Release notes
Version 3.7
- Released: September 10, 2025
- Compatible platform version: Work.AI 10.2.8.54+
The Evelyn v3.7 release enhances data validation, logging, and address matching to increase accuracy, simplify troubleshooting, and deliver more reliable results.
Detailed input data validation is designed to enhance error handling for Business Processes that use a CSV provider, offering more granular control over how Evelyn processes CSV files with data errors. You can now choose to process valid records while isolating invalid ones or stop the entire Business Process upon detecting the first error.
Errors in CSV input files are now logged in a standardized format that specifies the error type, column name, and exact line numbers, making troubleshooting faster and more precise.
Each Business Process log entry now includes a client ID and a transaction UUID, providing end-to-end transaction traceability. You can search by a transaction UUID to see all log entries related to a specific transaction or filter by client ID to isolate client-specific logs and quickly identify problematic transactions.
A new configuration option within the Model settings allows you to include or exclude street names in location matching. When street-level matching is enabled, the model applies additional logic to compare street information for more precise matches.
The address matching logic is reworked to remove unreachable conditions and ensure more consistent results. The model now evaluates addresses step by step: country → state → city → street. Matching stops immediately on a strong mismatch, while successful comparisons continue down to the lowest level available. Evelyn then returns the most specific strong match found. If no strong match exists, the address is reported as incomplete (
BadInfo).The classification model is upgraded to v4.3.0.
Bug fixes
Resolved the issue with the model failing when hit date fields were empty.
Resolved the issue where name enrichment by ticker symbols was disabled.
Resolved the issue with street addresses not being parsed into the model during CSV ingestion.
Resolved the issue when running an input file of names against the Thomson Reuters CLEAR API connector.