← sources

How do you handle agents failover?

source: reddit (reddit:AI_Agents)published 2026-09-27 06:03ingested 2026-09-28 02:36

retained text

We had a single-provider setup, and when that provider degraded, every agent run failed 

Now I'm looking at automatic fallback, so that if model A errors or times out, the request retries on an equivalent model from another provider. My questions are

1. Do you fall back to the same model on another host, or to a different model altogether? Prompts don't always carry over cleanly

2. Did you build this yourself, or are you using a gateway? I've seen LiteLLM, Portkey, and llmapi.ai all advertise fallbacks

3. How do you stop fallbacks from quietly running up costs (for example, landing on a pricier model)?

Thanks in advance!

document fetched 2026-09-28 00:19 · canonical https://reddit.com/r/AI_Agents/comments/1wrcefk/how_do_you_handle_agents_failover