New Features in This Release

The following new features are incorporated in this release CS 8.1.3.1.0.

The application now has the following features and capabilities:
  • Installer and Configuration
    • Installer defaults optimized for higher load: Default installation values for connection pools, thread pools, job cleanup, and session timeouts have been tuned to better support increased concurrency.
    • Standardized configuration properties (migration required): Configuration properties have been reorganized and reclassified with standardized prefixes (for example, MMG_ and AAI_) and grouped by service for clarity and stability.
  • Security and Governance
    • API Registry: Added support for basic authentication actions.
    • Unified session-based identity management: UI REST requests now derive user identity from the active authenticated session, removing legacy re-authentication prompts and strengthening governance controls.
    • More granular access control: Oversight capabilities (Health Dashboard, Model Report, Data Management) are now decoupled from broad administrative roles (for example, Health Dashboard no longer requires the Identity Admin role), improving separation of duties.
  • User Experience and Auditing
    • File Manager enhancements:
      • Enhanced auditing for all file operations.
      • Folder bookmarking (favorites) for quick access.
      • Opens the last accessed directory by default.
    • Workspace auditability: Audit Trail now captures workspace-related operations.
    • Notebook review improvements: Paragraph-level comments are now supported in notebooks and are captured in the audit/review process.
  • Pipelines, Datasets, and Model Lifecycle
    • Datasets: Push down filtering implemented in the dataset widget within pipelines.
    • Model Pipelines:
      • Pull/push objectives or drafts across workspaces.
      • User-level restriction for self-approval of model deployments.
      • Session-based filtering in executions with access to the session notebook view.
      • Canvas validation to detect cyclic loops with options to highlight cycle-causing links and flatten execution order based on notebook order.
      • New widget for windowed computations(tumbling/sliding) with configurable processing of Kafka event streams.
      • In the pipeline, EDA widgets are added for data profiling (distributions, missing values, outliers).
    • Bench marking and pipeline comparison expanded: compare any model against other pipelines within an objective/sub-objective and view latest execution results.
    • Traceable node-level execution: Execute individual pipeline nodes using a custom session for focused, auditable testing.
  • Operational Efficiency and Administration
    • Asynchronous Conda environment provisioning: Environment builds (YAML upload or UI) now run in the background, keeping the UI responsive.
    • Unified model management script: Consolidated shell utility for importing/exporting/retiring models and managing parameter sets.
    • Execution and monitoring utility: New internal script-based utility to execute and monitor processes without using a separate scheduler interface.
    • Scheduler time zone configuration: Installation now supports setting a preferred scheduler UI time zone (no longer UTC-only).
  • Performance Improvements
    • Reduced calls and time required for workspace attachment during notebook initialization, lowering latency.
    • Improved First Content Paint (FCP)by reducing synchronous client-server calls and using asynchronous AJAX calls.
    • GZIP compression: Up to 90% compression for eligible responses to reduce bandwidth usage. Images are also optimized across modules to reduce network overhead.
  • Technology Stack and Dependencies
    • The Conda environment registry now supports Python versions 3.10 through 3.12.
    • Python upgrade: Updated Python 3.12.11.
    • Model Catalog visualization resilience: Scoring/visualization flow improved as graphs still render even when some metrics (for example, ROC/AUC) cannot be computed; unavailable metrics show neutral values with clear messages.
  • Automated Scenario Calibration (ASC) Enhancements
  • ASC production workflow coverage: BTL analysis is now supported as part of the ASC production workflow.
  • ASC notebook enhancements:
    • Added a Download Recommendations notebook to provide a unified view of all available versions after analysis, with the ability to download recommendations.
    • Centralized threshold set management: Threshold set management is now incorporated into ASC centralize threshold versioning and scenario management with audit-ready lineage.
  • Expanded staggered tuning coverage: Staggered Tuning Analysis has extended support for BTL analysis.
  • Entity Resolution

    • Added support for using Oracle Text for out-of-box (OOB) and custom rules, eliminating the need to deploy additional Open Search infrastructure for candidate selection.

    • Improved logging for high-volume processing: Enhanced logging and run time improvements for ER Job 1 to provide clearer progress and diagnostics and shorter batch times when processing large data volumes.
    • ER Job 3 performance enhancements: Implemented performance improvements to reduce runtime and improve throughput for ER Job 3.
    • Persist ER job parameters: ER job parameters are now persisted to improve repeatability and reduce the need for reconfiguration between runs.
    • Skip Open Search index validation after refresh (CS8125): Added an option to skip Open Search index validation after an ER refresh run to reduce post-run overhead where validation is not required. (Verify this aligns with your organization’s operational and governance requirements before enabling.)
    • Expired record cleanup in staging tables: Improved handling of expired records as they are no longer retained unexpectedly in the following:
      • H$STG_PARTY_ADDRESS_MAP_PRE
      • H$STG_ADDRESS_MASTER_PRE
    • Data Survival job enhancement: Updated the ER Data Survival job to compute modified global parties, and improve downstream accuracy and traceability.
  • Data and Integration
    • Consolidated AML Event Layer: Introduced a consolidated AML event layer to standardize event handling across AML workflows.
  • Data Survival job enhancement: Updated the ER Data Survival job to compute modified global parties, and improve downstream accuracy and traceability.