      /* ignore */
+          fraction: 1,
-    }
+          bytesPerSecond: rate,
-    throw new ApiError(res.status, message);
+          secondsRemaining: 0,
-  }
+          finishing: true,
-  return res.status === 204 ? undefined : res.json().catch(() => undefined);
+        });
+      };
     }
 
-export function uploadIngest(name: string, file: File): Promise<unknown> {
+    xhr.onload = () => {
-  return uploadFile(`/ops/ingest/${encodeURIComponent(name)}`, file, name);
+      let parsed: unknown;
+      try {
+        parsed = xhr.responseText ? JSON.parse(xhr.responseText) : undefined;
+      } catch {
+        parsed = undefined;
+      }
+      if (xhr.status >= 200 && xhr.status < 300) {
+        resolve(parsed);
+        return;
+      }
+      const message =
+        (parsed as { message?: string } | undefined)?.message ??
+        `Error ${xhr.status}`;
+      reject(new ApiError(xhr.status, message));
+    };
+    xhr.onerror = () =>
+      reject(new ApiError(0, "Fallo de red durante la carga."));
+    xhr.onabort = () => reject(new ApiError(0, "Carga cancelada."));
+    xhr.ontimeout = () => reject(new ApiError(0, "Tiempo de carga agotado."));
+
+    xhr.send(body);
+  });
+}
+
+export function uploadIngest(
+  name: string,
+  file: File,
+  onProgress?: (p: UploadProgress) => void,
+): Promise<unknown> {
+  return uploadFile(
+    `/ops/ingest/${encodeURIComponent(name)}`,
+    file,
+    name,
+    onProgress,
+  );
 }
 
 export function deleteIngest(name: string): Promise<unknown> {
@@ -895,3 +987,192 @@ export function reportDownloadUrl(
   const tail = qs.toString();
   return `${API_ORIGIN}/reports/${slug}/${format}${tail ? `?${tail}` : ""}`;
 }
+
+/* ------------------------------------- Statement OCR intake (recibos) */
+
+/**
+ * Whether this deployment can ingest scans — automatic capture is hidden
+ * without it. OCR reads the page, object storage keeps it; both are required.
+ */
+export function getStatementStatus(): Promise<{
+  ocrAvailable: boolean;
+  storageAvailable: boolean;
+}> {
+  return apiFetch("/statements/status");
+}
+
+export function listStatementBatches(
+  page = 1,
+  pageSize = 25,
+): Promise<{
+  items: StatementBatch[];
+  total: number;
+  page: number;
+  pageSize: number;
+  pageCount: number;
+}> {
+  return apiFetch(`/statements/batches?page=${page}&pageSize=${pageSize}`);
+}
+
+export function getStatementBatch(id: string): Promise<StatementBatchDetail> {
+  return apiFetch(`/statements/batches/${id}`);
+}
+
+export function listStatementDocuments(
+  batchId: string,
+  status?: StatementDocumentStatus,
+): Promise<StatementDocument[]> {
+  const q = status ? `?status=${status}` : "";
+  return apiFetch(`/statements/batches/${batchId}/documents${q}`);
+}
+
+/** Multi-file upload — one batch is usually several multi-page scans. */
+export async function uploadStatementBatch(
+  files: File[],
+  serviceKind: ServiceKind,
+  label?: string,
+): Promise<StatementBatch> {
+  const body = new FormData();
+  for (const f of files) body.append("files", f, f.name);
+  const qs = new URLSearchParams({ serviceKind });
+  if (label) qs.set("label", label);
+
+  const res = await fetch(`${API_ORIGIN}/statements/batches?${qs}`, {
+    method: "POST",
+    credentials: "include",
+    body,
+  });
+  if (!res.ok) {
+    let message = `Error ${res.status}`;
+    try {
+      const b = await res.json();
+      if (b?.message) message = b.message;
+    } catch {
+      /* non-JSON error body */
+    }
+    throw new Error(message);
+  }
+  return res.json();
+}
+
+export function reviewStatementDocument(
+  id: string,
+  input: ReviewDocumentInput,
+): Promise<StatementDocument> {
+  return apiFetch(`/statements/documents/${id}`, {
+    method: "PATCH",
+    body: JSON.stringify(input),
+  });
+}
+
+export function rejectStatementDocument(id: string): Promise<StatementDocument> {
+  return apiFetch(`/statements/documents/${id}/reject`, { method: "POST" });
+}
+
+export function confirmStatementBatch(
+  batchId: string,
+  input: ConfirmBatchInput,
+): Promise<ConfirmBatchResult> {
+  return apiFetch(`/statements/batches/${batchId}/confirm`, {
+    method: "POST",
+    body: JSON.stringify(input),
+  });
+}
+
+/** The rendered page image. A plain <img src> — the cookie rides along. */
+export function statementPageUrl(documentId: string): string {
+  return `${API_ORIGIN}/statements/documents/${documentId}/page`;
+}
+
+/* ----------------------------------------------------- Policy OCR (GMX) */
+
+export function getPolicyOcrStatus(): Promise<{
+  ocrAvailable: boolean;
+  storageAvailable: boolean;
+}> {
+  return apiFetch("/policy-ocr/status");
+}
+
+export function listPolicyOcrBatches(
+  page = 1,
+  pageSize = 25,
+): Promise<{
+  items: PolicyOcrBatch[];
+  total: number;
+  page: number;
+  pageSize: number;
+  pageCount: number;
+}> {
+  return apiFetch(`/policy-ocr/batches?page=${page}&pageSize=${pageSize}`);
+}
+
+export function getPolicyOcrBatch(id: string): Promise<PolicyOcrBatchDetail> {
+  return apiFetch(`/policy-ocr/batches/${id}`);
+}
+
+export function listPolicyOcrDocuments(batchId: string): Promise<PolicyOcrDocument[]> {
+  return apiFetch(`/policy-ocr/batches/${batchId}/documents`);
+}
+
+export async function uploadPolicyOcrBatch(
+  files: File[],
+  label?: string,
+): Promise<PolicyOcrBatch> {
+  const body = new FormData();
+  for (const f of files) body.append("files", f, f.name);
+  const qs = new URLSearchParams();
+  if (label) qs.set("label", label);
+
+  const res = await fetch(
+    `${API_ORIGIN}/policy-ocr/batches${qs.toString() ? `?${qs}` : ""}`,
+    {
+      method: "POST",
+      credentials: "include",
+      body,
+    },
+  );
+  if (!res.ok) {
+    let message = `Error ${res.status}`;
+    try {
+      const b = await res.json();
+      if (b?.message) message = b.message;
+    } catch {
+      /* non-JSON error body */
+    }
+    throw new Error(message);
+  }
+  return res.json();
+}
+
+export function reviewPolicyOcrDocument(
+  id: string,
+  input: PolicyOcrReviewInput,
+): Promise<PolicyOcrDocument> {
+  return apiFetch(`/policy-ocr/documents/${id}`, {
+    method: "PATCH",
+    body: JSON.stringify(input),
+  });
+}
+
+export function rejectPolicyOcrDocument(id: string): Promise<PolicyOcrDocument> {
+  return apiFetch(`/policy-ocr/documents/${id}/reject`, { method: "POST" });
+}
+
+export function confirmPolicyOcrBatch(
+  batchId: string,
+  input: PolicyOcrConfirmInput,
+): Promise<PolicyOcrConfirmResult> {
+  return apiFetch(`/policy-ocr/batches/${batchId}/confirm`, {
+    method: "POST",
+    body: JSON.stringify(input),
+  });
+}
+
+/**
+ * URL for the source PDF of a parsed policy document. The endpoint returns
+ * the original upload (one PDF = one parsed policy), not a rendered page
+ * image, so the review screen embeds it in an iframe.
+ */
+export function policyOcrDocumentUrl(documentId: string): string {
+  return `${API_ORIGIN}/policy-ocr/documents/${documentId}/page`;
+}

apps/web/src/lib/labels.ts

@@ -45,6 +45,7 @@ export const SERVICE_KIND_LABELS: Record<string, string> = {
   PROPERTY_TAX: "Predial",
   FEDERAL_ZONE: "Zona Federal",
   ALARM: "Alarma",
+  TELEPHONE: "Teléfono",
   OTHER: "Otro",
 };
 
@@ -57,6 +58,7 @@ export const SERVICE_KIND_GLYPH: Record<string, string> = {
   WATER: "≈",
   ELECTRIC: "⚡",
   GAS: "◐",
+  TELEPHONE: "☎",
   CABLE: "▤",
   PROPERTY_TAX: "⌂",
   FEDERAL_ZONE: "⇲",

apps/web/src/lib/types.ts

@@ -12,6 +12,8 @@ export type Ability =
   | "policy:create"
   | "policy:update"
   | "policy:delete"
+  | "policy:ingest"
+  | "policy:ocr-review"
   | "property:create"
   | "property:update"
   | "property:delete"
@@ -20,6 +22,8 @@ export type Ability =
   | "bank:create"
   | "bank:void"
   | "bank:manage-accounts"
+  | "statement:ingest"
+  | "statement:review"
   | "lookup:manage"
   | "user:manage"
   | "db:manage";
@@ -134,6 +138,7 @@ export type ServiceKind =
   | "PROPERTY_TAX"
   | "FEDERAL_ZONE"
   | "ALARM"
+  | "TELEPHONE"
   | "OTHER"
   | string;
 
@@ -1204,3 +1209,226 @@ export interface ReportRunResult {
 export interface ReportCatalog {
   items: ReportDef[];
 }
+
+/* ------------------------------------- Statement OCR intake (recibos) */
+
+export type StatementBatchStatus =
+  | "UPLOADED"
+  | "PROCESSING"
+  | "READY_FOR_REVIEW"
+  | "COMPLETED"
+  | "FAILED";
+
+export type StatementDocumentStatus =
+  | "PENDING_OCR"
+  | "OCR_FAILED"
+  | "NEEDS_REVIEW"
+  | "MATCHED"
+  | "CONFIRMED"
+  | "POSTED"
+  | "REJECTED";
+
+export interface StatementBatch {
+  id: string;
+  serviceKind: ServiceKind;
+  status: StatementBatchStatus;
+  label: string | null;
+  fileCount: number;
+  error: string | null;
+  createdAt: string;
+  completedAt: string | null;
+  uploadedBy?: { name: string };
+  _count?: { documents: number };
+}
+
+export interface StatementBatchDetail extends StatementBatch {
+  byStatus: Partial<Record<StatementDocumentStatus, number>>;
+  /** Sum of the amounts still awaiting posting. */
+  pendingTotal: string;
+}
+
+export interface StatementDocument {
+  id: string;
+  pageNumber: number;
+  status: StatementDocumentStatus;
+  provider: string | null;
+  ocrConfidence: string | null;
+  extractedAccountRef: string | null;
+  extractedAmount: string | null;
+  extractedPeriod: string | null;
+  extractedDueDate: string | null;
+  extractedCadastralKey: string | null;
+  matchNote: string | null;
+  matchedCustomer: { id: string; name: string } | null;
+  matchedPropertyService: {
+    id: string;
+    kind: ServiceKind;
+    accountNumber: string | null;
+    meterNumber: string | null;
+    property: { id: string; addressLine1: string | null };
+  } | null;
+  postedTransactionId: string | null;
+}
+
+export interface ReviewDocumentInput {
+  accountRef?: string;
+  amount?: number;
+  period?: string;
+  dueDate?: string;
+  matchedPropertyServiceId?: string;
+  matchedCustomerId?: string;
+  status?: "MATCHED" | "NEEDS_REVIEW" | "CONFIRMED";
+}
+
+/** Check-level fields shared by every line posted from a batch. */
+export interface ConfirmBatchInput {
+  checkNumber: string;
+  transactionDate: string;
+  currency?: Currency;
+  typeId?: string;
+  outstanding?: boolean;
+  includeReviewed?: boolean;
+}
+
+export interface ConfirmBatchResult {
+  posted: number;
+  total: string;
+  checkNumber: string;
+}
+
+/* ------------------------------------------ Policy OCR intake (GMX) */
+
+export type PolicyOcrBatchStatus =
+  | "UPLOADED"
+  | "PROCESSING"
+  | "READY_FOR_REVIEW"
+  | "COMPLETED"
+  | "FAILED";
+
+export type PolicyOcrDocumentStatus =
+  | "PENDING_OCR"
+  | "OCR_FAILED"
+  | "NEEDS_REVIEW"
+  | "MATCHED"
+  | "CONFIRMED"
+  | "POSTED"
+  | "REJECTED";
+
+export interface PolicyOcrBatch {
+  id: string;
+  provider: string;
+  status: PolicyOcrBatchStatus;
+  label: string | null;
+  fileCount: number;
+  error: string | null;
+  createdAt: string;
+  completedAt: string | null;
+  uploadedBy?: { name: string };
+  _count?: { documents: number };
+}
+
+export interface PolicyOcrBatchDetail extends PolicyOcrBatch {
+  byStatus: Partial<Record<PolicyOcrDocumentStatus, number>>;
+}
+
+export interface PolicyOcrCoverage {
+  risk: string;
+  insuredAmount: number | null;
+  deductible: string | null;
+  lossParticipation: string | null;
+}
+
+export interface PolicyOcrMatchCandidate {
+  policyId: string;
+  customerId: string;
+  customerName: string;
+  policyNumber: string;
+}
+
+export interface PolicyOcrDocument {
+  id: string;
+  pageNumber: number;
+  status: PolicyOcrDocumentStatus;
+  provider: string | null;
+  ocrConfidence: string | null;
+  extractedPolicyNumber: string | null;
+  extractedInsuredName: string | null;
+  extractedAdditionalInsured: string | null;
+  extractedAgentName: string | null;
+  extractedLegalAddress: string | null;
+  extractedZip: string | null;
+  extractedPolicyFrom: string | null;
+  extractedPolicyTo: string | null;
+  extractedPolicyDate: string | null;
+  extractedCurrency: string | null;
+  extractedNetPremium: string | null;
+  extractedPolicyFee: string | null;
+  extractedBrokerFee: string | null;
+  extractedTotal: string | null;
+  extractedCoveragesJson: PolicyOcrCoverage[] | null;
+  extractedPremiumPayment: string | null;
+  matchedPolicy: {
+    id: string;
+    policyNumber: string | null;
+    customerId: string;
+    customer: { name: string };
+  } | null;
+  matchedCustomer: { id: string; name: string } | null;
+  matchCandidates: PolicyOcrMatchCandidate[] | null;
+  matchNote: string | null;
+}
+
+export interface PolicyOcrReviewInput {
+  policyNumber?: string;
+  insuredName?: string;
+  additionalInsured?: string;
+  agentName?: string;
+  legalAddress?: string;
+  zip?: string;
+  policyFrom?: string;
+  policyTo?: string;
+  policyDate?: string;
+  currency?: string;
+  netPremium?: number;
+  policyFee?: number;
+  brokerFee?: number;
+  total?: number;
+  premiumPayment?: string;
+  coveragesJson?: PolicyOcrCoverage[];
+  matchedPolicyId?: string;
+  matchedCustomerId?: string;
+  forceConfirm?: boolean;
+}
+
+export interface PolicyOcrConfirmDocument {
+  documentId: string;
+  policyId?: string;
+  customerId?: string;
+  policyNumber?: string;
+  insuredName?: string;
+  additionalInsured?: string;
+  agentName?: string;
+  legalAddress?: string;
+  zip?: string;
+  policyFrom?: string;
+  policyTo?: string;
+  policyDate?: string;
+  currency?: string;
+  netPremium?: number;
+  policyFee?: number;
+  brokerFee?: number;
+  total?: number;
+  premiumPayment?: string;
+  coveragesJson?: PolicyOcrCoverage[];
+  postPremium?: boolean;
+}
+
+export interface PolicyOcrConfirmInput {
+  documents: PolicyOcrConfirmDocument[];
+}
+
+export interface PolicyOcrConfirmResult {
+  applied: number;
+  policies: string[];
+  postedTransactions: number;
+}

docker/api.Dockerfile

@@ -46,7 +46,24 @@ ENV NODE_ENV=production
 #   ERROR 1045: Plugin caching_sha2_password could not be loaded
 # That breaks the pre-migrate deploy backup AND the whole "Operaciones" admin
 # panel (backup, restore, sync, re-import all shell out to these binaries).
-RUN apk add --no-cache python3 mdbtools mysql-client mariadb-connector-c openssl \
+#
+# mdbtools-utils, NOT mdbtools. Alpine splits the project: `mdbtools` is the
+# shared library only, and the command-line tools migration/extract.py shells out
+# to (`mdb-tables`, `mdb-export`) are in the -utils subpackage. Installing the
+# wrong one builds fine and fails at run time — the re-import in the "Operaciones"
+# panel dies with:
+#   RuntimeError: mdbtools not found on PATH (need mdb-tables and mdb-export)
+#
+# tesseract-ocr + tesseract-ocr-data-spa + poppler-utils drive the statement
+# OCR intake (RECEIPT_CAPTURE_SPEC §2): poppler's `pdftoppm` rasterises each
+# scanned page and tesseract reads it, with the Spanish traineddata for the
+# accented labels on CFE/CESPT/Telnor bills. These are external binaries rather
+# than a native npm addon so the pnpm workspace stays free of a compiled
+# dependency. If they are absent the API still boots — the statements module
+# reports itself unavailable and only that feature is disabled — but statement
+# ingest is the point of shipping them.
+RUN apk add --no-cache python3 mdbtools-utils mysql-client mariadb-connector-c openssl \
+      tesseract-ocr tesseract-ocr-data-spa poppler-utils \
  && apk add --no-cache --virtual .pybuild python3-dev build-base \
  && rm -rf /var/cache/apk/*
 

docs/DEPLOY_AND_MIGRATIONS.md

@@ -5,13 +5,20 @@ rollbacks possible.
 
 ## The short version
 
+Dispatch **Cut release** from the Actions tab and pick `patch`, `minor` or
+`major` (or `explicit` plus a number). It stamps every `package.json`, commits
+`chore(release): vX.Y.Z`, tags, and pushes both refs in one go. It refuses a
+version that already exists as a tag, and refuses a no-op bump.
+
+The equivalent by hand, if you would rather cut it locally:
+
 ```bash
 pnpm version:set 1.2.0                       # stamp every package.json
 git commit -am "chore(release): v1.2.0"
 git tag v1.2.0 && git push origin master v1.2.0
 ```
 
-That push triggers `.gitea/workflows/build.yml`, which builds **both** images in
+Either way that push triggers `.gitea/workflows/build.yml`, which builds **both** images in
 one matrix run and publishes:
 
 | tag pushed | image tags produced |
@@ -28,6 +35,13 @@ Then dispatch a deploy from the Actions tab:
 > `{{version}}` strips it. Git tag `v1.2.0`, dispatch `1.2.0`. Dispatching
 > `v1.2.0` deploys nothing that exists.
 
+**Cut release needs a `RELEASE_TOKEN` secret** — a Gitea PAT with
+`write:repository`. It does not use the built-in Actions token on purpose:
+whether a push made with that token re-triggers `build.yml` depends on the Gitea
+version, and a release that quietly publishes no images is worse than one that
+fails outright. If the build somehow does not start, `build.yml` has
+`workflow_dispatch` — run it against the new tag by hand.
+
 Because api and web are built from one matrix run, they cannot drift at build
 time. They *can* drift at deploy time if a stack is applied with only one image
 moved — the web footer shows both versions and flags a mismatch, and the deploy
@@ -280,10 +294,19 @@ a warning, which is what local development wants.
 Two more things the panel's dumps now do, for the same reasons the pre-migrate
 backup does them (see `deploy/scripts/pre-migrate-backup.mjs`):
 
-- **`--set-gtid-purged=OFF`.** galactus is the replication *source* with GTID
+- **`--set-gtid-purged=OFF`, but only when the dumper supports it.** galactus is
-  on, so without this every dump embeds `SET @@GLOBAL.GTID_PURGED` and cannot be
+  the replication *source* with GTID on, so on a MySQL client this flag is what
-  restored onto the server it came from — which is precisely what the restore
+  keeps every dump from embedding `SET @@GLOBAL.GTID_PURGED` and becoming
-  screen exists to do.
+  unrestorable onto the server it came from — which is precisely what the
+  restore screen exists to do. The panel, however, dumps from *inside the API
+  container*, where Alpine's `mysql-client` is MariaDB's: there `mysqldump` is a
+  shim over `mariadb-dump`, the flag does not exist, and passing it failed every
+  backup with `mysqldump: unknown variable 'set-gtid-purged=OFF'`. So the panel
+  probes `mysqldump --help` and passes the flag only if it is advertised,
+  invoking `mariadb-dump` directly otherwise (MariaDB writes no GTID state
+  unless asked with `--gtid`, so there is nothing to suppress). The pre-migrate
+  backup keeps the flag unconditionally — it runs in a real `mysql:8.4` image,
+  not in the API container.
 - **`set -o pipefail` and a `CREATE TABLE` count.** `mysqldump | gzip` reports
   gzip's exit status, and a `mysqldump` that dies on its first statement still
   produces a ~372-byte perfectly valid archive that passes `gzip -t`. Without

docs/RECEIPT_CAPTURE_SPEC.md

@@ -131,6 +131,179 @@ single-movement form.
 
 ## 2. PDF / OCR auto-capture
 
+> **BUILT — 2026-08-01.** Implemented and verified end to end against real
+> scanned statements. `apps/api/src/statements/` holds the module: a swappable
+> `OcrProvider` seam with a self-hosted Tesseract implementation, per-provider
+> parsers for CFE / CESPT / Telnor / gas / predial, a scoped matcher, and a
+> review queue that
+> posts through `BillingService.createBatch` with `source: "OCR"`. Web:
+> the "Captura automática (OCR)" tab of the Captura screen (upload + batch
+> list) and `/recibos/:id` (review queue with the page image beside the
+> extracted fields). New abilities `statement:ingest` / `statement:review`, both
+> STAFF.
+>
+> Auto-capture is a *mode of* §1.2's capture screen, not a separate menu entry:
+> it is the same daily job with a scanner instead of a keyboard, and both modes
+> post through the same ledger path. `/estado-cuenta/lote` opens the manual tab,
+> `/recibos` the automatic one; both render `components/Captura.tsx`.
+>
+> Requires object storage (`S3_ENDPOINT` + credentials): the scans are kept as
+> blobs. `GET /statements/status` reports `ocrAvailable` and `storageAvailable`,
+> and the upload card hides itself unless both hold.
+>
+> **Measured, not assumed.** Ten real scans (46 pages of CFE, CESPT and Telnor
+> bills) drove every decision below. Against them the shipped parser identifies
+> the provider on **46/46**, reads an account reference on **43/46**, an amount
+> on **42/46**, and a due date on **44/46**. Matched against the dev database
+> that is **39/46 (85%) exact auto-match, 40/46 (87%) identified**. The
+> remainder are legitimate review cases: one account number shared by two
+> services, three phone numbers not yet on file, one clave not in the book, and
+> one page too poorly scanned to read.
+>
+> **The OCR-provider question is closed: self-hosted Tesseract.** It clears the
+> bar for a queue where a human confirms every row, and at 300+ pages/month/
+> company a per-page API would carry real recurring cost for accuracy that is
+> not the bottleneck. `OcrProvider` keeps a managed API (Textract, Document
+> Intelligence, Document AI) a one-line swap in `statements.module.ts` with no
+> schema change.
+>
+> **Four things the samples proved that this spec had wrong or unknown:**
+>
+> 1. **Clave catastral ≠ predial — gap 2 below is resolved.** `DATMEX.clave` is
+>    934 rows of `[A-Z]{2}[0-9]{6}` (`MM000012`, `KH220204`), the exact format
+>    printed as `Cve. Cat.` / `CLAVE CATASTRAL` on real CESPT bills
+>    (`KB078025`, `KA903009`). `DATMEX.predial` — what
+>    `PROPERTY_TAX.accountNumber` actually holds — is 1135 numeric rows with
+>    only **663 distinct values**, so it is not a per-property key at all and
+>    appears on no statement. The clave was never migrated; it now lives on
+>    `Property.cadastralKey` (property-level, because two different services
+>    both print it) and is the matcher's secondary key. Predial is left
+>    untouched. Predial statements match on the clave alone.
+> 2. **Gas is not a dead end — gap 3 below was wrong.** `DATMEX.gas` has 334
+>    filled rows, of which **160 are real numeric account numbers**
+>    (`900004807`); the other 174 are tank descriptors (`ESTACIONARIO`,
+>    `CILINDRO`). All 334 went to `notes`. The 160 are recovered into
+>    `GAS.meterNumber`; only the descriptor rows start cold.
+> 3. **Phone is one line per property, not three.** Of 1518 properties, 534
+>    have `phone1`, 18 have `phone2` and exactly **1** has `phone3`. The
+>    secondaries are alternate contacts, so `TELEPHONE` backfills from `phone1`
+>    only rather than fanning out. This answers the open question below.
+> 4. **Statements arrive bundled, and their printed names are stale.** One PDF
+>    holds many customers, one per page (Telnor's own `Pág 3 de 6` refers to
+>    its internal pagination, not the office's scan). And the name on a utility
+>    bill is the account registrant, not the current owner: a CESPT receipt for
+>    account `5365218` prints `ARNAIZ ROSAS ELSA AURORA` where the office's
+>    book — corroborated by the clave — has `CATT, RANDY`. **The matcher never
+>    reads the name.**
+>
+> **Two OCR traps worth keeping in mind if the parsers are ever extended:**
+> scanned logos read badly (a CESPT header came back as `E BAJA ES PAGO /
+> EALIFORNIA`), so provider detection falls back to layout anchors — but only
+> after *every* brand check has run, since a Telnor page contains words a CFE
+> structural rule would otherwise claim. And amounts must be parsed by
+> separator position: a real Telnor bill OCR'd as `$ 649,00`, which naive
+> comma-stripping turns into $64,900.
+>
+> **Not covered:** handwritten folder numbers. Staff pencil a customer number on
+> each bill (`9`, `405`, `406`); Tesseract read `405` as `205`. Handwriting is
+> a review hint at best and is deliberately not an input to matching.
+
+> **EXTENDED — gas and predial, 2026-08-01.** A second corpus (14 documents,
+> 29 pages: five municipal predial batches and ten gas invoices) added four
+> parsers — `GAS TIJUANA` plus one per municipality, because Tijuana, Rosarito
+> and Ensenada issue three completely different documents. End to end against
+> the dev database that is **21/29 auto-matched, 22/29 identified**, with the
+> provider read on 29/29 and an amount on 26/29.
+>
+> The eight review cases are all legitimate: five Tijuana pages whose municipal
+> account is not yet on file (see below), one clave not in the book, one page
+> too poorly scanned to read a clave at all, and one gas account shared by two
+> services. Excluding the structural Tijuana case, that is 21/24.
+>
+> **Five things this corpus proved:**
+>
+> 1. **Not every statement is a scan.** The gas company sends born-digital CFDI
+>    invoices whose text layer is exact. Rasterising and re-recognising those
+>    can only lose information — one sample turned `MEDIDOR: VM01014426` into
+>    `ar (LTR): 014420` — so `OcrProvider.textPages` reads the embedded layer
+>    first (`pdftotext -bbox-layout`, same poppler package as `pdftoppm`) and
+>    OCR stays the fallback for real scans. Page images are still rendered and
+>    stored either way, because the reviewer needs to see the paper.
+> 2. **The clave catastral is not two letters and six digits.** Positions four
+>    through eight are digits in all 932 stored claves, but the third is a
+>    letter in fifteen of them (`MMB01041`, `CGH52121`). Digitising the whole
+>    tail maps that `B` to an `8` and produces a key matching no property.
+> 3. **Tijuana predial prints no clave catastral at all.** Its only identifier
+>    is an 8-digit municipal account, carried in a 32-digit payment barcode
+>    (`account(8) + DDMMYY + amount(9) + folio(9)`) that the legacy database
+>    never held. It goes in `PROPERTY_TAX.meterNumber` — the same column gas
+>    uses, and for the same reason: `accountNumber` holds `DATMEX.predial`,
+>    which is not a per-property key and overwriting it would destroy the only
+>    link back to the original records. So Tijuana pages start cold and are
+>    taught by the first confirm, exactly like gas.
+> 4. **On Rosarito and Ensenada the clave is the primary key, not a fallback.**
+>    Those receipts print nothing else, so a unique clave hit there is a real
+>    match and auto-matches; on a utility bill that merely happens to print one
+>    it stays a review hint, as before.
+> 5. **A misread `$` is the dangerous failure, not a missing one.** An Ensenada
+>    receipt for `$2,203.00` OCR'd as `82,203.00` — the dollar sign read as an
+>    8, which would post a charge 37× too large and look entirely ordinary in
+>    the ledger. Every predial amount therefore requires a literal `$`, and a
+>    page that cannot produce one reports no amount and goes to review. Two of
+>    the 29 pages take that path, which is the correct outcome for both.
+>
+> Regression cover for all of the above lives in
+> `statement-parser.spec.ts` and `tesseract.provider.spec.ts`; every fixture in
+> them is a verbatim OCR excerpt from a real receipt.
+
+> **EXTENDED — zona federal, 2026-08-01.** A third corpus (one document, 8
+> pages of Tijuana "Zona Federal Marítimo Terrestre" receipts — the federal
+> maritime-zone occupancy fee billed on beachfront lots) added the
+> `ZONA FEDERAL TIJUANA` parser. Provider read on 8/8, amount on 8/8 (all
+> eight verified against the paper), concession clave on 6/8, period on 8/8,
+> payment deadline on 2/8. Nothing auto-matched, and nothing could have — see
+> point 2.
+>
+> **Four things this corpus proved:**
+>
+> 1. **Tijuana bills predial and zona federal from the same treasury.** Same
+>    "Ayuntamiento de Tijuana" header, same Paseo del Centenario address, same
+>    `ATB-541201` RFC — every discriminator the predial parser uses matches a
+>    zona federal page too, so whichever rule is asked first wins. The words
+>    only this layout prints are `Marítimo Terrestre`, so its brand rule is
+>    asked ahead of all three predial ones.
+> 2. **`FEDERAL_ZONE.accountNumber` is an amount, not a reference.** It holds
+>    `DATMEX.zfed`, whose 77 values include `246.06`, `2369.09`, `22653.94` and
+>    a negative `-1679`; the concession claves the receipts are keyed by
+>    (`12-T -012`, `14-D -014`) appear nowhere in the database. Matching on that
+>    column could never hit — and because every row already has a value, the
+>    `[field]: null` guards on learning and on the blank-service fill would
+>    never fire either, so every page would return to review every bimester
+>    forever. The clave moves to `meterNumber`, joining gas and Tijuana predial,
+>    and the first confirm teaches the match. This is the same trap as
+>    `policies.total` and `PROPERTY_TAX.accountNumber`: a legacy column whose
+>    name promises an identifier and whose contents are something else.
+> 3. **The payable figure is not the printed subtotal.** The municipality rounds
+>    to whole pesos and prints the difference as its own `Ajuste Ley Hacienda
+>    Mpal` line — `-$0.05` against a 591.05 subtotal, `$0.21` against 2,872.79.
+>    The "Total a pagar" box that carries the rounded figure sits on a grey fill
+>    and OCR'd on 1 of 8 pages; the SubTotal row read on 8 of 8. So the amount
+>    is the rounded subtotal, cross-checked against the printed box wherever it
+>    survives (it agreed).
+> 4. **The office's own highlighter is an OCR failure mode.** Both pages that
+>    lost their clave lost it to a marker stroke drawn across the `Clave:` line
+>    — not to scan quality, which was otherwise fine. The clave is printed twice
+>    (receipt and stub), which rescued a third page whose heading was struck
+>    but whose stub was not; where both copies are struck, the page reports no
+>    clave and goes to review rather than guessing.
+>
+> **Not attempted:** deriving the payment deadline from the bimester. It is the
+> 17th of the month after the bimester closes on a current bill, but four of
+> these eight are late — they carry a $1,000 `Multa` — and print a
+> recalculated deadline a month out. A derived date would be wrong on exactly
+> the pages a human most wants to look at, so an unreadable deadline stays
+> null.
+
 ### Motivation (from the meeting)
 
 Each utility company (CFE, water, phone, gas...) sends 300+ individual
@@ -254,8 +427,8 @@ Per the meeting notes' own field list:
 | Agua — Número de cuenta | `WATER` | `accountNumber` | `AGUA` | ✅ populated today |
 | Zona Fed — Número de Zona Federal | `FEDERAL_ZONE` | `accountNumber` | `ZFED` | ✅ populated today |
 | Tel — Número de teléfono | `TELEPHONE` *(new)* | `accountNumber` | `Property.phone1/2/3` (currently on `Property`, not `PropertyService`) | ⚠️ schema gap — see below |
-| Impuesto — Clave Catastral | `PROPERTY_TAX` | `accountNumber` | migrated from `PREDIAL`, **not** `CLAVE` | ⚠️ needs verification — see below |
+| Impuesto — Clave Catastral | `PROPERTY_TAX` | `Property.cadastralKey`, plus `meterNumber` for Tijuana's municipal account | `CLAVE`; `PREDIAL` is left on `accountNumber` and never matched against | ✅ built — see the 2026-08-01 extension note |
-| Gas — Número de medidor | `GAS` | `meterNumber` | not populated — folded into free-text `notes` today | ⚠️ data gap — see below |
+| Gas — Número de medidor | `GAS` | `meterNumber` | not populated — folded into free-text `notes` today | ✅ 160/334 recovered from `notes` |
 
 Confidence rule of thumb once a field is confirmed populated, tune after
 seeing real statements:
@@ -810,16 +983,23 @@ action (`customer:purge`) taken well after release — not bundled into
 
 ## Open questions to take back to Jorge (collected)
 
-- OCR provider/budget for §2 (self-hosted vs. managed API, given 300+
+- ~~OCR provider/budget for §2~~ — **CLOSED**: self-hosted Tesseract, chosen on
-  pages/month/company).
+  measured accuracy against real scans (see §2's BUILT note). No per-page cost.
-- Whether source PDFs arrive pre-split per customer or as one bundled file
+- ~~Whether source PDFs arrive pre-split per customer or bundled~~ —
-  needing page-range detection (§2).
+  **CLOSED**: bundled, one customer per page. Split per page.
-- Whether "Clave Catastral" and the already-migrated `PREDIAL`-sourced
+- ~~Whether "Clave Catastral" and the `PREDIAL`-sourced
-  `PROPERTY_TAX.accountNumber` are the same number — blocks OCR matching
+  `PROPERTY_TAX.accountNumber` are the same number~~ — **CLOSED**: they are
-  for predial statements specifically until confirmed (§2).
+  different. `clave` is the cadastral key and is now on
-- Whether phone billing is really one service per phone number on file, or
+  `Property.cadastralKey`; `predial` is not unique and is not printed on
-  one per property regardless of how many numbers are recorded — decides
+  statements.
-  how the new `TELEPHONE` service kind gets backfilled (§2).
+- ~~Whether phone billing is one service per number or one per property~~ —
+  **CLOSED**: effectively one (534 / 18 / 1 across phone1/2/3), backfilled
+  from `phone1`.
+- **Still open (§2):** whether the CFE amount staff should owe is the rounded
+  headline (`$268`, what the barcode encodes and what is paid at the window) or
+  the exact `Total` in the breakdown (`$268.88`). The parser currently takes
+  the barcode figure, which matches what the office actually pays; worth one
+  confirmation from Jorge.
 - The actual bank name/currency/details for the Seguros USD account, and
   whether any historical Seguros bank data exists to migrate (§3).
 - Whether `BankAccount.businessLine` should be enforced or a soft hint

migration/backfill_statement_match_fields.py

@@ -0,0 +1,155 @@
+"""
+Closes the three data gaps the statement-OCR matcher depends on
+(docs/RECEIPT_CAPTURE_SPEC.md §2, "Matching logic").
+
+OCR matching is only as good as the field it matches against, and a
+field-by-field check of real scanned statements against what
+`transform_properties.py` actually loaded turned up three mismatches. This
+script fixes them on an existing database; `transform_properties.py` has been
+updated in the same commit so a full re-migration produces them directly.
+
+  1. CLAVE CATASTRAL — printed on both the CESPT water bill ("Cve. Cat.:
+     KB078025") and the predial statement, and held in `DATMEX.clave` (934
+     rows, format `[A-Z]{2}[0-9]{6}`). It was never migrated. What
+     `PROPERTY_TAX.accountNumber` carries instead is `DATMEX.predial`, a
+     different, purely numeric column that is *not* unique — 663 distinct
+     values across 1135 filled rows — and appears on no statement. So predial
+     is left exactly where it is, and the clave lands on `Property` (it is a
+     property-level key, which is why two different services both print it).
+
+  2. GAS — `GAS.meterNumber` is empty for all 334 rows because the transform
+     put `DATMEX.gas` into `notes`. That column is mixed: 160 rows hold a real
+     numeric account/meter number, the remaining 174 hold a tank descriptor
+     ("ESTACIONARIO", "CILINDRO"). The numeric ones are recoverable now; the
+     descriptors legitimately have no number, so those statements still start
+     cold and get their number from the first human confirmation.
+
+  3. TELEPHONE — no such `ServiceKind` existed, so a Telnor bill had nothing to
+     match against. One service row is created per property with a `phone1`.
+     Only phone1: 534 properties have one, 18 have a phone2 and exactly 1 has a
+     phone3, so the secondaries are alternate contacts rather than separately
+     billed lines.
+
+Idempotent — re-running updates nothing it has already done, and it never
+overwrites a value a human has since corrected.
+
+    ./.venv/bin/python backfill_statement_match_fields.py --env dev
+"""
+
+from __future__ import annotations
+
+import uuid
+from pathlib import Path
+
+import pandas as pd
+
+from dbenv import connect
+from sync import parse_mode
+
+STG = Path(__file__).parent / "output" / "stg_utilities"
+NULL = "∅"
+
+
+def s(v):
+    if v is None or pd.isna(v):
+        return None
+    v = str(v).strip()
+    return None if v in ("", NULL, "0") else v
+
+
+def main():
+    # `--sync` is accepted and ignored — the script is idempotent either way.
+    env, _sync_mode = parse_mode()
+    conn = connect(env)
+    c = conn.cursor()
+    print(f"[statement-match-fields] target env: {env}")
+
+    # --- 1. clave catastral -> properties.cadastralKey ----------------------
+    # Joined on provenance, the same key transform_properties.py writes, so a
+    # property that was re-created by a later sync still lines up.
+    dm = pd.read_parquet(STG / "datmex.parquet")
+    claves = []
+    for _, row in dm.iterrows():
+        clave = s(row["clave"])
+        if clave:
+            claves.append((clave, str(int(row["_row_num"]))))
+
+    updated = 0
+    for clave, legacy_id in claves:
+        c.execute(
+            "UPDATE properties SET cadastralKey = %s "
+            "WHERE legacySourceTable = 'DATMEX' AND legacyId = %s AND cadastralKey IS NULL",
+            (clave, legacy_id),
+        )
+        updated += c.rowcount
+    print(f"  cadastralKey: set on {updated} propert(ies) ({len(claves)} in source)")
+
+    # --- 2. gas account numbers out of notes -> GAS.meterNumber -------------
+    # REGEXP rather than a Python loop: the value is already sitting in the
+    # notes column, so this is one pass over 334 rows inside the database.
+    c.execute(
+        "UPDATE property_services SET meterNumber = notes "
+        "WHERE kind = 'GAS' AND meterNumber IS NULL "
+        "AND notes REGEXP '^[0-9]{5,}$'"
+    )
+    print(f"  GAS.meterNumber: recovered {c.rowcount} account number(s) from notes")
+
+    # --- 3. TELEPHONE service rows ------------------------------------------
+    # Digits only, matching how the transform now writes them: a scanned Telnor
+    # bill prints "664 609 3444" and reduces to the stored local 6093444 once
+    # the LADA is stripped, which is the matcher's job, not this script's.
+    c.execute(
+        "SELECT p.id, p.phone1 FROM properties p "
+        "WHERE p.phone1 IS NOT NULL AND p.phone1 <> '' "
+        "AND NOT EXISTS (SELECT 1 FROM property_services ps "
+        "                WHERE ps.propertyId = p.id AND ps.kind = 'TELEPHONE')"
+    )
+    rows = c.fetchall()
+    made = []
+    for pid, phone in rows:
+        digits = "".join(ch for ch in str(phone) if ch.isdigit())
+        if digits:
+            made.append((str(uuid.uuid4()), pid, digits))
+    if made:
+        c.executemany(
+            "INSERT INTO property_services "
+            "(id, propertyId, kind, accountNumber, active, notes) "
+            "VALUES (%s, %s, 'TELEPHONE', %s, 1, 'from DATMEX.telefono')",
+            made,
+        )
+    print(f"  TELEPHONE: created {len(made)} service row(s)")
+
+    conn.commit()
+
+    # --- validation ---------------------------------------------------------
+    c.execute("SELECT COUNT(*) FROM properties WHERE cadastralKey IS NOT NULL")
+    n_clave = c.fetchone()[0]
+    c.execute(
+        "SELECT COUNT(*) FROM property_services WHERE kind='GAS' AND meterNumber IS NOT NULL"
+    )
+    n_gas = c.fetchone()[0]
+    c.execute("SELECT COUNT(*) FROM property_services WHERE kind='TELEPHONE'")
+    n_tel = c.fetchone()[0]
+
+    # A clave that is not unique would silently make the secondary match key
+    # ambiguous, which is worse than not having one — surface it rather than
+    # letting the matcher discover it a statement at a time.
+    c.execute(
+        "SELECT COUNT(*) FROM (SELECT cadastralKey FROM properties "
+        "WHERE cadastralKey IS NOT NULL GROUP BY cadastralKey HAVING COUNT(*) > 1) d"
+    )
+    dupe_claves = c.fetchone()[0]
+
+    print("=== Statement match fields ready ===")
+    print(f"  properties with cadastralKey : {n_clave}")
+    print(f"  GAS services with meterNumber: {n_gas}")
+    print(f"  TELEPHONE services           : {n_tel}")
+    print(f"  duplicated cadastralKey values: {dupe_claves}"
+          + ("  (matcher treats these as ambiguous)" if dupe_claves else ""))
+    assert n_clave > 0 and n_tel > 0, "backfill produced nothing — check staging output"
+    print("  validation: OK")
+    conn.close()
+
+
+if __name__ == "__main__":
+    main()

migration/blob_extract.py

@@ -33,7 +33,7 @@ from pathlib import Path
 import boto3
 from botocore.config import Config
 
-from dbenv import connect, load_env
+from dbenv import connect, require, setting
 from extract import sanitize_column_name as san
 
 csv.field_size_limit(300_000_000)
@@ -102,12 +102,16 @@ def main():
     only = set(x.strip() for x in args.tables.split(",") if x.strip())
     sources = [s for s in SOURCES if not only or s["key"] in only]
 
-    env = load_env(args.env)
+    # Process environment first, deploy/.env.<env> second, with the same
+    # credential aliases the API uses — the "Operaciones" re-import runs this
+    # inside the API container, which has S3_ENDPOINT / MINIO_ROOT_* injected
+    # and no deploy/ directory at all.
     s3 = boto3.client(
-        "s3", endpoint_url=env["S3_ENDPOINT"],
+        "s3", endpoint_url=require(args.env, "S3_ENDPOINT"),
-        aws_access_key_id=env["MINIO_ROOT_USER"], aws_secret_access_key=env["MINIO_ROOT_PASSWORD"],
+        aws_access_key_id=require(args.env, "S3_ACCESS_KEY", "MINIO_ROOT_USER"),
+        aws_secret_access_key=require(args.env, "S3_SECRET_KEY", "MINIO_ROOT_PASSWORD"),
         config=Config(signature_version="s3v4"), region_name="us-east-1")
-    bucket = env["S3_BUCKET"]
+    bucket = setting(args.env, "S3_BUCKET") or "jorgecuadros-documents"
 
     conn = connect(args.env)
     cur = conn.cursor()

migration/dbenv.py

@@ -32,29 +32,53 @@ REPO = Path(__file__).resolve().parents[1]
 
 
 def load_env(env: str) -> dict:
+    """deploy/.env.<env> parsed to a dict, or {} when the file is absent.
+
+    Absent is normal, not an error: the API container runs these scripts with
+    DATABASE_URL / S3_* injected as real environment variables and ships no
+    deploy/ directory. Use `setting()` / `require()` rather than this — they
+    layer the process environment on top, which is what actually resolves."""
     f = REPO / "deploy" / f".env.{env}"
     if not f.exists():
-        raise SystemExit(
+        return {}
-            f"missing {f} — deploy the '{env}' DB stack and write its .env first "
-            f"(see dbenv.py header)."
-        )
     out = {}
     for line in f.read_text().splitlines():
         line = line.strip()
         if line and not line.startswith("#") and "=" in line:
             k, v = line.split("=", 1)
             out[k] = v
-    if "DATABASE_URL" not in out:
-        raise SystemExit(f"{f} has no DATABASE_URL")
     return out
 
 
+def setting(env: str, *keys: str):
+    """First non-empty value for `keys`, process environment first, then
+    deploy/.env.<env>. Several keys = fallback aliases (S3_ACCESS_KEY then
+    MINIO_ROOT_USER, as apps/api/src/storage/storage.service.ts does)."""
+    fromfile = load_env(env)
+    for k in keys:
+        v = os.environ.get(k) or fromfile.get(k)
+        if v:
+            return v
+    return None
+
+
+def require(env: str, *keys: str) -> str:
+    v = setting(env, *keys)
+    if not v:
+        raise SystemExit(
+            f"missing {' / '.join(keys)} — set it in the environment, or deploy the "
+            f"'{env}' stack and write {REPO / 'deploy' / f'.env.{env}'} "
+            f"(see dbenv.py header)."
+        )
+    return v
+
+
 def database_url(env: str) -> str:
     """Target DB URL. A DATABASE_URL in the process environment wins over
     deploy/.env.<env> — this is how the API container (which has its own
     DATABASE_URL and no deploy/.env files) drives a re-import against its own
     database."""
-    return os.environ.get("DATABASE_URL") or load_env(env)["DATABASE_URL"]
+    return require(env, "DATABASE_URL")
 
 
 def connect(env: str):

migration/run_all.py

@@ -41,6 +41,12 @@ PY = sys.executable  # the venv python running this orchestrator
 STEPS = [
     "transform_customers.py",
     "transform_properties.py",
+    # Statement-OCR match fields. transform_properties.py now produces these
+    # directly, so on a full rebuild this is a no-op that re-asserts they are
+    # there; on a database predating the OCR module it is what fills them in.
+    # Must follow transform_properties.py, which truncates both tables it
+    # touches.
+    "backfill_statement_match_fields.py",
     "transform_policies.py",
     "transform_transactions.py",
     "prune_empty_customers.py",
@@ -54,6 +60,12 @@ STEPS = [
 SYNC_STEPS = [
     "transform_customers.py",
     "transform_properties.py",
+    # Statement-OCR match fields. transform_properties.py now produces these
+    # directly, so on a full rebuild this is a no-op that re-asserts they are
+    # there; on a database predating the OCR module it is what fills them in.
+    # Must follow transform_properties.py, which truncates both tables it
+    # touches.
+    "backfill_statement_match_fields.py",
     "transform_policies.py",
     "transform_transactions.py",
     # Manual-safe prune: drops legacy-owned empties that the customer upsert

migration/transform_properties.py

@@ -147,7 +147,7 @@ def main():
         props.append((
             pid, cust_id, s(row["direccion"]), ", ".join(addr2_parts) or None,
             s(row["telefono"]), s(row["telefono2"]), s(row["telefono3"]),
-             s(row["zona"]), "DATMEX", legacy_id,
+             s(row["zona"]), s(row["clave"]), "DATMEX", legacy_id,
 
         ))
 
@@ -178,9 +178,16 @@ def main():
                 svc("ELECTRIC", account=s(row[rc]),
                     notes=s(row["luz_tipo"]) if rc == "rpu" else None,
                     active=flag("electric"))
-        # GAS
+        # GAS — the column mixes two things: an account/meter number for 160 of
+        # the 334 filled rows, and a tank descriptor ("ESTACIONARIO",
+        # "CILINDRO") for the rest. Only the numeric form can be matched
+        # against a scanned gas statement, so it is promoted to meterNumber;
+        # the descriptor stays a note, as before.
         if s(row["gas"]) or flag("gas1", False):
-            svc("GAS", due=s(row["gas_vence"]), notes=s(row["gas"]), active=flag("gas1"))
+            gas_val = s(row["gas"])
+            gas_meter = gas_val if gas_val and gas_val.isdigit() and len(gas_val) >= 5 else None
+            svc("GAS", meter=gas_meter, due=s(row["gas_vence"]),
+                notes=s(row["gas"]), active=flag("gas1"))
         # CABLE
         if s(row["cable_num"]) or (s_keep0(row["cable_sky"]) or "0") in _TRUE:
             svc("CABLE", account=s(row["cable_num"]), route=s(row["cia_cable"]),
@@ -194,6 +201,19 @@ def main():
         if s(row["zfed"]) or flag("federalzone", False):
             svc("FEDERAL_ZONE", account=s(row["zfed"]), notes=s(row["zfed_t"]),
                 active=flag("federalzone"))
+        # TELEPHONE — DATMEX never had a phone *service*, only the contact
+        # numbers unpivoted into Property.phone1/2/3 above, even though the
+        # legacy ledger billed phone as its own transaction type. A Telnor bill
+        # can only be matched against a service row, so the primary number
+        # becomes one. Only phone1: of 1518 properties, 534 have phone1, 18
+        # phone2 and exactly 1 phone3 — the secondaries are alternate contacts,
+        # not additional billed lines. Stored as the bare local number, which is
+        # how DATMEX holds it and what a printed bill reduces to once the 664
+        # Tijuana LADA is stripped.
+        tel = s(row["telefono"])
+        if tel:
+            svc("TELEPHONE", account="".join(ch for ch in tel if ch.isdigit()) or None,
+                notes="from DATMEX.telefono")
         # ALARM
         if s(row["alarm_system"]):
             svc("ALARM", notes=s(row["alarm_system"]))
@@ -218,7 +238,7 @@ def main():
         cur.execute("DELETE ps FROM property_services ps JOIN properties p ON p.id=ps.propertyId WHERE p.legacyId IS NOT NULL")
         cur.execute("DELETE ta FROM trust_accounts ta JOIN properties p ON p.id=ta.propertyId WHERE p.legacyId IS NOT NULL")
         cur.executemany(
-            "INSERT INTO properties (id,customerId,addressLine1,addressLine2,phone1,phone2,phone3,zone,legacySourceTable,legacyId) VALUES (%s,%s,%s,%s,%s,%s,%s,%s,%s,%s) ON DUPLICATE KEY UPDATE customerId=VALUES(customerId),addressLine1=VALUES(addressLine1),addressLine2=VALUES(addressLine2),phone1=VALUES(phone1),phone2=VALUES(phone2),phone3=VALUES(phone3),zone=VALUES(zone),archivedAt=NULL", props)
+            "INSERT INTO properties (id,customerId,addressLine1,addressLine2,phone1,phone2,phone3,zone,cadastralKey,legacySourceTable,legacyId) VALUES (%s,%s,%s,%s,%s,%s,%s,%s,%s,%s,%s) ON DUPLICATE KEY UPDATE customerId=VALUES(customerId),addressLine1=VALUES(addressLine1),addressLine2=VALUES(addressLine2),phone1=VALUES(phone1),phone2=VALUES(phone2),phone3=VALUES(phone3),zone=VALUES(zone),cadastralKey=VALUES(cadastralKey),archivedAt=NULL", props)
         delete_missing(cur, "properties", ("legacySourceTable", "legacyId"), prop_keys, "WHERE legacyId IS NOT NULL")
         cur.executemany("INSERT INTO property_services (id,propertyId,kind,accountNumber,meterNumber,route,dueDay,active,notes) VALUES (%s,%s,%s,%s,%s,%s,%s,%s,%s)", services)
         cur.executemany("INSERT INTO trust_accounts (id,propertyId,bankName,trustNumber,bankFee,dueDate1,dueDate2) VALUES (%s,%s,%s,%s,%s,%s,%s)", trusts)
@@ -227,7 +247,7 @@ def main():
         for t in ("property_services", "service_documents", "trust_accounts", "properties"):
             cur.execute(f"TRUNCATE TABLE {t}")
         cur.execute("SET FOREIGN_KEY_CHECKS=1")
-        cur.executemany("INSERT INTO properties (id,customerId,addressLine1,addressLine2,phone1,phone2,phone3,zone,legacySourceTable,legacyId) VALUES (%s,%s,%s,%s,%s,%s,%s,%s,%s,%s)", props)
+        cur.executemany("INSERT INTO properties (id,customerId,addressLine1,addressLine2,phone1,phone2,phone3,zone,cadastralKey,legacySourceTable,legacyId) VALUES (%s,%s,%s,%s,%s,%s,%s,%s,%s,%s,%s)", props)
         cur.executemany("INSERT INTO property_services (id,propertyId,kind,accountNumber,meterNumber,route,dueDay,active,notes) VALUES (%s,%s,%s,%s,%s,%s,%s,%s,%s)", services)
         cur.executemany("INSERT INTO trust_accounts (id,propertyId,bankName,trustNumber,bankFee,dueDate1,dueDate2) VALUES (%s,%s,%s,%s,%s,%s,%s)", trusts)
     conn.commit()

package.json

@@ -1,6 +1,6 @@
 {
   "name": "jorgecuadros-platform",
-  "version": "1.0.1",
+  "version": "1.0.6",
   "private": true,
   "workspaces": [
     "apps/*",

packages/database/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@jorgecuadros/database",
-  "version": "1.0.1",
+  "version": "1.0.6",
   "private": true,
   "main": "generated/client/index.js",
   "types": "generated/client/index.d.ts",

packages/database/prisma/migrations/20260731235721_statement_ocr_intake/migration.sql

@@ -0,0 +1,79 @@
+-- AlterTable
+ALTER TABLE `properties` ADD COLUMN `cadastralKey` VARCHAR(191) NULL;
+
+-- AlterTable
+ALTER TABLE `property_services` MODIFY `kind` ENUM('WATER', 'ELECTRIC', 'GAS', 'CABLE', 'PROPERTY_TAX', 'FEDERAL_ZONE', 'ALARM', 'TELEPHONE', 'OTHER') NOT NULL;
+
+-- CreateTable
+CREATE TABLE `statement_batches` (
+    `id` VARCHAR(191) NOT NULL,
+    `serviceKind` ENUM('WATER', 'ELECTRIC', 'GAS', 'CABLE', 'PROPERTY_TAX', 'FEDERAL_ZONE', 'ALARM', 'TELEPHONE', 'OTHER') NOT NULL,
+    `status` ENUM('UPLOADED', 'PROCESSING', 'READY_FOR_REVIEW', 'COMPLETED', 'FAILED') NOT NULL DEFAULT 'UPLOADED',
+    `uploadedById` VARCHAR(191) NOT NULL,
+    `label` VARCHAR(191) NULL,
+    `fileCount` INTEGER NOT NULL DEFAULT 0,
+    `error` TEXT NULL,
+    `createdAt` DATETIME(3) NOT NULL DEFAULT CURRENT_TIMESTAMP(3),
+    `completedAt` DATETIME(3) NULL,
+
+    INDEX `statement_batches_status_createdAt_idx`(`status`, `createdAt`),
+    PRIMARY KEY (`id`)
+) DEFAULT CHARACTER SET utf8mb4 COLLATE utf8mb4_unicode_ci;
+
+-- CreateTable
+CREATE TABLE `statement_documents` (
+    `id` VARCHAR(191) NOT NULL,
+    `batchId` VARCHAR(191) NOT NULL,
+    `pageNumber` INTEGER NOT NULL,
+    `storageKey` VARCHAR(191) NOT NULL,
+    `status` ENUM('PENDING_OCR', 'OCR_FAILED', 'NEEDS_REVIEW', 'MATCHED', 'CONFIRMED', 'POSTED', 'REJECTED') NOT NULL DEFAULT 'PENDING_OCR',
+    `ocrRawText` TEXT NULL,
+    `ocrConfidence` DECIMAL(4, 3) NULL,
+    `provider` VARCHAR(191) NULL,
+    `extractedAccountRef` VARCHAR(191) NULL,
+    `extractedAmount` DECIMAL(12, 2) NULL,
+    `extractedPeriod` VARCHAR(191) NULL,
+    `extractedDueDate` DATETIME(3) NULL,
+    `extractedCadastralKey` VARCHAR(191) NULL,
+    `matchedPropertyServiceId` VARCHAR(191) NULL,
+    `matchedCustomerId` VARCHAR(191) NULL,
+    `matchNote` VARCHAR(191) NULL,
+    `reviewedById` VARCHAR(191) NULL,
+    `reviewedAt` DATETIME(3) NULL,
+    `postedTransactionId` VARCHAR(191) NULL,
+    `createdAt` DATETIME(3) NOT NULL DEFAULT CURRENT_TIMESTAMP(3),
+
+    UNIQUE INDEX `statement_documents_postedTransactionId_key`(`postedTransactionId`),
+    INDEX `statement_documents_status_idx`(`status`),
+    INDEX `statement_documents_matchedCustomerId_idx`(`matchedCustomerId`),
+    UNIQUE INDEX `statement_documents_batchId_pageNumber_key`(`batchId`, `pageNumber`),
+    PRIMARY KEY (`id`)
+) DEFAULT CHARACTER SET utf8mb4 COLLATE utf8mb4_unicode_ci;
+
+-- CreateIndex
+CREATE INDEX `properties_cadastralKey_idx` ON `properties`(`cadastralKey`);
+
+-- CreateIndex
+CREATE INDEX `property_services_kind_accountNumber_idx` ON `property_services`(`kind`, `accountNumber`);
+
+-- CreateIndex
+CREATE INDEX `property_services_kind_meterNumber_idx` ON `property_services`(`kind`, `meterNumber`);
+
+-- AddForeignKey
+ALTER TABLE `statement_batches` ADD CONSTRAINT `statement_batches_uploadedById_fkey` FOREIGN KEY (`uploadedById`) REFERENCES `users`(`id`) ON DELETE RESTRICT ON UPDATE CASCADE;
+
+-- AddForeignKey
+ALTER TABLE `statement_documents` ADD CONSTRAINT `statement_documents_batchId_fkey` FOREIGN KEY (`batchId`) REFERENCES `statement_batches`(`id`) ON DELETE CASCADE ON UPDATE CASCADE;
+
+-- AddForeignKey
+ALTER TABLE `statement_documents` ADD CONSTRAINT `statement_documents_matchedPropertyServiceId_fkey` FOREIGN KEY (`matchedPropertyServiceId`) REFERENCES `property_services`(`id`) ON DELETE SET NULL ON UPDATE CASCADE;
+
+-- AddForeignKey
+ALTER TABLE `statement_documents` ADD CONSTRAINT `statement_documents_matchedCustomerId_fkey` FOREIGN KEY (`matchedCustomerId`) REFERENCES `customers`(`id`) ON DELETE SET NULL ON UPDATE CASCADE;
+
+-- AddForeignKey
+ALTER TABLE `statement_documents` ADD CONSTRAINT `statement_documents_reviewedById_fkey` FOREIGN KEY (`reviewedById`) REFERENCES `users`(`id`) ON DELETE SET NULL ON UPDATE CASCADE;
+
+-- AddForeignKey
+ALTER TABLE `statement_documents` ADD CONSTRAINT `statement_documents_postedTransactionId_fkey` FOREIGN KEY (`postedTransactionId`) REFERENCES `transactions`(`id`) ON DELETE SET NULL ON UPDATE CASCADE;
+

packages/database/prisma/migrations/20260801000000_policy_ocr_intake/migration.sql

@@ -0,0 +1,80 @@
+-- CreateTable
+CREATE TABLE `policy_ocr_batches` (
+    `id` VARCHAR(191) NOT NULL,
+    `provider` VARCHAR(191) NOT NULL DEFAULT 'GMX',
+    `status` ENUM('UPLOADED', 'PROCESSING', 'READY_FOR_REVIEW', 'COMPLETED', 'FAILED') NOT NULL DEFAULT 'UPLOADED',
+    `uploadedById` VARCHAR(191) NOT NULL,
+    `label` VARCHAR(191) NULL,
+    `fileCount` INTEGER NOT NULL DEFAULT 0,
+    `error` TEXT NULL,
+    `createdAt` DATETIME(3) NOT NULL DEFAULT CURRENT_TIMESTAMP(3),
+    `completedAt` DATETIME(3) NULL,
+
+    INDEX `policy_ocr_batches_status_createdAt_idx`(`status`, `createdAt`),
+    PRIMARY KEY (`id`)
+) DEFAULT CHARACTER SET utf8mb4 COLLATE utf8mb4_unicode_ci;
+
+-- CreateTable
+CREATE TABLE `policy_ocr_documents` (
+    `id` VARCHAR(191) NOT NULL,
+    `batchId` VARCHAR(191) NOT NULL,
+    `pageNumber` INTEGER NOT NULL,
+    `storageKey` VARCHAR(191) NOT NULL,
+    `status` ENUM('PENDING_OCR', 'OCR_FAILED', 'NEEDS_REVIEW', 'MATCHED', 'CONFIRMED', 'POSTED', 'REJECTED') NOT NULL DEFAULT 'PENDING_OCR',
+    `ocrRawText` TEXT NULL,
+    `ocrConfidence` DECIMAL(4, 3) NULL,
+    `provider` VARCHAR(191) NULL,
+    `extractedPolicyNumber` VARCHAR(191) NULL,
+    `extractedInsuredName` VARCHAR(191) NULL,
+    `extractedAdditionalInsured` VARCHAR(191) NULL,
+    `extractedAgentName` VARCHAR(191) NULL,
+    `extractedLegalAddress` TEXT NULL,
+    `extractedZip` VARCHAR(191) NULL,
+    `extractedPolicyFrom` DATETIME(3) NULL,
+    `extractedPolicyTo` DATETIME(3) NULL,
+    `extractedPolicyDate` DATETIME(3) NULL,
+    `extractedCurrency` VARCHAR(191) NULL,
+    `extractedNetPremium` DECIMAL(12, 2) NULL,
+    `extractedPolicyFee` DECIMAL(12, 2) NULL,
+    `extractedBrokerFee` DECIMAL(12, 2) NULL,
+    `extractedTotal` DECIMAL(12, 2) NULL,
+    `extractedCoveragesJson` JSON NULL,
+    `extractedPremiumPayment` VARCHAR(191) NULL,
+    `matchedPolicyId` VARCHAR(191) NULL,
+    `matchedCustomerId` VARCHAR(191) NULL,
+    `matchCandidates` JSON NULL,
+    `matchNote` VARCHAR(191) NULL,
+    `reviewedById` VARCHAR(191) NULL,
+    `reviewedAt` DATETIME(3) NULL,
+    `createdPolicyId` VARCHAR(191) NULL,
+    `postedTransactionId` VARCHAR(191) NULL,
+    `createdAt` DATETIME(3) NOT NULL DEFAULT CURRENT_TIMESTAMP(3),
+
+    UNIQUE INDEX `policy_ocr_documents_createdPolicyId_key`(`createdPolicyId`),
+    UNIQUE INDEX `policy_ocr_documents_postedTransactionId_key`(`postedTransactionId`),
+    INDEX `policy_ocr_documents_status_idx`(`status`),
+    INDEX `policy_ocr_documents_matchedCustomerId_idx`(`matchedCustomerId`),
+    UNIQUE INDEX `policy_ocr_documents_batchId_pageNumber_key`(`batchId`, `pageNumber`),
+    PRIMARY KEY (`id`)
+) DEFAULT CHARACTER SET utf8mb4 COLLATE utf8mb4_unicode_ci;
+
+-- AddForeignKey
+ALTER TABLE `policy_ocr_batches` ADD CONSTRAINT `policy_ocr_batches_uploadedById_fkey` FOREIGN KEY (`uploadedById`) REFERENCES `users`(`id`) ON DELETE RESTRICT ON UPDATE CASCADE;
+
+-- AddForeignKey
+ALTER TABLE `policy_ocr_documents` ADD CONSTRAINT `policy_ocr_documents_batchId_fkey` FOREIGN KEY (`batchId`) REFERENCES `policy_ocr_batches`(`id`) ON DELETE CASCADE ON UPDATE CASCADE;
+
+-- AddForeignKey
+ALTER TABLE `policy_ocr_documents` ADD CONSTRAINT `policy_ocr_documents_matchedPolicyId_fkey` FOREIGN KEY (`matchedPolicyId`) REFERENCES `policies`(`id`) ON DELETE SET NULL ON UPDATE CASCADE;
+
+-- AddForeignKey
+ALTER TABLE `policy_ocr_documents` ADD CONSTRAINT `policy_ocr_documents_matchedCustomerId_fkey` FOREIGN KEY (`matchedCustomerId`) REFERENCES `customers`(`id`) ON DELETE SET NULL ON UPDATE CASCADE;
+
+-- AddForeignKey
+ALTER TABLE `policy_ocr_documents` ADD CONSTRAINT `policy_ocr_documents_reviewedById_fkey` FOREIGN KEY (`reviewedById`) REFERENCES `users`(`id`) ON DELETE SET NULL ON UPDATE CASCADE;
+
+-- AddForeignKey
+ALTER TABLE `policy_ocr_documents` ADD CONSTRAINT `policy_ocr_documents_createdPolicyId_fkey` FOREIGN KEY (`createdPolicyId`) REFERENCES `policies`(`id`) ON DELETE SET NULL ON UPDATE CASCADE;
+
+-- AddForeignKey
+ALTER TABLE `policy_ocr_documents` ADD CONSTRAINT `policy_ocr_documents_postedTransactionId_fkey` FOREIGN KEY (`postedTransactionId`) REFERENCES `transactions`(`id`) ON DELETE SET NULL ON UPDATE CASCADE;

packages/database/prisma/schema.prisma

@@ -53,6 +53,12 @@ enum ServiceKind {
   PROPERTY_TAX
   FEDERAL_ZONE
   ALARM
+  /// Telephone was never unpivoted out of DATMEX — the numbers sat on
+  /// `Property.phone1/2/3` as contact fields even though the legacy ledger
+  /// billed phone as its own transaction type. OCR matching needs a real
+  /// service row to match a Telnor bill against, so it becomes one; see
+  /// `migration/backfill_statement_match_fields.py`.
+  TELEPHONE
   OTHER
 }
 
@@ -115,6 +121,9 @@ model Customer {
   vehicles     Vehicle[]
   transactions Transaction[]
 
+  statementDocuments StatementDocument[]
+  policyOcrDocuments PolicyOcrDocument[] @relation("PolicyOcrDocumentCustomer")
+
   @@map("customers")
 }
 
@@ -202,6 +211,9 @@ model Policy {
   properties     Property[]
   renewalNotices RenewalNotice[]
 
+  ocrMatchedDocuments PolicyOcrDocument[] @relation("PolicyOcrDocumentPolicy")
+  ocrCreatedDocuments PolicyOcrDocument[] @relation("PolicyOcrDocumentCreatedPolicy")
+
   @@unique([legacySourceDb, legacySourceTable, legacyId])
   @@index([policyNumber])
   @@map("policies")
@@ -355,6 +367,115 @@ model PolicyDocument {
   @@map("policy_documents")
 }
 
+/// Insurance OCR intake (mirrors statement_batches / statement_documents for
+/// the utility side). One upload session of policy PDFs from a provider
+/// portal (GMX, etc.) — the parser proposes policyNumber → existing Policy
+/// (or "new, pick customer"), staff confirms, and the system attaches the
+/// 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.
+  provider     String               @default("GMX")
+  status       PolicyOcrBatchStatus @default(UPLOADED)
+  uploadedById String
+  uploadedBy   User                 @relation("PolicyOcrBatchUploader", fields: [uploadedById], references: [id])
+  label        String?
+  fileCount    Int                  @default(0)
+  /// Set when the pipeline fails as a whole (bad PDF, OCR binaries missing).
+  error        String?              @db.Text
+  createdAt    DateTime             @default(now())
+  completedAt  DateTime?
+
+  documents PolicyOcrDocument[]
+
+  @@index([status, createdAt])
+  @@map("policy_ocr_batches")
+}
+
+enum PolicyOcrBatchStatus {
+  UPLOADED
+  PROCESSING
+  READY_FOR_REVIEW
+  COMPLETED
+  FAILED
+}
+
+/// One parsed policy page — one Policy → one Customer (after staff confirms).
+model PolicyOcrDocument {
+  id         String                  @id @default(uuid())
+  batchId    String
+  batch      PolicyOcrBatch          @relation(fields: [batchId], references: [id], onDelete: Cascade)
+  pageNumber Int
+  /// The rendered page image in object storage. Source PDF kept too on the
+  /// batch (statement pattern) so re-running a corrected parser is possible.
+  storageKey String
+  status     PolicyOcrDocumentStatus @default(PENDING_OCR)
+
+  ocrRawText    String?  @db.Text
+  ocrConfidence Decimal? @db.Decimal(4, 3)
+  /// Which parser claimed the page ("GMX" today).
+  provider      String?
+
+  // Extracted header fields, all staff-editable in review.
+  extractedPolicyNumber      String?
+  extractedInsuredName       String?
+  extractedAdditionalInsured String?
+  extractedAgentName         String?
+  extractedLegalAddress      String?   @db.Text
+  extractedZip               String?
+  extractedPolicyFrom        DateTime?
+  extractedPolicyTo          DateTime?
+  extractedPolicyDate        DateTime?
+  extractedCurrency          String?
+  extractedNetPremium        Decimal?  @db.Decimal(12, 2)
+  extractedPolicyFee         Decimal?  @db.Decimal(12, 2)
+  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.
+  extractedCoveragesJson     Json?
+  extractedPremiumPayment    String?
+
+  // Match by `Policy.policyNumber` → existing Policy / Customer.
+  matchedPolicyId   String?
+  matchedPolicy     Policy?   @relation("PolicyOcrDocumentPolicy", fields: [matchedPolicyId], references: [id])
+  matchedCustomerId String?
+  matchedCustomer   Customer? @relation("PolicyOcrDocumentCustomer", fields: [matchedCustomerId], references: [id])
+  /// All policies carrying the same number, with their customer. One is
+  /// normal; >1 means the policy number is shared across customers and a
+  /// human must pick.
+  matchCandidates   Json?
+  matchNote         String?
+
+  reviewedById String?
+  reviewedBy   User?     @relation("PolicyOcrDocumentReviewer", fields: [reviewedById], references: [id])
+  reviewedAt   DateTime?
+
+  createdPolicyId     String?      @unique
+  createdPolicy       Policy?      @relation("PolicyOcrDocumentCreatedPolicy", fields: [createdPolicyId], references: [id])
+  postedTransactionId String?      @unique
+  postedTransaction   Transaction? @relation("PolicyOcrDocumentTransaction", fields: [postedTransactionId], references: [id])
+
+  createdAt DateTime @default(now())
+
+  @@unique([batchId, pageNumber])
+  @@index([status])
+  @@index([matchedCustomerId])
+  @@map("policy_ocr_documents")
+}
+
+enum PolicyOcrDocumentStatus {
+  PENDING_OCR
+  OCR_FAILED
+  NEEDS_REVIEW
+  MATCHED
+  CONFIRMED
+  POSTED
+  REJECTED
+}
+
 // ---------------------------------------------------------------------------
 // Utilities domain
 // ---------------------------------------------------------------------------
@@ -373,6 +494,14 @@ model Property {
   phone2            String?
   phone3            String?
   zone              String?
+  /// Clave catastral (DATMEX.clave) — the cadastral key, format `KA903009`.
+  /// Property-level, not per-service: it is printed on both the CESPT water
+  /// bill and the predial statement, which is exactly why it is a useful
+  /// secondary match key when a bill's account number does not OCR cleanly.
+  /// Distinct from the numeric DATMEX.predial that `PROPERTY_TAX.accountNumber`
+  /// carries — that column is not unique (663 distinct across 1135 rows) and
+  /// is not what any statement prints.
+  cadastralKey      String?
   // Soft-delete marker (see Customer.archivedAt).
   archivedAt        DateTime?
   legacySourceTable String?
@@ -384,6 +513,7 @@ model Property {
   trustAccount TrustAccount?
 
   @@unique([legacySourceTable, legacyId])
+  @@index([cadastralKey])
   @@map("properties")
 }
 
@@ -401,6 +531,13 @@ model PropertyService {
   active        Boolean     @default(true)
   notes         String?     @db.Text
 
+  statementDocuments StatementDocument[]
+
+  // The OCR matcher looks a service up by (kind, accountNumber) — always
+  // scoped to one kind, never fuzzily across every identifier column, so a
+  // water account number cannot collide with an unrelated phone number.
+  @@index([kind, accountNumber])
+  @@index([kind, meterNumber])
   @@map("property_services")
 }
 
@@ -415,6 +552,120 @@ model ServiceDocument {
   @@map("service_documents")
 }
 
+// ---------------------------------------------------------------------------
+// Statement OCR intake (RECEIPT_CAPTURE_SPEC §2)
+//
+// Each utility company mails 300+ paper statements a month, one per customer,
+// which staff key in by hand. These two tables are the intake side of removing
+// that: a batch of scanned PDFs is split per page, OCR'd, matched to a
+// PropertyService by its scoped account number, and queued for review. Nothing
+// here writes to the ledger — confirming a document posts it through
+// `BillingService.createBatch`, the same path hand-keyed batches take.
+//
+// Everything ingested is a CHARGE (a bill awaiting payment), never a proof of
+// payment: the office scans what it must pay, and settles it by check through
+// the existing capture flow.
+// ---------------------------------------------------------------------------
+
+enum StatementBatchStatus {
+  UPLOADED
+  PROCESSING
+  READY_FOR_REVIEW
+  COMPLETED
+  FAILED
+}
+
+enum StatementDocumentStatus {
+  PENDING_OCR
+  OCR_FAILED
+  /// No confident match, or the extraction itself was low-confidence.
+  NEEDS_REVIEW
+  /// Confident auto-match, awaiting a human confirm.
+  MATCHED
+  /// Staff confirmed; not yet posted.
+  CONFIRMED
+  POSTED
+  /// Duplicate, unreadable, or wrong batch.
+  REJECTED
+}
+
+/// One upload session — e.g. "October CFE statements".
+model StatementBatch {
+  id           String               @id @default(uuid())
+  /// What kind of service every statement in this batch bills. The parser
+  /// still detects the provider per page and flags any page that disagrees,
+  /// rather than trusting the uploader's label.
+  serviceKind  ServiceKind
+  status       StatementBatchStatus @default(UPLOADED)
+  uploadedById String
+  uploadedBy   User                 @relation("StatementBatchUploader", fields: [uploadedById], references: [id])
+  label        String?
+  fileCount    Int                  @default(0)
+  /// Set when the pipeline fails as a whole (bad PDF, OCR binaries missing).
+  error        String?              @db.Text
+  createdAt    DateTime             @default(now())
+  completedAt  DateTime?
+
+  documents StatementDocument[]
+
+  @@index([status, createdAt])
+  @@map("statement_batches")
+}
+
+/// One statement — one customer, one period — after splitting the batch.
+model StatementDocument {
+  id         String                  @id @default(uuid())
+  batchId    String
+  batch      StatementBatch          @relation(fields: [batchId], references: [id], onDelete: Cascade)
+  /// 1-based page of the source PDF this was split from.
+  pageNumber Int
+  /// The rendered page image in object storage. The source PDF is kept too, so
+  /// a reviewer can always see exactly what the parser read.
+  storageKey String
+  status     StatementDocumentStatus @default(PENDING_OCR)
+
+  /// Raw OCR text, kept even after a manual correction so a mismatch between
+  /// what the machine read and what staff entered stays auditable.
+  ocrRawText    String?  @db.Text
+  /// Mean per-word confidence reported by the OCR engine, 0..1.
+  ocrConfidence Decimal? @db.Decimal(4, 3)
+  /// Which parser claimed the page ("CFE", "CESPT", "TELNOR").
+  provider      String?
+
+  // Extracted, then staff-corrected in place. `extractedAccountRef` is already
+  // normalised for matching (CFE leading zeros stripped, Telnor LADA removed).
+  extractedAccountRef   String?
+  extractedAmount       Decimal?  @db.Decimal(12, 2)
+  extractedPeriod       String?
+  extractedDueDate      DateTime?
+  /// Clave catastral when the statement prints one — a second key to match on
+  /// when the account number is unreadable.
+  extractedCadastralKey String?
+
+  matchedPropertyServiceId String?
+  matchedPropertyService   PropertyService? @relation(fields: [matchedPropertyServiceId], references: [id])
+  matchedCustomerId        String?
+  matchedCustomer          Customer?        @relation(fields: [matchedCustomerId], references: [id])
+  /// Why this landed where it did — "exact account match", "no candidate",
+  /// "2 candidates". Shown in the review queue so staff can trust or distrust
+  /// the suggestion without opening the image.
+  matchNote                String?
+
+  reviewedById String?
+  reviewedBy   User?     @relation("StatementDocumentReviewer", fields: [reviewedById], references: [id])
+  reviewedAt   DateTime?
+
+  postedTransactionId String?      @unique
+  postedTransaction   Transaction? @relation(fields: [postedTransactionId], references: [id])
+
+  createdAt DateTime @default(now())
+
+  @@unique([batchId, pageNumber])
+  @@index([status])
+  @@index([matchedCustomerId])
+  @@map("statement_documents")
+}
+
 /// From TRUSTVENCE.
 model TrustAccount {
   id          String    @id @default(uuid())
@@ -483,13 +734,17 @@ model Transaction {
   legacyId          String?
   createdAt         DateTime                  @default(now())
 
+  /// Set only on OCR-posted rows — the statement page this came from.
+  statementDocument StatementDocument?
+  policyOcrDocument PolicyOcrDocument? @relation("PolicyOcrDocumentTransaction")
+
+  @@unique([legacySourceDb, legacySourceTable, legacyId])
   @@index([customerId, transactionDate])
   // By-check reconciliation (billing.byCheck / the cheque-count report) looks
   // rows up by check number alone — the legacy EDITA CHEQUE COUNT lookup.
   @@index([checkNumber])
   // Drives the duplicate-post guard in BillingService.createBatch.
   @@index([captureRef])
-  @@unique([legacySourceDb, legacySourceTable, legacyId])
   @@map("transactions")
 }
 
@@ -600,6 +855,12 @@ model User {
   updatedAt    DateTime      @updatedAt
   activityLogs ActivityLog[]
 
+  statementBatches   StatementBatch[]    @relation("StatementBatchUploader")
+  statementsReviewed StatementDocument[] @relation("StatementDocumentReviewer")
+
+  policyOcrBatches  PolicyOcrBatch[]    @relation("PolicyOcrBatchUploader")
+  policyOcrReviewed PolicyOcrDocument[] @relation("PolicyOcrDocumentReviewer")
+
   @@map("users")
 }