CLI

Strøm provides two CLI binaries:

• stroem — Local workspace tool for exploring and running tasks without a server
• stroem-api — Remote server client for triggering jobs and querying a running Strøm server

stroem (Local)

Local workspace commands. No server required.

# Set the workspace path (default: current directory)
stroem --path /path/to/workspace <command>

run

Run a task locally without a server. Loads the workspace, walks the task’s DAG, and executes each step via the local shell runner. Only type: script with runner: local (default) is supported — docker, pod, task, agent, and approval steps are rejected upfront.

# Run a task in the current directory
stroem run my-task

# Run from a specific workspace path
stroem run my-task --path /path/to/workspace

# Run with input
stroem run deploy --input '{"env": "staging"}'

# A file input field takes a local path
stroem run import-report --input '{"data_file": "./report.csv"}'

For a file input field the value in --input is a local path. The CLI checks that it is a regular file and copies it read-only to {temp}/stroem-run-inputs-{uuid}/{file_id}/{safe_name}, so the path in the script has the same shape as on a server. The directory is removed when the run ends. A required file field with no value fails with Input field '…': a file is required.

Supports:

• DAG dependency ordering
• Template rendering ({{ input.* }}, {{ step.output.* }}, {{ secret.* }})
• when conditions (skip steps based on expressions)
• for_each loops (iterate over arrays)
• The same dependency gate as the server: a step runs only when its whole depends_on tree is satisfied — each edge’s accept set (default: must complete) names which of the dependency’s outcomes let it through, independent of any flag on the dependent itself. An untolerated failure no longer aborts the whole run — independent branches keep going, and unreached dependents are skipped unreachable, matching server jobs
• A when or for_each evaluation error fails that step, same as the server, instead of aborting the run
• Masked step context for downstream templates: a completed step’s output, null for a skipped one, null plus error for a failed one — a failed loop’s partial array is never exposed to a step that reads it
• Ctrl+C graceful cancellation (still aborts the run and exits non-zero — there is no step-level cancellation locally)
• OUTPUT: {json} parsing for step outputs
• Exit code reflects the run’s outcome, not a raw failure count: 0 when every failed step’s own continue_on_failure catches it, 1 when at least one failed step escaped uncaught (a downstream step’s continue_on_failure/accept never excuses someone else’s failure). When every failure was caught, the run prints one extra line after the summary counts: Every failure was caught by continue_on_failure.

Limitations:

• Steps execute sequentially, even when the DAG allows parallelism
• for_each iterations always run sequentially regardless of the sequential setting
• Only type: script with local runner is supported — docker, pod, task, agent, and approval steps are rejected
• Summary counts are diagnostic only: a successful loop counts as one step in the tally, while a failed loop’s instances are each counted individually

validate

Validate workflow YAML files before deploying.

# Validate current directory
stroem validate

# Validate a specific path
stroem validate /path/to/workspace/.workflows/

The validator checks:

• YAML syntax and structure
• Action type validity
• Runner field validity
• Flow step and dependency references
• DAG cycle detection
• Trigger cron expression syntax
• Hook action references
• Dependency shape: a leftover continue_when_skipped flag, or a malformed depends_on tree (duplicate siblings, empty all/any groups, empty or duplicate-entry accept lists) — a named, actionable error, not a silent warning (see the 0.18 upgrade guide). stroem run and job creation on the server enforce the same check; loading a workspace on its own does not.
• ref: values (Git Refs): the ref’s syntax, and that it appears only where it is supported (flow-step actions, type: task actions, scheduler/webhook triggers). A reference with ref: is not resolved offline — it is skipped with a warning, like a library action

tasks

List all tasks in the workspace.

stroem tasks
stroem --path /path/to/workspace tasks

actions

List all actions in the workspace.

stroem actions

triggers

List all triggers (scheduler and webhook) in the workspace.

stroem triggers

inspect

Show detailed information about a single task: input schema, flow steps with dependencies, hooks.

stroem inspect deploy-pipeline

stroem-api (Remote)

Remote server commands. Requires a running Strøm server.

# Set the server URL (default: http://localhost:8080)
export STROEM_URL=http://localhost:8080

# Set authentication token (optional)
export STROEM_TOKEN=strm_your_api_key

trigger

Execute a task and create a new job.

# Trigger with default input
stroem-api trigger hello-world

# Trigger with input
stroem-api trigger hello-world --input '{"name": "CLI"}'

# Trigger in a specific workspace
stroem-api trigger etl-pipeline --workspace data-team --input '{"date": "2025-01-01"}'

# Give a file to a file input field (repeatable)
stroem-api trigger import-report --file data_file=./report.csv

--file FIELD=PATH uploads the file (streamed), then executes the task with the file ID as the value of FIELD. A key that appears in both --input and --file is an error. See File Inputs.

status

Check the status of a job.

stroem-api status <job-id>

logs

View the logs of a job. Prints the last 256 KiB of the log (whole lines) by default and, when the log is longer, a note on stderr:

note: showing the last 256.0 KiB of up to 83.3 MiB; use --full for the whole log

# Tail (default: last 256 KiB)
stroem-api logs <job-id>

# Stream the whole log
stroem-api logs <job-id> --full

# Ask for a different tail size (max log_storage.read.tail_max_bytes, 4 MiB by default)
stroem-api logs <job-id> --tail-bytes 1048576

--full and --tail-bytes are mutually exclusive.

Against a server older than this release, --full receives the old full-log JSON body (no streaming) and holds it in memory before printing, instead of streaming NDJSON.

jobs

List recent jobs.

# List last 10 jobs
stroem-api jobs --limit 10

tasks

List tasks from the server (across workspaces).

# List all tasks
stroem-api tasks

# Filter by workspace
stroem-api tasks --workspace data-team

cancel

Cancel a running or pending job.

stroem-api cancel <job-id>

workspaces

List all configured workspaces on the server: name, task / action / trigger counts ((off) when the server config disables the workspace’s triggers), and the display name and description from the server config (- when unset).

stroem-api workspaces