Every message in Altrum says what failed and what to do. This page lists the ones you are most likely to see, word for word, so you can search for them. Each answer also carries a request reference: quote it when you contact support.
Signing in and access
Members
AI systems
Assessments
Guardrails
AI Gateway
This is in development. Details may change before it is available.
When your application calls the AI Gateway and it refuses or cannot complete a request, the error comes back in the format of the API you called, so your SDK reads it as it reads any other error: OpenAI style on the OpenAI routes, Anthropic style on the Anthropic routes, Google style on the Google routes. Every error carries a request reference, in the x-request-id header and in the body where the format allows, and a code that stays the same between releases.
A guardrail error names the guardrails that fired and never your text, and says “this AI system” or “this organization” for whoever’s guardrails applied. For example:
An error that comes from your own model provider keeps the provider’s message.
Something else
If a message is not here, it still says what to do. If it asks you to contact support, raise a ticket with the request reference: see Support.