Skip to content

v0.71.0

Choose a tag to compare

@hhff hhff released this 20 Nov 06:59
408ef05

What

  • The box's gateway signs people in through a table of sign-in methods. One person can have several methods (a passkey, and later others). Each sign-in runs through start and finish routes tied to a short-lived __Host- flow cookie, so a finish from another browser, a replay, or a flow started on the old gateway is refused. Invites are stored as a hash and checked for expiry. Existing sessions and passkeys keep working (kit-254 #445).
  • The machine's hub records memory use every minute and serves the last day at /memory.json (kit-591 #444).
  • Shift+clicking something that already has an open thread opens that thread, and the kit refuses to start a second thread on the same subject (kit-97 #442).
  • Side peeks list their artifacts. A goal's, feature's, ticket's and thread's peek links the artifacts that belong to it (kit-320 #443).
  • Upkeep's peek lists its recurring jobs, built-ins and proposed jobs, and each job's peek shows its recent runs once the data exists (kit-166 #446).
  • Touches can name parts of any file, not only files listed in split_files (#423).
  • A new priority-inversion finding: the metrics flag a higher-priority goal's ticket waiting on a lower-priority one, and D4 orders dependencies by goal priority (#420).

Why

  • Approved goals: org access with several sign-in methods (kit-254), box memory monitoring (kit-591), and a cleaner, more consistent dashboard (kit-97, kit-320, kit-166).
  • Scheduling: finer touches let more tickets run in parallel (#423), and the scheduler weighs priority (#420).

Migration

This is a minor, and the box redeploys, because the gateway and the hub's code changed.

After adopting:

  • A sign-in that was half-finished when the gateway restarted must start again. Signed-in sessions are unaffected.
  • Invites written by the old gateway are still accepted.
  • Starting a second thread on a subject that already has an open one is refused. Scripts that did that should reply in the existing thread instead.
  • validate may report new priority-inversion findings on existing graphs.