Commands¶

The reference for the bio CLI's two commands, suite and config, with their subcommands and flags.

Command	CLI	Description
suite	`bio suite run <spec.yaml> [--out DIR] [--dry]`	Run a declared experiment (ABIO Expr Spec § experiments): drafts the grid, runs every trial, writes `records.jsonl`, `manifest.json`, `map.json`/`map.csv`, `report.txt`, and the run's one key figure `key.png` (+ `key.json`: readout and caption). `--dry` prints the grid, the no-peeking verdict and the cost estimate without running.
	`bio suite resume <DIR>`	Continue a run from its record store (crashed or cost-stopped).
	`bio suite aggregate <DIR>`	Rebuild the reliability map from `records.jsonl` alone.
	`bio suite report <DIR>`	Render the text report and the key figure (`key.png`, `key.json`) from the record store alone.
	`bio suite models`	Refresh the recorded `models.list` snapshot (`suite/models_snapshot.json`, id → `created_at`) that lets an undated generation id count as pinned; a free call.
	`bio report [--open] [--no-examples]`	Run the suite once and write what it tested, and whether it passed, as one page: the capability matrix with each proving test's sentence, the broader suites, the examples run fresh, the runs on disk — each with its key figure embedded (`reports/report.md` + `.html`, the HTML self-contained; `just report`).
config	`bio config [show \\| set KEY VALUE]`	The framework configuration: provider keys (read from the environment), the pinned model.

The M1 command set (build, run, report, fetch, store, hydrate, dehydrate, cd, agent, experiment, sim, scenario, lookup) went with the M1 scenario runtime in M47.7 — an experiment is now one declared file run through the suite harness (ABIO Suite Runtime).

Every catalog experiment (catalog/experiments/*.yaml) is a bio suite run target; the twelve scripted zeros are pinned as golden regressions in CI.