CLI Reference
The chia console script (chia/cli/main.py) manages clusters, visualizes
flows, and proxies Ray’s CLI (with optional per-cluster targeting).
chia <command> [options]
Cluster management
chia up <config.yaml>
Bring up a Chia/Ray cluster from a YAML config.
Flag |
Meaning |
|---|---|
|
Skip the confirmation prompt. |
|
DEBUG logging. |
|
Print the plan and generated scripts without executing. |
|
Only add new nodes to an existing cluster (skip ones already up). |
chia down <config.yaml>
Tear down the cluster. Flags: -y/--yes, -v/--verbose.
Visualization
chia viz <source.py>
Render the @ChiaFunction call graph of a flow statically from source.
Flag |
Meaning |
|---|---|
|
Orchestrator function (auto-detected if omitted). |
|
Output format (default |
|
Output directory (default: alongside the source). |
chia viz-profile <log...>
Visualize a ChiaProfileCollector JSONL log. See the
Profiling guide for how to record one.
Flag |
Meaning |
|---|---|
|
Output format (default |
|
Output dir (ignored for |
|
( |
|
( |
|
Run index to visualize (default: last). |
|
Gap that segments runs (default 600). |
|
Horizontal inches per second of wall-clock (default 0.5). |
Cluster Status
chia status [...]
Print cluster status. Proxies to ray status.
Flag |
Meaning |
|---|---|
|
Specify the Chia cluster YAML file to target. |
|
all other args (including |
chia list [...]
List all states of a given resource. Proxies to ray list (e.g. chia list nodes).
Flag |
Meaning |
|---|---|
|
Specify the Chia cluster YAML file to target. |
|
all other args (including |
Job Management
chia job <subcommand> [...]
Submit, stop, delete, or list Ray jobs.
Proxies to ray job — submit, status, logs, list, delete,
etc. all forward, so chia job --help shows Ray’s own job help. Supports
--chia-cluster. The one exception is stop, which
chia overrides with an augmented implementation; use chia ray job stop for
Ray’s unmodified behavior.
Flag |
Meaning |
|---|---|
|
Specify the Chia cluster YAML file to target. |
|
Stop a job by its ID. |
|
all other args (including |
chia job stop <job_id>
By default a thin wrapper over ray job stop. With --kill-tracked-pids,
it first asks the PID registry actor to kill every tracked subprocess (SIGTERM,
then SIGKILL for any that linger past the grace period), then stops the job. The
kill returns as soon as the processes have actually exited, so it normally
completes in well under the grace period.
Flag |
Meaning |
|---|---|
|
Kill tracked subprocesses (via the PID registry) before stopping the job. Off by default. |
|
Seconds to wait for each tracked subprocess to exit after SIGTERM before
escalating to SIGKILL (only used with |
Fallback to Ray (chia ray ...)
chia ray <anything> forwards to ray <anything> verbatim, so every
current and future Ray command (ray memory, ray timeline, ray
attach …) is reachable — not just the promoted status/list/job.
Using an explicit gateway also means chia’s own up/down never shadow
ray up/ray down, and chia ray --help shows Ray’s complete command
list.
Supports --chia-cluster. Unlike the promoted
commands, the fallback applies no overrides — chia ray job stop runs
Ray’s stop, not chia’s augmented one:
chia ray memory --chia-cluster=cluster_a.yaml
chia ray timeline
chia ray job stop <job_id>
FireSim
FPGA build/run orchestration:
chia firesim-build <config.yaml> --recipe NAME [--instance-type ...]chia firesim-run <config.yaml> [--hw-config ... --workload ... | --suite ...] [...]chia firesim-upload-workload [config.yaml] --marshal-config ... --images-dir ... --suite-name ...chia firesim-cleanup <config.yaml>— terminate orphaned chia EC2 instances.
Run chia <command> --help for the full, current option list.