Change History

Date	Description
2026-04-01	This issue is the fifth official release. Separated the OmniAdvisor Feature Guide into an independent document from the OmniRuntime Feature Guide. Separated the OmniAdvisor Release Notes into an independent document from the OmniRuntime Release Notes.
2025-12-30	This issue is the fourth official release. OmniAdvisor 2.0.0: Added the notice on Spark SQL job data collection and security usage specifications in Architecture. Added an important risk warning for disabling password complexity verification function in Initializing the Feature. Deleted section "(Optional) Enabling the Background Data Management Page." Added Environment Dependency Installation Failed and "Error: Failed building wheel for pyrfr" Displayed. Update the OmniAdvisor 2.0 system framework package name. For details, see Environment Requirements.
2025-09-30	This is the third official release, which incorporates the following change: Released OmniAdvisor 2.0.0: This version is independent of OmniAdvisor 1.0.0 and 1.1.0, and provides the following new features: This feature samples task parameters and recommends optimal configurations through AI iterative tuning, expert rule–based tuning, transfer generalization tuning, and native operator acceleration, enabling end-to-end parameter tuning for Spark tasks. It supports foreground user-unaware tuning and background retesting.
2024-03-30	This issue is the second official release. Added the functions of parsing Spark application tasks and tuning task parameters. Added the functions of parsing tasks and tuning parameters in Spark Yarn cluster mode. Added the functions of parsing SQL tasks (Spark and Hive) that run multiple SQL statements in the same session and tuning task parameters. Added the function of sampling and tuning parameters of the string type. Enhanced parameter sampling algorithms. Added the method of using the OmniAdvisor CLI.
2024-01-11	This issue is the first official release. OmniAdvisor uses AI algorithms to intelligently tune parameters of Spark and Hive tasks running in online systems throughout the whole process.