Skip to content

Provider Strategy

Provider-specific code should remain isolated behind stable product contracts so the application can adapt to provider changes without rewriting user-facing flows.

The provider layer is a production risk area because model APIs, latency, disclosure requirements, and output behavior can shift.

  • Keep provider details behind a narrow service boundary.
  • Normalize provider outputs before they reach product routes.
  • Track latency and timeout behavior separately from product success.
  • Make queueing a first-class architecture concern before payment hardening.

This decision connects directly to the Vercel timeout and queue-required risks.