rebase run

Run a Rebase function, workflow, or model from local source without deploying it.

Synopsis

rebase run <file.py>::<target>
rebase run <file.py>::<target> --mode job --param a=1 --param b=2
rebase run <file.py>::<target> --isolation dedicated
rebase run <file.py> --parameters-json '{"a": 1, "b": 2}'
rebase run <file.py> --no-wait
rebase run <file.py> --local

Arguments

ArgumentTypeDescription
<target_ref>strFunction, workflow, or model reference. Use file.py::target_name, or omit ::target_name when the file contains exactly one runnable Rebase target.

Options

OptionTypeDescription
--param, -pname=valueTarget parameter. Values are parsed as JSON when possible. Repeat for multiple parameters.
--parameters-jsonjsonJSON object with target parameters.
--modestrExecution mode: interactive (default) or job.
--isolation, -istrInteractive isolation: shared (default) or dedicated.
--run-type, -rstrDeprecated compatibility alias: quick, quick_shared, or long.
--module, -mboolInterpret the target source as a Python module path instead of a file path.
--wait / --no-waitboolWait for the result before exiting. Defaults to --wait.
--timeoutintMaximum seconds to wait for the result. Defaults to 600.
--poll-intervalfloatSeconds between status polls. Defaults to 1.0.
--localboolExecute the function or workflow in this Python process instead of submitting a cloud run. Workflow steps execute eagerly in-process. Models are not supported. Cannot be combined with cloud execution options or --no-wait.

Output

When waiting for a result, rebase run streams run progress, backend submission events, and workflow step status. Successful waited runs end with:

✓ Run completed in 1.23 seconds.

Then the command prints the JSON result.

When using --no-wait, the command prints the submitted run ID and status.

Behavior

rebase run creates an ephemeral run record and sends the source for that run. It defaults to interactive/shared execution for the lowest cloud-loop latency, regardless of whether you also plan to deploy the target with another mode. It does not create or update a reusable Function, Workflow, or Model target. Use rebase deploy or rebase model deploy when you want a versioned target.

With --local, no run record is created at all: the file is imported and the function or workflow executes in the current Python process, using your local environment. This is the fast inner loop for iterating on code before submitting cloud runs. Output has the same shape as a cloud result ({} for None, dicts as-is, other values wrapped as {"value": ...}), but local execution has access to your full module globals, which packaged cloud source does not.

Use the inspection commands to look up submitted runs:

CommandPurpose
rebase run cancelCancel a submitted run.
rebase run getShow run metadata.
rebase run listList submitted runs.
rebase run logsShow events, workflow steps, and stdout/stderr logs.

cancel

Cancel a submitted run.

Synopsis

rebase run cancel <run-id>

Arguments

ArgumentTypeDescription
<run-id>strRun ID or a unique prefix of one. Omit it for the workspace's latest run.

Behavior

Cancels the run on its execution infrastructure and marks it cancelled. Workflow runs and job-mode function/model runs are cancellable. Interactive function and model runs execute synchronously and cannot be cancelled; use mode="job" when cancellation is required. See rebase run cancel for the full matrix.

get

Show run metadata.

Synopsis

rebase run get
rebase run get <run-id>
rebase run get <run-id> --json

Arguments

ArgumentTypeDescription
<run-id>strRun ID or a unique prefix of one. Omit it for the workspace's latest run.

Options

OptionTypeDescription
--jsonboolPrint machine-readable JSON output.

Output

Shows run metadata including target type, status, execution mode and isolation, concrete execution backend, target IDs, backend run IDs, parameters, result, error, timings, and timestamps. It is followed by a chronological timeline of run events and workflow steps, a "Where the time went" breakdown, and the run's cost when it touched credits.

list

List submitted runs.

Synopsis

rebase run list
rebase run list --target-type workflow --limit 25
rebase run list --project default
rebase run list --json
rebase run list --include result,parameters --json

Options

OptionTypeDescription
--projectstrFilter by project name.
--target-typestrFilter by target type. Supported values are function, workflow, and model.
--limitintMaximum number of runs to list. Defaults to 100. Maximum is 500.
--include, -istrComma-separated bodies the rows leave out by default: result, parameters. Naming either returns each run's whole record.
--jsonboolPrint machine-readable JSON output.

Output

The table output includes run ID, target type, status, execution (mode/isolation), project, created timestamp, and finished timestamp. Rows are summaries: --json rows carry result_bytes and parameters_bytes instead of the result, parameters and trigger_context bodies. Use --include result,parameters for whole records, or rebase run get for one run.

logs

Show run events, workflow step state, and captured stdout/stderr logs.

Synopsis

rebase run logs
rebase run logs <run-id>
rebase run logs <run-id> --no-follow
rebase run logs <run-id> --json

Arguments

ArgumentTypeDescription
<run-id>strRun ID or a unique prefix of one. Omit it for the workspace's latest run.

Options

OptionTypeDescription
--follow / --no-followboolFollow until the run reaches a terminal state. Defaults to --follow.
--poll-intervalfloatSeconds between run status polls when following. Defaults to 1.0.
--timeoutintMaximum seconds to follow the run. Defaults to 600.
--jsonboolPrint raw event, step, and log JSON output.

Output

Shows persisted run events, workflow step state (for workflow runs), and the run's stdout/stderr log lines. When following, new log lines stream in as they arrive. See rebase run logs for log availability per run type.

On this page