AutoCoder CodeCartography chat tasks were still failing with:
Detached worker task ... has no issue scope and repo_path is not a Git checkout: /workspace/repos/BodyLog/Bodylog.dev
The backend deployment had rolled to gitea.app.lan/autocoder/autocoder-engine:sha-29fefa35af55, but AUTOCODER_WORKER_IMAGE still pointed to gitea.app.lan/autocoder/autocoder-engine:sha-f2bd937302e5. Detached worker pods were therefore still running the stale worker runtime.
Patched deploy/autocoder-engine in namespace autocoder-engine:
AUTOCODER_WORKER_IMAGE=gitea.app.lan/autocoder/autocoder-engine:sha-29fefa35af55
The engine rollout completed and reconciliation replaced stale detached worker pods. Replacement worker pods were confirmed running image sha-29fefa35af55.
AutoCoder-Engine main was updated so deployment classification treats nested src/worker_runtime/ changes as worker-runner changes and runs the backend rollout lane when worker image changes are required.
Commits:
03e2aaa Deploy nested worker runtime changes0e99730 Run backend rollout for worker image changesThe already failed chat task chat-1781531172761983797 remains terminal failed; it should be retried by queueing a new chat task.