Skip to content

Releases

v1.2.0

Choose a tag to compare

@sakuga-agent sakuga-agent released this 26 Oct 10:17
55a3d4d
  • A pull request that a sub-agent of a Claude Code session opened or changed shows that sub-agent after the name of the session, with its task. The spinner follows the sub-agent, and ✖ marks a sub-agent that failed. --json gives agent.subAgent. (#49, #59)
  • --repo <repo> (repeatable) and the repos setting watch only some repositories: owner/name, or name in the organization. (#60)
  • A workflow run of the head commit that is queued, or between two jobs, counts as a running check: a pull request no longer shows ready to merge before its CI starts. (#63)
  • A stacked pull request that GitHub moved to a new base after its parent merged shows parent merged · rebase until a new commit. --json gives stack.restack, and --events gives a restack event. (#65)
  • Merge policy:
    • answeredReviewRequests holds ready to merge while a review asked of a bot (@sakuga-review review balanced) has no answer. (#66)
    • minScore.deepest reads the deepest review level, so a light review after a deep one no longer passes the rule. (#67)
    • resolvedBotThreads: { "goneMarker": "…" } counts a thread as resolved if its last comment is a bot comment that says the finding is gone. (#64)
  • --watch --events:
    • The left event of a merged pull request says who merged it (mergedBy, mergedAt), and so does --json. (#61)
    • A comment event comes for each new comment, review or line comment of a person. commentEvents in the settings ignores some logins and the replies that carry a marker. (#68)
    • A pull request goes to error only after two failed reads in a row, and the event gives the message of GitHub. (#62)

v1.1.0

Choose a tag to compare

@Mheaus Mheaus released this 25 Oct 12:48
ef81c0e
  • boo --json prints the state of each pull request for a script: the state, the checks, each reviewer with its score, the stack, the agent and the notes. The keys stay the same from one version to the next. (#40)
  • boo --watch --events streams each change as one line of NDJSON: a state, a review, a check, a pull request that joins or leaves the list. (#42)
  • A red check tells its cause: the check fails here only, it also fails on the base branch, or its run was cancelled. (#41)
  • A red pull request that no session works on, with no commit for 30 minutes, is stalled (zZ on its row). The setting stalledAfterMinutes changes the time. (#43)
  • Two runs of boo share their reads of GitHub for 30 seconds, and only one of them posts a retry. With less than 20 % of the GraphQL points left, boo reads every 5 minutes. (#44)
  • A merge policy (mergePolicy in the settings) adds rules before ready to merge: a minimum score, a proof in the description, resolved bot threads, files that only their owner merges. m refuses a pull request that misses a rule. (#45)
  • The footer counts the pull requests that GitHub did not answer. (#47)

v1.0.1

Choose a tag to compare

@sakuga-agent sakuga-agent released this 21 Oct 11:28
be7f77e

The first release by the publish workflow: npm install --global @sakuga-software/boo.

  • Each pull request takes one line, with the coding agent that writes it (✻ Claude Code in its orange, ◇ the others) and the Claude Code session on this machine that works on it.
  • e or → unfolds a pull request: the size of the change, the agent, each reviewer with its state and its confidence score, the checks and the notes.
  • a sends rebase, review, fix or test to the local session of the pull request; w opens its cloud session online.
  • Enter opens the pull request, b the review bots.

The code is the same as 1.0.0, which was published by hand to create the package. This version has a provenance.