Release Notes
2026.04
Feature Updates
| Feature | Overview | Related Document |
|---|---|---|
| Add users to offline development tasks | A task can now be assigned to multiple users. Users who have Subprocess permissions in other Offline Dev tasks can reference the task. | Offline Development Permission Management |
| Draft status and version management added to offline development | Versions can now be managed through Draft and Published states, with support for reviewing historical versions and rolling back to a previous version. | Offline Development Task Version Management |
| Support for parameter assignment | Dynamic values from datasets or database tables can now be used as parameter values and passed to downstream nodes. | Parameter Assignment |
More database types supported for DB Dataflow | Added support for MySQL, StarRocks, Doris, and SelectDB. | DB Dataflow |
More database types supported for Real-time Sync | Added Oracle and SQL Server as source databases, and Hologres as a target database. | Real-time Sync Overview |
Product Behavior Changes
Added users to offline development tasks
Before the upgrade: In Offline Dev, only the task owner could edit and run a task, and each task could have only one owner.
After the upgrade: Offline Dev now includes a collaborator role, allowing multiple collaborators to edit and run the same offline development task together.
Added draft status and version management to offline development
Before the upgrade: Saving an offline development task immediately published it, and scheduling always ran the most recently saved version.
After the upgrade: Offline Dev now supports Save as Draft and Publish. Incomplete changes can be saved as drafts, scheduling runs the latest published version, and each published version is recorded in version history.
2026.01
Feature Updates
| Feature | Overview | Related Document |
|---|---|---|
| Add collaborators to offline development tasks | A task can now be assigned to multiple collaborators, each of whom has edit permissions for the offline development task. | Offline Development Permission Management |
Support DB Dataflow | Dataflow computation logic can now be pushed down to enterprise data warehouses. | DB Dataflow |
2025.12
Feature Updates
| Feature | Overview | Related Document |
|---|---|---|
| Support workflow task migration | Supports rapid migration and overwrite of tasks, typically used when updating test and production versions. | Offline Development Task |
| Support creating data models | Output datasets can now generate CREATE TABLE statements based on the schema, which users can further customize before creating the table. | Create a Data Model |
Product Behavior Changes
Support creating data models in the dataflow Database Output node
Before the upgrade: If the target table name did not already exist in the input database, the table was created automatically. Users could not modify the generated table creation statement, which limited downstream usage.
After the upgrade: A new Create Table capability is available. The platform automatically generates a table creation template based on the table name, primary key, and field structure, and users can modify it according to warehouse design requirements.
2025.11
Feature Updates
| Feature | Overview | Related Document |
|---|---|---|
Support custom image configuration for Python Node | Users can build images based on their required Python versions and packages. After configuring the image in the admin backend, it can be used in Python Node. | Python Node |
| Support SAP ERP data input in dataflow | Table data can now be retrieved from SAP systems through RFC interfaces. | SAP ERP Input Support |
| Improved usability of monitoring instances | Instance nodes now expose input and output data resources to help users quickly locate resource details. | [Task Monitor](08-Task Monitor.md) |
Product Behavior Changes
The high-performance switch for StarRocks extraction and output in dataflow is only displayed after high-performance mode is enabled in the backend
Before the upgrade: High-performance mode could be enabled directly without backend configuration.
After the upgrade: The high-performance switch is displayed only after it has been enabled in the backend.
Support custom image configuration for Python Node
Before the upgrade: At runtime, Python Node could only use images provided by Guandata. Users could not configure the Python version or installed packages themselves.
After the upgrade: Users can build images based on the Python versions and packages used in their own environments, configure the image URL and secret in Management Center, and then use their own Python image at runtime in Python Node.
2025.09
Feature Updates
| Feature | Overview | Related Document |
|---|---|---|
| High-performance extraction for StarRocks in dataflow | When the account type of Database Input is StarRocks, dataflow can use high-performance extraction mode to improve task execution efficiency. | Database Input |
| Batch backfill | Supports one-time, batch scheduling and execution of historical time ranges for specified offline computing tasks. | Offline Development Task |
| Push down dataflow computation logic to databases | ETL operators normally use Spark compute resources. After direct-connection pushdown is enabled, ETL computation logic is automatically pushed down to the database, and warehouse compute resources are used during execution and preview. | Dataflow Supports Spark JDBC Pushdown |
| Support FTP/SFTP Excel files | The File Input node in dataflow now supports Excel files from FTP and SFTP and can process them in downstream steps. | File Input |
Real-time Sync | Supports synchronizing source-side data changes to the target database in real time, so the target remains consistent with the source. | Real-time Sync Overview |
Product Behavior Changes
High-performance extraction for StarRocks in dataflow
Before the upgrade: StarRocks extraction took a relatively long time, and performance still needed improvement.
After the upgrade: You can enable high-performance mode in Database Input to improve extraction efficiency and reduce execution time by more than half in most cases.
Batch backfill
Before the upgrade: Multi-day backfill scenarios had to be completed by manually running the task multiple times.
After the upgrade: The batch backfill capability allows you to enter multiple sets of parameter values at once, generating multiple backfill instances to complete multi-day or multi-month backfills.
Push down dataflow computation logic to databases
Before the upgrade: In direct-connection database scenarios, data was first extracted to the data development platform and then processed using Spark compute resources.
After the upgrade: In direct-connection database scenarios, part of the computation logic is pushed down to the database and executed using database compute resources.
Support FTP/SFTP Excel files
Before the upgrade: File Input in dataflow did not support Excel files.
After the upgrade: File Input in dataflow now supports Excel files.
Real-time Sync
Before the upgrade: High-frequency data synchronization scenarios were implemented through near real-time scheduling in Offline Dev.
After the upgrade: Real-time Sync can now be used to achieve real-time synchronization of business data.