Skip to content

Announce Flink Agents 0.1.0 #809

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Open
wants to merge 2 commits into
base: asf-site
Choose a base branch
from
Open

Announce Flink Agents 0.1.0 #809

Changes from all commits
Commits
Show all changes
2 commits
Select commit Hold shift + click to select a range
01d75aa
Announce Flink Agents 0.1.0
xintongsong Oct 2, 2025
0e8500a
address comments
xintongsong Oct 4, 2025
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
66 changes: 66 additions & 0 deletions docs/content/posts/2025-10-01-release-flink-agents-0.1.0.md
View file
Open in desktop
Original file line number Diff line number Diff line change
@@ -0,0 +1,66 @@
---
title: "Apache Flink Agents 0.1.0 Release Announcement"
date: "2025-10-01T08:00:00.000Z"
authors:
- xtsong:
name: "Xintong Song"
aliases:
- /news/2025/10/01/release-flink-agents-0.1.0.html
---
\<todo: update release date in: filename, header-date, header-aliases\>

The Apache Flink Community is excited to announce the first preview release of Apache Flink Agents (0.1.0).

## What is Apache Flink Agents

Apache Flink Agents is a new sub-project of Apache Flink for building event-driven AI agents directly on Flink’s streaming runtime. It unifies stream processing and autonomous agents in one framework, combining Flink's proven strengths -- scale, low latency, fault tolerance, and state management -- with agent capabilities -- LLMs, tools, memory, and dynamic orchestration.

### Why Apache Flink Agents Matters

While AI agents have made rapid progress in interactive applications like chatbots and copilots, these systems typically operate in synchronous, one-off interactions. Many business use cases cannot wait for a user to prompt them into action. In industrial settings such as e-commerce, finance, IoT, and logistics, critical decisions must be made instantly in response to live events like a payment failure, a sensor anomaly, or a user click.

To succeed in production, enterprise agents must be capable of:
- Operating on live, high-volume event streams such as transactions, sensor anomalies, or user clicks.
- Running continuously and autonomously, not only when prompted.
- Guaranteeing safety, auditability, and recovery when something goes wrong.

These workloads require more than intelligence alone. They demand massive scale, millisecond latency, fault tolerance, and stateful coordination. These are exactly the strengths of Apache Flink.

Until now, there has been no unified framework to bring agentic AI patterns into Flink’s proven streaming ecosystem. Apache Flink Agents bridges this gap by treating agents as event-driven microservices that are always-on, reliable, and scalable.

### Key Features

Building on Flink's battle-tested streaming engine, Apache Flink Agents inherits distributed, at-scale, fault-tolerant structured data processing and mature state management, and adds first-class abstractions for Agentic AI building blocks and functionalities -- large language models (LLMs), prompts, tools, memory, dynamic orchestration, observability, and more.

The key features of Apache Flink Agents include:
Copy link

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

We should mention replayability here.

Replayability for Testing and Safety: Supports replay of historical event streams, enabling regression testing, debugging, drift detection, and safe evaluation of agents without impacting live systems.

Copy link
Contributor Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

I'm not sure about this one.

It depends on what do we mean by "replay the historical event streams".

  • If we are replaying from a Flink's checkpoint, then it can hardly be used for the mentioned testing / debugging / evaluation purposes. Because with durable execution, we will skip all the actions (model invocation, tool use, etc.) that are previously executed and directly reuse their outcomes / results. The models / tools won't be actually called during the replay.
  • If we are not replaying from a Flink's checkpoint, then it's no different from using any other agent framework, as long as it reads from a replayable input like kafka.

Copy link

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Flink Agents offer this Replay-based Per Action State Consistency feature(implemented by @letaoj) which is related to replayability I think, is this what @thefalc suggested to call out?

Copy link
Contributor Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

@alnzng, that one is mentioned as "Exactly-Once Action Consistency". Here I didn't mention "replay-based", because I want our users to focus on WHAT capability (the consistency) we provide, rather than HOW (replay-based) we deliver that capability. And as I said, that seems to be conflict against the testing / debugging / evaluation purposes that @thefalc mentioned.

Copy link

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

@xintongsong I understand and agree with your point about focusing on WHAT rather than HOW. Let me clarify my original thoughts:

  • I thought the Exactly-Once Action Consistency section mainly focuses on ensuring unique (exactly-once) action execution, rather than deterministic action behavior (i.e., deterministic recovery despite the inherently non-deterministic nature of LLMs).
  • The non-deterministic LLM output is actually one of the core challenges addressed by the replay-based per-action state consistency design. Making LLM actions deterministic is also a fundamental requirement for features like regression testing and debugging.
  • With that in mind, I was thinking this aspect might deserve to be highlighted separately.

That said, your side effects statement may already implicitly cover this, since non-deterministic LLM behaviors can be viewed as side effects of agent actions. If that’s the case, perhaps we could make it explicit (e.g., “side effects such as non-deterministic LLM outputs”) to make the intention clearer.

- **Massive Scale and Millisecond Latency:** Processes massive-scale event streams in real time, leveraging Flink's distributed processing engine.
- **Seamless Data and AI Integration:** Agents interact directly with Flink's DataStream and Table APIs for input and output, enabling a smooth integration of structured data processing and semantic AI capabilities within Flink.
- **Exactly-Once Action Consistency:** Ensures exactly-once consistency for agent actions and their side effects by integrating Flink's checkpointing with an external write-ahead log.
- **Familiar Agent Abstractions:** Leverages well-known AI agent concepts, making it easy for developers experienced with agent-based systems to quickly adopt and build on Apache Flink Agents without a steep learning curve.
- **Multi-Language Supports:** Provides native APIs in both Python and Java, enabling seamless integration into diverse development environments and allowing teams to use their preferred programming language.
- **Rich Ecosystem:** Natively integrates mainstream LLMs, vector stores from diverse providers, and tools or prompts hosted on MCP servers into your agents, while enabling customizable extensions.
- **Observability:** Adopts an event-centric orchestration approach, where all agent actions are connected and controlled by events, enabling observation and understanding of agent behavior through the event log.

## Release 0.1.0

Flink Agents 0.1.0 is available for downloading [here](https://flink.apache.org/downloads/). \<todo: add flink agents to flink website download page\>

Documentations and quickstart examples can be found [here](https://nightlies.apache.org/flink/flink-agents-docs-main/). \<todo: update doc url to `.../flink-agents-release-0.1`\>

**Note:** Flink Agents 0.1.0 is a preview version, which means:
- Some functionalities may contain known or unknown issues. You can check the list of known issues and their resolution status via the [Github Issues](https://github.com/apache/flink-agents/issues).
- The current APIs and configuration options are experimental and may undergo non-backward compatible changes in future versions.
Therefore, we do not recommend using this version in production environments with high stability requirements.

We greatly appreciate any feedback you can provide, whether it's sharing how you're using our product, suggesting new features, helping us identify and fix bugs, or anything else that comes to mind. Your insights are invaluable to us.

You may reach out to us via:
- [Join the Apache Flink Slack](https://flink.apache.org/what-is-flink/community/#slack) and ask for help in [#flink-agents-user](xxx). \<todo: create channel and update link\>
- Post feature requirements and bug reports at [Github Issues](https://github.com/apache/flink-agents/issues).
- Share your use case and ideas at [Github Discussions](https://github.com/apache/flink-agents/discussions).

## List of Contributors

The Apache Flink community would like to thank each and every one of the contributors that have made this release possible:

Adem Amen Thabti, Alan Z., Eugene, Hao Li, HuangXingBo, Kavishankarks, KeGu-069, Letao Jiang, Qingsheng Ren, Richard, Wenjin Xie, Xintong Song, Xu Huang, Xuannan, twosom, zhaomin1423