> ## Documentation Index > Fetch the complete documentation index at: https://docs.arcuserp.com/llms.txt > Use this file to discover all available pages before exploring further. # Import bulk records > Accepts up to 1000 records for one of the supported resource types named in the `resource` path parameter, returning a `job_id` for async processing, or, with `dry_run=true`, a synchronous preview with no database writes. Deduplicates concurrent submissions via the `Idempotency-Key` header. ## OpenAPI ````yaml /openapi.yaml post /entities/{entity_id}/migration/{resource}/bulk openapi: 3.1.0 info: title: Arcus ERP Public API version: 1.0.0 description: > Arcus ERP public REST API. Designed for external integrations and data migration. **Authentication.** Bearer token (API key) via the `Authorization` header. Format: `Authorization: Bearer ark_live_ent__` (or `ark_test_*` for sandbox). API keys are issued per-entity in **Settings > Developers > API Keys**. **Entity scoping.** The entity is encoded in the API key prefix; routes are flat (e.g. `/v1/accounts`, `/v1/orders`, `/v1/products`). A small set of platform endpoints (migration, reconciliation, events, webhook endpoints, API keys) use the `/v1/entities/{entity_id}/...` form -- those are noted in their tags. **Key capabilities.** - Related-resource hydration via `?expand[]=` (see `x-arcus-expand` on each resource). - Cursor-based pagination (`starting_after` / `ending_before` / `limit`). - Idempotency via the `Idempotency-Key` header. - Webhook events for asynchronous notification. - Conditional requests / ETag for cache validation. **Changelog.** Entries are dated and name every published contract whose MEANING moved, not only the ones whose field names changed. The narrative version of the same entries, written for integrators, is published at https://arcuserp.mintlify.app/changelog. **2026-09-22 (planned 2026-09-21), REORDER-BUYER-TRUTH: demand changed what it MEANS on three published contracts, with no field renamed.** An integrator that pins field names sees no breakage and different numbers, which is why this entry exists. - **Demand now counts build consumption.** `demand_avg_per_day` on the public product, kit and inventory-balance objects, and `daily_demand` / `demand_basis` / `net_suggested_qty` on `GET /v1/purchasing/reorder-report`, are composed from fulfilled sales lines PLUS posted `build_consume` inventory draws: each physical decrement of a product counts exactly once. The previous rule adopted an internal-consumption basis only when the sales blend was exactly zero, so a product both sold AND consumed into work orders planned as if the build draws did not exist. Products drawn into work orders move; on one production-shaped dataset five did, the largest from 0.05/day to 14.95/day. - **`demand_basis` now carries four values, not two:** `sales`, `sales_and_builds`, `builds` and `consumption`. A consumer with a two-branch reader (anything that is not `consumption` is `sales`) silently hides the two new ones. - **`current_demand_units` on `GET /v1/inventory/balances/:id` moved, by a second rule.** It counts committed-but-unshipped CUSTOMER demand, and it now counts a kit component's own line rather than its parent kit line, and excludes non-sales documents. On one production dataset 269 of 1,043 balance rows changed, 227 of them downward; the largest single move was 1,941 to 25, on a product whose open PURCHASE order line had been reported as customer demand. Re-baseline anything that alerts or reorders off this field. - `run_rate_30/90/180/365` and `blended_daily` on the same balance object move for the first reason above. - **Added, not changed:** `GET /v1/purchasing/reorder-report` accepts `demand_window` and returns the unit, cover-through and reorder-point-provenance fields documented on that operation; `order_multiple` is accepted and returned on the product-vendor doors. - **Volume note.** The `inventory.low_stock`, `inventory.out_of_stock` and `inventory.back_in_stock` webhook events are population-gated on `on_hand <= reorder_point`, and reorder points move with the demand above, so subscribers should expect a one-time step change in event volume around the release. servers: - url: https://api.arcuserp.com/v1 description: Arcus ERP API (accepts both live `ark_live_*` and test `ark_test_*` keys) - url: https://dev-api.arcuserp.com/v1 description: >- Dev sandbox API (test-only data, accepts `ark_test_*` keys against dev RDS) security: [] paths: /entities/{entity_id}/migration/{resource}/bulk: post: tags: - Migration summary: Import bulk records description: >- Accepts up to 1000 records for one of the supported resource types named in the `resource` path parameter, returning a `job_id` for async processing, or, with `dry_run=true`, a synchronous preview with no database writes. Deduplicates concurrent submissions via the `Idempotency-Key` header. operationId: postMigrationBulk parameters: - name: entity_id in: path required: true schema: type: string format: uuid - name: resource in: path required: true schema: type: string enum: - accounts - products - orders - journal-entries - vendor-bills - ap-payments - product-categories - product-variants - product-pricing-policies - product-vendors - product-prepackaged-presets - customer-payments - inventory-balances-opening-stock - inventory-balances-set-opening-stock - inventory-transactions - inventory-movements - order-tax-lines - account-addresses - account-contacts - purchase-orders - returns - vendor-credits - name: dry_run in: query required: false schema: type: boolean description: > If true, runs validation synchronously and returns results without any DB writes. - $ref: '#/components/parameters/Idempotency' requestBody: required: true content: application/json: schema: $ref: '#/components/schemas/MigrationBulkRequest' responses: '200': description: Dry-run result (no writes made) content: application/json: schema: $ref: '#/components/schemas/MigrationDryRunResult' '202': description: Job accepted for async processing content: application/json: schema: $ref: '#/components/schemas/MigrationBulkAccepted' '400': $ref: '#/components/responses/Error' '401': $ref: '#/components/responses/Error' '403': $ref: '#/components/responses/Error' '409': description: Idempotency key collision (different request body for same key) content: application/json: schema: $ref: '#/components/schemas/ErrorEnvelope' security: - ApiKeyAuth: [] components: parameters: Idempotency: name: Idempotency-Key in: header required: false description: > Client-generated unique key for idempotent POST/PATCH/DELETE operations. Max 255 chars. On retry with the same key, the original response is returned without re-executing the operation. Keys expire after 24 hours. Pattern: or --. schema: type: string maxLength: 255 schemas: MigrationBulkRequest: type: object required: - records - external_source properties: records: type: array minItems: 1 maxItems: 1000 items: type: object description: > Resource payload matching the canonical create handler's body for the URL `{resource}`. See the canonical create endpoint (POST /v1/accounts, /v1/products, /v1/orders, /v1/journal-entries, /v1/vendor-bills, /v1/ap-payments) for the full accepted field set. **Always present (auto-injected by the dispatch loop):** - `entity_id` is silently dropped from the record; the value is derived from the API key's bound entity (Layer 1 isolation). - `external_source` from the top-level request body is stamped onto every record. - `migration_batch_id` from the top-level request body is stamped onto every record. - `imported_at` defaults to the server's now() but is overridden by any caller-supplied value (e.g. preserve the source ERP's original `created_at` for audit fidelity). **Migration-mode flags (auto-applied by the dispatch loop per-resource -- callers do NOT set):** - `resource=journal-entries`: `bypass_period_check=true` (allows posting to closed historical periods; gated by `migration:write` scope at the route layer). - `resource=vendor-bills`: `suppress_gl=true` (prevents double-posting JEs already loaded via `resource=journal-entries`; the canonical `postJournalEntry` handler is the SSOT for JE writes). - `resource=ap-payments`: `suppress_gl=true` AND `skip_plaid=true` (the latter critical: historical ACH payments must NEVER trigger live Plaid Transfer attempts on entities with Plaid configured). - `resource=purchase-orders`: `is_historical_import=true` is injected per record by the dispatch loop; `createPurchaseOrder` branches on it into its historical sister `createMigratedPurchaseOrder`, which preserves the source PO number / dates / status and writes the line set atomically (no native counter advance, no draft->open auto-transition). **Resource-specific record shape -- `resource=purchase-orders`:** Imports a settled purchase order from a prior ERP as an Arcus PO (`orders.document_type='purchase_order'` + `order_items` rows). A PO carries a VENDOR account, not a customer -- distinct from `resource=orders` (sales orders). Per-record fields: - `account_id` (required, UUID): the Arcus vendor account. - `order_number` (optional): the source PO number (e.g. Versa `PO-42122`). Preserved verbatim on `orders.order_number`; the native PO counter is NOT advanced. Omit to let the server assign a counter number. - `order_date` (optional, ISO 8601): source PO date; defaults to `imported_at`. - `expected_date` (optional, date): expected receipt date. - `order_status` / `fulfillment_status` / `payment_status` (optional): source statuses; preserved verbatim (lower-cased for enum cast). - `subtotal`, `tax_total`, `order_total`, `amount_paid`, `balance_due` (optional, numeric): source header totals. When `order_total` is supplied it is kept verbatim; otherwise the header totals are recalculated from the line set. - `location_id`, `shipping_address_id` (optional, UUID): source-document links. - `notes`, `internal_notes` (optional): free text. - `items` (array): PO lines. Each: `product_id` (required, UUID), `quantity` (required, positive integer), `unit_cost` (optional, numeric -- per-unit cost; `sell_rate` accepted as an alias), `variant_id` (optional, UUID), `product_title` (optional, snapshot name), `tax_amount` (optional, numeric), `notes` (optional), `sort_order` (optional, integer). **Required per record:** - `external_id`: source-system primary key for this record (e.g. Versa's product_id). Required for idempotency: re-running the same bulk request with the same external_id is a no-op when `conflict_mode=skip`. The partial UNIQUE on `(entity_id, external_source, external_id) WHERE external_source IS NOT NULL` enforces this DB-side. **Optional per record:** - `external_url`: deep-link back to the source-system row (e.g. https://versa.example.com/products/1019471) - All other fields documented on the canonical create endpoint for the resource. **Resource-specific record shape -- `resource=orders` (STATUS<->MONEY COHERENCE gate G28):** Migrated sales orders / invoices carry the source header money (`amount_paid` + `balance_due`/`remaining_balance`) VERBATIM. An ingest coherence gate runs on each header BEFORE the write (visible in `?dry_run=true` as per-record `errors[]`): - A DRAFT / unconfirmed document (`document_type=quote`, or `order_status` in draft/unconfirmed) carrying `balance_due > 0` is REJECTED (`code=draft_carrying_balance`, HTTP 422 per record): a draft is not a receivable. Confirm the document in the source, or import it with `balance_due=0`. NOT overridable. - A PHANTOM-PAID header (`amount_paid` greater than the committed payment evidence for the order) is QUARANTINED (`code=phantom_paid_header`, HTTP 422 per record) with the delta to the cent. This surfaces a polluted source for adjudication; it never batch-aborts (other records still load). - `accept_header_money` (boolean, optional): set `true` on a record to OVERRIDE a phantom-paid quarantine and load the header money verbatim -- for a deliberate migration whose source payment detail is genuinely incomplete. The override is AUDITED (activity_log action `migration.header_money_accepted`). It does NOT override a draft-carrying-balance reject. - A paid-but-open header (committed payments cover the open balance) LOADS with an informational annotation (the derive path corrects it); a fulfilled-but-never-invoiced header LOADS with an advisory annotation. **Resource-specific record shape -- `resource=inventory-transactions`:** This resource imports the per-product movement HISTORY (sales + purchases) that the demand/reorder engine reads, NOT current on-hand (current on-hand is owned by `resource=inventory-balances-opening-stock`). It has no FIFO / balance / GL side effects. Because the `inventory.inventory_transactions` table has no `external_source` / `external_id` columns, idempotency for this resource is keyed on `(entity_id, source_transaction_id, reference_type)`; `external_id` is ignored. Per-record fields: - `product_id` (required, UUID): Arcus product. - `location_id` (required, UUID): Arcus location. - `type` (required): an `inventory_txn_type` enum value (`sale`, `receiving`, `rma`, `transfer_out`, ...). - `quantity` (required, signed integer): negative = outbound (sale/consumption), positive = inbound (receipt). - `posted_at` (required, ISO 8601): the source system's posting date. This is preserved verbatim onto BOTH `posted_at` and `created_at` (the demand engine reads `created_at` inside a 90-day window). - `source_transaction_id` (required, string ≤100 chars): the source system's transaction id. Idempotency key + stored in `transaction_id` + `reference_number`. - `created_at` (optional): source created timestamp; defaults to `posted_at`. - `reference_type` (optional): defaults `versa_migration`. - `unit_price`, `total_price` (optional, numeric): per-unit and extended movement value. - `supply_demand` (optional): `supply` / `demand` / `neutral`; inferred from `type` if absent. - `account_id`, `order_id`, `order_item_id`, `sales_channel_id` (optional, UUID): source-document links. - `notes` (optional): free text (source memo). external_source: type: string maxLength: 60 description: > Source system identifier (e.g. "versa_cloud", "quickbooks_online", "netsuite", "shopify_export"). Required. Persisted on every record's `external_source` column. Used by GET /v1/entities/{entity_id}/reconciliation/counts?external_source= to filter counts to a specific source. conflict_mode: $ref: '#/components/schemas/MigrationConflictMode' migration_batch_id: type: string format: uuid description: > Optional batch identifier you generate once per logical migration "wave" and pass on every record. Persisted on every record's `migration_batch_id` column. Used to group records that were imported together for later reconciliation or rollback. dry_run: type: boolean default: false description: > If true, the records are validated end-to-end but no DB writes occur. Useful for catching missing required fields and FK violations before committing a real import. MigrationDryRunResult: type: object description: Dry-run result (200 response for ?dry_run=true). No DB writes made. properties: object: type: string enum: - migration_dry_run_result dry_run: type: boolean enum: - true resource: type: string external_source: type: string total_records: type: integer succeeded_records: type: integer failed_records: type: integer errors: type: array items: type: object MigrationBulkAccepted: type: object description: > Async bulk import accepted (HTTP 202). The job has been queued and a background runner is kicked immediately, so processing typically begins within a second and all records in the batch are processed back-to-back. Poll `GET /v1/entities/{entity_id}/migration/jobs/{id}` for status, error array, and aggregate counts. required: - object - id - status - resource - dry_run - conflict_mode - external_source - total_records - created_at properties: object: type: string enum: - migration_job id: type: string format: uuid description: The migration job ID. Use this to poll `/migration/jobs/{id}`. status: type: string enum: - pending - in_progress - succeeded - failed - partial description: >- Initial value is `pending`; transitions to `in_progress` when the runner picks it up, then to `succeeded` / `failed` / `partial` based on per-record outcomes. resource: type: string enum: - accounts - products - orders - journal-entries - vendor-bills - ap-payments - product-categories - product-variants - product-pricing-policies - product-vendors - product-prepackaged-presets - customer-payments - inventory-balances-opening-stock - inventory-balances-set-opening-stock - inventory-transactions - order-tax-lines - account-addresses - account-contacts - purchase-orders - returns description: The resource path segment this job is loading. dry_run: type: boolean description: >- Reflects the `dry_run` flag from the request. When `true`, no DB writes; the job's final report shows what WOULD have happened. conflict_mode: $ref: '#/components/schemas/MigrationConflictMode' external_source: type: string description: The source-system identifier stamped on every record in this batch. total_records: type: integer description: Number of records queued in this batch. created_at: type: string format: date-time ErrorEnvelope: type: object description: | Canonical error response envelope. All API errors use this shape. required: - error - code properties: error: type: string description: Machine-readable error key example: not_found code: type: string description: Machine-readable error code (often same as error) example: not_found type: type: string enum: - validation_error - permission_error - not_found - conflict - rate_limit - internal - expand_error - not_implemented example: not_found hint: type: string description: Human-readable one-sentence explanation (English) example: >- The requested order does not exist or does not belong to this entity. param: type: string description: The parameter that caused the error, if applicable example: expand[0] required: type: string description: The scope required (only on insufficient_scope errors) example: accounts:read request_id: type: string description: >- Unique request ID for support tracing (maps to CloudWatch log stream) example: req_abc123 MigrationConflictMode: type: string enum: - skip - update - replace - error default: skip description: > How to handle a record whose (external_source, external_id) already exists in Arcus. - skip: leave the existing record unchanged, count as success - update: call the canonical UPDATE handler with the incoming payload - replace: NULL out the existing record's provenance UNIQUE, then INSERT new record - error: abort the entire batch on the first collision (default if not specified: skip) responses: Error: description: Error response (400/401/403/404/409/422/429/500) content: application/json: schema: $ref: '#/components/schemas/ErrorEnvelope' securitySchemes: ApiKeyAuth: type: http scheme: bearer description: | API key issued per entity via Settings > Developers > API Keys. Each key carries scopes (e.g. orders:read, products:write). Bearer token format: Authorization: Bearer ark_live_ent__ Test keys use ark_test_ent__. Both are issued per entity via Settings > Developers > API Keys. ```` This documentation is built and hosted on [Mintlify](https://mintlify.com), a developer documentation platform.