How chainapsis/keplr-app-registry works
This repository is not a runtime service; it’s a content + validation + build pipeline for the Keplr Dashboard “Explore” app registry.
Primary purpose inferred from README + tree:
• Repo provides TypeScript CLI/scripts to:
• validate each app’s metadata (app.json)
• enforce ordering rules via sort-config.json
• generate the canonical output artifact: app-list.json
• CI (GitHub Actions) runs these checks on push/PR.
So your “system” is mainly a repository-driven data pipeline: Inputs (PR files) → validation → generation → published artifact → consumed by Keplr Dashboard.
“Registry data” (source of truth)
• apps/**/app.json — per-app metadata
• sort-config.json — pinned ordering configuration (managed by Keplr team)
• app-list.json — generated output (explicitly “DON’T UPDATE directly”)
“Build/validation toolchain” (execution logic)
• src/make-app-list.ts — reads apps + sort config, outputs app-list.json
• src/validate-app.ts — validates app metadata constraints (category enum, URLs, required fields, etc.)
• src/validate-sort-config.ts — validates pinning config format/contents
“Automation / CI”
• .github/workflows/pull-request.yml
• .github/workflows/push.yml
“Runtime environment / packaging”
• package.json — scripts, dependencies, entry points
• tsconfig.json — TypeScript build/runtime settings
• Yarn Berry / PnP:
• .yarnrc.yml, .pnp.cjs, .pnp.loader.mjs, .yarn/**
What the README adds (important relationships & rules to show)
• Schema rules for app.json:
• appCategory must be one of a fixed list (case-sensitive)
• externalUrls requires at least one of twitter/github/discord
• optional appWebsiteUrlsByChainId
• app-list.json ordering algorithm:
1) pinned apps first (from sort-config.json)
2) all others alphabetically by appName
• app-list.json is generated, not manually edited
• Downstream consumer: Keplr Dashboard Explore page
For each apps/<app>/
• read app.json → validate schema/fields
Aggregate apps → apply ordering logic → write app-list.json
Make the per-app loop explicit (e.g., a “For each app folder” annotation), because that’s a core architectural behavior.
Repository data components
• apps/<app>/app.json (metadata)
• sort-config.json (pin list)
• app-list.json (generated registry output)
Tooling components (Node/TS)
• Node.js runtime
• Yarn 3 (Berry) with Plug’n’Play (PnP) resolver
Pipeline components
• GitHub Actions workflow: PR validation (pull-request.yml)
• GitHub Actions workflow: push/build (push.yml)
Downstream consumer
• Keplr Dashboard Explore page feature/module: “App directory renderer”
• “Adds/updates app submission” (Submitter → Repo)
• “Triggers CI” (PR/push → GitHub Actions)
• “Validates JSON schema & required fields” (CI → validate-app.ts)
• “Validates pinning config” (CI → validate-sort-config.ts)
• “Generates ordered list (pinned + alpha)” (CI → make-app-list.ts)
• “Produces artifact: app-list.json” (Toolchain → Repo output)
• “Consumes registry to render Explore page” (Keplr Dashboard → app-list.json)
Architectural patterns/design principles present (call these out
These are worth labeling because they explain “why it’s shaped this way”:
• Repository-as-database: Git stores the canonical registry entries.
• Build-time generation: app-list.json is a derived artifact to simplify downstream consumption.
• CI as gatekeeper: PR checks prevent invalid entries from merging.
GitHub Repo (main branch)
• contains: apps/, sort-config.json, generated app-list.json
Consumer: Keplr Dashboard Explore
And show:
• apps/** feeding both validation steps and generation
• sort-config.json feeding sort validation + generation
• app-list.json feeding Keplr Dashboard
This captures the essential architecture with maximal fidelity to the repository structure.