Save, load, and replay runs

Persist an individual optimization, inspect its stored results, check its integrity, or explicitly reproduce it in a compatible environment.

Stable public API. The 1.x compatibility policy defines the supported surface; import from the public facade shown below.

from vamos import save_result, load_run, load_result, verify_run, reproduce

Run artifacts guide · Stored run and report models · Results

Loading and verification are data-only operations. reproduce is a separate executable operation: it creates a new run and does not overwrite the source. Exact replay is limited to reconstructable built-ins in a materially matching environment and backend.

save_result(result, path, *, requested_spec=None, resolved_spec=None, labels=None, limits=None)

Persist a result as one immutable, relocatable v1 run directory.

load_run(path, *, verify='required', limits=None)

Load immutable manifest access without resolving or executing code.

load_result(path, *, verify='required', limits=None)

Load the canonical numerical result; this never reruns optimization.

verify_run(path, *, require_level=None, limits=None)

Fully verify a canonical run without executing or resolving code.

reproduce(path, *, output=None, limits=None)

Verify, exactly execute, compare, and store a new built-in replay.