Story

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.

https://www.bsub.io

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.

Example (PDF extraction):

  bsubio submit -w pdf/extract *.pdf

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.

Current processors:

  SST/Whisper -- speech-to-text

  Typography -- Typst, Pandoc

  PDF extraction -- Docling

  Video transcoding -- FFmpeg

Looking for testers! CLI is open source: https://github.com/bsubio/cli. Installers available for Linux/macOS; Windows testing is in progress. Free during early testing; pricing TBD.

If you’re on Windows, feedback is especially helpful: contact@bsub.io

If you try it, I’d appreciate feedback on API design, latency, missing processors, or anything rough around the edges.