What to include
- Run id — from
compute runorcompute runs list. - Request id — from the API error body or
compute doctor --json. - What you ran (
file.py::fn,--gpu, whether you used--wait). - Approximate UTC time.
COMPUTE_API_KEY, config.toml, or Checkout session ids.
Where to send it
There is no public issue tracker and nostatus.compute.cx page yet. The dashboard is not a ticketing UI.
Privacy, export, and deletion: privacy@theoric.com (same address as Privacy). Include your Compute user id (usr_…) when you have one.
For a failed or stuck run, keep the run id and request id until there is a published operator contact. Terms still tell you to include those ids.
Self-check before you write
machines is empty, teardown did its job even if the function failed. Send the run id anyway if the receipt looks wrong.
