You'll noticeTier: how much it should matter to you
2Useful: my rating, 1 to 5
1Signal: worth watching, 1 to 5
LLM GatewayArea: what it touches
ImprovementsKind: in v2.1.295,
ImprovementsSection of the release
What
When Bedrock rejects a request because a model ID cannot be invoked with on-demand throughput, the apps gateway now logs that the model runs only through an inference profile and advises mapping it to one. An inference profile is Bedrock's way of routing a model that cannot be called directly. The response headers from Bedrock's original error are kept.
Why
This is a common Bedrock misconfiguration, and the log now says how to fix it.
