Inspect Flow

Overview

pip install "inspect-steward[flow]"

Inspect Flow specs are fully supported as Steward eval set definitions. Name the spec config.py in your workspace directory and the rest of the documentation applies unchanged.

Worker Overrides

Steward runs flow run with three options set:

Override	Why
--set execution_type=inproc	One process per task is Steward’s own isolation model, and a venv per worker would put the eval in a grandchild process that Steward cannot track.
--log-dir <scratch>	Flow writes artifacts and scans for prior logs before eval_set() is reached. Each worker gets its own scratch directory for that; the run’s log directory is set at the eval_set() call.
--no-store-read --no-store-write	Flow’s reuse store would race under fan-out, and buys nothing, since a worker runs the task it was given either way.

These are all options flow run already ships.

Declarations That Do Not Survive

Importantexecution_type: venv loses dependencies and python_version

Flow applies those declarations when it builds the virtualenv. Steward forces inproc, so the environment Steward runs in is the environment the eval runs in, and provisioning it is up to you.

A spec that builds its FlowSpec in code loses them silently, because Steward reads the definition without executing it. The warning Steward prints for this covers only a top-level declaration in a YAML spec.

Noteflow run --resume will not resume a Steward run

Every flow run records its log directory as a global pointer, and Steward hands Flow a scratch directory, so a bare flow run --resume would resume nothing and start over into a directory nobody is watching.

Under Steward, the resume path is steward launch against the workspace, which is also the amend path.

Scanners

A spec carrying scanner: cannot be run by Steward yet. This is refused when the definition is read, rather than discovered when every worker fails.