Skip to content

b8e qa audit

Run QA journeys as an ordinary user, as many at once as your plan’s concurrent runs allow: each step’s outcome prints as it finishes, then the report

OptionTakesDefaultDescription
--project <id>valueproject (default: the one b8e project selected)
--journeys <id|name,...>value“QA journey ids or names (see b8e qa journeys list): each value is an id or exact name, otherwise a comma list; repeat the flag to combine a name with a comma in it with others
--group [id|name]valuerun every journey in a QA group, by id or exact name; alone, the project’s only group, or asks which
--environment <name>valuewhose web target the journeys start on; needed when the project has more than one
-f, --format <format>valueprettyoutput format
--timeout <duration>valuebudget per journey (e.g. 10m, default 30m)
--cloudbooleanrun on Binclusive’s cloud runners instead of this machine: prints the run ids, then follows the runs to a verdict (—timeout and —headless apply to local runs only)
--build-target <release/dist>valueandroid —cloud runs only, and required there: the uploaded build the cloud device installs (e.g. 1.4.0/a1b2c3)
--keep-app-databooleanandroid only, local only: keep the data the app under test already holds, instead of clearing it before each journey
--cibooleannon-interactive CI mode (requires —journeys or —group, and a selected project)
--headlessbooleanrun the browser headless (default)
--no-headlessnegationshow the browser window
--fixtures <dir>valuefolder of files the tester may attach when a step says to upload one

Identical to what the terminal prints — both are rendered from the same commander definition.

Usage: b8e qa audit [options] [command]
Run QA journeys as an ordinary user, as many at once as your plan's concurrent
runs allow: each step's outcome prints as it finishes, then the report
Options:
--project <id> project (default: the one `b8e project`
selected)
--journeys <id|name,...> QA journey ids or names (see `b8e qa journeys
list`): each value is an id or exact name,
otherwise a comma list; repeat the flag to
combine a name with a comma in it with others
(default: [])
--group [id|name] run every journey in a QA group, by id or exact
name; alone, the project's only group, or asks
which
--environment <name> whose web target the journeys start on; needed
when the project has more than one (choices:
"pr", "staging", "production", "unknown")
-f, --format <format> output format (choices: "pretty", "json",
default: "pretty")
--timeout <duration> budget per journey (e.g. 10m, default 30m)
--cloud run on Binclusive's cloud runners instead of
this machine: prints the run ids, then follows
the runs to a verdict (--timeout and --headless
apply to local runs only)
--build-target <release/dist> android --cloud runs only, and required there:
the uploaded build the cloud device installs
(e.g. 1.4.0/a1b2c3)
--keep-app-data android only, local only: keep the data the app
under test already holds, instead of clearing
it before each journey
--ci non-interactive CI mode (requires --journeys or
--group, and a selected project)
--headless run the browser headless (default)
--no-headless show the browser window
--fixtures <dir> folder of files the tester may attach when a
step says to upload one
-h, --help display help for command
Commands:
stop [options] <cloud-run-id> Cancel a cloud QA request by its cloud run
request ID. A queued request is removed from
the queue. For a dispatched request, active
runs end and VM teardown is attempted; teardown
failure is reported but does not change the
cancelled outcome. A request that already ended
is left unchanged.
report [options] <run-id> A QA run by id, with its owner-computed verdict
and joinedSteps, including pending steps. Null
when it does not exist or belongs to another
organization.
list|ls [options] The project's QA runs, newest first. Runs
minted under the organization that owned the
project before a move stay with that
organization. List a project's Agentic QA runs,
newest first. Each run has its QA run ID
(qar_...), journeyId and name, state (running,
completed, failed, abandoned) and verdict
(clean, findings, failed, incomplete,
site_blocked). Pages with first/after; pass
pageInfo.endCursor as after to load more.
JOURNEYS
Each --journeys value is matched whole first, as a journey id or an exact
journey name. Only a value that is neither is split on commas, and each
piece is matched the same way. So a name that contains a comma selects
that journey; to run it with others, give each in its own --journeys.
A journey given twice runs once.
GROUPS
--group <id|name> runs every journey in one of the project's QA groups, by the
group's id or exact name. --group alone takes the project's only group, or asks
which one when there are several and the terminal can ask. It cannot be combined
with --journeys. A --cloud group run is cut into runs of up to 5 journeys, each on
its own VM; every one that starts is followed, and the report covers the whole
group. A group run on this machine runs every journey.
Every step prints a text label: PASSED, FAILED, INCONCLUSIVE, COMPLETED, BLOCKED,
AGENT ERROR, INFRA ERROR or SKIPPED. Only a FAILED step is a Finding.
The report also lists what the browser saw fail on the site (5xx and failed
requests, console errors); these are stored on the run and are never Findings.
CONCURRENCY
Journeys run at once up to your plan's concurrent QA runs, and one at a time
when that limit cannot be read. Each progress line names its journey; the
report lists journeys in the order you gave them.
TIMEOUT
--timeout is a budget per journey, not for the whole batch. On heavy pages a
step takes about 25–30 s, so size it at 30 s per step: a 25-step journey used
631 s of a 900 s (15m) budget in a measured run.
CLOUD
--cloud runs the journeys on a Binclusive cloud runner instead of this machine,
one after another on one VM, up to 5 per run. It prints each journey's run id as
the VM starts it, follows the runs to their end, and exits the same way a local
run does. --timeout and --headless do not apply to it, and --fixtures is refused.
An Android project's cloud run needs --build-target <release>/<dist>, a build
uploaded with `b8e build upload`; the cloud device installs it before the first
journey. A web project's run refuses the flag.
APP DATA
An Android journey starts the way a fresh install does: before it grants the
app's permissions and launches it, the run clears the app under test's data
(adb shell pm clear), so a language, sign-in or onboarding an earlier run left
does not carry over. --keep-app-data skips the clear and keeps everything the
app already holds, such as a signed-in account; use it for a journey that
relies on that. Each journey's output says whether its app data was cleared or
kept. A cloud run always starts fresh, so the flag is refused with --cloud.
FIXTURES
A step that says to upload or attach a file names it, and the tester attaches
the file of that name from the --fixtures folder. It takes a name, never a
path: a name that resolves outside the folder (a symlink out included) is
refused, and so is every upload in a run given no --fixtures.
EXIT CODES
0 every step passed or completed
1 at least one Finding
3 a journey failed to run, or its results are unavailable
8 ran, but some steps could not be checked
Ctrl-C stops the browser but not the run: it closes itself once it goes quiet, and
`b8e qa audit report <run-id>` prints what it recorded. A --cloud run keeps going on
its VM until `b8e qa audit stop <cloud-run-id>` stops it.
EXAMPLES
$ b8e qa audit --journeys qaj_123
$ b8e qa audit --journeys qaj_123 --environment staging
$ b8e qa audit --journeys qaj_123 --keep-app-data
$ b8e qa audit --journeys qaj_123,qaj_456 --cloud
$ b8e qa audit --group Smoke
$ b8e qa audit --group qag_123 --cloud --ci
$ b8e qa audit --journeys qaj_123 --cloud --build-target 1.4.0/a1b2c3
$ b8e qa audit --journeys "Checkout as a guest,qaj_456" --format json > qa-report.json
$ b8e qa audit --journeys "Checkout: card, then refund" --journeys qaj_456
$ b8e qa audit list
$ b8e qa audit report qar_123
$ b8e qa audit stop 0b6f2c1e-6d0a-4c8e-9a51-3f2d7f1c9e42
INHERITED FLAGS
--org <id|slug|name> organization context [$BINCLUSIVE_ORG_ID]
-y, --yes skip all prompts and accept defaults
--non-interactive never prompt; fail instead of asking
--progress <mode> how progress is reported
-q, --quiet suppress progress output (same as --progress quiet)
--no-color disable color
-v, --verbose verbose diagnostics
-V, --version print the CLI version
-h, --help show help for command