Release grouped saved-Scenario Run support

CHANGELOG.md

@@ -1,6 +1,10 @@
 # Changelog
 
-## Unreleased
+## 1.1.0
+
+- Track one grouped saved-Scenario Run with `test-run-id`, replacing retired sweep IDs.
+- Wait through scenario preparation and approval before evaluating results.
+- Document saved YAML Scenarios as the repeatable PR testing rules and report cleanup separately.
 
 - Assign preview URLs to a Composal project and track the PR number or source branch.
 - Auto-detect PR, workflow-run, and branch-push context; register branch previews even before an open PR exists.

README.md

@@ -11,7 +11,7 @@ Test the product journeys affected by your PR against its deployed preview. Veri
     preview-url: ${{ steps.deploy.outputs.url }}
 ```
 
-Configure the Composal project in **Verify → Projects → your project → PR Testing**. Select its source repository and **Your CI**, your scenario pack, and an external Verify environment template containing your personas and safety policy. Add a Composal administrator API token as the `COMPOSAL_TOKEN` Actions secret and your organization slug as the `COMPOSAL_ORG` variable.
+Configure the Composal project in **Verify → Projects → your project → PR Testing**. Select its source repository, **Your CI**, and an external Verify environment template containing your personas and safety policy. Verify selects relevant saved YAML Scenarios from that project and launches them together in one parallel Run. Add a Composal administrator API token as the `COMPOSAL_TOKEN` Actions secret and your organization slug as the `COMPOSAL_ORG` variable.
 
 Run this step after deploying **`github.event.pull_request.head.sha`** and waiting for the preview to become reachable. GitHub's `github.sha` can be a synthetic merge commit; Verify deliberately uses the PR head. Your existing deployment provider supplies the URL; this action does not build an arbitrary app.
 
@@ -98,7 +98,7 @@ jobs:
 
 The fork condition avoids running a secret-dependent step when GitHub withholds secrets. Keep your deployment and secret use on trusted workflow events; do not switch to `pull_request_target` just to expose secrets to fork code.
 
-By default the action waits up to 15 minutes. Passed or explicitly skipped requests succeed. Failed, blocked, cancelled, superseded, cleanup-blocked, and timed-out requests fail the step. A wait timeout also fails the step, with a link to the run that continues in Composal. Set `wait: 'false'` for an asynchronous handoff; a successful handoff means accepted, not passed.
+By default the action waits up to 15 minutes. Passed or explicitly skipped requests succeed. Failed, blocked, cancelled, superseded, and timed-out scenario outcomes fail the step. Cleanup is reported separately and does not turn a passing product outcome into a failure. A wait timeout also fails the step, with a link to the run that continues in Composal. Set `wait: 'false'` for an asynchronous handoff; a successful handoff means accepted, not passed.
 
 ## Inputs and outputs
 
@@ -116,7 +116,7 @@ By default the action waits up to 15 minutes. Passed or explicitly skipped reque
 | `timeout`     | `'15'`                 | Wait limit in minutes, between 1 and 60. Set the job timeout higher.                                             |
 | `api-url`     | `https://composal.ai`  | HTTPS API origin, for installations using another endpoint.                                                      |
 
-Outputs: `url` (PR history or registered branch environment), `pull-request-id`, `run-id`, `sweep-id` (when started), `environment-id` (branch registration), and `status`. The workflow summary links to the full results. Critical/high issue recordings remain authenticated links in Verify.
+Outputs: `url` (PR history or registered branch environment), `pull-request-id`, `run-id`, `test-run-id` (when started), `environment-id` (branch registration), and `status`. The workflow summary links to the full results. Critical/high issue recordings remain authenticated links in Verify.
 
 Retries within the same workflow attempt reuse the same verification request. Rerunning the workflow creates a fresh request. A workflow for an obsolete head cannot start a run for the newer head.
 

action.yml

@@ -40,8 +40,8 @@ outputs:
     description: Composal PR identity.
   run-id:
     description: This verification request's identity.
-  sweep-id:
-    description: Browser sweep identity, when started.
+  test-run-id:
+    description: Grouped scenario Run identity, when started.
   status:
     description: Verification state or terminal outcome. queued means accepted, not passed.
   environment-id:

verify.mjs

@@ -182,7 +182,7 @@ export async function verify(
     url: url.href,
     'pull-request-id': pr.id || '',
     'run-id': pr.requested_run_id || '',
-    'sweep-id': '',
+    'test-run-id': '',
     status: pr.requested_run_id ? 'queued' : pr.status || 'skipped',
     'environment-id': pr.environment_id || '',
   }
@@ -202,12 +202,20 @@ export async function verify(
     if (run.id !== pr.requested_run_id)
       throw new Error('A newer verification request superseded this handoff.')
     outputs.status = run.display_state
-    outputs['sweep-id'] = run.verify_sweep_id || ''
+    outputs['test-run-id'] = run.run_id || ''
     await publish(outputs)
   }
   if (!config.wait) return outputs
   const deadline = now() + config.timeoutMs
-  const active = new Set(['queued', 'waiting_for_preview', 'running', 'cancelling', 'terminal'])
+  const active = new Set([
+    'queued',
+    'waiting_for_preview',
+    'preparing',
+    'prepared',
+    'running',
+    'cancelling',
+    'terminal',
+  ])
   while (now() < deadline) {
     const page = await api(`${requests}/${encodeURIComponent(pr.id)}`)
     const run = page.runs.find((candidate) => candidate.id === pr.requested_run_id)
@@ -219,7 +227,7 @@ export async function verify(
     )
       throw new Error('A newer PR revision or verification request superseded this run.')
     outputs.status = run.display_state
-    outputs['sweep-id'] = run.verify_sweep_id || ''
+    outputs['test-run-id'] = run.run_id || ''
     await publish(outputs)
     if (!active.has(outputs.status)) {
       if (!['passed', 'skipped'].includes(outputs.status))

verify.test.mjs

@@ -32,7 +32,7 @@ const run = (status = 'passed') => ({
   id: 'vprun_1',
   head_sha: sha,
   display_state: status,
-  verify_sweep_id: 'vsweep_1',
+  run_id: 'run_grouped_1',
 })
 function transport(responses) {
   const calls = []
@@ -180,10 +180,13 @@ test('validates input authority and refuses ambiguous or missing PR context', ()
   assert.equal(config({ pr: '#45', sha }, {}).number, 45)
 })
 
-test('hands a deployed URL to the exact request, waits, and publishes the sweep link', async () => {
+test('hands a deployed URL to the exact request, waits, and publishes the grouped run identity', async () => {
   const api = transport([
     pr,
     run('waiting_for_preview'),
+    { pull_request: pr, runs: [run('preparing')] },
+    { pull_request: pr, runs: [run('prepared')] },
+    { pull_request: pr, runs: [run('queued')] },
     { pull_request: pr, runs: [run('running')] },
     { pull_request: pr, runs: [run()] },
   ])
@@ -209,11 +212,11 @@ test('hands a deployed URL to the exact request, waits, and publishes the sweep
   assert.equal(api.calls[0].headers.Authorization, 'Bearer private-token')
   assert.equal(api.calls[0].redirect, 'error')
   assert.equal(result.status, 'passed')
-  assert.equal(result['sweep-id'], 'vsweep_1')
+  assert.equal(result['test-run-id'], 'run_grouped_1')
   assert.equal(result.url, 'https://composal.ai/acme/verify/pull-requests/vpr_1')
   assert.deepEqual(
     updates.map((value) => value.status),
-    ['queued', 'waiting_for_preview', 'running', 'passed']
+    ['queued', 'waiting_for_preview', 'preparing', 'prepared', 'queued', 'running', 'passed']
   )
 })