Confluent Cloud for Apache Flink has been overhauled to serve both developers writing code and data teams working in SQL/dbt. The Table API is now Generally Available in Java and matches or beats the DataStream API for most workloads, while new Process Table Functions (PTFs) give developers stateful, timer-based control for cases plain SQL can't express. UDFs now support Java on GCP and Python on AWS, closing a multi-cloud gap. On the data side, a GA dbt adapter and Materialized Tables let data engineers manage Flink pipelines with familiar dbt workflows, and new Snapshot Queries unify batch and streaming queries over Iceberg/Parquet data. The platform also adds AI-oriented features (Flink AI Model Inference, Multivariate Anomaly Detection, Streaming Agents, MCP support), tighter Kafka/Flink governance and security, and new Flink Advantage Packs pricing offering up to 75% savings for committed usage.
Table of contents
Mission-Critical Power for DevelopersBuilding Always-Fresh Data Products for Analytics EngineersAI-Ready Processing: Enriching Data for the AI EraA Unified Platform: Kafka and Flink Co-DesignedPredictable Flink Pricing with Advantage PacksGetting StartedQuestions this post answers
Is the Flink Table API generally available in Java on Confluent Cloud?
Yes, the Flink Table API is now Generally Available in Java on Confluent Cloud, giving developers a code-first way to build streaming applications while keeping the operational simplicity of a serverless runtime. For roughly 95% to 99% of streaming use cases, it matches or outperforms the lower-level DataStream API. Track Flink API changes like this on daily.dev before migrating streaming workloads off DataStream.
What are Process Table Functions (PTFs) in Confluent Cloud Flink used for?
Process Table Functions (PTFs) let developers maintain custom state and configure precise timers within Flink pipelines for logic that standard SQL cannot express, such as fraud detection, dynamic pricing, or multi-metric anomaly detection. PTFs are generally available in Java across AWS, Azure, and GCP, working alongside the Table API rather than replacing it. Developers evaluating stateful stream processing options can follow updates like PTFs on daily.dev.
Can I manage Confluent Cloud Flink pipelines using dbt?
Yes, the dbt adapter for Confluent Cloud Flink and Materialized Tables are now generally available, letting data teams manage the entire Flink pipeline lifecycle using the same dbt run workflow already used for warehouses like Snowflake or Databricks. This removes the need to learn separate CLI tools or manual streaming workflows. Data engineers weighing dbt-based streaming workflows can keep up with releases like this on daily.dev.