Jobs
Run-to-completion workloads with argv, stdin, input files, collected outputs and streamed logs.
postCancel Job
Deletes the job's pod and marks the job CANCELLED. The current state is not checked, so an already-finished job is re-marked CANCELLED too. **Access:** operator or admin role in the namespace (via Xeonr team role or a namespace API key), or a platform admin.
postGet Job
Returns a job record, including its persisted spec and, once terminal, its result (captured stdout/stderr are capped at 256 KiB each). **Access:** any role in the namespace (viewer, operator or admin — via Xeonr team membership or a namespace API key), or a platform admin.
postList Jobs
Lists jobs newest first, optionally filtered by `environment_id` and `state` (UNSPECIFIED = any). `page_size` 0 means 100 and is clamped to 1000; `page_token` is the opaque offset from `next_page_token`. An empty `namespace` lists every namespace and is refused (InvalidArgument) for anyone but a platform admin. **Access:** any role in the namespace (viewer, operator or admin — via Xeonr team membership or a namespace API key), or a platform admin.
postRun Job
Runs a batch job to completion and streams its progress: an `accepted` event carrying the job id first, then state changes, log chunks and a final `result`. `environment_id` must name an environment in the same namespace. Empty `argv` falls back to the environment's default argv, then the image entrypoint. `timeout_sec` 0 means 300. A `pool_lane` with a configured pool and a pool-compatible shape (argv set; no workspace, inputs, outputs, adapter or secrets) is served from a warm pod; otherwise the job cold-starts. A repeated `idempotency_key` within the namespace re-attaches to the original job instead of starting a new one. `secrets` require an `adapter`; `inputs`/`outputs` require a writable mount (one is synthesised at /work when no mounts are declared); at most 8 `workspaces`; env names in the CP_ prefix or set by the platform are rejected. The job keeps running if the stream is dropped. **Access:** operator or admin role in the namespace (via Xeonr team role or a namespace API key), or a platform admin.
postTail Logs
Streams a job's logs. For a running job this follows the pod's output live (stdout/stderr split where the job was supervised); for a terminal job it returns the captured stdout stored on the record, starting at byte `from_offset`. Returns Unimplemented for a live job when the api has no cluster access. **Access:** any role in the namespace (viewer, operator or admin — via Xeonr team membership or a namespace API key), or a platform admin.
Write Container File POST
Client-streaming write into a running workload's container: the first message must be the header (InvalidArgument otherwise), later messages are data. Overwrites by default (mode 0 means 0644); APPEND and PATCH_AT_OFFSET are also supported. The whole write is capped at 64 MiB. The rootfs is read-only unless the environment sets writable_rootfs, so writes usually land only in mounted volumes. The target must be a RUNNING sandbox or job (FailedPrecondition otherwise); errno is mapped to a status (ENOENT NotFound, EEXIST AlreadyExists, EACCES/EPERM PermissionDenied, EROFS FailedPrecondition, ENOSPC ResourceExhausted). /.cp is hidden. Unavailable when the in-container handler is not running; Unimplemented without cluster access. **Access:** namespace OPERATOR or above (write) on the target's namespace. Workspace-scoped credentials (a workload's own agent key) are refused.
Cancel Job POST
Deletes the job's pod and marks the job CANCELLED. The current state is not checked, so an already-finished job is re-marked CANCELLED too. **Access:** operator or admin role in the namespace (via Xeonr team role or a namespace API key), or a platform admin.