A Go service built to be operated from day one

The challenge

The team needed an integration service quickly but had been burned before by contract work that arrived as a bare repository — no deployment story, no observability, and no way for the in-house team to operate it without a rewrite.

What we did

1. 1

   Contract first

   Agreed the API surface and error semantics up front so the client's frontend team could build against a stub while the service was implemented.

2. 2

   Operations built in, not bolted on

   Structured logging, Prometheus metrics, health and readiness endpoints, and graceful shutdown were part of the first working version rather than a later milestone.

3. 3

   Tested the failure paths

   Focused test effort on timeout, retry, and partial-failure behaviour against the third-party APIs, which is where integration services actually break.

4. 4

   Shipped it, not just wrote it

   Delivered the containerised build and the pipeline that deploys it, so handover was a walkthrough rather than a migration.