展示HN：Bsub.io – 命令行工具的零配置批量执行
我创建 bsub 是因为每次需要从网络应用程序运行重型命令行工具时，我都厌倦了配置 Docker 镜像、Python 环境、GPU、沙箱和资源限制。我想要的是：发送文件 -> 在云中运行作业 -> 获取输出 -> 完成。
[bsub 官网](https://www.bsub.io)
bsub 让你可以像远程批处理作业一样执行 Whisper、Typst、Pandoc、Docling 和 FFmpeg 等工具，而无需进行环境设置。你可以通过命令行界面（CLI）在本地尝试它们，或者通过简单的 REST API 进行集成。
示例（PDF 提取）：
```
bsubio submit -w pdf/extract *.pdf
```
这就像在本地运行工具一样，但计算和隔离是在云中进行的。
技术细节：
- 每个作业在一个隔离的容器中运行，具有定义的 CPU/GPU/RAM 限制。
- 文件在作业期间临时存储，完成后会被删除。
- REST API 返回作业状态、日志和结果。
- 轻量级处理器（Typst、Pandoc）的冷启动时间较短；Whisper/FFmpeg 由于模型加载/编码时间需要更长。
- 后端可以水平扩展；在负载高峰期间可以增加更多工作节点。
当前处理器：
```
SST/Whisper -- 语音转文本
排版 -- Typst, Pandoc
PDF 提取 -- Docling
视频转码 -- FFmpeg
```
更多处理器正在开发中；欢迎对那些在本地设置起来麻烦的工具提出建议。
寻找测试者！CLI 是开源的：[GitHub 链接](https://github.com/bsubio/cli)。Linux/macOS 的安装程序已可用；Windows 测试正在进行中。早期测试期间免费；定价待定。
如果你使用 Windows，反馈尤其重要：contact@bsub.io
如果你尝试了这个工具，我会很感激你对 API 设计、延迟、缺失的处理器或任何粗糙之处的反馈。
查看原文
I built bsub because I was tired of wiring up Docker images, Python environments, GPUs, sandboxing, and resource limits every time I needed to run heavy command-line tools from web apps. I wanted: send files -> run job in the cloud -> get output -> done.<p><a href="https://www.bsub.io" rel="nofollow">https://www.bsub.io</a><p>bsub lets you execute tools like Whisper, Typst, Pandoc, Docling, and FFmpeg as remote batch jobs with no environment setup. You can try them locally via the CLI or integrate via a simple REST API.<p>Example (PDF extraction):<p><pre><code> bsubio submit -w pdf/extract \*.pdf
</code></pre>
Works like running the tool locally, but the compute and isolation happen in the cloud.<p>Technical details: - Each job runs in an isolated container with defined CPU/GPU/RAM limits. - Files are stored ephemerally for the duration of the job and deleted after completion. - REST API returns job status, logs, and results. - Cold start for light processors (Typst, Pandoc) is low; Whisper/FFmpeg take longer due to model load/encoding time. - Backend scales horizontally; more workers can be added during load spikes.<p>Current processors:<p><pre><code> SST/Whisper -- speech-to-text
Typography -- Typst, Pandoc
PDF extraction -- Docling
Video transcoding -- FFmpeg
</code></pre>
More coming; suggestions welcome for tools that are painful to set up locally.<p>Looking for testers! CLI is open source: <a href="https://github.com/bsubio/cli" rel="nofollow">https://github.com/bsubio/cli</a>. Installers available for Linux/macOS; Windows testing is in progress. Free during early testing; pricing TBD.<p>If you’re on Windows, feedback is especially helpful: contact@bsub.io<p>If you try it, I’d appreciate feedback on API design, latency, missing processors, or anything rough around the edges.