AI UNDERDOGSDAILY PICK
AI UNDERDOGS
Agent 拓扑重启即失？
Topology dies on reboot?
OpenRig
花了半天搭的多智能体架构
重启一次全没了——每个会话都是孤岛
OpenRig 说这不行
You spend half a day wiring up a
multi-agent graph, reboot once, it's gone — every
session an island. OpenRig says that's broken
智能体的 Terraform
Terraform for agents
OpenRig 是多智能体编排的基础设施层——一个 YAML 文件声明你所有 agent 的拓扑关系
一条命令启动全部
持久身份、共享内存、快照恢复
这些云计算用了十年的原语
终于有人觉得 coding agent 也该有
OpenRig is an infra layer for multi-agent orchestration
— one YAML file declares your full fleet
topology, one command boots everything. Persistent identity, shared
memory, snapshot-restore — infra primitives the cloud world
used for a decade, finally brought to coding
agents
★ SIGNAL 1
拓扑图肉眼可见
Graph you can inspect
它不是 spawn 几个进程就叫编排——OpenRig 会自动发现
启动、快照、恢复
并且把你的会话拓扑可视化
打开 localhost:7433
Claude Code 和 Codex 在同一个拓扑图里
点进去就能看到实时连接
It's not just 'spawn some processes and call
it orchestration' — OpenRig discovers, boots, snapshots, restores
and visualizes the graph your sessions form. Hit
localhost:7433, click to inspect — Claude Code and
Codex in the same topology, live
★ SIGNAL 2
基础设施级原语
Infra-grade primitives
作者的品味在于：他不是在做一个「Agent 管理面板」
而是把 Terraform 的思路搬过来——声明式定义
可复现部署、状态持久化
这些在云基础设施里是基本功
但在 agent 世界里几乎没人做
The taste here: he's not building an 'agent
manager dashboard' — he's bringing the Terraform mindset
Declarative config, reproducible deploys, state persistence. These are
table stakes in cloud infra, but almost nobody
applied them to agents
作者自己说了：「这个拓扑能扛住真实工作
而不是在重启时崩溃」——他不是在做一个 demo 级别的玩具
而是认真想解决 agent 会话状态丢失这个问题
开源项目，还在早期，但方向对了
The author said it: 'The topology survives real
work instead of collapsing at reboot time.' He's
not shipping a demo — he's solving agent
session state loss for real. Open source, still
early, but the direction is right
AI UNDERDOGS
GitHub 开源，欢迎 star
Open source on GitHub — go give it a star
OpenRig
关注 · 每天发现更多 AI 神作
www.openrig.dev