DQC Logo
|

v2.39

This release speeds up big workloads with asynchronous previews and high-throughput batching—cutting run times, reducing timeouts, and boosting stability. It also strengthens AI rule generation, segmentation, and connectors for smoother operations across the DQC Platform.

undefined New Features

A variety of new features has made it into the 2.39 release - see below!

Find / rules

Sample segmentation for validation

  • Enables segmentation of samples and rule-specific “Valid” samples

  • Helps pinpoint where quality drifts occur and compare outcomes across cohorts

Export issues for a segment

  • Download issues filtered by any chosen segment to speed up triage and hand-offs to business owners

Ruleset import/export

Preserve metadata + confirmation toast

  • Exported and imported rulesets now keep their tags and quality dimensions, so filters, dashboards, and governance reporting work immediately after migration

  • An import confirmation toast provides instant feedback on success/failure of imported tags and dimensions, reducing guesswork and speeding up troubleshooting

Improvements

Python node: context files

  • Attach context files to Python nodes so remediation code can reference mapping files or samples

  • Improves reproducibility of fixes within DQC Platform improvement workflows

Asynchronous previews

  • Preview node in the improvement workflows now run asynchronously on the backend, keeping the UI responsive and reducing timeouts on large datasets

General

High-throughput batch processing

  • Improved parallelism with concurrent batches per worker and smarter pool limits to maximize throughput

  • Stabilized batching (fewer DB calls, cleaner logs, safer cleanup) for long-running workflows

  • Benefits: shorter end-to-end runtime, better resource utilization, and higher reliability under load

  • Scope: LLM rule, Document (PDF) processing, and improvement workflow

undefined Small improvements and bug fixes

  • Better error display across the platform to shorten troubleshooting

  • Fixed CSV UTF8-SIG import in improvement workflows

  • Python node coding assistant errors resolved; improved prompts

Connector & execution hardening

  • OData: OAuth2 for OData v4 and a more robust connector view (simplified parameters)

  • PDF connector refinements (linking, orientation/long names, safer deletes, Parquet write fixes)

AI rule generation: reliability & fallbacks

  • Adds a fallback SQL rule when the DQ-AI Assistant cannot confidently predict a rule, ensuring checks still run

  • Consolidates multiple rule agents on a shared base for consistent behavior and simpler evolution

  • Prompt improvements for custom SQL/Python/Regex rules reduce retries and clarify expected outputs

undefined Breaking Changes

None in this release