Hermes keeps reporting 429, usually not Hermes itself is broken, but the model provider you are connecting with triggered the current limit. Don't rush to change a bunch of configurations, the most effective order is: first confirm whether the provider is throttling, then change the model or provider, and finally turn on fallback.
First, determine which layer is limiting the flow
The official FAQ is straightforward: the reason for 429 is that you have exceeded the provider's rate limits. That said, the problem tends to be at the OpenRouter, Nous Portal, OpenAI, Anthropic, or natively compatible service layer, rather than Hermes' conversational layer. You can run hermes config show to see the current provider first, and then use hermes model to check the available models.
The most practical processing sequence
- Wait a while and try again, temporary peaks sometimes resume on their own.
- Switch to another provider or model to avoid continuing to hit the same throttling line.
- If you are a high-frequency user, turn on the fallback model to automatically switch to the alternate model when the main model fails.
Hermes' fallback is not a decorative function, but is specially prepared for this type of problem. The official documentation divides it into two layers: credential pools will rotate multiple keys from the same provider first; The primary model fallback cuts to the alternate provider:model when the primary model fails. This layer is useful for people who often run long sessions and call tools frequently.
When to switch providers
If you only hit 429 occasionally, wait a while; If you run into a throttling almost every day, it's time to switch to a more stable provider or upgrade your plan. Hermes supports OpenRouter, Nous Portal, OpenAI, Anthropic, z.ai, Kimi, MiniMax, and native model services. For beginners, OpenRouter and Nous Portal are often the easiest options to get through.
Summary in one sentence: Hermes reported 429, the core is not "repairing Hermes", but "changing providers, turning on fallbacks, and controlling the frequency of calls".