Download Latest Version flow-next v8.4.0 source code.zip (14.1 MB) Google Add to Preferred Sources
Home / flow-next-v8.3.0
Name Modified Size InfoDownloads / Week
Parent folder
flow-next v8.3.0 source code.tar.gz < 13 hours ago 11.4 MB
flow-next v8.3.0 source code.zip < 13 hours ago 14.1 MB
README.md < 13 hours ago 1.8 kB
Totals: 3 Items   25.5 MB 0

Teams on a local model or a rate-limited provider can now get the full three-reviewer review: one setting runs the reviewers one after another instead of all at once, still merged into one fix pass. make-pr also fits enterprise repositories better. It works when gh is logged in only to GitHub Enterprise, and its spec-close commit follows the repository's own commit-message rules, so an unattended run no longer stops on a hook.

Added

  • Run the review panel one reviewer at a time. Set review.fanoutExecution to sequential and the first review round runs its reviewers one after another instead of all at once, which suits local models and rate-limited providers. Running them together stays the default, and the round still counts once against the cap. On Claude Code, raise BASH_MAX_TIMEOUT_MS so the review call can wait for all of them. Thanks to @jcarosi for the contribution (#524).

Fixed

  • make-pr works when gh is logged in only to GitHub Enterprise. The preflight still checks github.com first and now accepts a login to any host gh knows, so the pull request opens on the Enterprise host your origin points at. A GHE-only machine used to stop at the preflight with "gh CLI not authenticated". Thanks to @georgkeller for the report (#532).
  • The spec-close commit follows your repo's commit rules. When your project instructions state commit-message rules, the agent writes make-pr's close commit to match, for example chore(flow): close fn-4 AB#1234. The fixed chore(flow): close <spec-id> message broke subject length limits and required work-item references, so hooks or reviewers refused it and flow --auto could not finish. Without rules, or when a value the rules need is unknown, the message stays the same. Thanks to @georgkeller for the report (#534).
Source: README.md, updated 2026-10-09