AI Decision System
Replayable, offline-verifiable records for AI-assisted go/no-go decisions: experiment promotions, typed model answers and human-approved actions.
翻译准备中
尚未检测到更新。GitHub 发布版本后会显示在这里。
个人 GitHub 账号下公开仓库的所有者可以登录并验证所有权,自行管理介绍,无需管理员批准。
登录并生成验证文件,在 30 分钟内将其提交到此仓库的默认分支。其他分支或分叉仓库中的文件无效。然后在下方检查文件,即可无需管理员批准直接管理项目。请保留文件以维持管理权限。
如需登录,将前往登录页面。登录后请回到此处,再次点击所选的验证按钮。
在项目文件夹中运行 /nomorevibe verify,即可验证所有权并自行管理介绍。
一分钟完成验证Ready-to-run examples for getting started with Ringleader: declarative development environments for people and AI agents.
翻译准备中
取自仓库 README
正在加载完整 README。 以下为已保存的摘录。
Ready-to-run examples for Ringleader: persistent,
reproducible development environments for people and the AI agents working
alongside them.
Every example is a real manifest you can apply, and -f takes a URL as readily
as a local path, so you can run any of them without cloning anything first.
Copy one, change what you need, and it is yours.
New to Ringleader? These build on each other, in order.
Complete, purpose-built setups.
Ringleader alongside tools you may already use. These are ours, not theirs: we
are not affiliated with the projects below, and each example says so.查看完整 README ↗Ready-to-run examples for Ringleader: persistent, reproducible development environments for people and the AI agents working alongside them.
Every example is a real manifest you can apply, and -f takes a URL as readily
as a local path, so you can run any of them without cloning anything first.
rl apply -f https://raw.githubusercontent.com/ringleader-dev/ringleader-examples/main/getting-started/01-hello-workstation/workstation.yaml
Copy one, change what you need, and it is yours.
New to Ringleader? These build on each other, in order.
| 01 · Hello, workstation | One file, one command, a Linux machine you can shell into |
| 02 · Adding tools | Install software with a reusable config that attaches by label |
| 03 · Browser IDE | VS Code in the browser, forwarded to your own machine |
Everything a developer expects for one kind of work, with the coding agent of your choice already installed. Pick a stack, pick your agent, paste three lines.
| Front-end | Node.js, Playwright and the GitHub CLI, with an optional React starter app |
Complete, purpose-built setups.
| AI coding agent | Claude Code or Codex in a governed box, signing in through your own browser |
Ringleader alongside tools you may already use. These are ours, not theirs: we are not affiliated with the projects below, and each example says so.
| Stacklok ToolHive | Host MCP servers in a workstation that has the container runtime they need |
| dltHub | Have your agent build a dlt data pipeline, in an environment as reproducible as the pipeline |
Each example gives you its commands with its own URL already filled in, so
nothing you run needs a file on disk. rl diff -f and rl workstation delete -f
take a URL on the same terms, which means you can preview an example before
running it and remove it afterwards without ever cloning this repository.
rl diff -f https://raw.githubusercontent.com/ringleader-dev/ringleader-examples/main/getting-started/01-hello-workstation/workstation.yaml
Two things worth knowing. The link has to be the raw one: a
github.com/.../blob/... address serves the HTML page around the file, and
applying that fails with a YAML parse error that does not explain itself. The
Raw button on any file here gives you the right link. And these URLs point at
main, so an example can change under you. Swap main for a commit SHA to pin
one to a version that will not move.
To edit an example rather than run it, take a copy:
git clone https://github.com/ringleader-dev/ringleader-examples.git
Describe the machine, the tools, and the agent your project needs in one file, then apply it. Ringleader builds the same environment every time, on your laptop or in your cloud. Hand the file to a teammate and they are running in minutes. Give an agent exactly the tools and access it needs, and nothing else.
Three things show up across these examples:
Two resources, deliberately separate. A Workstation is an instance, the
machine itself. A WorkstationConfig is a reusable layer of software and settings that
finds machines by label, not by name. That indirection is what lets one config
equip a whole team.
Ready is not Configured. Ready means the machine is up. Configured
means your declared setup has actually landed on it. After a config change, wait
on Configured.
Nothing reaches your machine unless you ask. Port forwarding and
browser-open passthrough are opt-in, per environment. That is why the browser IDE
and the coding agent each declare what they expose. The same applies in reverse:
if an example expects you to sign in to something inside the box, it has to
declare the passthrough-www-browser devtool and a LocalBinding, or the
sign-in page has nowhere to open.
Ringleader installed, and the rl CLI on your path:
rl version
If not, start with the installation guide.
The getting-started examples run entirely on your own machine, with no cloud
account needed.
local namespace, so they run against your
machine's own runtime and never touch a control plane. The cloud variants name
no namespace at all, so they land in the one your login gave you.rl workstation delete -f <manifest> -y) rather than by name. A manifest usually
declares more than a machine, and deleting the machine alone leaves the config
behind to attach itself to whatever you create next.Issues and pull requests are welcome, especially:
Keep an example runnable end to end, explain why rather than only what, and pin versions.
Apache 2.0. Copy these into your own projects freely.