Realtime Compute for Apache Flink:November 30, 2025

AI features

New AI functions are added for text processing in Flink SQL:

• Text classification: Classifies text based on preset categories and returns a confidence score. This function is suitable for ticket categorization and content tagging.

• Sentiment analysis: Analyzes the sentiment of text and returns a quantitative score with a positive, negative, or neutral label. This function is suitable for comment analysis and public opinion monitoring.

• Information extraction: Extracts data, such as names and phone numbers, from unstructured text using a JSON schema and outputs the data in a structured format.

• Text summarization: Generates a summary of long text. You can set a maximum length for the summary.

• Text vectorization: Creates high-dimensional vector representations of text. These vectors enable similarity calculations and semantic retrieval.

• Text translation: Translates text between more than 10 languages and automatically detects the source language.

• Data masking: Automatically masks sensitive information, such as phone numbers and ID card numbers, in text.

Data ingestion

• Compatibility: Aligns with Apache Flink CDC 3.5.

• PostgreSQL CDC (public preview): Supports PostgreSQL sources.

• MongoDB: Supports MongoDB sources.

Connectors

• Kafka:

  • The Canal-JSON format now supports parsing source database index events. You can also fetch the original key and value as metadata.

  • Supports Protocol Buffers v3 (proto3).

• Hologres:

  • Dimension table lookup joins support filter pushdown.

  • Stream Copy supports the BinaryRow format.

  • Supports configuring primary-key conflict strategies for INSERT operations.

• JDBC (PostgreSQL): Adds the ignore delete option to ignore delete events from the source table.

• Paimon:

  • Supports cross-partition updates.

  • Supports a username prefix to easily locate jobs.

• Redis: Enhances support for cluster instances and batch writes.

Catalog

• PostgreSQL catalog: Introduces the PostgreSQL catalog to manage PostgreSQL metadata.

• StarRocks catalog: Supports treating nullable primary key columns in StarRocks as non-nullable.

• MaxCompute catalog: Supports mapping a database to a MaxCompute project or schema.

Core features

1. Namespace cloning: Copy a namespace to other workspaces. This feature is suitable for cluster migration and dual-run testing.

2. Monitoring service change: Manually change the monitoring service for your workspace in the console.

3. Access control:

   • Configure 'disabled' permission points.

   • Set granular permissions for Git integration.

4. Log archiving: Archive logs to SLS through RAM rather than AccessKey pairs to improve security.

User experience

• O&M: The O&M page now displays the status and results of dynamic updates and automatic tuning.

• Billing tags: In a hybrid billed workspace, resources are automatically tagged subscription or pay-as-you-go for easier split billing.

• Resource search: Search for resources by workspace, workspace ID, or namespace.