You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
ADR-0013 replaced CodeRabbit with the fleet's own Haiku reviewer. All 17 operating repos now carry a working claude-review.yml caller (step 2 of #276), so CodeRabbit's central-config repo has nothing left to configure. This PR retires lentago/coderabbit by the procedure in terraform/README.md, "Retiring a repo": archive the repo, then remove it from repos.json, required-checks.json and brand/fleet.json, and terraform state rm its resources right before the merge.
What changes:
fleet-ops/repos.json, fleet-ops/required-checks.json, brand/fleet.json: the coderabbit entries are removed. Only those blocks are cut, with no reformatting.
brand/generated/coderabbit/: the generated banner, OG card and readme block are deleted, since the repo leaves the brand map.
terraform/README.md: the per-repo resource count in the transfer note is corrected from twelve to fourteen, because ADR-0013 added two labels.
ADR-0013 Decision 5 also lists turning off auto-review in the central .coderabbit.yaml. That step is superseded: uninstalling the app stops all CodeRabbit activity at once, and the archived repo is read-only, so editing its config first would only add a PR to a repo that is closing.
This PR's plan fails until the state removal, by design
github_repository.fleet carries prevent_destroy. Until the coderabbit resources leave state, this PR's plan proposes destroying them and stops on that guard, so the gate check fails and the PR can't merge early. That's the interlock the README describes.
Steps for an org admin, in this order
Uninstall the CodeRabbit app. Org settings → GitHub Apps → CodeRabbit → Uninstall. This can be done any time.
Archive the repo:gh repo archive lentago/coderabbit --yes
Right before merging, from a shell with this repo's terraform backend credentials, remove its 14 resources from state (the repository, the merge gate, the ruleset, and 11 labels):
Review of PR #284 (Retire lentago/coderabbit, ADR-0013 step 3)
I could not run ci/validate.py or read issue #283 (both were blocked by the approval prompt), and I did not run a Terraform plan. The JSON edits were checked by reading the file tails, not with a parser. The findings below come from reading the diff and the surrounding config.
minor · terraform/README.md and the PR body, step 2: The body archives lentago/coderabbit now, but main still declares it in repos.json until this merges. Between the archive and the merge, a plan on main reads an archived repo that config still declares. If that plan proposes an update to the repo, the apply on merge of any other tf-touching PR fails. The retirement path has precedent (office-presence and others), so this is probably fine. Confirm that a plan after the archive shows no change for coderabbit, or move the archive after the merge.
minor · docs/adr/0013-haiku-reviews-every-pr-coderabbit-retires.md:117-118: Decision 5 still says auto-review goes off in the central .coderabbit.yaml. The PR body says that step is superseded, but the ADR records it as the decision and is not amended. A future reader of the ADR will see a step that was deliberately dropped.
minor · PR body, final line: The claude.ai/code/session_… link is in the body of a public repo. It is not a credential, but it points at a session transcript that the repo's readers do not need.
The repos.json, required-checks.json and brand/fleet.json entries are removed together, and the brand/generated/coderabbit/ directory is removed with its source.
terraform/ has no imports.tf entry or hard-coded ruleset id for coderabbit, so no import block is left dangling.
The terraform state rm list has 14 entries: the repository, the merge gate, the ruleset and 11 labels. This matches fleet-ops/labels.json (11 ensure: true labels, with model:* opt-in and empty for this repo) and the README's new count.
The required-checks.json removal leaves no required context without a reporting workflow.
The ADR-0013 link in the PR body points to a file that exists in the tree.
Archiving before merge leaves an archived repo declared on main
Accepted as designed, with the window kept short. This is the order terraform/README.md "Retiring a repo" prescribes (archive, then remove the rows and state rm in the same PR). The repo was archived 2026-10-29, the state removal and this merge follow immediately, and no other tf-touching .github PR is open, so no apply runs against the archived repo in between.
2
ADR-0013 Decision 5 still lists the .coderabbit.yaml step
Fixed (9bef47e). The step is struck through with a dated "superseded" note, and the status line records the amendment.
3
Session link in the body
Declined. It's the attribution footer the fleet's PR attribution rule requires.
Checked.#283 is "Retire lentago/coderabbit (step 3 of #276)" and asks for exactly this: the app uninstall, the archive, and removal from fleet config and terraform state.
The plan fails only on prevent_destroy ("Instance cannot be destroyed"), as expected, until the terraform state rm in the body runs.
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Closes #283. Part of #276, step 3.
Summary
ADR-0013 replaced CodeRabbit with the fleet's own Haiku reviewer. All 17 operating repos now carry a working
claude-review.ymlcaller (step 2 of #276), so CodeRabbit's central-config repo has nothing left to configure. This PR retireslentago/coderabbitby the procedure interraform/README.md, "Retiring a repo": archive the repo, then remove it fromrepos.json,required-checks.jsonandbrand/fleet.json, andterraform state rmits resources right before the merge.What changes:
fleet-ops/repos.json,fleet-ops/required-checks.json,brand/fleet.json: thecoderabbitentries are removed. Only those blocks are cut, with no reformatting.brand/generated/coderabbit/: the generated banner, OG card and readme block are deleted, since the repo leaves the brand map.terraform/README.md: the per-repo resource count in the transfer note is corrected from twelve to fourteen, because ADR-0013 added two labels.python3 -I ci/validate.pypasses (configs, fleet, brand, census, register).ADR-0013 Decision 5 also lists turning off auto-review in the central
.coderabbit.yaml. That step is superseded: uninstalling the app stops all CodeRabbit activity at once, and the archived repo is read-only, so editing its config first would only add a PR to a repo that is closing.This PR's plan fails until the state removal, by design
github_repository.fleetcarriesprevent_destroy. Until the coderabbit resources leave state, this PR's plan proposes destroying them and stops on that guard, so thegatecheck fails and the PR can't merge early. That's the interlock the README describes.Steps for an org admin, in this order
gh repo archive lentago/coderabbit --yesterraformworkflow. The plan must show no changes. Then merge, and the apply is a no-op.🤖 Generated with Claude Code
https://claude.ai/code/session_01SjEL1KkpsJ63gCUyCqLWiv