Release notes (2026)

Feature

Description

References

Sales

• Dataphin now supports the EMR Serverless Spark compute engine for greater flexibility.

• You can now purchase the data query acceleration feature. This feature lets you configure SelectDB, Doris, or StarRocks as an acceleration engine to query MaxCompute data directly, eliminating data migration and reducing both cost and latency.

• Purchase Guide - Fully Managed

• Billing Description - Fully Managed

• Activate Dataphin

Feature

Description

References

Knowledge graph

This release introduces the knowledge graph feature. A knowledge graph is a technology that abstracts real-world entities and their relationships into a graph structure for storage and computation. It moves beyond traditional data warehouses and indicator systems, which only handle structured data. By integrating unstructured data, such as documents and text, with structured data, it builds a comprehensive and interconnected knowledge network. This helps enterprises understand data semantics and uncover complex relationships.

Knowledge graph overview

Data source management

Dataphin now supports Milvus 2.6.x as a data source.

Create a Milvus data source

Batch integration

• Adds batch integration support for Milvus data sources.

• The DataHub reader now supports Blob type topics and automatically outputs the shardId, systemTime, and a Base64-encoded Base64Data field to facilitate subsequent parsing.

• The Feishu Bitable reader is optimized for data writing. The number of records in a batch write operation is increased to 1,000. The reader now supports fetching over 100 fields, and the record synchronization limit of 10,000 is removed.

• The Kafka reader now supports sequential writing to MaxCompute Delta tables. This ensures that only one record is retained for each primary key in the destination table, preventing data duplication.

• When a Date field in the Elasticsearch reader is null, the reader now writes a null value to the downstream table to ensure data consistency.

• Configure Milvus reader

• Configure Milvus writer

• Configure DataHub reader

• Configure Feishu Bitable reader

• Configure Kafka reader

• Configure Elasticsearch reader

Batch development

• You can now manage permissions for batch compute tasks by first-level directory, granting public access or specifying which members can view tasks within that directory.

• Batch physical tables now support StarRocks, Doris, and SelectDB as data sources.

• Database SQL now supports TDSQL for MySQL, TDSQL for PostgreSQL, and ClickHouse v20/v21 (versions earlier than v23+) as data sources.

• You can now set global SQL parameters (set parameters).

• Compute task

• Create and edit a batch physical table

• Create a Database SQL task

Real-time development

Dataphin now supports reading Kafka messages in JSON format, improving its ability to process semi-structured data.

None

Data quality

• You can now configure runtime settings for quality rules, including an automatic termination timeout and automatic retries for failed rules.

• You can now configure the maximum number of archived records for custom exception tables when MaxCompute is used as the source for computations.

• Create a quality rule for a Dataphin table

• Create a quality rule for a global data table

• Create a quality rule for a data source

• Create a real-time meta table rule

• Create a quality rule for an indicator

Metadata center

• Dataphin now supports TDSQL for PostgreSQL data sources and DataWorks (public cloud) application system collection sources.

• You can now define custom field attributes to add business context and better describe the meaning of fields.

• Supported operations for different types of collection sources

• Create and manage properties of asset objects, APIs, and dashboards

Data service

• New Dataset APIs are now available. You can create APIs from a dataset by using a configuration wizard or writing SQL. These APIs natively support vector retrieval and file download link generation, providing an out-of-the-box knowledge query service for AI agents.

• Direct Connection APIs now support AnalyticDB for MySQL 3.0, TDSQL for MySQL, and TDSQL for PostgreSQL data sources.

• The number of physical tables that can be added to a multi-physical table service unit is increased to 10.

• Create a dataset API in SQL mode

• Create a dataset API in wizard mode

• Data sources and compute engines supported by data service

• Create a multi-physical table service unit

OpenAPI

• New: ListOperationRecord. Queries operation records by page.

• New: GetOperationRecordRunCode. Gets the execution code for an operation record.

• New: GetOperationRecordById. Gets a single operation record by its ID.

• New: GetOperationRecordDetail. Gets the details of an operation record, including logs and results.

• New: SubmitBatchTask. Submits a task by using a specified batch compute template.

OpenAPI list

Feature

Description

References

Sales

The unstructured data feature is now available for purchase. You can choose from Basic, Standard, and Enterprise editions, and select Credit resources as needed.

• Basic: 5,000

• Standard: 20,000

• Enterprise: 50,000

• Purchase Guide - Fully Managed

• Billing Description - Fully Managed

• Activate Dataphin

Cross-tenant deployment

Database SQL now supports batch replacement of schemas through replacement rules in cross-tenant deployment.

Import a deployment package

Dataset

Dataphin now supports unstructured datasets. You can specify storage locations based on the dataset type: File Dataset, Table Dataset, or Hybrid Dataset.

Datasets

Workflow

The new workflow feature is the core engine for processing unstructured data. You can use it to orchestrate operators into a complete pipeline for automated batch processing of audio, video, images, text, and other unstructured data.

Workflow

Batch development

Database SQL now supports Lindorm data sources, including the Lindorm compute engine, wide table engine, and column-oriented storage engine.

Create a Database SQL task

Batch integration

• Adds batch integration support for InfluxDB 2.x and TDSQL for PostgreSQL.

• Optimized reading from and writing to Iceberg tables in the Hive component. You can now set range partitions and multi-level range partitions. This update also ensures compatibility with quoted partition values and case-sensitive fields during write operations.

• Configure InfluxDB reader

• Configure InfluxDB writer

• Configure dataset reader

• Configure TDSQL for PostgreSQL reader

• Configure TDSQL for PostgreSQL writer

Real-time development

• When Kafka is selected as the data source for a meta table, SHA-256 is now supported as an encryption method.

• The Paimon data source is deprecated. Use the Hive data source instead by selecting Aliyun EMR 5.x Hive 3.1.x and enabling the Paimon lake format.

• Create and manage a meta table

• Create a Paimon data source (deprecated, use Hive-Paimon instead)

Task O&M

You can now trigger a downstream node to run after an upstream node fails or completes.

Instance diagnostics

Asset inventory

You can now view the details of dataset assets, including basic information, version information, file lists, metadata previews, and data lineage.

Dataset details

Data quality

Permission configuration for quality rule scheduling is optimized. New "Project-level Quality Rule Management" and "Quality Rule Management" permissions let users view and select production tasks from specific or all projects.

Create a quality rule for a single monitored object

Asset catalog

You can now view metadata change history for business and technical indicator assets, including information such as indicator definitions and business owners, to facilitate auditing and backtracking.

View business indicator details

Catalog management

A new automatic delisting rule is available. When an asset is deleted, the system delists it sequentially based on the rules in a specified rule group.

• Manage quasi-assets

• Auto-listing and Delisting

Data service

You can now apply for API permissions in batches.

• Data Service Marketplace

• Manage API Permissions

• Permission Management for Data Services

Feature

Description

References

Sales

The X-Data Quality feature is now available for purchase in a Standard edition. This feature addresses data quality issues by integrating multi-source information, including data lineage, task code, and anomalous data. It uses large language models to analyze problems, form key evidence chains, and provide rectification suggestions. This creates a closed-loop governance process and significantly reduces manual costs.

• Purchase Guide - Fully Managed

• Billing Description - Fully Managed

• Activate Dataphin

Intelligence

The X-O&M Assistant intelligent application now supports configuring a subdomain to access the mobile O&M assistant.

X-O&M Assistant

Security settings

A new data source permission setting feature is available. You can now control whether project administrators or domain architects are allowed to create data sources.

Security Settings

Permission management

You can now use table classification and sensitivity levels to determine the approval template and dynamically control permission requests.

Apply for, renew, and return table permissions

Data source management

• Adds support for GBase 8c and Data Lake Formation (DLF) data sources.

• Data source improvements: Microsoft SQL Server is now compatible with earlier versions to resolve metadata acquisition issues; Oracle data source queries for table metadata are now case-insensitive; Salesforce data source adds support for the JWT authentication method.

• Create a GBase 8c data source

• Create a Salesforce data source

• Create a Data Lake Formation data source

Cross-tenant deployment

• Exporting and importing Flink Template SQL tasks and real-time task templates is not supported.

• For real-time objects, dependencies are not automatically added to the deployment package. When you deploy for the first time, you must manually add the projects for Flink DataStream and Flink SQL tasks, and the data sources for meta tables and mirror tables to the deployment package.

• Export a deployment package for specified objects

• Manage global objects to be deployed

Batch integration

• A new feature is available for collecting unstructured data. It consolidates unstructured files from different locations into a specified storage location. This feature supports OSS, S3, and FTP data sources.

• Adds batch integration support for GBase 8c and Data Lake Formation (DLF). Adds support for the OpenSearch reader and the DolphinDB writer.

• For full database migration, GBase 8c is now supported as a source, and Data Lake Formation (DLF) is supported as a destination.

• You can now hide global password variables in integration task run logs to enhance security.

• The Salesforce reader now supports the bulk read method.

• The FTP reader now supports reading escaped JSON.

• The one-click table creation feature for writers is optimized. To reduce operational risk, tables are now created during the deployment process instead of directly from the development environment.

• The multi-table filtering option for batch integration readers now supports native database regular expressions. The system uses these expressions to match and read tables in the target database.

• The MaxCompute writer now supports preparatory and concluding statements.

• Configure FTP reader

• Configure Salesforce reader

• Configure GBase 8c reader

• Configure GBase 8c writer

• Configure OpenSearch reader

• Configure Data Lake Formation reader

• Configure Data Lake Formation writer

• Configure DolphinDB writer

Batch development

• You can now add tags to compute tasks for filtering in O&M operations.

• You can now migrate compute tasks and integration tasks across projects.

• You can now apply for table permissions in batches from the compute task submission details page.

• MaxCompute adds support for the create snapshot table and add column if not exists syntax.

• Compute task

• batch integration

• Development Platform Settings

Task O&M

• O&M for scheduled tasks is improved. You can now modify the scheduling type, perform a data backfill for multiple tasks at once, download a list of nodes at a specified upstream or downstream level, and upload an Excel file to search for tasks.

• Data backfill instances now provide a flat view to see instance information and support downloading a list of instances.

• When creating a data backfill task, you can select custom business dates for a large-scale backfill to run the task for any time period.

• Overview of general O&M for scheduled tasks

• Overview of O&M for data backfill instances

Data standard

• Standard implementation assessment now supports field type compatibility, such as high-precision compatibility with low-precision, compatibility between similar types, and text compatibility with string, varchar, and other types.

• You can now manually add mapping relationships between data standards and asset objects, such as Dataphin fields and indicators.

• Mapping relationship overview

• Create and manage data standards

Data security

You can now import and export data categories in batches for more efficient configuration.

Batch import and export data categories

Metadata center

Sampling settings now support a disabled period. During this period, automatically triggered data sampling query tasks fail immediately instead of running.

Sampling configuration

Data service

• When the network is configured to use an Alibaba Cloud API gateway, you can now customize the AccessKey ID and AccessKey Secret.

• Direct Connection APIs now support DolphinDB, GBase 8c, OpenSearch, and GBase 8a data sources. Service unit APIs now support GBase 8c and GBase 8a data sources.

• When creating an API, you can now choose whether to include the executed SQL statement in the API response.

• When an API is called without specifying return fields, the response includes all fields that you have permission to access.

• When you test and debug an API, the system now retains the most recent test data to facilitate result comparison.

• Create a query-type API by using a direct connection to a data source

• Create an operation-type API by using a direct connection to a data source

• Create a service unit API in SQL mode

• Create a service unit API in wizard mode

• Create and manage my applications

• Appendix: API call examples

• Test and publish an API

Feature

Description

References

Super X (AI-powered applications)

• Built-in models now include support for Qwen3-Next-80B-A3B-Instruct/thinking and qwen3-235b-a22b-thinking/instruct-2507.

• X-Analysis is optimized. You can now choose whether to use sample data, which can significantly improve the accuracy of SQL generation. This update also adds support for smart repair and an option to open in a notebook.

• X-Analysis

• Built-in models supported by intelligent applications

Permission management

• You can now apply for row-level permissions. You can filter associated row-level permissions based on indicator tables and configure control rules.

• You can now apply for row-level permissions for multiple data tables in batches.

• Apply for, renew, and return table permissions

• Apply for and manage row-level permissions

• Add and manage row-level permissions

Batch integration

• Dataphin now provides batch integration support for Easysearch, PolarDB-X 2.0, Feishu Bitable, and Snowflake.

• The API reader and writer now support custom parameter delimiters and key-value connectors. New signature functions are supported: SHA1HEX, SHA256, SHA256HEX, SHA512HEX, and HMAC_SHA512. Custom signature strings are also supported.

• The FTP reader now supports reading tar.gz compressed files and text data in JSONL format. When the compression format is tar.gz or zip, you can configure file matching rules to filter files within the compressed package. When the compression format is tar.gz, you can specify a starting line for the data.

• Adds support for reading and writing custom data types in PostgreSQL, as well as range, GaussDB, openGauss, AnalyticDB for PostgreSQL, and Greenplum data types.

• Configure FTP reader

• Configure API reader

• Configure API writer

• Configure Easysearch reader

• Configure Easysearch writer

Batch development

• Database SQL now supports Trino and PolarDB-X 2.0 data sources.

• Database SQL supports writing -- @@dataphin_sql_begin -- @@dataphin_sql_end to manually specify an independent SQL statement and avoid incorrect SQL splitting by the system.

Create a Database SQL task

Metadata center

The new sampling configuration feature helps users understand data structures and develop SQL. You can configure data sampling tasks for data previews, security and standard identification, and intelligent applications. This feature also lets you configure automatic data sampling ranges for Dataphin tables and data source tables.

Sampling configuration

Data quality

• You can now import and export quality rules created from custom SQL templates in batches.

• Adds quality rule monitoring for GoldenDB, PolarDB, PolarDB-X 1.0, and PolarDB-X 2.0 data sources.

• Batch import quality rules

• Batch export quality rules

• Create a quality rule for a Dataphin table

• Create a quality rule for a data source

Catalog management

• Introduces the asset tags feature. You can now centrally manage tags for the asset catalog to optimize resource allocation.

• The definitions for business indicators now support referencing other business indicators and calculation symbols, which can be automatically populated into the indicator relationship expression in the indicator relationship graph.

• Manage quasi-assets

• Create and manage asset tags

Data service

• You can now create operation-type APIs for direct data source connections, including Create, Update, and Delete operations.

• Direct Connection APIs now support pagination parameters and optional parameters.

• Direct Connection APIs (synchronous and asynchronous) and service unit APIs (SQL and wizard modes) now support Doris and SelectDB data sources.

• For GET-type registered APIs, a smart add feature is now available for return parameters, allowing you to extract them from test results.

• For LIST APIs, you can use the returnTotalNum parameter to control whether to return the total number of results. When returnTotalNum=true, the total number of results (totalNum) is returned in the response.

• Data service improvements: API names can now be up to 100 characters long, and parameters for registered external and composite APIs now support Object and Array(Object) data types.

• Create a query-type API by using a direct connection to a data source

• Create an operation-type API by using a direct connection to a data source

• Create a registered API

• Create an API based on a model

• Call an API