Skip to content

Release 2.1.19 (2026-06-08)

Bug fixes

  • defer spec-drift check to post-start so stopped VMs can start (#1536) vrg-vm start could not start any stopped dedicated VM: it aborted with a spurious "no longer meets spec / Rebuild it" error while vrg-vm list reported SPEC: ok for the same VM.

The spec fingerprint is stamped inside the guest (/etc/vergil/vm-spec.fingerprint) and is only readable over limactl shell while the VM runs. _preflight_target ran the drift check as a pre-start gate, so against a stopped VM the shell failed, the fingerprint read as None, and vm_spec_status returned needs-rebuild — making every stopped dedicated VM un-startable and recommending a destructive rebuild of a healthy VM. vrg-vm list was unaffected because it only evaluates drift for running VMs.

Documentation

  • capture paywalled security cost as a migration driver (#1531) Record GitHub Advanced Security pricing (Secret Protection $19, Code Security $30 per active committer/month) as a strategic driver for the forge migration: negligible for a funded individual but a real barrier to entry for the small open-source teams the project serves. Add the open research question of whether Forgejo/Codeberg offer equivalent security at all, the near-term posture (stay private, park security, go public when ready), and a billing-docs footnote. Ref #1530.

  • add Vergil GUI vision & feasibility study (#1532) (#1537) North-star vision and feasibility assessment for a Tauri-based desktop GUI that thin-orchestrates the existing vrg-* CLIs, limactl/Lima sandboxes, and the #1521 forge abstraction. Captures the window-management motivation, per-component feasibility ratings, and a value-first v1/vNext/Later roadmap. Strategy artifact, not an implementation plan.