Process Intelligence
Version 2025/10
Release Date: October 28, 2025
Process Intelligence CSV Export (Process Mining Compatible)
A new export feature has been added to the ProcessMaker Process Intelligence (PMPI) Admin Panel, enabling users to manually export process event logs in a CSV format compatible with industry-standard process mining platforms.
Users can now extract structured event data—including timestamps, task names, case IDs, and application context—and upload it to external tools such as ARIS Process Mining for detailed analysis.
Applies to all collected event logs in the PMPI Admin Panel. The exported CSV includes:
Case_ID – Unique identifier of each process instance
Activity – Name of the task (e.g., screen or window)
Start – Start timestamp of the task
End – End timestamp of the task
Improves interoperability with third-party process mining tools, enhances data transparency, and supports enterprise-grade analytics workflows.
Test Runner – Clearer Error Messages
Improved test case error handling with more descriptive, user-friendly messages across all mismatch scenarios.
Key Improvements
Overlapping Application Rules → “Extra Application Matched: AppName.”
Overlapping Window Rules → “Extra Window Name Matched: WindowName.”
No Rule Match → “Test case doesn’t match any rule.”
Unexpected Identifiers → “Process identifier ID detected for process Name.”
Missing Identifiers → “Identifier not found.”
Wrong Identifier Value → “Expected Value1 for Process, but got Value2.”
Training Data – Frontend Enhancements
Improved the Training Data interface to give users better control over exports and clarity about data limits.
Key Improvements
Added a Start Date filter for CSV downloads, allowing users to define the time range of training data exported.
Added a UI message indicating data scope: “Training data only includes the last 20,000 collected events. For more events, use the training data download option.”
Fixed an issue where updates made to the JSON editor were not reflected immediately in the UI. The editor now reliably displays the latest changes after edits are saved or applied.
Resolved an issue where the download URL for training data CSV exports was not correctly handling query parameters, resulting in broken links. CSV downloads now use the proper parameters and format.
Transaction metrics retrieved from the Process API are now consistently ordered by the selected metric (e.g., throughput time or active work time) before applying the 2000-record limit. This ensures that users receive the most relevant data in a predictable order, even when large datasets are involved.
Resolved an issue in the training data matching logic where records with
Nonevalues disrupted sorting behavior. Sorting now handlesNonevalues gracefully, ensuring consistent and accurate data alignment.Added a safeguard to limit the volume of training data retrieved from MongoDB, improving performance and preventing excessive memory usage during data loading and processing.
Restricted the sequence lengths in multi-system step windows to prevent excessive memory usage and ensure consistent performance across long-running workflows.
Resolved an issue where calendar data beyond a certain date was not appearing in the PI dashboard due to aggregation pipeline failures. The problem was caused by large session payloads exceeding internal processing limits. The fix improves data handling during aggregation, ensuring that recent data is correctly processed and displayed in the dashboard.
Last updated