7.3 MySQL HeatWave GenAI Roles and Privileges

To perform the following MySQL HeatWave GenAI functions, ask the admin user to grant you the required roles and privileges:

• To create a vector store using asynchronous load, the following role and privileges are required:

  • The mysql_task_user role:

    mysql>GRANT 'mysql_task_user' TO 'user_name'@'%';

  • The VECTOR_STORE_LOAD_EXEC privilege:

    mysql>GRANT VECTOR_STORE_LOAD_EXEC ON *.* TO 'user_name'@'%';

  • The SELECT privilege on the following performance_schema and sys schema tables:

    mysql>GRANT SELECT ON 'performance_schema'.'rpd_nodes' TO 'user_name'@'%';                    
    mysql>GRANT SELECT ON 'performance_schema'.'rpd_table_id' TO 'user_name'@'%';                 
    mysql>GRANT SELECT ON 'performance_schema'.'rpd_tables' TO 'user_name'@'%';                   
    mysql>GRANT SELECT ON 'sys'.'vector_store_load_metadata' TO 'user_name'@'%';                  
    mysql>GRANT SELECT ON 'sys'.'vector_store_load_tables' TO 'user_name'@'%';

  • The EXECUTE privilege on the following stored procedures in the sys schema:

    mysql>GRANT EXECUTE ON PROCEDURE 'sys'.'vector_store_load_current_schema' TO 'user_name'@'%'; 
    mysql>GRANT EXECUTE ON PROCEDURE 'sys'.'vector_store_load' TO 'user_name'@'%';

  • The CREATE, ALTER, and EVENT privileges on the schema where you want to set up the vector store table:

    mysql>GRANT CREATE, ALTER, EVENT ON 'schema_name'.* TO 'user_name'@'%';

• To run the batch queries using ML_GENERATE_TABLE, ML_RAG_TABLE, and ML_EMBED_TABLE, the following privileges are required:

  • SELECT and ALTER privileges on the input table:

    mysql>GRANT SELECT, ALTER ON input_schema.input_table TO 'user_name'@'%';

  • SELECT, INSERT, CREATE, DROP, ALTER, UPDATE privileges on the schema where the output table is created.

    mysql>GRANT SELECT, INSERT, CREATE, DROP, ALTER, UPDATE ON output_schema.* TO 'user_name'@'%';

For more information, see Privileges Provided by MySQL and Default MySQL Privileges.

What's Next

Learn about the available LLMs, embedding Models, and languages.