basic-custom-app-build（基本的自定义应用程序构建）

Domo Basic App Build Playbook

When to use

• Use at project kickoff for a brand-new Domo custom app.
• Use when taking over an existing app and normalizing it to platform best practices.

Progress checklist

Copy this checklist and update it as you work

Build Progress:
- [ ] Phase 0: Rules loaded
- [ ] Phase 1: Manifest & contracts
- [ ] Phase 2: App shell (domo.js)
- [ ] Phase 3: Data access
- [ ] Phase 4: App storage (if needed)
- [ ] Phase 5: Toolkit clients (if needed)
- [ ] Phase 6: Feature skills (AI / Code Engine / Workflow — as needed)
- [ ] Phase 7: Performance review
- [ ] Phase 8: Build & publish
- [ ] Phase 9: Verification

Phase 0 — Load rules (always-on)

Apply before writing any code:

• rules/core-custom-apps-rule.md
• rules/custom-app-gotchas.md

Phase 1 — Manifest & contracts

Use manifest.

Define all external resource mappings first — datasets, collections, workflows, Code Engine packages. Everything else depends on this. Use alias-safe names only (^[A-Za-z][A-Za-z0-9]*$), no spaces or special characters.

In addition to creation of the manifest.json, check root folder for an existing thumbnail.png to copy into the public folder of the new app.

Existing-app takeover? Audit the current manifest.json against actual code usage before changing anything.

Phase 2 — App shell

Use domo-js.

Set up the baseline: ryuu.js import, navigation via domo.navigate(), event listeners, environment info.

Advanced users using DA CLI? Ask the agent to also use da-cli for scaffolding and manifest instance workflows. Should not be used unless user explicitly asks agent to use it.

Phase 3 — Data access (if domo datasets need to be queried)

Use dataset-query (primary) and data-api (routing overview).

Build queries with @domoinc/query. Use the Query API for all dataset reads — it respects page filters and does server-side aggregation.

Before writing UI/data-mapping logic, fetch the real schema for every dataset in manifest.json datasetsMapping and create an explicit field map (date field, primary metric(s), dimension fields). Do not rely on guessed names like sales, category, or order date unless confirmed in schema.

If required fields are missing for the requested visualization, fail fast with a clear mapping error instead of rendering zeros.

Need raw SQL? Use sql-query, but know that SQL ignores page filters.

Phase 4 — App storage (if appdb, or any user data entry is needed)

Use appdb and appdb-collection-create when storage must be created.

Skip if the app only reads datasets. Use AppDB when you need to persist app-specific state, user preferences, or document-style data.

If required collections do not exist yet, run appdb-collection-create first to provision datastore+collection before wiring document CRUD.

Phase 5 — Toolkit clients (if appdb, domo workflows, or domo sql query is needed)

Use toolkit.

Move to typed @domoinc/toolkit clients where they add value (structured responses, type safety). Not required for simple apps.

Phase 6 — Feature skills (as needed)

Only load the skills your app actually requires (3 examples are listed here but you have access to many more skills):

Decision guide: If the user hasn't mentioned AI, Code Engine, or Workflows, skip this phase entirely. Don't add complexity the app doesn't need.

Non-optional trigger for server functions: If app functionality requires server-side behavior (secrets, external API calls, privileged transformations), create or update a Code Engine package first, then sync manifest.json packagesMapping, and only then wire runtime invocation in app code.

Phase 7 — Performance review

Use performance.

Review all queries before finalizing. Check for full-dataset fetches, missing aggregations, and unnecessary columns.

Phase 8 — Build & publish

Use publish.

npm run build → cd dist → domo publish. On first publish, copy the generated id back to your source manifest.

Phase 9 — Verification

After publishing, confirm:

• App loads without console errors in Domo
• All dataset aliases resolve (no 404s on data calls)
• AppDB collections are wired in the card UI (if used)
• Page filters propagate correctly (if app is embedded in a dashboard)
• Navigation uses domo.navigate(), not <a href>
• Thumbnail has been copied into public folder

Build-time guardrails

• Client-side only: no SSR/server routes/server components.
• Use Vite base: './'.
• Prefer HashRouter unless rewrites are intentionally handled.
• Treat domo.env.* as convenience only; use verified identity for trust decisions.