TL;DR
The most important story behind RisingWave 3.0 isn’t the sheer number of new features—it’s that RisingWave has evolved into a much stronger real-time data platform for Agentic AI.
As AI moves beyond chatbots toward autonomous agents, the underlying data infrastructure becomes just as important as the models themselves. AI agents need continuous access to fresh data, the ability to retrieve and update context in real time, integrate with operational systems, and reliably trigger downstream actions. They also need open data formats that support analytics, fine-tuning, and long-term memory.
RisingWave 3.0 is built for this new generation of AI applications.
This release introduces Apache Iceberg V3 support to strengthen real-time lakehouse architectures, native pgvector ingestion for AI and semantic search workloads, new WebSocket and HTTP connectors for low-latency event ingestion and action execution, exactly-once delivery for reliable data pipelines, and DataFusion as the default analytical engine for faster batch processing. At the same time, SQL continues to evolve beyond a query language into the operational interface for managing streaming systems, allowing developers—and increasingly AI agents themselves—to automate tasks such as backups, schema evolution, connector management, and pipeline operations directly through SQL.
Together, these capabilities make RisingWave more than a streaming database. They establish it as a unified platform for continuously ingesting, processing, analyzing, and acting on live data—the foundation required to build reliable, production-grade Agentic AI applications.
A Streaming Data Platform Built for AI Agents
Every AI agent follows a similar lifecycle: ingest new information, maintain context, reason over current state, and trigger downstream actions. RisingWave is designed to power each stage of that lifecycle.
With Apache Iceberg V3 support, organizations can continuously stream operational data into open lakehouse tables that are immediately available for analytics, feature engineering, and model training. Combined with exactly-once delivery and native Iceberg integration, RisingWave enables reliable real-time pipelines without sacrificing the openness of the modern data stack.
The connector ecosystem has also expanded significantly. RisingWave 3.0 introduces a new WebSocket source for persistent low-latency ingestion and a new HTTP sink capable of dynamically routing records to different endpoints. These capabilities make it easier for AI agents to consume live events and trigger downstream services, APIs, or automation workflows in real time.
At the same time, PostgreSQL CDC, MySQL CDC, SQL Server CDC, Kafka, Google Pub/Sub, MQTT, JDBC, and Delta Lake integrations have all received major improvements, including PostgreSQL 18 support, native pgvector ingestion, enhanced observability, online configuration updates, and exactly-once delivery guarantees. Together, these enhancements make it easier to build reliable, production-ready AI data pipelines.
SQL Is Becoming the Operational Interface
One of the biggest themes in RisingWave 3.0 is making SQL the operational interface for streaming systems.
Traditionally, managing streaming infrastructure required a combination of SQL, command-line tools, REST APIs, and custom scripts. With RisingWave 3.0, many of these operational workflows can now be performed directly through SQL.
Developers can create metadata backups, inspect cursors across the entire cluster, tune backfill parallelism, modify subscription retention, rotate connector credentials, and evolve schemas without rebuilding streaming pipelines. These capabilities simplify day-to-day operations while making streaming infrastructure easier to automate.
For Agentic AI, this opens an interesting possibility: AI agents can not only query streaming data using SQL, but also safely manage and operate streaming workloads using the same familiar interface.
Faster Analytics with a Smarter Query Engine
Real-time applications increasingly combine streaming data with interactive analytics. RisingWave 3.0 delivers significant improvements to both.
Apache DataFusion is now the default execution engine for batch queries, bringing broader SQL compatibility, improved memory management, additional SQL functionality, and stronger analytical performance out of the box.
The optimizer has also become considerably smarter. RisingWave can now automatically recognize when an existing materialized view can answer a batch query and transparently rewrite the query to use the precomputed results. Developers define a materialized view once, and the optimizer automatically reuses it whenever possible, reducing query latency while lowering compute costs.
Built for Production
Production systems demand more than performance—they require reliability, security, and operational visibility.
RisingWave 3.0 introduces stronger observability through richer system catalogs, improved connector metrics, and enhanced tracing capabilities. Operators gain finer control over cluster behavior with configurable join cache eviction, adaptive parallelism improvements, and additional runtime configuration options.
Security has also improved with support for secrets inside user-defined functions, dependency tracking for secrets, and simplified secret lifecycle management.
Meanwhile, continued investments in PostgreSQL compatibility, online schema evolution, connector upgrades without downtime, and hundreds of bug fixes make upgrading to RisingWave 3.0 an easy choice for production deployments.
Looking Ahead
The future of data infrastructure isn’t simply faster stream processing—it’s enabling intelligent systems to continuously understand and react to the world around them.
With RisingWave 3.0, we’re bringing together real-time streaming, SQL, open lakehouse architectures, and a mature connector ecosystem into a single platform that powers both modern data applications and Agentic AI.
Whether you’re building AI copilots, autonomous agents, recommendation systems, fraud detection, operational analytics, or streaming lakehouses, RisingWave 3.0 provides the foundation to ingest, process, analyze, and act on data as it happens.
This is our biggest release yet, and we’re excited to see what you’ll build with RisingWave 3.0.
