Skip to content

August 2026

The following enhancements, fixes, and completed work shipped in Aunsight Data Services during August 2026. Changes are summarized below by feature area.

Mapping Mode — Map-Assist & QA-Assist in Dataflow Builder

A new Mapping Mode brings Map-Assist and QA-Assist directly into the Dataflow Builder, giving data engineers AI-assisted mapping and quality review in the same place they build dataflows.

  • New — DF-Builder bottom panel and Assist menu: a new bottom panel (resizable by dragging) and an Assist menu in the DF-Builder header host the Map-Assist and QA-Assist experiences; both are hidden in custom-project dataflow builders.
  • New — User artifacts supported as a top-level argument on solution-installation action runs, underpinning the mapping-mode workflows.
  • New — Mapping candidate approval lifecycle: an API plus DF-Builder UI to view and transition a mapping candidate's status, with mapped-model table status and candidate counts shown on the Mapping tab.
  • New — Run QA-Assist from the DF-Builder and read the report in the bottom panel; retrieve the most recent report when reopening a dataflow; download the report as Markdown.
  • New — Findings are highlighted in the dataflow by data path and severity, and users can navigate directly from a finding to its operation on the canvas.
  • New — Issues tab with separate Structural and QA-Assist sections; QA-Assist outputs adjusted to a new human/machine report contract; Map-Assist outputs adjusted to the new artifact contract.
  • New — Review tooling: review-mode notes and an output-field checklist API for candidate review.

Next-Generation Automapping

A rebuilt automapping pipeline that narrows candidates with ERM and domain context, extracts and distills source documentation, and generates mapping plans and candidate dataflows with confidence scoring.

  • New — Candidate narrowing: use industry domain and ERM key candidates to narrow candidate tables, then use the narrowed tables to narrow candidate fields, and reduce the input sent to the LLM.
  • New — LLM tooling and prompt development: dedicated narrowing and association tools with tuned prompts.
  • New — Document extraction tool plus new Extract and Distill/Synthesis Solution Actions to bring source documentation into the mapping process.
  • New — Mapping generation: associate Mapped Model fields with Local Source Model tables/fields, generate a mapping plan, a mapping-logic description, a candidate dataflow script, and a summary report — with confidence scoring and defined weighting factors.
  • New — New Auto-mapping Solution Action wiring the next-gen Map-Assist together, with support for the case of no passed source fields and ongoing iteration.
  • New — Extraction of items in different formats into the knowledge base to feed the mapping process.

Mapped-Model Validation Framework

A framework and scenario library for validating Mapped Model output, supporting QA-Assist conformance checks.

  • New — Framework built to execute a QA test suite, execute custom metrics in the warehouse query profilers, and evaluate conformance and bound checks from profile-request settings.
  • New — Validation test scenarios created around datatype constraints and field-level constraints for the Mapped Model.
  • New — Test environments built for multiple client solutions; testing milestones defined and validation markdowns built for the remaining Mapped Model tables.

Data Controls & Pipeline Status Reporting

A new model-metrics and pipeline-status reporting capability that surfaces completeness, integrity, and schema-quality signals and can email clients after a successful pipeline run.

  • New — Completeness & Counts: per-deliverable-model table row-count change checks, plus staging completeness confirmation and Local Source Model finalization checks.
  • New — Duplicates & Integrity: duplicate-rate detection and identification of affected delivery tables.
  • New — Schema & Quality Checks: detect source and deliverable schema changes and related quality issues.
  • New — Automated client emails after a successful pipeline run, reporting pipeline status.
  • New — Custom selective-publish script for ERM tier changes on accelerated warehouses.

Sync Agent Blackout Windows

Configurable blackout windows let operators pause Sync Agent activity during defined periods.

  • New — Blackout Window DTO and CRUD API; the Sync Agent now honors configured blackout windows and manual pauses.
  • New — Read-only display of Sync Agent blackout status and an edit form for configuring a blackout window.

Data Acquisition — Plugins

New and expanded source connectors.

  • New — Salesforce plugin: read capability for Financial Services Cloud.
  • New — New Sync-Agent plugin to read from AU2 (Drill plugin).

Warehouses, Buckets & Toolbelt

Enhancements to warehouse write modes, local buckets, export, and the toolbelt.

  • New — Upsert-delete write mode added for Local Warehouse tables, with support for upsert-delete on Store Warehouse operations in dataflows.
  • New — Remote Bucket → User Bucket transfer task implemented.
  • New — File Browser now supports Local Buckets on the Solution Config UI.
  • New — Omit tables/views with no columns during warehouse discovery.
  • New — Public API commands exposed as MCP tools via the toolbelt.
  • New — New duckdb export format.
  • New — Task ProgressManager refactored to autocomplete progressor keys.

StarRocks Accelerated Warehouse (SRDB)

Continued StarRocks support completed this cycle.

  • New — Lib-solution / IDW-F support for a StarRocks AWHS as a delivery target.
  • New — Solution images updated for additional client solutions to add StarRocks support.

Data Lineage

Data-lineage tooling and delivery.

  • New — Dataflow Trace added to the lineage export, and a Data Lineage report delivered for Account tables.

Bug fixes

Defects resolved during August 2026.

  • Fix — AU3 YearsBetween and MonthsBetween expressions corrected.
  • Fix — Golden Records failing to form on publish resolved.
  • Fix — Ingestion task failure fixed where NVARCHAR(MAX) was converted to an invalid VARCHAR.
  • Fix — Compute Gateway now finds the latest Sync Agent version.
  • Fix — Oracle plugin discovery no longer populates an empty warehouse (no tables/views).
  • Fix — TaskLifecycleError fixed where the task executor ended while a task was still running.
  • Fix — Expanded-combinations warning no longer fires incorrectly on local-to-accelerated transfers.
  • Fix — Dataflow operation IDs past dfop-999 no longer generate with a comma.
  • Fix — Discrepancy between bucket docs and auto-generated write settings resolved.
  • Fix — StarRocks delivery transfers no longer stall for tens of minutes in "gathering summary statistics".
  • Fix — Map-Assist per-table analyze no longer fails with a type-comparison error.
  • Fix — DF-Builder: opening the operation panel no longer hides the contents of the Save and Run sub-menus.
  • Fix — Mapping-candidate transition status issues resolved.
  • Fix — Discovered warehouse data storage adjusted to handle large payloads.