Configuration¶

Environment variables¶

N4CLUSTER_SERVER: default coordinator URL for the client/worker (http://127.0.0.1:8765). Overridden by --server.
N4CLUSTER_TOKEN: default bearer token for the server, client and worker. Overridden by --token.

Precedence everywhere is flag > environment variable > built-in default.

Server settings (`ServerConfig`)¶

Set through n4cluster server flags (see CLI reference); the defaults below are the trusted-LAN posture.

Setting	Default	Meaning
`token`	`None`	bearer token; `None` ⇒ no auth (dev)
`allow_python_jobs`	`False`	permit `python_entrypoint` pipelines
`lease_ttl_s`	`60.0`	task lease TTL (renewed on heartbeat)
`worker_dead_after_s`	`45.0`	mark a silent worker dead after this
`max_artifact_mb`	`2048`	streaming cap for artifact uploads
`max_request_mb`	`16`	cap for JSON request bodies (uploads exempt)
`cors_origins`	`[]`	allowed browser origins (opt-in)

A request whose JSON body exceeds max_request_mb is rejected with 413; multipart artifact uploads keep their own streaming max_artifact_mb limit.

State directory layout¶

<state_dir>/
├── store.sqlite        # queue + metadata (WAL mode)
└── objects/aa/bb/<sha256>   # content-addressed blob store

Logging¶

The server and worker use Python logging. Choose verbosity with --log-level (debug/info/warning/…) and optionally tee to a file with --log-file. CLI command output (status, logs, jobs, …) is printed directly and is independent of logging.