v0.6.0

Minor Changes

• 2e8a234: Uploads now carry the build's commit sha and branch, alongside the per-story ids scry-sbcov writes into metadata.json, so hosted search can report freshness — which commit a component came from and whether it is from the current build. The commit and branch are resolved from SCRY_COMMIT_SHA/SCRY_BRANCH, then the GitHub Actions environment, then the working copy, and either is omitted rather than defaulted when it cannot be determined, so an unknown commit reads as unknown instead of as one that cannot be looked up. Requires @scrymore/scry-sbcov 0.5.0, which is what emits the per-story fields.

@scrymore/scry-deployer v0.6.0

Release @scrymore/scry-deployer@0.6.0

See CHANGELOG.md for details.

v0.5.2

Patch Changes

• 708486c: Bump @scrymore/scry-sbcov to ^0.4.0: screenshot framing options (captureMode: root, captureViewport, captureScale) read from the project's scry-sbcov.config.js, so component stories are captured at the size of the Figma component they are compared against and phone-sized screens are no longer clipped at 720 px. No deployer flags change.

@scrymore/scry-deployer v0.5.2

Release @scrymore/scry-deployer@0.5.2

See CHANGELOG.md for details.

v0.5.1

Patch Changes

• 9fb43a2: A failed coverage upload no longer prevents your components being indexed.

  The coverage retry was unguarded, so when the second attempt also failed it threw
  out of uploadBuild and the metadata upload after it never ran. Coverage is a
  report; the metadata archive is what makes components searchable — so a failure in
  the optional artifact silently took down the essential one.

  Seen on a real 467-story design system: every story captured, the archive built,
  and nothing indexed — twice in a row.

  The retry is now guarded and prints a warning that says plainly which part failed
  and that indexing is unaffected.

@scrymore/scry-deployer v0.5.1

Release @scrymore/scry-deployer@0.5.1

See CHANGELOG.md for details.

v0.5.0

Minor Changes

• 7819a50: A deploy that was asked for --with-analysis now fails when analysis produces
  nothing.

  Previously it printed ✅ Upload successful! and exited 0. No metadata archive
  meant nothing was queued, so no component ever became searchable — and because
  the indexing notice only prints when metadata was sent, there was no output at
  all to distinguish it from a healthy run. CI stayed green. The first sign of
  trouble was search returning nothing, days later.

  This is a behaviour change: the command now exits non-zero in that state. The
  Storybook is still uploaded and hosted, so "failure" overstates it slightly — but
  the job asked for was to make components searchable, and a green build there means
  search silently returns nothing.

  Common causes, both seen in practice: a missing Playwright browser (fixed for
  generated workflows in 0.4.1), and a TypeScript resolution error in the analyzer
  on a plain npm install tree.

v0.4.1

Patch Changes

• a6ce9b3: scry init no longer reports success for steps that failed.

  Setting up a real repository end to end produced "✅ Changes committed and pushed"
  and "✅ Repository secret (SCRY_API_KEY)" while having done neither. CI then failed
  at the deploy step with no credentials, and the only clue was a warning printed
  twenty lines above the success banner that contradicted it.

  Three fixes:

  • git add no longer aborts the commit. It throws on a .gitignore'd path, and
    one throw skipped the workflow files entirely. A leftover .storybook-deployer.json
    ignore rule from the pre-0.4.0 workaround was enough to prevent CI ever being set up.
  • gh variable is no longer assumed. It arrived in gh 2.21; Ubuntu 22.04 ships
    2.4.0. On older gh the first call threw and the secret after it was never reached,
    leaving the repository with no variables and no secret. There is now a capability
    check and a gh api fallback.
  • The closing summary reports what happened, including a distinct message for
    "not attempted" when GitHub setup was skipped.

• 92bc27e: The GitHub Actions workflow scry init generates now installs a browser.

  Without it, screenshot capture failed for every story, no metadata archive was
  produced, and nothing was ever indexed — while the deploy exited 0 and the
  workflow went green. Proved on a real repository: CI passed and the project
  recorded storybook_uploaded and nothing else.

  The install command follows the project's package manager. npx is not safe to
  assume: under pnpm it resolves against the pnpm-managed environment and reports
  playwright: not found (exit 127) even with --yes.

  If you ran scry init before this release, add the step by hand before your
  deploy step, or re-run init — otherwise CI will keep passing without indexing
  anything.

@scrymore/scry-deployer v0.5.0

Release @scrymore/scry-deployer@0.5.0

See CHANGELOG.md for details.

@scrymore/scry-deployer v0.4.1

Release @scrymore/scry-deployer@0.4.1

See CHANGELOG.md for details.