Release Notes

A summary of changes to Kensho Link.

July 2026

Reliability: Increased Service Capacity

Service capacity was increased to better handle traffic spikes to the on-demand linking endpoint, reducing the likelihood of intermittent errors under load.

Improved Model Training Pipeline

Introduced an automated pipeline that samples and enriches production traffic for use in model training, enabling continuous improvement of match quality over time.

June 2026

Country-Aware Matching Is Now the Default

The improved country-aware matching model (introduced in May) is now the default for all requests to the /generic endpoint. No changes are required for existing integrations.

The /generic_country_filter endpoint has been retired — callers using it should migrate to /generic.

May 2026

Improved Matching: More Accurate Results for Uniquely-Spelled Company Names

The matching pipeline now considers both original and spell-corrected forms of company name tokens when finding candidates. This improves match accuracy for companies with intentional or unconventional spellings that could previously be overlooked.

Improved Matching: Country-Aware Subsidiary Resolution

The matching model now factors in country information earlier in the resolution process. When country data is provided, the service returns local subsidiaries rather than foreign parent companies — improving match accuracy for multinational organizations.

This capability is available via the existing country, country_iso2, and country_iso3 input fields. When no country is provided, behavior is unchanged.

Automated Knowledge Base Validation

Introduced an automated validation pipeline that verifies knowledge base quality before updates are promoted to production. This ensures users always receive results from a validated, tested dataset.

April 2026

Security: Runtime Upgraded to Python 3.11

The service runtime was upgraded from Python 3.9 (end-of-life, no longer receiving security patches) to Python 3.11. This upgrade addresses security vulnerabilities present in the unsupported runtime and improves overall service stability.

Reliability: Reduced Errors Under High Load

Fixed an issue that caused intermittent request failures during traffic spikes. The service now handles concurrent requests more reliably.

Reliability: Improved Monitoring Response Times

Internal service health monitoring was significantly optimized, ensuring faster detection and alerting on any service degradation.