Skip to content

Reconcile durable worker lifecycles - #175

Draft
akgohain wants to merge 1 commit into
agent/durable-inference-commandfrom
agent/durable-worker-lifecycle
Draft

Reconcile durable worker lifecycles#175
akgohain wants to merge 1 commit into
agent/durable-inference-commandfrom
agent/durable-worker-lifecycle

Conversation

@akgohain

@akgohain akgohain commented Aug 4, 2026

Copy link
Copy Markdown
Collaborator

Summary

Makes durable model operations represent the worker lifecycle rather than only submission acceptance.

  • Accepted training and inference commands remain running.
  • Adds an authenticated, generic reconciliation endpoint for correlated PyTC runtime snapshots.
  • Strict workflow/command/run correlation prevents process-global worker state from completing the wrong operation.
  • Terminal snapshots idempotently produce succeeded, failed, or cancelled operation state and matching workflow/model-run evidence.
  • Worker snapshots retain canonical correlators, exit state, errors, and discovered artifacts for both training and inference.

Scope boundary

Reconciliation is an explicit backend primitive in this layer. The next stacked layer will own cancellation transport, polling/supervision, and restart recovery.

Verification

  • Full backend suite: 340 passed, 9 subtests passed
  • Focused runtime/operation/artifact/worker suites: 70 passed, 9 subtests passed
  • Black and git diff --check passed.

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant