@decocms/parity 0.17.8 → 0.18.0
This diff represents the content of publicly available package versions that have been released to one of the supported registries. The information contained in this diff is provided for informational purposes only and reflects changes between package versions as they appear in their respective public registries.
- package/AGENTS.md +1 -0
- package/CHANGELOG.md +21 -0
- package/README.md +4 -0
- package/dist/cli.js +2311 -322
- package/package.json +1 -1
package/AGENTS.md
CHANGED
|
@@ -31,6 +31,7 @@ Do **not** run parity for:
|
|
|
31
31
|
| "explain this specific issue" | `parity explain <runId> <issueId>` (needs `ANTHROPIC_API_KEY`) |
|
|
32
32
|
| "does THIS single site actually work?" | `parity e2e --url ... --flows ...` (no comparison — pre-launch / partner audit) |
|
|
33
33
|
| "exercise search / cart / login on a site" | `parity e2e --url ... --flows=search,cart-interactions,login` |
|
|
34
|
+
| "migrate a live storefront to a new stack" | `parity migrate --url ... --target faststore` (phased capture → agent-ready bundle; see `docs/migrate.md`) |
|
|
34
35
|
|
|
35
36
|
### `parity run` vs `parity e2e` — decision tree
|
|
36
37
|
|
package/CHANGELOG.md
CHANGED
|
@@ -7,6 +7,27 @@ and this project adheres to [Semantic Versioning](https://semver.org/).
|
|
|
7
7
|
|
|
8
8
|
## [Unreleased]
|
|
9
9
|
|
|
10
|
+
## [0.18.0] — 2026-08-17
|
|
11
|
+
|
|
12
|
+
### Added
|
|
13
|
+
|
|
14
|
+
* **`parity migrate` — phased, target-agnostic migration capture (new command).** Single-site (no prod×cand) pipeline that produces a token-lean, agent-ready bundle + prompt for migrating a live storefront to a new stack. Three resumable phases: **theme + assets** (color/typography/spacing/radii token election, breakpoints from `@media`, motion tokens, plus brand assets — logo/favicon/apple-touch/OG/manifest/fonts — downloaded **through the browser** to bypass Akamai/CF bot 403s, and an icon inventory); **sitemap** (browser-routed page discovery + classification); **components** (reusing `extract`'s detector, structural dedupe, deterministic **Tailwind IR** + raw-CSS fallback, light interaction hints via CSSOM, and suggested e2e selectors). Token economy is first-class (lean tier vs full `manifest.json`, `×N` collapse, per-component files). (#194)
|
|
15
|
+
* **`--target faststore` playbook + `custom-theme.scss`.** Appends a FastStore v4 playbook (CLI commands, doc links, section/CMS structure, Phosphor icons) and emits a deterministic `custom-theme.scss` mapping brand tokens → `--fs-*` global tokens. FastStore v4 styles with SCSS design tokens + `data-fs-*` (not Tailwind). (#196)
|
|
16
|
+
* **Multi-viewport theme + per-viewport site screenshots** (`--viewports mobile,desktop`). (#197)
|
|
17
|
+
* **VTEX IO block-tree source + deterministic FastStore mapper.** For VTEX IO stores, reads the real declarative block tree from `window.__RUNTIME__` → `blocks.json` + `component-map.json` (`product-summary→ProductCard`, custom blocks → `custom-component`). Verified live on storetheme.vtex.com. (#198)
|
|
18
|
+
* **Platform-aware PDP discovery** for `migrate` — Salesforce Commerce `.html`/`Product-Show` product pages, with an action-endpoint rejection guard (Wishlist-Add/Cart-Add never treated as a PDP). (#200)
|
|
19
|
+
* **`salesforce-commerce` (Demandware) platform detection**, checked before the loose `fs-`/`vtex-` class heuristics (fixes Sephora BR mis-detected as `vtex`). (#199)
|
|
20
|
+
|
|
21
|
+
### Fixed
|
|
22
|
+
|
|
23
|
+
* **403 hardening: `migrate` Phase 2 discovery routes through the browser** instead of a bare node fetch that bot-protected stores 403 (was silently degrading to home-only). (#195)
|
|
24
|
+
* **`--fail-on` now gates the exit code on its own (issue #178).** The blocking-issue exit-1 check was gated behind `--ci`, which had no other effect in the codebase — a run with double-digit criticals exited 0 unless the caller also remembered `--ci`, even though `--fail-on` (default `critical`) was already parsed and ready. The check now always runs; `--ci` has been removed as dead weight (`parity audit` never had it and always gated on `--fail-on` alone, which is the pattern `parity run` now follows too).
|
|
25
|
+
* **`--pages`/`--pages-file` scoping documented, and no longer silently disable visual-diff in no-LLM environments (issue #178).** These flags only ever scoped the visual-diff/vitals-extra-pages passes, never the `flows` crawl — now documented in `--help`/`docs/cli.md`, and `parity run` prints a one-line warning when both an explicit page list and a flows crawl are active. Separately, `applySmartDefaults` (issue #71) used to force `noVisualDiff = true` whenever no LLM provider was configured, even when `--pages`/`--pages-file` was explicitly set — making the flag silently inert. Explicit page selection now still gets the prod/cand screenshot + pixelmatch heatmap capture without an LLM verdict.
|
|
26
|
+
* **`banner-aspect-ratio` no longer double-counts carousel boundary clones (issue #178).** Images marked `aria-hidden="true"` or `data-slider-clone` (or nested inside such a wrapper) — the decorative clones a hydration-safe infinite carousel renders at its boundaries — are now excluded from the banner census in `src/diff/dom.ts`, which previously produced false "candidate has N extra banners" findings.
|
|
27
|
+
* **`lazy-section-presence` no longer reports a spurious "missing: render" on every page of a TanStack Start site (issue #178).** The last-path-segment fallback in `extractSectionIds` (used when a lazy-render network entry has no `x-deco-section` header) now ignores known infra/admin route names (`render`, `meta`, `invoke`) that carry no real section-identifying information as a bare fallback id.
|
|
28
|
+
* **Visual-diff LLM prompt now hedges on small-text/count reads (issue #178).** Numeric counts and small text read off a (possibly downscaled) full-page screenshot were occasionally misread by the vision model and reported as confident high-severity regressions. The prompt now caps such claims at `low` severity and requires an explicit hedge in the description; `LLM_PROMPT_VERSION` bumped to invalidate cached verdicts from the old prompt.
|
|
29
|
+
* **`--vitals-pages` noise/scoping trade-off surfaced in `--help`/`docs/cli.md` (issue #178).** `--vitals-pages 0` can dramatically cut unrelated auto-sampled noise (one run went from 53 to 13 `high` findings) but was previously discoverable only by reading source — now called out in both the flag's help text and a `docs/cli.md` callout.
|
|
30
|
+
|
|
10
31
|
## [0.17.3](https://github.com/decocms/parity/compare/v0.17.2...v0.17.3) (2026-07-28)
|
|
11
32
|
|
|
12
33
|
### Fixed
|
package/README.md
CHANGED
|
@@ -96,6 +96,10 @@ Without any provider, the CLI still runs and outputs raw check results — only
|
|
|
96
96
|
|
|
97
97
|
Individual flags always override the preset. See `parity run --help` for the full default behavior breakdown.
|
|
98
98
|
|
|
99
|
+
### A note on Web Vitals numbers
|
|
100
|
+
|
|
101
|
+
By default parity measures TTFB/FCP/LCP with the browser's HTTP cache enabled, on a warm connection — a "repeat visit" scenario. Lighthouse and CrUX field data, by contrast, simulate/measure cold first-time visits (cache disabled). Neither is wrong, but they answer different questions, so don't compare parity's default vitals 1:1 against Lighthouse or CrUX. Pass `--bypass-cache` to approximate a cold-visit measurement instead (disables the browser cache via CDP in addition to bypassing CDN/edge caches).
|
|
102
|
+
|
|
99
103
|
## CI usage (GitHub Actions)
|
|
100
104
|
|
|
101
105
|
```yaml
|