Helper v1.0.11

Two fixes that affected any real job:

• The node's heartbeat stopped for the entire duration of a running job, because job execution blocked the poll loop. The scheduler drops a node after 90 seconds without one and the stuck-job reaper fails its running jobs after five minutes, so any job longer than that had its own node declared offline and the work reaped out from under it. Liveness now continues throughout a job.
• Workload images are pulled before the job's runtime budget starts. These images are several GB; previously a first job could spend its entire paid allowance downloading and then be killed at the cap having produced nothing.

Helper v1.0.10

Fixes a job crash on Windows and macOS Docker Desktop hosts: containers running under the unmapped fallback UID (65532) had no /etc/passwd entry, so any Python library calling getpass.getuser() deep in an import chain (observed via torch/diffusers/transformers) crashed the job before it did any work. USER/LOGNAME are now set explicitly so that lookup never has to happen.

Helper v1.0.9

One-time --start/--resume enrollment, --autostart persistence across reboots, and inline Docker/GPU-container readiness checks — the resident agent no longer requires a provider to see or paste a long-lived secret, or to keep a terminal window open.

Helper v1.0.8

Just open the app: double-clicking with no flags now prompts once for whatever code the website shows, and figures out on its own whether it is a hardware-detection code or an activation code -- no terminal, no --start flag needed for the common path. Advanced/manual flags (--start, --resume, --autostart, --code) are unchanged.

Helper v1.0.7

--start now actually runs in the background: on Windows it launches a hidden, detached copy immediately (not just at next login) and tracks its PID so --autostart off can stop it right away; on Linux it relies on systemd's --now to do the same. Closing the terminal after 'Connected' no longer pauses job-receiving until the next reboot.

Helper v1.0.6

Adds --start (redeem a one-time setup code and save your credential locally), --resume (reuse the saved setup with no arguments), and --autostart on/off (per-user login autostart on Windows/Linux, no admin rights required). The long-lived agent token no longer needs to be copied by hand.

Helper v1.0.5

Fixes prerequisite checks appearing to hang: each check (NVIDIA driver, Docker, GPU containers) now prints as soon as it finishes instead of after all three, and every check now has a timeout so an unresponsive Docker daemon or a stalled image pull can't block the others indefinitely.

Helper v1.0.4

Adds local provider prerequisite checks (NVIDIA driver, Docker, GPU container access) reported alongside hardware detection.

Helper v1.0.3

Report VRAM and RAM to one decimal place instead of full float precision.

Helper v1.0.2

Pairing codes can now carry the server that issued them (CODE@host), so the paste-the-code flow works outside production without --api-base. Connection failures now report the actual address and cause instead of a generic message.