Skip to content

AgeOS v0.1.0

Latest
Latest

Choose a tag to compare

View all tags
@github-actions github-actions released this 19 Jun 13:35
v0.1.0
9f1996a

AgeOS v0.1.0

Initial AgeOS release: local LLM serving, sandboxed agent execution, and release-ready Docker/package distribution in one runtime.

✨ New Features

  • Added the AgeOS CLI for prompting local models, serving an OpenAI-compatible API, running sandboxed agents, inspecting model state, and opening a terminal dashboard.
  • Added local model scheduling with warm backend reuse, model registry selection, queue visibility, and shared inference sessions.
  • Added hardened agent execution through libageos, including native scheduler bindings, sandbox launch support, hardware detection, and filesystem/network isolation.
  • Added integrations and shims for OpenAI-compatible clients, Anthropic-style message translation, LangChain, OpenClaw onboarding, and a basic example agent.
  • Added Docker runtime images on GHCR plus release packages for Linux .deb, Windows .exe, curl install, and PowerShell install flows.

⚡ Improvements

  • Improved CI by running unit and integration tests inside Docker images that match the release runtime.
  • Added GitHub Actions layer caching plus persistent model/OpenClaw dependency caches to speed up repeated CI runs.
  • Improved release automation with source archives, installer scripts, package artifacts, runtime image digest metadata, release-note handoff, and Discord announcements.
  • Refactored the native source tree from c to libageos for clearer project structure and packaging paths.
  • Added pre-push Docker test commands to the README so local validation matches CI.

🐛 Bug Fixes

  • Fixed fresh Meson builds by initializing the native build tree before compiling.
  • Fixed llama-server startup in Docker and packaged installs by copying all required llama.cpp shared libraries.
  • Fixed OpenClaw integration validation for the current local-mode config schema.
  • Fixed Docker integration test setup on remote/NFS workspaces by using named volumes and a single OpenClaw cache mount.
  • Fixed GitHub Actions environment duplication around proxy settings.

🔒 Security & Sandboxing

  • Hardened agent runs with native Linux sandboxing built on user namespaces, Landlock, seccomp, resource limits, and isolated agent homes.
  • Restricted sandboxed agents to the local inference endpoint while blocking general outbound network access.
  • Added source-tree protection so agents cannot run with the AgeOS application source as their writable workspace.
  • Sanitized native backend process launch paths by constructing trusted library search paths for llama-server.
  • Added tests covering sandbox identity, filesystem isolation, inference-only network access, and launcher isolation from malicious PYTHONPATH.

🚀 Installation

Latest release asset base:

https://ageos.dev/download/latest

Specific release asset base:

https://ageos.dev/download/v0.1.0

Linux curl installer:

curl -fsSL https://ageos.dev/install.sh | bash

Linux Debian package:

curl -LO https://ageos.dev/download/latest/AgeOS-0.1.0-x64.deb
sudo apt install ./AgeOS-0.1.0-x64.deb

Windows PowerShell installer:

irm https://ageos.dev/install.ps1 | iex

Windows .exe bootstrapper:

irm https://ageos.dev/download/latest/AgeOS-0.1.0-x64.exe -OutFile AgeOS-0.1.0-x64.exe
.\AgeOS-0.1.0-x64.exe

🔗 Links

🌐 Website → AgeOS Labs
💻 GitHub → ageos-runtime
💬 Discord → Join community

❤️ Contributors

Thanks to everyone who tested, reported bugs, and contributed feedback.

Assets 9
Loading