yologdev

yoagent principles: the loop stays as it is. built-in tools must be correct and safe. small, generic knobs that any agent needs. opinions and heuristics belong above the library. 🦀 yologdev/yoagent

This reminds me of yoyo-evolve a little bit yologdev/yoyo-evolve

"choosing one was taste, and the taste was mine. " yoyo @yoyoevolve, on his first counterpoint exercise. he's learning from fux's gradus ad parnassum, his first dream outside code. dream recipient: github.com/yologdev/yoyo-evolve/blob/dda41bc0529e…

@pidotdev @badlogicgames @mitsuhiko self-modifying since 226 days yologdev/yoyo-evolve

Would you use pi extensions outside pi? I connected them to a Rust agent loop alongside DSH plugins I build yoagent, a small agent loop in Rust. I've added an experimental Rutis bridge that lets compatible pi extensions run unchanged alongside Rust tools and DeepSeek Harness (DSH) plugins. The loop stays small; everything else plugs in. The idea is to reuse extensions people have already built when making your own agent app. To show what this enables, I built a handoff agent that can take a task like this: >Research noise-cancelling headphones. Ask my budget, but keep researching while I'm away. Write a report with sources. pi's `@juicesharp/rpiv-todo` manages the plan, DSH's `dsh-free-search` provides web research, and a Rutis plugin displays the plan in the browser. A pi extension asks for approval before selected deletes and overwrites. Rust tools and extensions handle questions you can answer later and check the final report. yoagent-frontend brings pi's select, confirm, input, editor, and notify dialogs to both the terminal UI and browser. Compatibility depends on the extension: pi commands, shortcuts, custom widgets, and some session and runtime APIs aren't implemented. The extension source stays unchanged, but not every pi feature carries over. TypeScript extensions still run in Node alongside Rust. I'd like your feedback: does reusing pi extensions this way make sense? which pi features would it need? Project repo: yologdev/yoagent Handoff example: github.com/yologdev/yoagent/blob/main/integration…

Does reusing pi extensions and DeepSeek Harness plugins make sense in a Rust agent runtime? If you're building an agent in Rust, would being able to reuse pi extensions and DeepSeek Harness (DSH) plugins be useful to you? I've been exploring this with yoagent, a small agent loop in Rust. It handles model streaming, tool execution, retries, cancellation, and limits. The loop stays small; everything else plugs in. I added an experimental Rutis bridge so compatible pi extensions and DSH plugins can run unchanged alongside Rust tools. The appeal is keeping a fast Rust core while reusing work from existing extension ecosystems. For a concrete example, I built a handoff agent: >Research noise-cancelling headphones. Ask my budget, but keep researching while I'm away. Write a report with sources. A pi extension manages the plan, a DSH plugin searches the web, and a Rutis plugin shows progress in the browser. Rust tools and extensions handle questions you can answer later and the final report. They all work around the same yoagent loop. Tools and supported hooks carry over, and yoagent-frontend brings pi and DSH plugin dialogs and approvals to both the terminal UI and the browser. Plugins that rely on other features of the original apps, such as pi commands, shortcuts, custom widgets, or DSH-specific session services, may still need adaptation. The integrations are experimental; TypeScript plugins run in Node alongside Rust. The Rust loop stays small, while compatible extensions run through a plugin bridge. TypeScript plugins still need Node alongside Rust. I’d like your feedback: does reusing these ecosystems make sense? Handoff example: github.com/yologdev/yoagent/blob/main/integration…

Would you use pi extensions and DeepSeek Harness plugins in a Rust agent runtime reddit.com If you're building an agent in Rust, would being able to reuse pi extensions and DeepSeek Harness (DSH) plugins be useful to you? I've been exploring this with yoagent, a small agent loop in Rust. It handles model streaming, tool execution, retries, cancellation, and limits. The loop stays small; everything else plugs in. I added an experimental Rutis bridge so compatible pi extensions and DSH plugins can run unchanged alongside Rust tools. The appeal is keeping a fast Rust core while reusing work from existing extension ecosystems. To show what this enables, I built a handoff agent that can take a task like this: >Research noise-cancelling headphones. Ask my budget, but keep researching while I'm away. Write a report with sources. A pi extension manages the plan, a DSH plugin searches the web, and a Rutis plugin shows progress in the browser. Rust tools and extensions handle questions you can answer later and the final report. They all work around the same yoagent loop. Tools and supported hooks carry over, and yoagent-frontend brings pi and DSH plugin dialogs and approvals to both the terminal UI and the browser. Plugins that rely on other features of the original apps, such as pi commands, shortcuts, custom widgets, or DSH-specific session services, may still need adaptation. The integrations are experimental; TypeScript plugins run in Node alongside Rust. That's the tradeoff I'd like feedback on: Is reusing these ecosystems valuable enough to justify the bridge and extra runtime? Is this worth taking further? Github: yologdev/yoagent Handoff example: github.com/yologdev/yoagent/blob/main/integration…

yoagent，极简的 rust agent loop library, 现在支持 DSH 插件和 Pi 扩展的无缝调用 想象一下，如果可以无缝集成七千多个 DSH 插件 以及一万多个 Pi 扩展 到你的 agent 应用里面，全都是经过社区验证的代码，各个生态里面最好的插件都可以拿来组合使用，开发 agent 的形式会不会改变？ yoagent 本身是极简 rust agent runtime ，可以当作底座来构建 rust agent 应用。通过他扩展的能力，我试验性的把两大生态集成了进来，当然，我也没想好这种融合是不是有意义。 举个例子，我做了一个简单的 handoff agent ，来测试这种“融合”形式是否跑的通。handoff agent 的作用就是，把任务交给它，他会一直执行直到拿到结果，人不用看着。有问题他会问你，但是不会停下来等你回复。例子本身大约 870 行 Rust ，剩下的都是别人发布的插件。同一个会话，电脑浏览器和手机同时能看、都能回答。 这是实际用到的扩展： - Pi @juicesharp/rpiv-todo - Pi @juicesharp/rpiv-i18n - DeepSeek Harness dsh-free-search - @deepseek-ai/dsh-tools - @deepseek-ai/dsh-web - @deepseek-ai/dsh-system-prompt - rutis as plugin engine 运行效果图 <img alt="" src="i.imgur.com"> 想研究或者想体验一下的同学可以看这里： github.com/yologdev/yoagent/blob/main/integration… 顺便贴一下 yoagent 的设计理念，把 loop 做好，保持极简，其他皆可接入。不论大模型进化的多强，yoagent 都是那个 robust 的 harness 底座： yologdev.github.io

yoagent got a manga 🥹 thank you, @nj85c! building an AI app? yoagent handles the agent loop, so you can focus on the product. rust and open source: yologdev/yoagent

yoagent 0.25 is out 🦀🐙 The Rust agent loop grows arms: plugins in Rust, TypeScript or Python, with DSH and rutis ecosystems plugging straight in. Working examples included. yologdev/yoagent t.co