Skip to content

Releases

0.68.0

Choose a tag to compare

@timmo001 timmo001 released this 16 Nov 14:26
6e2f15a

Server

  • Failures of a job that's no longer in the catalogue, such as one that's been renamed, are settled when the server starts: they leave the Failures list as cancelled, along with their heals and anything they asked of you. Queued runs of such a job are dropped, since no worker would take them.
  • Nothing new is queued for a job that's no longer in the catalogue, such as a heal retrying one of its old failures.

Full Changelog: 0.67.0...0.68.0

0.67.0

Choose a tag to compare

@timmo001 timmo001 released this 16 Nov 14:05
0d6e48b

Server

  • Retrying a run of a job that's no longer in the catalogue, such as one that's been renamed, is refused. Before, the retry was queued for a job no worker runs, so it sat in the queue for good and its failure stayed at retrying.
  • Dropping a queued retry puts the failure it was retrying back to failed, rather than leaving it at retrying with a link to a run that's gone.

Jobs

  • Dependencies plans its updates without checking out each repository's submodules.

Internals

  • Updated docs-kit and the groundwork packages to 0.50.0.

Full Changelog: 0.66.0...0.67.0

0.66.0

Choose a tag to compare

@timmo001 timmo001 released this 16 Nov 12:46
da5525a

Server

  • Agent reviews from a run that handed work off now leave the Failures list once they're done. A heal that pushed a fix or found nothing to fix settles itself, as does one asking you once you've dismissed it. One that broke or is still asking is settled by a later heal of the same repository, where a drift review covers drift.
  • A failure retried by a run that handed off is settled once every heal of that handoff is done.
  • A retry that doesn't pass now sends back every failure it was retrying, not only the first, so none is left stuck at retrying.
  • A release job's run with no release, which could happen before that was refused, is cancelled along with its heals, since it never could have run.

Dashboard

  • The Failures and default runs lists leave out commands that failed, as Still open already did. They were seen as they happened and are never healed.
  • A heal that settled itself no longer links to itself as what settled it.

Jobs

  • A run that couldn't reach a server, such as Upkeep's while it restarts for an upgrade, is queued again once, as a run the worker stopped is, rather than failing a heal that nothing then heals.
  • Dependencies shares its lookups across a worker's runs, and each run reads only its own repositories.

Full Changelog: 0.65.0...0.66.0

0.65.0

Choose a tag to compare

@timmo001 timmo001 released this 16 Nov 10:50
337e85e

Dashboard

  • The Jobs page shows how a job's latest batch is getting on under its last run, such as "Batch 10:21: 33 of 40 done, 1 didn't pass". deps runs as a batch of runs, one per repository, since 0.64.0 (#103).
  • A job's page has a Latest batch note: when it was queued, how many of its runs are done, running, waiting or didn't pass, and when the last one ended.
  • On the Runs page, each run that's one of a batch says "in a batch" under its repository, linking to its job's page.

Server

  • The jobs list sends each job that fans out with its latest batch's progress, and says it fans out.

Full Changelog: 0.64.0...0.65.0

0.64.0

Choose a tag to compare

@timmo001 timmo001 released this 16 Nov 10:16
4f3a537

Jobs

  • deps no longer runs as one long job. A scheduled or manual run for every repository is queued as a batch: one short run per repository tracked for deps. Each takes a few minutes, a run that's interrupted, fails or is retried only covers its own repository, and a waiting event run for a repository joins the batch rather than running twice (#103).
  • deps watches CI once per repository, on its last push, instead of after every group. Each group is still checked and pushed on its own, and Renovate's pull requests are closed once that CI passes. A CI failure names every group in the batch. Watching CI took most of a long run's time.
  • A regression decision is reused for a week: the same model's answer on the same update, from any run on any worker, judged against the current threshold. Repositories moving the same dependency no longer each ask the model and use up its daily limit.
  • deps keeps the GitHub responses other runs read, dropping only those no run has read for a week, so one run per repository doesn't throw away the others' free 304 reads.

Server

  • The lane takes a batch's runs after its other runs, so bump, sync and drift no longer wait behind deps.
  • A batch's failures wait for the whole batch to end, then share one heal over every repository that failed, reading each one's error and log, as one run of the whole job did. Its retry runs each of those repositories again. This also happens when the last run of the batch is lost or cancelled.
  • New /api/work/regression-decisions/latest endpoint, for a worker to find an earlier decision on an update.

Internals

  • The CLI and worker keep the original error as the cause when narrowing one.
  • Updated @timmo001/oxlint-rules to 0.7.1.

Full Changelog: 0.63.1...0.64.0

0.63.1

Choose a tag to compare

@timmo001 timmo001 released this 15 Nov 19:17
442773d

Dashboard

  • Tables that fill the window, such as Runs, Jobs and Agents, now leave the same gap under them as at the sides, and every page's bottom gap matches its sides.

Jobs

  • A worker that's draining waits up to 2 hours for its runs to finish.

Full Changelog: 0.63.0...0.63.1

0.63.0

Choose a tag to compare

@timmo001 timmo001 released this 15 Nov 19:10
56d9f77

Jobs

  • Heal and change agents now read the standing notes kept for the repositories they work on (and Upkeep's own notes when working on Upkeep), and are told how to keep new ones. A line of Note for <owner/name>: <the note> in what an agent says last keeps a note marked as that run's, and Note out of date: <id>: <why> flags one for you to keep or drop.

Dashboard

  • Each repository's page has a Notes for agents card, the Jobs page has one for Upkeep's notes, and the Attention page's Agent rules are now notes too. Add a note with an optional expiry date, edit or delete it, and see whether you or an agent wrote it, with a link to the agent's run.
  • A note an agent flags as out of date shows with its reason and Keep it or Delete it, and counts under what needs you.
  • The Runs, Jobs and Agents tables reach down to the bottom of the window instead of stopping at half the screen.

Server

  • One store for notes, each scoped to a repository, upkeep or attention. The attention agent's rules move into it, and the old rule endpoints are gone in favour of /api/notes.
  • Flagged notes count toward the warning health, and the status summary says how many notes to check.
  • Workers read unexpired notes from /api/work/notes and add or flag them for the run they hold.

CLI

  • upkeep notes list|add|edit|delete|stale|keep, with --expires YYYY-MM-DD or never.
  • MCP and Code Mode gain notes, note_add, note_edit, note_delete, note_stale and note_keep.

Full Changelog: 0.62.2...0.63.0

0.62.2

Choose a tag to compare

@timmo001 timmo001 released this 15 Nov 18:49
c16dc74

Dashboard

  • Every table takes about four rows' height even when it has fewer, with the space left blank under the last row, so a short table doesn't jump as rows come and go. With more rows it grows up to its usual height and then scrolls.

Internals

  • Groundwork updated to 0.49.0.

Full Changelog: 0.62.1...0.62.2

0.62.1

Choose a tag to compare

@timmo001 timmo001 released this 15 Nov 18:43
95983bc

Dashboard

  • A table now scrolls once it's half the screen tall, rather than 70%, so a long one such as Runs leaves room for the rest of the page. It always shows about four rows before it scrolls, even on a short screen. Tables with a height of their own, such as those on a run's page, keep it.

Full Changelog: 0.62.0...0.62.1

0.62.0

Choose a tag to compare

@timmo001 timmo001 released this 15 Nov 18:29
e8d310d

Dashboard

  • Runs have their own page, Runs, beside Jobs in the menu. It has the latest runs of every job with all their filters and presets. A run's page now highlights Runs rather than Jobs. The Jobs page keeps the jobs and each worker's agents.

Server

  • A repository's drift heal that's waiting on you settles once a later drift heal for the same repository finds nothing to fix or pushes a fix. A drift review looks at all of a repository's drift as it stands, so the older question is done or out of date. It leaves Needs you and its Inbox entry goes. Ones already stuck that way are settled on upgrade.

Jobs

  • The worker reports its agents' schedules even when OpenCode can't read one session's, such as a heal session whose worktree has been cleaned up. Before, one such session hid every schedule from the run page and upkeep schedules list.
  • An agent woken by "Ask me in the Inbox" first fetches and checks the default branch and reads the Inbox for a thread on the same work. If it's already done, it says so and dismisses the run instead of asking again.

Internals

  • Shared workflows updated to timmo001/workflows 20261115.0.

Full Changelog: 0.61.0...0.62.0