openEuler releases 24.03 LTS SP4 and 26.09 innovation versions with AI agent focus

The openEuler project has announced two significant updates: the 24.03 LTS Service Pack 4 (SP4) and the 26.09 innovation release. The LTS track continues to provide feature updates to the 24.03 series, while the 26.09 release offers newer technologies on a shorter support cycle ending in March 2027.

These releases emphasize deep integration of AI agents and developer tools. Key additions include the Witty framework for natural language system tuning, the PolyMind self-hosted AI platform, and the epkg user-space package manager designed for project-specific software isolation. The updates also bring substantial kernel-level improvements to scheduling, file system performance, and memory management.

Technical breakdown of AI and kernel enhancements

• AI Frameworks: The Witty framework enables intelligent diagnosis and tuning via natural language, supporting both local models and OpenAI-compatible APIs, complemented by an Agent sandbox for isolated code execution.
• Kernel Subsystems: Integration of the sched_ext eBPF-based scheduler framework allows runtime loading of custom CPU schedulers, while FUSE file systems now utilize io_uring for improved performance.
• Memory and I/O: A programmable page cache allows applications to define custom caching policies to accelerate large AI model loading, and ARM architectures can now opt into 64K page-size kernels.
• Package Management: The 26.09 DevStation edition introduces epkg, a user-space package manager that installs software into home directories to maintain separate environments without affecting system packages.
• Agent Security: The SkillHub AgentCensor provides a security layer between AI agents and the system by logging and controlling tool calls, file changes, and system calls.

Why it matters

These releases signal a shift toward operating systems that treat AI agents as first-class citizens, providing the necessary sandboxing and security controls. For developers and sysadmins, the combination of eBPF-based scheduling and user-space package management offers more granular control over performance and environment isolation.