@memberjunction/connector-propfuel

MemberJunction Integrations

Vendor integration connectors for MemberJunction, each a self-contained mini Open App — its own npm package, schema, metadata-seed migrations (SQL Server + Postgres), and independent release. Every connector lives under <Category>/<Connector>/ and installs on demand:

mj app install https://github.com/MemberJunction/Integrations/CRM/HubSpot

Nothing is seeded into a MemberJunction database until you install the connector you want, and installing one connector pulls only that connector's package — connectors are fully decoupled.

How it works (full Open App model)

Each <Category>/<Connector>/ is a real Open App, exactly like bizapps-common:

• Its own package @memberjunction/connector-<name> — the @RegisterClass(BaseIntegrationConnector, …) class + tests. Framework packages are peer dependencies (the host supplies them).
• Its own schema mj_connector_<name> — the app's namespace + flyway_schema_history anchor. It holds no tables of its own; it only exists so the Open App migration runner runs the connector's migrations.
• Metadata-seed migrations — migrations/ (SQL Server) + migrations-pg/ (Postgres). The migration body runs EXEC __mj.spCreateIntegrationObject … to seed the core integration catalog (MJ: Integrations / Integration Objects / Integration Object Fields + MJ: Actions). Installing the app runs the standard path: create schema → run migrations.
• metadata/ is the human-editable source of truth; the migrations are generated from it.

The framework (@memberjunction/integration-engine, the Integration/IO/IOF tables, runtime discovery, credential types, the bizapps Action-Category tree) stays in core MJ. The three-way invariant holds per connector (Integration.ClassName == @RegisterClass driver == IntegrationName), enforced by scripts/validate-invariants.mjs.

Repository layout

Integrations/
├── CRM/  AMS/  LMS/  Marketing/  Finance/  Events/  Platform/      one mini Open App per connector
│   └── <Connector>/
│       ├── mj-app.json          schema: mj_connector_<name>, migrations: skyway, metadata, bootstrap package
│       ├── package.json  src/   @memberjunction/connector-<name> + tests
│       ├── metadata/            mj-sync source of truth (integration/ + actions/) — author here
│       ├── migrations/          GENERATED SS seed migrations (from `mj sync push`)
│       └── migrations-pg/       GENERATED PG variants (from `mj migrate convert`)
├── scripts/                     new-connector, wrap-migration, build-pg-migrations, validate-*, sync-manifest-versions
├── .github/workflows/           pr.yml (build/test/lint/validate-migrations/changeset) · release.yml (changesets publish)
├── .changeset/                  non-fixed → per-connector independent versions
└── turbo.json

RelationalDB / FileFeed are framework-generic primitives with no vendor catalog → code-only packages (no mj-app.json). Workspaces glob: ["*/*"].

Creating / maintaining a connector

You are only involved in migration creation; CI does the rest.

# 1. Scaffold (one command → buildable, lint-passing skeleton)
npm run new-connector -- CRM Acme --name "Acme CRM"

# 2. Implement CRM/Acme/src/AcmeConnector.ts and author CRM/Acme/metadata/ (the catalog + actions)

# 3. Migration creation (needs a dev MJ DB) — capture the seed SQL, then Flyway-name it
cd CRM/Acme && mj sync push --dir metadata           # writes raw seed SQL into ./migrations
cd ../.. && npm run migrations:wrap -- CRM/Acme       # → migrations/V<ts>__acme__Metadata.sql
npm run migrations:pg -- CRM/Acme                     # → migrations-pg/…pg.sql  (deterministic, no DB)

# 4. Add a changeset (drives this connector's independent version bump) and open a PR to `next`
npm run changeset

pr.yml then builds/tests the changed connector, runs the invariant + migration lint, checks PG parity, and requires the changeset. On merge of a release, release.yml (changesets) publishes only the changed connector packages, tags each <pkg>@<version>, and syncs each mj-app.json version/mjVersionRange.

Versioning & install

• Per-connector independent versions (changesets non-fixed) — a Salesforce fix never bumps HubSpot.
• mj app install …/CRM/HubSpot --version X.Y.Z resolves the connector's subpath-scoped git tag CRM-HubSpot@X.Y.Z (created by scripts/tag-apps.mjs in the release flow; the install derives the tag prefix from the subpath, so no naming guesswork).
• Connectors declare mjVersionRange (>=5.43.0 <6.0.0) and depend on the published MJ framework as peer deps.

Requirements

Requires a MemberJunction host ≥ 5.43.0 (multi-app-per-repo subpath installs in @memberjunction/open-app-engine).

Development

npm install
npm run build            # turbo build of every connector package
npm test                 # vitest unit tests
npm run lint             # invariants + migrations

Notes

Connector Actions @lookup the bizapps Action-Category tree and credential types, both core-seeded MJ metadata with stable IDs (so the baked migration SQL resolves on any install).