Database: Autonomous Database Serverless

Autonomous AI Database extends external authentication support to Database Tools. You can sign in to Database Actions and supported tools using Oracle Cloud Infrastructure IAM, Microsoft Entra ID, Google Cloud Platform, AWS Cognito, Okta, as a globally identified database user, eliminating the need for database passwords. This support for external authentication in Database Actions (SQL Developer Web) and Data Studio enables you to leverage centralized identity management and single sign-on capabilities.

Oracle Autonomous AI Database now supports a managed Agent2Agent (A2A) server that exposes Select AI agent teams through the open standard A2A protocol. This feature allows external A2A-compatible client agents to discover and call published Select AI agent teams by using standard HTTPS interfaces and protocol-based communication, reducing the need for custom integrations. This allows Oracle Autonomous AI Database to participate in broader multi-agent workflows while keeping the agent logic and data access inside the database.

Oracle Deep Data Security in Autonomous AI Database provides identity-aware, fine-grained authorization by enforcing row-level and column-level access controls directly within the database.

Oracle Data Science Agent is supported with Oracle Autonomous AI Database Serverless 26ai. With Oracle Data Science Agent, you can profile data, prepare features, train models, evaluate results, and generate inference SQL through a guided chat experience in Oracle Machine Learning.

The DBMS_DCAT PL/SQL 26ai package enables Oracle Autonomous AI Database to connect, synchronize, and query external tables based on Iceberg REST-compatible catalogs like Databricks Unity Catalog and Polaris.

Lake Cache in Oracle Autonomous AI Database supports mounted catalog tables.

Lake Cache in Oracle Autonomous AI Database supports mounted catalog tables.

Oracle Autonomous AI Database Serverless is available on Oracle AI Database@AWS, enabling you to deploy Autonomous AI Database workloads in selected AWS Regions while continuing to manage and operate them through Oracle Cloud Infrastructure (OCI).

Oracle Autonomous AI Database provides DBMS_SHARE subprograms to simplify delta sharing setup and subscription workflows. These subprograms support share profile generation, credential creation, profile validation, and provider subscription management.

You can read and list files directly from SharePoint using the DBMS_CLOUD package, making it easier to load data into the Autonomous AI Database. A DOWNLOAD pipeline type also enables you automate ingestion from these sources, with support for extended metadata in SharePoint.

Cloud Operator access for troubleshooting and issue mitigation now includes an additional option. Permissions can be granted for a defined duration or as unlimited access, which remains in effect until explicitly revoked.

For Autonomous Data Guard, zero data loss protection (RPO = 0) is provided for a local standby database. Autonomous Data Guard performs automatic failover to a local standby database when a standby is available and the system guarantees zero data loss. If a data loss limit is specified (0 to 3600 seconds), failover occurs within the defined limit.

Autonomous AI Database accepts absolute paths in CREATE DIRECTORY statements and creates directories under managed locations for migration compatibility.

You can use DBMS_CLOUD_IMPORT to import data into Autonomous AI Database. DBMS_CLOUD_IMPORT supports Oracle source databases as well as non-Oracle databases, for example, MySQL, PostgreSQL, and Amazon Redshift. You can import either the entire or the subset of the data stored in a supported source database.

Autonomous AI Database allows seamless switching from Private endpoint access to Secure access from allowed IPs and VCNs only without any TLS changes, preserving your existing configuration.

When you define a private endpoint for your Autonomous AI Database 26ai instance, you can provide enhanced security by setting a database property to enforce that all outgoing connections to a target host are subject to and limited by the private endpoint's egress rules.

Autonomous AI Database 26ai supports Multifactor Authentication (MFA) to strengthen database access security by requiring an additional verification factor in addition to the user password. MFA can be configured for database login, for running protected SQL statements, or for both.

Lake Cache in Autonomous AI Database includes the following enhancements:

Data Transforms includes the following new features and changes:

Autonomous AI Database automatically converts or suppresses unsupported grant and revoke operations, maps roles to their supported equivalents, and maintains a transparent audit trail in the DBA_CONVERTED_STATEMENTS view.

The Translate feature enables you to translate text across multiple languages using SQL or PL/SQL, based on the AI provider configured in your AI profile. This feature supports multiple providers, including OCI, Google, AWS, and Azure.

Oracle Spatial Studio provides a no-code user interface to access the spatial features of Oracle Database. The interface supports spatial data loading, preparation, visualization, and analysis. Spatial Studio is a fully integrated feature of Autonomous AI Database and appears in the Tool Configuration Tab on the Autonomous AI Database Details page. It is available only for the Elastic CPU (ECPU) model.

The actiontype variables in Autonomous AI Database enables precise control over sub-operations like adminPassword, scheduledOperations, manageEncryptionKeys, and more during database creation or updates. This supports separation of duties, least-privilege access, and compliance, without disrupting existing broad permissions.

With secret- based password authentication, you can create and manage database users whose passwords are stored as secrets in a cloud provider’s vault and referenced through a vault secret credential. The database fetches the password from the vault during authentication, generates verifiers only in memory, and does not persist passwords or verifiers on disk.

Database Identity Network Headers in Autonomous AI Database adds database identity metadata as a JSON X-header to outbound UTL_HTTP calls so remote endpoints can verify which Autonomous AI Database (and tenant) is calling them.

Two new memory utilization metrics for monitoring SGA and PGA Utilization are now available an Autonomous AI Database.

You can access built-in tools in your Autonomous AI Database with single sign-on. You must log in once with your database credentials to switch between tools such as Oracle APEX, Database Actions, Graph Studio, Oracle Machine Learning, and Oracle Data Transforms without re-authenticating.

Select AI supports natural language to SQL (NL2SQL) generation for SQL Property Graphs. You can include one or more property graphs in an AI profile object list and issue natural language prompts that Select AI converts into PGQ (Property Graph Query) statements using the GRAPH_TABLE operator.

You can use the DBMS_CLOUD_ADMIN_SEC package to clone the privileges and permissions assigned to the ADMIN user for a specified user account in Autonomous AI Database. This enables administrators to create named user accounts with the same privileges as the ADMIN user. It provides an alternative to using the shared ADMIN account improving security.

Data Transforms includes the following new features and changes:

Autonomous AI Database supports Multifactor Authentication (MFA) to strengthen database access security by requiring an additional verification factor in addition to the user password. MFA can be configured for database login, for running protected SQL statements, or for both.

The ConnectionDropsDetected event is generated when a significant number of connection drops are observed for your Autonomous AI Database.

With incremental refresh of materialized views over Cloud Links, the Autonomous AI Database refreshes only changed data from remote tables instead of recomputing the entire materialized view.

Resource utilization metrics are available for monitoring Autonomous AI Databases.

Use an Autonomous AI Database as a Select AI Proxy to securely integrate with Oracle and non-Oracle databases and bring natural language queries to these databases by routing requests through Autonomous AI Database.

Oracle Autonomous AI Database provides two restart options: a standard Restart, and an Online restart that reduces downtime. With Online restart, the Autonomous AI Database instance is restarted with minimal impact on database availability.

When you define a private endpoint for your Autonomous AI Database instance you can provide enhanced security by setting a database property to enforce that all outgoing connections to a target host are subject to and limited by the private endpoint's egress rules.

With MY$POOL scope in cloud links, you can register datasets once in a provider Autonomous AI Database, making them discoverable and queryable from any other Autonomous AI Database in the same Elastic Pool. Scope checks are enforced on the consumer side to ensure proper access control.

Autonomous AI Database now includes a built-in MCP Server that provides a managed, per-database MCP server for Oracle Autonomous AI Database on Serverless Exadata Infrastructure (26ai and 19c).

Data Transforms includes the following new features and changes.

You can create dedicated elastic pools in Autonomous AI Database, where the pool leader and all members reside on the same infrastructure.

You can use the Transportable Tablespace feature in Autonomous AI Database to efficiently migrate data and metadata from on-premises Oracle databases or Oracle Database Cloud Service (DBCS) databases to Autonomous AI Database.

Autonomous AI Database allows you to rotate wallets for an Autonomous AI Database instance or for all instances that a cloud account owns in a region, with a grace period of 1 hour to 72 hours.

You can use the Autonomous AI Database API for DynamoDB to work seamlessly with Autonomous AI Database using your existing DynamoDB tools and SDKs.

Providers define scopes (tenancy, compartment, schema, or object level) to authorize specific consumers to generate Table Hyperlinks without manual sharing.

When a source database uses customer-managed encryption keys, you have several options for specifying the encryption key type and location for a cross tenancy clone.

You can use cs_resource_manager subprograms in an Autonomous AI Database to define and manage your Database Resource Manager plans.

With Chart View in Table Hyperlinks, you can switch between tabular and chart views for better analysis experience.

The Catalog tool of the Data Studio Suite of tools provides PL/SQL APIs to manage catalogs using the DBMS_CATALOG package.

Oracle Machine Learning UI has been enhanced to support direct interaction of OML Notebooks with your external GitHub repositories.

External systems and developers can use Data Transforms REST APIs to retrieve information about Data Transforms such as projects, connections, data loads, data flows, workflows, schedules, and variables.

You can enable Autonomous Data Guard on Autonomous AI Databases with the JSON and APEX workloads for both local and cross-region standby databases.

Data Transforms includes the following new features and changes.

You can use Data Lake Accelerator with Autonomous AI Database 26ai instances to support external data processing.

For a database in an elastic pool, you can enable a vanity URL on your Autonomous AI Database and configure a vanity URL custom domain for your Autonomous AI Database instance using an API Gateway.

You can use live workload capture-replay to capture workloads running on a source Autonomous AI Database instance and simultaneously replay them on a target refreshable clone, as the workload runs on the source database.

You can invoke Python scripts with the WITH_CONTEXT attribute from an Autonomous AI Database instance.

External Table Cache in Autonomous AI Database enables you to cache frequently accessed data from external tables in your database.

Autonomous AI Database for Developers supports the backup and restore features available on Autonomous AI Database.

Oracle AI Database architects Artificial Intelligence (AI) into its core, seamlessly integrating AI across all major data types and workloads.

Autonomous Data Guard can be enabled for elastic pools with local standbys.

You can integrate with a broader Iceberg REST catalog ecosystem with the support of Databricks Unity and Snowflake Polaris, in addition to the existing integration with Amazon Glue and location-based Iceberg root metadata files.

Select AI Agent (autonomous agent framework) introduces a framework for building agentic workflows enabling developers to create agents, tools, tasks, and teams that reason, act, and collaborate using generative AI within the database.

The Translate feature enables you to translate your text across multiple languages using SQL or PL/SQL, as supported by your AI provider.

You can run a mixed workload for data analytics and critical transaction processing by connecting to Autonomous AI Lakehouse databases using the TP and TPURGENT predefined database services.

You can test your workloads in Database Actions by capturing a workload on production system and replay it on test system.

Use the DBMS_CLOUD.CREATE_EXTERNAL_TABLE procedure to create implicit partitioned external tables.

You can now use Select AI for Python to access DBMS_CLOUD_AI capabilities in Autonomous AI Database from Python applications.

Select AI supports customizable conversations enabling you to use Select AI with different topics without mixing context, improving both flexibility and efficiency.

Select AI enables you to give feedback on generated NL2SQL results to improve query generation accuracy.

Select AI enables you to generate a summary of your text directly in a prompt or from a URI.

You can generate a Table Hyperlink with ACL checking enabled.

Data Transforms includes the following new features and changes.

You can use directory files, including those in network file systems, with data pipelines and the SELECT AI vector index in Autonomous Database.

You can use wildcard patterns to select multiple files in both object storage and directory objects when working with DBMS_CLOUD, DBMS_CLOUD_PIPELINE, and DBMS_CLOUD_AI in Autonomous Database.

You can read and access data with a Table Hyperlink in a Google Sheet using Oracle Autonomous Database Add-on for Google Sheets.

You can create a password-protected Table Hyperlink by specifying a Table Hyperlink password.

Autonomous Database provides cloning where you can choose to create a full clone of the active instance, create a metadata clone, create a refreshable clone, or create a refreshable metadata clone.

You can create Cloud Tables where table data resides on Oracle managed Cloud Storage and the table data does not consume database storage.

When you create a Table Hyperlink or a Table Hyperlink Group, you can set the consistent parameter to specify that the producer provides data consistently across different pages when a consumer access data.

Autonomous Database introduced new dataplane events, UnplannedMaintenanceBegin and UnplannedMaintenanceEnd.

Query the DBA_NONJSON_OBJECTS view to assess the space consumed by non-JSON objects on an Autonomous Database instance.

View the available elastic pool capacity.

If your Autonomous Database instance is using Oracle Database 19c, you can upgrade to Oracle Database 23ai with scheduled upgrade.

External table cache in Autonomous Database enables you to cache frequently accessed data from external tables in your database.

You can view or copy connection code samples for different connectivity options for Autonomous Database.

Substitution variables are pre-defined placeholders which are resolved and substituted with the corresponding actual values based on the runtime context of your Autonomous Database.

When you have heavy read workloads using cloud links you can configure an elastic pool leader or member as a cloud links provider, where the provider enables ProxySQL query offload to offload queries (reads) to any number of refreshable clones.

You can invoke generic scripts, of type BASH, C, or Python from an Autonomous Database instance.

The ResourceUtilizationWarning event is generated as a warning for compute usage, storage usage, and number of sessions when certain thresholds are exceeded.

Select AI extends its support to the following AI providers:

The view V$DIAG_ALERT_EXT is not available for alert log data.

The Data Studio is updated with new features and changes.

ProxySQL enables unified access to multiple Autonomous Databases by routing statements for DML and queries, and eliminates the need to physically consolidate all the data into a single database.

Table Hyperlink Groups allow you to access data from multiple objects or multiple SQL SELECT statements through a single URL.

Autonomous Database supports using customer-managed encryption keys with refreshable clones.

Autonomous Database patches are applied during a scheduled maintenance window as rolling patches.

Data Transforms is updated with new features and changes in the following areas.

You can combine Organization Management with the benefits of elastic pools.

You can generate a Table Hyperlink with ACL checking enabled.

You can send messages from Autonomous Database to an OCI Notifications service topic.

Critical events and Information events include an eventCategory attribute.

You can access external data by mounting catalogs and querying them using a simplified SQL syntax.

Select AI includes the following new features.

A cross-region standby can be added to an elastic pool or can be an elastic pool leader.

Use Autonomous Database Oracle-managed heterogeneous connectivity to create database links to Google Drive.

In the ECPU compute model, the built-in database tools that use a VM, including: OML, Graph Studio, and Data Transforms have default configuration changes and billing changes.

Elastic pool cost and usage details are recorded in the Oracle Cloud Infrastructure cost reports and in the OCI_COST_DATA view.

When you have heavy read workloads, where a relatively small amount of data is scanned by multiple queries, including from PL/SQL or from Oracle Scheduler Jobs, you can offload queries (reads) from either an elastic pool leader or from an elastic pool member to Refreshable Clones.

You can read and access data with a Table Hyperlink in a Microsoft Excel spreadsheet using Oracle Autonomous Database Add-in for Excel.

The OCPU compute model option is removed when you provision or clone a new Autonomous Database instance with Data Warehouse or Transaction Processing workloads.

DBMS_CLOUD supports loading and querying data using external tables as well as exporting and importing of data from Autonomous Database to object storage.

You can specify default values for one or more bind variables in a SQL statement when the default_bind_values parameter is included with DBMS_DATA_ACCESS.GET_PREAUTHENTICATED_URL.

Data Transforms is updated with new features and changes in the following areas.

Select AI includes the following new features.

You can use in-database transformers (also known as embedding models) that you imported into your Oracle Database 23ai instance with Select AI RAG to generate vector embeddings for document chunks and user prompts.

The Data Studio is updated with new features and changes.

The LongPendingTransaction event is generated when a transaction is open for longer than 24 hours.

The Recovery Time Objective (RTO) for cross-region Autonomous Data Guard Standby databases is lowered to less than ten (10) minutes.

Automatic Regression Detection in Autonomous Database provides proactive handling of regressions and enables automated issue detection, diagnosis, and mitigation.

The AverageActiveSessions metric shows the average number of sessions that are actively executing or waiting for resources.

The DBMS_CLOUD.CREATE_CREDENTIAL procedure supports comments when you use the comments parameter.

You can migrate SQL statements from PostgreSQL, MySQL, or SQL Server to Oracle SQL and run the statements in Autonomous Database.

You can generate and manage Table Hyperlinks to access data in Autonomous Database.

Select AI Retrieval Augmented Generation (RAG) supports multiple file formats, including PDF, DOCX, JSON, XML, and HTML among others.

When you create an elastic pool, the workload type of the leader is not restricted.

If a SQL parse error occurs several times in a certain time window, for example 200 times or more in an hour, this generates Too Many Parse Errors entries in the alert log.

When you create a refreshable clone from an Oracle Database 19c source, you can select Oracle Database 23ai for the refreshable clone.

When you have heavy read workloads, where a relatively small amount of data is scanned by multiple queries, including from PL/SQL or from Oracle Scheduler Jobs, you can offload queries (reads) from an elastic pool leader to Refreshable Clones.

Two new metrics are introduced for the Database service in the oci_autonomous_database namespace.

The APEX instances in Autonomous Database Serverless (including APEX Service) are upgraded to Oracle APEX 24.2.

The Data Studio is updated with new features and changes.

You can use real-time SQL Plan Management when you migrate from a source Oracle Database to Autonomous Database.

When you have heavy read workloads, where a relatively small amount of data is scanned by multiple queries, you can offload queries (reads) from an elastic pool leader to Refreshable Clones.

You can create an external table over Pre-Authenticated Request (PAR) URLs.

With Cloud Links, a data owner registers a table or view for remote access for a selected audience as defined by the data owner.

Autonomous Database introduced a new data plane event, DatabaseUserPasswordWarning.

You can create custom metrics to collect, publish, and analyze metrics.

You can disconnect an Autonomous Data Guard standby or a Backup-Based Disaster Recovery peer.

When you provision or clone an Autonomous Database instance you can select a patch level to apply to upcoming patches.

You can set the service name mapping to use when Cloud Links consumers access data from a data set owner.

When you access ORC, Parquet, or Avro data stored in object store using external tables, you can either automatically or explicitly set the maximum size of text columns.

You can invoke Google Cloud Platform Cloud Run functions as SQL functions.