Automate

Command-line tool

qa-ai.mjs: one file, Node 18 or later, no dependencies. Runs tests, waits for the verdicts, writes JUnit and uploads app builds.

Get it#

curl -fsSL https://app.testsy.dev/cli/qa-ai.mjs -o qa-ai.mjs
export QA_AI_API_KEY=qa_…          # Settings → API keys
node qa-ai.mjs version

Variable
`QA_AI_API_KEY`	Required.
`QA_AI_URL`	The app's address. Default `https://app.testsy.dev`.

Commands#

qa-ai run --project <p_id> [--test <t_id,…>] [--mode auto|cold|warm] [--timeout 45m]
          [--junit file.xml] [--fail-on failed,error] [--no-wait]
          [--build <app.apk | MyApp.app.zip | ab_id>]… [--label <text>]
qa-ai build upload <app.apk | MyApp.app.zip> --project <p_id> [--label <text>] [--json]
qa-ai builds --project <p_id>
qa-ai status <j_id…>
qa-ai projects
qa-ai tests --project <p_id>

run starts every test in the project, or the tests you name (--project and --test can be combined), then checks every 15 seconds until each has a verdict, printing each result with its link, and a summary.

Flag
`--mode`	`auto` (default), `cold` or `warm`: see learned paths.
`--timeout`	How long to wait: `30s`, `45m` (default), `1h`; a bare number is minutes.
`--junit <file>`	Write a JUnit report: one test case per run; a failure as `<failure>`, any other non-pass as `<error>`.
`--fail-on`	Verdicts that make it exit 1. Default `failed,error`. Also `blocked`, `inconclusive`.
`--no-wait`	Print the run ids as JSON and exit 0.
`--build`	App tests run these builds. A file is uploaded first; an `ab_` id is used as is. Repeat it for several apps: each app test gets the build of its own app.
`--label`	A label for uploaded builds. Default: the commit (`GITHUB_SHA` or `CI_COMMIT_SHORT_SHA`).

build upload uploads a build in parts (three at a time) and prints its id, or the whole build with --json. On GitHub it also sets the step output build-id.

builds, tests and projects list ids; status prints runs as JSON.

In CI#

The tool recognises GitHub Actions, GitLab, CircleCI, Bitbucket, Buildkite, Azure Pipelines and Jenkins, and sends the commit, branch, job and job link with each run. On GitHub it adds a results table to the job summary. Each start sends an Idempotency-Key, and requests are retried on network errors and 5xx answers, so a retry never starts a run twice.

Exit codes#

0 every run passed (or none ended in a --fail-on verdict) · 1 a run ended in a --fail-on verdict · 2 a setup or API problem · 3 timed out.