PRODUCTION SCENARIO
A developer sends a 12,000-token API documentation reference on every request and wants to cache it with the Anthropic SDK.
How is the cache breakpoint declared in the request?
Answering here is anonymous. Nothing is saved unless you sign in.
Show answer and explanation
Answer: Attach a cache_control block of type ephemeral to that content block
Caching is requested by attaching cache_control with type ephemeral to a content block in tools, system or messages, up to four breakpoints per request. An HTTP Cache-Control header governs intermediary caching of the HTTP response and never reaches the tokenized prefix. The request-level control that does exist is a top-level cache_control field for automatic caching, not a cache boolean, so that option's name and shape are not real. And the Files API avoids re-uploading bytes, but the tokens are still processed on every request unless a breakpoint marks them.