Prompt caching for foundational models

[edmundhighcock]

Hi everyone,

I'm thinking about implementing prompt caching when query e.g. Bedrock, to reduce cost by using cached tokens.

Is anyone working on this? Is this feature already available and I have just missed it?