Command line reference for the worker agent

Run the worker

Go to https://metapage.io/settings/queues to get the terminal command to run a worker agent.

Use the command line argument --gpus=x to set the number of GPUs to x. This should not exceed the number of gpus on your machine.

These are the environment variables that can be used instead of cli arguments, also some advanced options are only available via env vars:

Env var	cli command example	Description
METAPAGE_IO_CPUS	—cpus 1	The max number CPU cores the worker will claim for jobs
METAPAGE_IO_GPUS	—gpus 1	The max number GPUs the worker will claim for jobs
METAPAGE_IO_MODE	—mode `local`	Default: `remote`. `remote` mode means the queue is in the cloud. `local` means all compute runs on the local machine, no scaling possible, but does not require any external API.
METAPAGE_IO_QUEUE		The queue the worker connects to for jobs to complete
METAPAGE_IO_PORT	—port 8000	Default: 8000. When in `--mode local`, this is the port the worker uses to connect.
METAPAGE_IO_API_ADDRESS	—api-address https://container.mtfm.io	Advanced or development use only. Allows connecting the worker to a different API queue server
(tbd)	--data-directory	Default: `/tmp/worker-metapage-io`. The base directory where jobs and the file cache are stored. Use your home directory if you want this persisted.
METAPAGE_IO_GENERATE_WORKER_ID		If set to `true` then the worker will always generate a new unique id on every start.
METAPAGE_IO_WORKER_RUN_STANDALONE		If set to true, will start a docker daemon process as a subprocess.
METAPAGE_IO_DEBUG	—debug	Default: `false`. Adds a lot of extra logging.

A deno command to automatically upgrade the worker to the latest version:

deno run -A --unstable jsr:@metapages/cli@0.2.0 worker upgrade