Hi everybody! I built Scrollcase after repeatedly running into the same problem with AI and scientific software: getting an environment working on my machine was manageable, but giving that exact environment to someone else was often a completely different problem. Scrollcase turns the environment itself into an artifact. You describe what should be inside, lock the dependencies, and build a target-specific box containing the runtime, dependencies, native libraries, code and optional assets such as model weights. A box can use Python, Node, or a native compiled binary. The receiving machine verifies it and runs it without needing Python, Node, pip, npm, a compiler or Docker installed. I also wanted the artifact to be something you could actually reason about and trust, so boxes are signed, dependencies are locked, assets can be hash-checked, builds are deterministic, and declared tests can run before an artifact is produced. Scrollcase deliberately stops at building the artifact. It doesn't require a registry or hosting service, and it doesn't decide how you deploy or update your software. You can distribute boxes through GitHub Releases, object storage, your own backend, a desktop app, or whatever already fits your infrastructure. It's fully open source. I'd especially love feedback from people distributing AI/ML, scientific, desktop or native software where asking users to recreate the development environment isn't a realistic option. GitHub: https://github.com/suffro/scroll... Docs & demos: https://scrollcase.dev/

Scrollcase

Self-contained signed boxes for Python, Node and native envs

Self-contained signed boxes for Python, Node and native envs