June 2026

June 2026 release notes for Inventory Management.

June 15, 2026

Optional columns in standard schemas

Standard-schema columns can now be declared optional. Each column's validations must include exactly one of { "type": "required" } or { "type": "optional" }: a required column's key must be present in every ingest payload, while an optional column can be omitted and is stored as NULL (or its defaultValue).

Primary-key columns must stay required, and nonNull may only be combined with required. Applies to the save-schema, add-column, and custom-table endpoints. See Schema lifecycle.

Data Quality Dashboard redesign

The Data Quality Dashboard has been rebuilt as a single scrolling overview page with a table-level drill-down. It now opens with a Monitoring filters bar (start date, end date, solution) and a Current health snapshot of five KPI cards — total requests, rows processed, validation failures, rows ingested, and pass rate — each with period-over-period comparisons. Below that are three trend charts (error-code trend, validation results over time, and pass rate over time), a ranked Top error codes table, searchable Breakdown by table and Breakdown by request views, and a Current slice summary. Selecting a table opens a detail view scoped to that table, solution, and date range.

June 3, 2026

New: Integration walkthrough

A new Integration walkthrough page in the API Guide walks customer IT teams, solution engineers, and support engineers through the four stages of integrating with the Data Ingestion API: extract, transform, align field names, and batch + send. Each stage has the relevant API calls embedded inline — GET /api/v2/schema/solutions to confirm what's rolled out, POST /api/v2/objects/{objectName} for ingesting batches (with a dryRun variant), and POST /api/v2/schema/{objectName}/add-attribute for adding tenant-specific columns — alongside Snowflake COPY INTO and Redshift UNLOAD recipes for the extract step. Designed as a single one-stop guide rather than scattered references.

Bulk-ingest guide: folder-load sort order spelled out

The Historical data ingestion guide's "File or folder?" section now spells out the folder-load sort-order rules. Lexicographic sort gives the right order when filenames encode order as zero-padded numbers (part-0001.csv) or ISO-style date prefixes (2024-06-01.csv) — both forms warehouse exports produce by default. It gives the wrong order for non-padded part numbers, DD-MM-YYYY or MM-DD-YYYY date names, or month-name files like January_report.csv. The doc now lists each failure case in a comparison table and recommends the safe naming pattern.

June 2, 2026

Historical data ingestion guide

A new guide covers loading large historical datasets (millions to billions of rows) into the Data Ingestion API. It explains the design considerations for at-scale loads — streaming, dynamic batching, rate-limiting, foreign-key ordering, checkpointing, and failed-row handling — and provides a self-contained Python reference script you can configure and run. See Historical data ingestion.

Bulk-ingest script

The Python reference script in the historical-data-ingestion guide:

• Loads one table per run with a minimal CONFIG block (csv_path, object_name, solution_name, operation_type, parallel). To load several tables, run the script once per table.
• Accepts a folder of CSV part-files as the source, in addition to a single CSV file. Chunked exports are loaded in sorted filename order.
• Reads gzip-compressed CSVs transparently, detected from the .gz extension.
• Recognises both empty string (Redshift UNLOAD default) and \N (Snowflake COPY INTO default) as null markers.
• Fails fast if a CSV row has more cells than the header, instead of silently shifting data into the wrong columns.
• Tolerates uncastable cells (e.g. "abc" in an integer column) by passing them through to the API as raw strings, so one bad cell becomes one rejected row, not a stopped load.
• Exposes a parallel toggle for UPSERT loads where a primary key can appear more than once in the source.

The guide also includes warehouse-specific UNLOAD / COPY INTO recipes for Snowflake and Redshift, and a section on re-running the script (resume behaviour, where checkpoint and failed-row files live, how to start fresh).

Operation type guidance

The API guide now frames the choice between APPEND and UPSERT around what the downstream solution needs from the data — current state vs full history — rather than around data type.

Add-column endpoint: nullability guidance

The Add a new column section now has a Controlling required-ness and nullability on the new column subsection. It documents that required and nonNull are independent validators in the add-column path and lays out the four valid combinations as a small table:

• Key required + value non-null
• Key required + value nullable
• Key optional + value non-null
• Key optional + value nullable

The subsection also calls out the API-enforced mutual exclusion of defaultValue and nonNull — sending both returns 400 Bad Request with the exact error message the API produces ("Cannot use 'nonNull' validation with defaultValue. defaultValue ensures the field always has a value."), so customers can match the failure in their integration logs.

These guarantees apply equally to the Inventory Management solution; the doc clarification is shared across all Supply Chain & Retail solutions that use the Data Ingestion API.