v0.12.50: @newsroom.writer: sgit in GitHub Actions side by side, and agent workflows on ephemeral vaults

Two new articles written here from two voice memos, each with a brief.

sgit in GitHub Actions, side by side: a dedicated showcase repository and site that moves the same build through four lanes in parallel on every run (a vault server, the vault as an artefact, the vault as published files, append lanes), reporting time, bytes, keys and a screenshot per lane on growing vaults, with a token budget for LLM steps across providers.

Agent workflows on ephemeral vaults: seven architectures from one agent to a swarm, four questions each must answer (memory, communication, asynchrony, visibility), vault apps to watch them, and a repository of recipes that runs as a gate before sgit's main branch.

Both use measurements run for them locally (sgit 0.21.0, vault server 0.34.0, four cores): a one-folder clone took 0.59 s at 100 files and 0.77 s at 2,000, a full clone 0.95 s to 5.55 s; a hundred append-lane writers landed 1,000 entries in 4.9 s with in-memory storage, none lost; ten scoped clones at once took 1.9 s and ten full clones 21.1 s. The runs found four defects: disk storage scanning the whole store on every append, files over 4 MB needing cloud storage, a headless clone failing to fetch a file, and full clones queuing on one server process. Four figures.