Releases: agentscope-ai/agentscope-runtime-java
v1.0.2
What's New
Refactor: Modularized Sandbox Architecture
- Extracted the sandbox functionality into a dedicated, standalone module:
agentscope-runtime-sandbox. - Moved extensible capabilities (e.g., FC sandbox, AgentRun Sandbox) from the core module into new extension modules to reduce dependencies.
Added: Flexible AgentApp Configuration
- You can now configure
AgentAppparameters using either:- Environment variables, or
- Command-line arguments
This enables easier deployment across different environments (e.g., dev, staging, production) without code changes.
New Contributors
- @chiangzeon made his first contribution in #84
- @Shen Junkun made his first contribution in #83
v1.0.1
What's New
- CORS support added to AgentApp — configurable allowed origins, methods, and headers.
- Zero-copy read-write mount for sandbox file mounting — improves I/O performance with no data duplication.
v1.0.0
AgentScope Runtime Java V1.0 Release Note
AgentScope Runtime Java V1.0 builds on a robust foundation of efficient agent deployment and secure sandbox execution. It now delivers a unified "Agent as API" experience spanning the entire agent development lifecycle — from local deployment and testing to production rollout — while featuring expanded sandbox type support and enhanced protocol compatibility.
At the same time, the integration method of agents with runtime services has evolved from black-box module replacement to a white-box adapter pattern — empowering developers to retain the native interfaces and behaviors of their existing agent frameworks, while natively embedding runtime capabilities (including state management, session history, and tool registration) directly into the application lifecycle. This delivers enhanced flexibility and seamless cross-framework interoperability.
Key Improvements in V1.0
- Unified Dev/Prod Paradigm — Consistent agent functionality across development and production environments
- Native Multi-agent Support — Full compatibility with AgentScope’s multi-agent paradigms
- Mainstream SDK & Protocol Integration — Support for OpenAI Responses API SDK and Google A2A protocol
- Visual Web UI — Out-of-the-box web chat interface available immediately after deployment
- Expanded Sandbox Types — GUI, Browser, FileSystem, Mobile, and Cloud (most visualized via VNC)
- Flexible Deployment Modes — Local web application, Docker, Kubernetes, or hosted cloud deployment
In V1.0, these capabilities are exposed via an adapter pattern, enabling seamless integration with native modules of different agent frameworks while preserving their original interfaces and behaviors. This ensures both broad compatibility and maximum flexibility for developers.
For detailed change logs and migration guidance, please refer to the CHANGELOG.
What's Changed
- fix(A2A): Fix ignored update artifact event in A2A executor by @KomachiSion in #30
- fix(AgentScope): Resolve message duplication issue in AgentScope adapter by @chickenlj in #31
- fix(test): Fix broken AgentScope unit tests by @chickenlj in #33
- chore(version): Bump version to V0.2.1 by @chickenlj in #34
- fix(A2A): Add active task cancellation support and optimize agent configuration by @KomachiSion in #37
- fix(A2A): Correct local IP retrieval in multi-container deployments by @alickreborn0 in #37
- feat(deployment): Add
agentrunand Kubernetes deployer support by @xuehuitian45 in #38 - refactor(runtime): Refactor properties keys for
runtime-a2a-spring-boot-starterby @KomachiSion in #39 - feat(a2a): Add support for A2A blocking requests with direct message responses by @KomachiSion in #39
- fix(core): Fix message handling bugs and adapt A2A to new protocol format by @xuehuitian45 in #43
- fix(starter): Ensure starter mode works properly in V1.0.0-beta by @chickenlj in #44
- feat(doc): Add cookbook and fix source code bugs identified during documentation by @xuehuitian45 in #45
- docs: Update documentation for V1.0 release by @chickenlj in #46
- feat(openai): Integrate OpenAI Responses API protocol by @xuehuitian45 in #47
- feat(example): Add simple sandbox demonstration example by @xuehuitian45 in #48
- fix(core): Resolve source code issues blocking cookbook sample execution by @xuehuitian45 in #49
- refactor(a2a): Enhance streaming request handling and context state management by @KomachiSion in #50
- feat(sandbox): Add AgentBay sandbox support by @xuehuitian45 in #53
- feat(sandbox): Introduce mobile and custom sandbox types by @xuehuitian45 in #54
- docs: Add English documentation for AgentScope Runtime Java V1.0 by @chickenlj in #55
- fix(test): Update unit tests to align with V1.0 changes by @chickenlj in #56
- chore(version): Bump version to V1.0.0 by @chickenlj in #57
New Contributors
- @KomachiSion made their first contribution in #30
- @alickreborn0 made their first contribution in #37
Contributors
Special thanks to chickenlj, xuehuitian45, and 2 other contributors for their efforts in making this release possible.
v1.0.0-BETA1
first 1.0 beta release
v0.1.3
v0.1.1
v0.1.0
We are honored to announce the first official release of Agentscope Runtime - 0.1.0.
Core Features
- Deployment Infrastructure: Built-in services for session management, memory, and sandbox environment control
- Sandboxed Tool Execution: Isolated sandboxes ensure safe tool execution without system compromise
- Developer Friendly: Simple deployment with powerful customization options
- Framework Agnostic: Not tied to any specific framework. Works seamlessly with popular open-source like AgentScope and Spring AI Alibaba.
Learn More
Please learn more about how to use Agentscope Runtime by reading our documentation.