FDE ForgeBeta
Start free
Week 5 · Enterprise AI

Have a fallback provider

0:00 / 3:00

15 seconds

Provider status: red.

The only model Helix called is offline. The copilot page shows a spinner. Ward rounds do not wait.

Provider APrimaryProvider BFallbackGatewayFail overClinicianStill answers
Lesson notesHave a fallback provider: the short version+

Week 5 · LLM gateways · about 3 minutes

What you will be able to do

Keep Helix answering when the primary LLM provider is down by failing over to a second provider.

The idea

One provider outage should not stop the customer. Keep a second provider (or a self-hosted model) wired with the same output schema.

Failover is not a nice-to-have for clinical workflows — it is part of the MVV.

def complete(prompt: str, primary, secondary):
    try:
        return primary.complete(prompt)
    except ProviderDown:
        return secondary.complete(prompt)

If both providers share one account and one region, you do not have a fallback.

Key takeaways

  • Wire a real secondary path.
  • Same output schema on both.
  • Log failovers for the postmortem.

Part of the Forward Deployed Engineer roadmap. New to the role? Start with what a Forward Deployed Engineer does.