Skip to content

Barista documentation

Barista runs long-lived workloads whose process state can rest without keeping a sandbox active. Start with the current guides below; planned behavior is labelled where it appears.

Use Barista today

  • Getting started — run a memory-capable node, create an instance, pause it, and verify that it did not reboot.
  • CLI reference — every implemented command, flag, and exit code.
  • Examples — direct instances, fleet desired state, scheduled wake, retained snapshots, and scripting.
  • Local developmentfake versus hypeman, guest binary builds, tests, and macOS limitations.
  • Best practices — image, hook, caller, and fleet guidance.

Concepts

Concept pages describe implemented behavior by default. Sections headed Planned explain product direction rather than an available interface.

Reference

The protobuf packages barista.node.v1alpha1 and barista.guest.v1alpha1 remain the wire-contract source of truth.

Platform

Product, decisions, and specifications

These are governance and design inputs, not a getting-started path:

Ratified capability requirements live in the repository's OpenSpec specifications.

Evidence and upstream work

Performance evidence consolidated for users is in Limits and performance; Git history retains superseded measurement narratives.