Changelog

This page tracks changes to the data exports API. Additive changes (new datasets, new fields, new optional query parameters) stay on the same schema version. Breaking changes — renames, removals, semantic shifts — land at a new endpoint version (e.g. /exports/v2/...) so existing pipelines continue to work unchanged.

​

v1.2 — 2026-06-11

​

Added

• Activated 9 datasets that previously appeared in the catalog as available: false:
  • S1 Customers & Accounts: D01 Borrowers, D02 Borrower Accounts, D03 Borrower Sub Accounts, D05 Borrower Account Events.
  • S3 Collections Cases: D16 Borrower Deals, D17 Borrower Deal Activity Logs, D18 Borrower Deal Logs, D19 Deal Activity Logs, D20 Deal Views.
• public_id field exposed on the newly activated datasets that lacked one (D03, D17, D18, D19, D20) — string-cast Mongo _id, the same convention as the v1.1 additions to D15/D21/D22/D23. D05 exposes its native public_id.

​

Changed

• D01 Borrowers and D02 Borrower Accounts — the per-row identifier is now exposed as public_id sourced from the string-cast Mongo _id. (Earlier internal drafts surfaced a business identifier; the shipped field is the _id-derived value.)
• D16 Borrower Deals — the per-row identifier is now exposed as public_id (previously case_public_id, while D16 was unavailable). The underlying value is unchanged — it’s still user_lender_detail_public_id from the source — only the response field name is now consistent with the rest of the API. Note this is the case/link id and may not be unique per deal row; see the dataset page. If you wrote a client against an internal preview that used case_public_id, rename the column in your warehouse mapping.
• D01 Borrowers — phones array elements now contain key_masked (the phone-middle-masked number) and dnc. The source array on the underlying collection is phone (singular); the response surfaces it as phones. preferred_language is sourced from pref_lang in the model — only the rendered name changes.
• D16 Borrower Deals — dropped two fields from the allowlist (first_due_date_timestamp, everReachedAttempted); they aren’t present in the underlying Mongoose model or any sampled doc.
• Pagination doc updated — TIER 2 (variable page size) datasets now include D03, D17, D18, D19 in addition to D21, D22, D23.

No breaking changes. meta.schemaVersion stays at "v1".

​

v1.1 — 2026-06-08

​

Added

• public_id field on the following datasets: D15 Borrower Settlements, D21 Communication Logs, D22 Conversations, and D23 Manual Communication Logs. Use it as the stable per-row identifier for warehouse upserts. On D21/D22/D23 it is derived from the underlying Mongo _id; on D15 it is the record’s native public_id.
• body.schemaVersion: "v1" on the catalog endpoint response.
• meta.schemaVersion: "v1" on every fetch-dataset response.
• meta.tenantPathKind on fetch-dataset responses — "direct" or "via". Tells the client whether to expect constant or variable page sizes (see Pagination → Variable page size).
• meta.readFrom on fetch-dataset responses — "primary" or "secondary". Indicates which replica served the request.
• Optional query parameter ?includeTotal=true on fetch-dataset. Returns pagination.total (filter-wide row count) when set, with guarded semantics for via (TIER 2) datasets and large/slow counts. See Fetch Dataset → Query parameters and Pagination → Filter-wide totals.
• pagination.total / pagination.totalExact / optional pagination.totalNote keys, populated only when includeTotal=true.

​

Fixed

• via (TIER 2) datasets (D21, D22, D23) no longer truncate exports silently on a full final page. Previously, a page where limit <= candidateCount < overfetchLimit could return nextCursor: null even when more matching rows existed beyond the page. The cursor-advance logic now correctly fires the “page is full → advance to last returned row” branch regardless of whether the candidate window was completely filled.
  If you already ingested data under the v1.0 behavior, consider running a full backfill on D21, D22, and D23 to pick up any rows that may have been truncated.

​

v1.0 — initial Phase 1 release

​

Added

• Two new endpoints under the API Reference tab: GET /exports/catalog and GET /exports/{sectionCode}/{datasetCode}.
• Cursor-based pagination via opaque nextCursor tokens, with support for initial load (omit since) and incremental load (pass the last batch’s max cursor-field value as since).
• 10 datasets available for fetch: D07 Borrower Payments, D08 Borrower Payment Plans, D09 Borrower Payment Plan Installments, D12 Scheduled Payments, D14 Payment Proofs, D15 Borrower Settlements, D21 Communication Logs, D22 Conversations, D23 Manual Communication Logs, and D26 User Lender Details.
• 15 additional dataset codes reserved in the catalog (available: false). Phase 2 will activate codes for S1 Customers & Accounts and S3 Collections Cases.
• Per-dataset field allowlists with masking applied at the database layer (id_number and passport_number masked to last 4, dob reduced to year, ip_address reduced to /16).
• Two-layer tenant isolation: tenant DB resolution via JWT, plus lender_id filtering inside the tenant DB.
• Strict requireLenderMatch middleware on export endpoints, enforcing that the URL lenderPublicId equals the JWT’s lenderId.

---