Changes in MySQL HeatWave 9.4.0 (2025-07-22, General Availability)

Note

MySQL HeatWave User Guide has been restructured and improved to provide better visibility of high-level tasks.

Note

These release notes were created with the assistance of MySQL HeatWave GenAI.

• MySQL HeatWave AutoML

• MySQL HeatWave GenAI

• MySQL HeatWave Lakehouse

• MySQL HeatWave

MySQL HeatWave AutoML

• MySQL HeatWave AutoML now supports the recording of hyperparameters of trained models, enhancing the transparency and reproducibility of machine learning processes. This update allows for more detailed insights into model training, enabling better optimization and management of machine learning workflows. With this feature, you can access and utilize hyperparameter information to refine your models, leading to improved performance and accuracy.

  For more information, see Model Metadata. (WL #16946)

• MySQL HeatWave now supports a new feature called NL2ML, which provides an intuitive natural-language interface to machine learning (ML). With NL2ML, you can ask questions about MySQL HeatWave AutoML and receive step- by-step guidelines on how to use ML for specific business problems. This innovative feature allows you to generate AutoML queries that can be copied and executed, making it easier to get started with ML. Additionally, NL2ML offers a chat-bot interface for follow-up questions and leverages your schemas, tables, and columns to create personalized AutoML queries.

  For more information, see Learn About MySQL HeatWave AutoML with Oracle Cloud Infrastructure Generative AI. (WL #16785)

• MySQL HeatWave AutoML now supports enhanced memory management capabilities, enabling more efficient resource utilization. MySQL HeatWave can now check memory usage and release resources based on severity levels, ensuring optimal performance and reliability. This enables better management of memory between MySQL HeatWave GenAI and MySQL HeatWave AutoML. (WL #16759)

• The rpd_nodes performance schema table includes new columns that enable you to monitor memory usage by MySQL HeatWave AutoML on MySQL HeatWave nodes.

  For more information, see The rpd_nodes Table. (WL #16761)

MySQL HeatWave GenAI

• MySQL HeatWave GenAI now supports using Dedicated Clusters from OCI Generative AI Service, providing you with more flexibility and capabilities to leverage the power of generative AI. With this addition, MySQL HeatWave GenAI continues to expand its feature set, offering a more comprehensive and robust solution for you to explore and innovate.

  Important

  Support for mistral-7b-instruct-v1 and llama3-8b-instruct-v1 in-MySQL HeatWave LLMs has now been deprecated.

  MySQL 9.4.1 onwards, if you use any of the deprecated in-MySQL HeatWave LLMs: llama2-7b-v1, mistral-7b-instruct-v1, or llama3-8b-instruct-v1, then MySQL HeatWave GenAI will automatically replace it with llama3.2-3b-instruct-v1.

  For more information, see Supported Models and Languages. (WL #16836)

MySQL HeatWave Lakehouse

• MySQL HeatWave Lakehouse now supports creation and modification of Lakehouse tables using SQL syntax for setting table options, in addition to the existing way of using JSON strings for setting the ENGINE_ATTRIBUTE. This provides another way for you to define and load data from Object Storage for Lakehouse tables.

  For more information, see Lakehouse External Table SQL Syntax. (WL #16837)

• The CREATE EXTERNAL TABLE syntax in MySQL HeatWave Lakehouse has been enhanced to provide a more streamlined way of creating tables. The primary and secondary engines for MySQL HeatWave Lakehouse tables are now determined by session variables, eliminating the need to specify engines explicitly.

  For more information, see Lakehouse External Table Syntax. (WL #16831)

MySQL HeatWave

• MySQL HeatWave now supports the use of the URI keyword in addition to the URL keyword for loading data with Bulk Load and exporting query results.

  For more information, see Bulk Ingest Data and Export Query Results to Object Storage. (WL #16988)

• Bulk Load now supports loading tables without primary keys. (WL #16805)

• The syntax for creating MySQL HeatWave temporary tables now uses the ENGINE=RAPID option instead of SECONDARY_ENGINE=RAPID making RAPID the primary engine for these tables. This enhancement also lets you specify the primary engine for MySQL HeatWave temporary tables using the default_tmp_storage_engine variable.

  For more information, see Create MySQL HeatWave Temporary Tables. (WL #16898)

• MySQL HeatWave now supports enhanced offload debuggability by introducing significant enhancements, including unified error handling and improved integration with MySQL Shell. This enhancement gives you access to more informative error messages and workarounds, making it easier to diagnose and resolve offload errors. (WL #16409)