augray
augray
If the user's code has already set a SIGALRM handler set in the resolver process (perhaps due to a non-standalone function which registered one and didn't clean up), [this code](https://github.com/sematic-ai/sematic/blob/a8db5a5bbaf56bfe4a74fdd46847576e73dc6244/sematic/resolvers/state_machine_resolver.py#L229-L240)...
Currently the db stores this column as an actual string, which happens to contain serialized json. We should type it as serialized json to allow better interaction via the DB...
Once a user has specified search criteria for the runs page (ex: search for runs tagged with "example"), the url should be updated such that it is copy-pastable to share...
Very rarely, it is possible to end up in a state where a *run* exists, but a corresponding *resolution* does not. This is because the root run gets created before...
When viewing a ray cluster in k8s, it would be helpful to have an easy way to map that back to a resolution or run id
Currently to work with Sematic, Kuberay must be installed to a specific namespace within a cluster (same one as Sematic itself), and not installed in any others. Kuberay is coming...
This way all the relevant logs from the head/workers are still usable even after the ray cluster is terminated (assuming you have log ingestion in your cluster).
Currently, once you have executed a run, there is no way to remove it (even as an admin). This applies to pipelines as well (pipelines are technically just derived from...
When a retry happens where one run is the parent of other runs, the child runs should be stopped if they were not already in a terminal state. If the...