New Features, Enhancements and Limitations in this Release

To improve administrative clarity and system stability, we have implemented a structured reorganization of the configuration properties.

Enhancement: We have introduced a standardized prefixing system, for example, MMG_ for Model Management and Governance and AAI_ for Analytical Infrastructure, and have logically grouped all properties by their respective services.
Migration Requirement: This update includes a structural property realignment. When upgrading to version 26.0.0, the installation process requires you to input values by using these updated property names.
Business Impact: This ensures that the origin and functional intent of every parameter is immediately identifiable, thus reducing configuration errors during deployment. To support a smooth transition, refer to the Configuration Property Names section in the Oracle Financial Services Model Management and Governance Installation Guide, which maps the legacy property names to the new standardized names.

Governance and Security Framework

Unified Session-Based Identity Management
OFS MMG has transitioned to an automated, session-based identification process for all the user interface REST requests, significantly strengthening the platform’s security posture.
- Enhancement: The system now programmatically derives your identity directly from the active authenticated session, thus eliminating legacy re-authentication prompts.
- Business Impact: Previously, manual re-validation steps presented a governance risk, as you could potentially enter alternative credentials during a session to intercept access levels. By centralizing identity management within the internal security framework, we have closed this gap. This ensures that every action is strictly verified against the original trusted session, thus providing a more seamless and secure user experience.
Granular Access Control and Functional Decoupling
- Enhancement: We have refined our access model to provide role-specific precision, ensuring that you have access to the exact tools required for their functions without over-extending permissions. This update specifically addresses modules where access was previously tied to broad roles, including the Health Dashboard, Model Report, and Data Management. We have now decoupled these oversight functions from broad administrative roles; for example, accessing the Health Dashboard no longer requires the Identity Admin role, which previously carried the sensitive power to modify user roles and permissions.
- Business Impact: This granular approach allows organizations to grant visibility into system health, data management, and model reporting to a wider group of stakeholders while keeping administrative identity controls restricted to a small number of core administrators. By applying these specific controls to the designated modules, teams can operate with higher security and clearer separation of duties.

Operational Efficiency and Experience

Intelligent Workspace Recovery (Retry Capability)
A new recovery mechanism for OFS STSA and simulation workspaces has been introduced to eliminate the need for manual technical cleanups during provisioning interruptions.
- Enhancement: If a workspace setup is interrupted by transient issues, such as network restrictions or missing metadata, you previously had to manually identify and delete orphaned background processes, database schemas, and information domains before restarting.
- Business Impact: The system now offers a dedicated Retry option. On failure, the system provides clear diagnostic logs and, when triggered, automatically manages the backend cleanup before resuming the setup from the point of progress. This streamlines the path to a functional sandbox environment and reduces the need for technical support intervention.
Asynchronous Environment Provisioning (Conda UI)
The Conda environment creation workflow has been optimized to support continuous productivity.
- Enhancement: The Conda environment creation workflow has been optimized to support continuous productivity. Whether initiating a build through a YAML file upload or directly through the user interface, the system now processes the environment setup and package installations in the background. Previously, the interface would remain locked during this period, but the process is now fully asynchronous, allowing the user interface to remain responsive throughout the duration of the build.
- Business Impact: You can initiate an environment build and immediately navigate to other modules to continue their work without interruption. This allows you to return to the environment screen at your convenience to verify the status, ensuring that the environment setup no longer leads to idle user time or workflow bottlenecks.

Analytical Power and Model Lifecycle Management

In-Pipeline Exploratory Data Analysis (EDA)
EDA capabilities are now integrated directly into the model pipeline, facilitating data validation at any development stage.
- Enhancement: You can now insert EDA widgets directly into the pipeline to visualize data distribution, missing values, and outliers by using standard statistical metrics.
- Business Impact: This allows for real-time assessment and data cleaning immediately before model training or feature engineering, ensuring high-quality inputs without requiring you to exit the pipeline interface.
Model Status and Versioning Alerts (OFS STSA Ecosystem Only)
To support rigorous governance, an automated alerting mechanism now tracks the lifecycle of approved models and their downstream impacts. Please note that this functionality is currently applicable only to the OFS STSA ecosystem.
- Enhancement: While approved models remain locked to prevent unauthorized changes, the system now triggers email notifications when a new version is available or when a status change occurs, for example when a new champion model is designated.
- Business Impact: These alerts provide essential visibility for stakeholders who manage macromodels or higher-level definitions. This intelligence allows you to proactively decide whether to uptake a new version or maintain a specific snapshot, ensuring informed operations while preserving analytical flexibility.
Benchmarking and Pipeline Comparison
The platform’s comparative analytics have been expanded to support rigorous benchmarking across different modeling approaches.
- Enhancement: You are no longer limited to comparing different versions of the same model. You can now compare a specific model against any other pipeline within an objective or sub-objective. Additionally, the system allows you to view and compare results from the latest execution of these models.
- Business Impact: This enables direct benchmarking between different methodologies, for example, comparing different mathematical approaches) and scripts. By viewing the latest execution results side-by-side, you can objectively validate which approach or coding logic yields the best outcome, leading to more confident deployment decisions.
Traceable Custom Session Node Execution
We have introduced the ability to execute granular nodes within a pipeline, providing developers a focused testing environment within the workflow.
- Enhancement: Previously, you were unable to execute individual granular nodes in isolation. The system now enables this by prompting you to create or assign a custom session for node-level runs.
- Business Impact: This allows developers to test specific segments of a pipeline, evaluate outcomes, and gain confidence in the logic before committing to full progression. Because these executions are traceable and tracked against the model, it maintains a permanent governance trail. This flexibility allows you to test and pivot as needed during development while keeping a high standard of oversight.

Backend Utility Enhancements

Unified Model Management Script
- Enhancement: We have consolidated several core maintenance tasks into a single, streamlined shell utility. This unified script simplifies the model lifecycle by allowing for the import, export, and retirement of models, as well as the management of parameter sets, through one central command-line tool.
- Business Impact: This consolidation significantly reduces administrative overhead and simplifies system maintenance. By providing a single point of control for model lifecycle tasks, your technical teams can perform migrations and updates more quickly and with a lower risk of manual error.
Utility For Execution and Monitoring
- Enhancement: We have introduced a new utility that allows the execution and monitoring of processes directly through the platform's internal scripts. This eliminates the requirement to log into a separate scheduler interface to track backend tasks, providing all the necessary status information within the primary environment.
- Business Impact: Implementation of monitoring tools directly into the workflow increases operational speed and reduces tool fatigue. You can now oversee their executions more efficiently, leading to faster troubleshooting and a more focused experience without the friction of switching between multiple platforms.
Personalized Scheduler Time Zone Configuration
- Enhancement: We have updated the installation process to allow for greater regional flexibility by enabling the configuration of a preferred time zone for the scheduler user interface. While the system previously defaulted all scheduled activities and timestamps to UTC, you can now select and set a specific time zone during the initial setup phase.
- Business Impact: This update ensures that scheduling data and execution logs align with your local operational hours, making it easier for teams to coordinate tasks and monitor workflows. By displaying time in a familiar local format, you can more accurately track time-sensitive operations without the need for manual time conversions.
Model Metrics and Visualization Refinement
- Enhancement: We have improved the Model Catalog's scoring process to ensure that visualizations and graphs are always generated, even if specific metrics like ROC and AUC cannot be calculated due to your data's format. The system now intelligently separates numerical scores from status messages; if a metric cannot be computed, it is assigned a neutral value to allow the rest of the report to load, while a clear message explains why that specific score is unavailable.
- Business Impact: This update ensures that your performance screens always remain fully functional and visible. By preventing a single calculation error from blocking the entire report, you can consistently access their model graphs and analysis, thus saving time and providing immediate clarity on data-specific limitations.

Technical Foundation

Platform Updates: The product now includes OFS Data Studio 25.4.4-1.
Library Maintenance: All third-party libraries have been upgraded to their latest stable versions to ensure peak performance and infrastructure security.

Limitations from Past Releases that have been Remediated in this Release

Notebook initialization no longer fails even when the logged-in username is in lowercase.
Graceful cleanup of data model jobs in case of abrupt shutdown of services is now handled.

Known Issues and Limitations

Table deletion sync-up between schemas is not supported during workspace edits.
Error log table creation fails if the column data types are LONG, CLOB, BLOB, BFILE, and ADT during workspace data population.
Unable to perform the dataset cache action with the model library.
The PDF of the model report does not contain data in the output section.
Deployment of models that are published from the model summary screen will not promote the associated dependencies such as graphs, parameter sets, datasets, and models. However, the same works fine, if the models are published from within the pipeline UI.
As of now, Python 3.12 does not support apache-flink completely, hence, installing Python 3.12 might display a few errors.
Oracle-guardian-ai is no longer a mandatory library installed by mmg-python library. Instead it needs to be installed and configured in a dedicated conda environment. This is due to the limitation of the oracle-guardian-ai version compatibility and no support for Python 3.12.

3 New Features, Enhancements and Limitations in this Release