Catalogs
Reference the OpenPets pet and plugin catalogs, generated artifacts, ZIP hosting, sidecars, validation gates, and desktop fetch behavior.
Catalogs
The app discovers and installs pets and plugins from versioned JSON catalogs
served at openpets.dev, with ZIPs hosted on R2 behind zip.openpets.dev. This
doc describes those catalogs as contracts and how the desktop consumes
them. It is intentionally scoped to the app-facing data of web/ - catalogs,
ZIP hosting, pet/plugin metadata - and not the marketing site/frontend.
This doc is the consumer-and-contract view. Producer commands and release gates are covered in Testing and validation and the package scripts they reference; do not treat older website-local notes as authoritative when they disagree with the generated catalog.
Direction (read this first)
- Pet catalog v3 is the source of truth. v2 exists only for old app versions and as a fallback. New work targets v3.
- Plugin catalog v2 is active. v1 is kept as an empty compatibility shim
for older desktop builds. (Verified URLs in
apps/desktop/src/plugin-catalog.ts.) - This matches the forward-only direction in
AGENTS.md: don’t optimize new behavior for legacy catalog versions.
Endpoints the app fetches
| Data | URL | Owner in app |
|---|---|---|
| Pet catalog v3 (index) | https://openpets.dev/pets/catalog.v3.json |
apps/desktop/src/catalog.ts |
| Pet catalog v3 pages | …/pets/catalog.v3/page-NNN.json |
catalog.ts |
| Pet catalog v3 search | …/pets/catalog.v3/search.json (+ search pages) |
catalog.ts |
| Pet catalog v2 (legacy/fallback) | https://openpets.dev/pets/catalog.v2.json |
catalog.ts |
| Pet ZIPs | https://zip.openpets.dev/pets/{slug}/{installId}.zip |
pet-installation.ts |
| Plugin catalog v2 (active) | https://openpets.dev/plugins/catalog.v2.json |
plugin-catalog.ts |
| Plugin catalog v1 (empty compat) | https://openpets.dev/plugins/catalog.v1.json |
plugin-catalog.ts |
| Plugin ZIPs | https://zip.openpets.dev/plugins/{plugin-id}.zip |
plugin-package.ts |
Pet catalog v3 contract
v3 is paginated to keep each runtime fetch small. The flow the app follows:
- Fetch the root index (
catalog.v3.json):version: 3,generatedAt,total,pageSize, asearchURL,filters(categories with counts,originalsCount,featuredCount), and apages[]array of page URLs. - Fetch pages on demand. Each page entry carries install + render data:
id,displayName,description,thumbnail,spritesheet,zip,category, optionalsubcategory,featured,original. - Use search pages for lightweight lookup:
id,displayName,searchText,category,catalogPage,featured,original.
Only pets with a valid category (western or asian) appear in v3 - the
generator drops the rest and logs a warning. To keep the app UI clean, the
Control Center browsing/search indexes surface only “curated” (original or
featured) pets. However, explicit lookup/installation by ID allows installing
any valid v3 catalog pet. The validators on the app side live
in catalog-validation.ts (validateCatalogV3Index, validateCatalogV3Page,
validateCatalogV3SearchIndex, validateCatalogV3SearchPage, plus
validateCatalogV2). Treat the generator output plus those validators as the
contract, not any hand-written copy.
Fetch fallback chain
catalog.ts resolves the catalog V3 → V2 → bundled fixture. The fixture
(catalog.v2.fixture.json) keeps the app usable offline / in tests. The fixture
should never be the path real users hit online; it is a last-resort floor, not a
shipping catalog.
Pet generated artifacts
The publishing scripts (web/scripts/) treat web/public/pets/manifest.json as
the canonical generated state and derive everything else from it:
web/public/pets/
manifest.json canonical generated catalog state
install.json install metadata
catalog.v2.json legacy/fallback catalog
catalog.v3.json v3 index
catalog.v3/
page-000.json … paginated pages
search.json
search-page-NNN.json
{slug}/
spritesheet.webp
thumb.webp
thumb.webp.meta.json
Regenerating artifacts from the manifest is generate:catalog-artifacts;
verifying them is the verify:catalog* family (see
Testing and validation). The web build also emits
app/lib/pets.generated.js / pets.preview.js for the site - out of scope here.
Plugin catalog contract
The plugin catalog (plugin-catalog-validation.ts) is validated strictly:
schema version, unique ids, semver + SHA fields, canonicalized permissions, and
an optional minimum-OpenPets-version gate. Catalog cards may carry an
iconDataUrl (base64 SVG) so the Plugins UI renders an icon without an extra
fetch. Each entry’s downloadUrl must point at zip.openpets.dev/plugins/.
Catalog v2 also carries publisherType: "official" | "community"; older
catalogs without the field are treated as official by the desktop validator.
Community entries are public catalog plugins but cannot be bundled/default-on.
Sidecars: plugin provenance and submissions (website-only)
To secure community-contributed plugins without changing the app-facing
catalog.v2.json schema, the website serves sidecar metadata files:
| File | Purpose |
|---|---|
https://openpets.dev/plugins/provenance.json |
Reviewed provenance for installable community plugins. |
https://openpets.dev/plugins/submissions.json |
Pending external GitHub submissions shown on the website but not installable. |
These files are only used by the website/CI environment for provenance display,
validation, and automated owner-publishing policy. provenance.json is keyed by
plugin ID and defines:
publisher: GitHub user/organization owner.sourceUrl: Upstream GitHub repository.sourceSubdirectory: Optional subdirectory under the repository root.sourceCommit: The reviewed and approved commit SHA.reviewedAt: ISO date/time of review.updatePolicy:safe-auto(safe for automated release updates) ormanual-review.
submissions.json is also keyed by plugin ID, but entries are candidates only:
they must not appear in the installable catalog until promoted into
plugins/community/, packaged, uploaded, and release-validated.
The desktop fetch (plugin-catalog.ts) is hardened: timeout, redirect
rejection, response-size cap, and caching with refresh. Install/verification of
the downloaded ZIP (SHA-256, host/path allowlist, entry restrictions, manifest
↔ catalog consistency) is plugin-package.ts. See Plugin platform.
ZIP hosting (R2)
Both pet and plugin ZIPs live on the R2 bucket (default openpets) backing
zip.openpets.dev. The hard rule: never ship a catalog entry whose ZIP isn’t
live on R2. The verification commands HEAD-check every ZIP for exactly this
reason. Override the bucket with OPENPETS_R2_BUCKET; --skip-r2 is for local
testing only.
How the app uses all this
- Browsing: the Pets page in the Control Center pages through v3 and uses the search index for filtering.
- Installing: see the install flow in Pets - catalog lookup (which allows installing any valid v3 catalog pet by ID, even if not original or featured) → ZIP download → validated extraction → state update → tray refresh. (Control Center UI surfaces only curated original/featured pets, but explicit install by ID allows any valid v3 pet).
- Plugins: the Plugins page lists catalog v2 entries filtered by app version and install state; install downloads + verifies the plugin ZIP. See Plugin platform.
Related docs
- Official plugins lists the current plugin catalog lineup and bundling defaults.
- Plugin platform explains plugin packaging and publishing behavior.
- Testing and validation lists the catalog and plugin release gates.
