Request-reply, events, batch, middleware — pick the integration pattern before the tool, and design for the failure first.
What you'll learn
The four patterns that cover nearly every Salesforce integration: synchronous request-reply behind named credentials, fire-and-forget with platform events, nightly batch sync on external IDs, and when the middleware question actually becomes a middleware answer.
Prerequisites
Comfortable Apex (the Apex Triggers module or equivalent) and a basic grasp of REST. The video section builds a callout live if you have never written one.
How the sections chain
Four written sections plus a build-along video: synchronous first because it is the default everyone reaches for, then events for when the caller should not wait, then batch for when nobody is waiting at all — closing with the architecture question that decides how the next integration gets built.
What you'll be able to do
- Stand up a named-credential callout with a tested mock
- Design idempotent subscribers for at-least-once delivery
- Build reruns-are-safe batch loads on external IDs
- Argue middleware vs point-to-point with edge counts
- Choose a pattern from latency and ownership, not fashion
The sections
- 01 · Request-reply with named credentials The synchronous callout, done properly: named credentials for auth, a timeout you chose on purpose, and a mock for the test. 15m
- 02 · Fire-and-forget with platform events When the caller does not need an answer, publish an event and move on — decoupling, replay, and the delivery guarantees you actually get. 15m
- 03 · Batch sync without tears Nightly loads still run the world. Upsert on external IDs, order parents before children, and make reruns safe. 15m
- 04 · Middleware or point-to-point? Two systems talking directly is simple right up until it is three systems, then five. Where the line actually sits. 10m
- 05 · Watch: a callout built end to end Twenty minutes from a blank class to a tested, mocked, named-credential callout — narrated, mistakes left in. 20m