struct

Anthropic::CacheControl

Inherits JSON::Serializable < Struct < Value < Object

Cache control configuration for prompt caching

Prompt caching reduces costs by up to 90% and latency by up to 85% for prompts with large amounts of repeated content.

# Default 5-minute ephemeral cache
cache = Anthropic::CacheControl.ephemeral

# Extended 1-hour cache (requires beta header)
cache = Anthropic::CacheControl.one_hour

# Use in content
content = Anthropic::TextContent.new(
  text: "Large context...",
  cache_control: CacheControl.ephemeral
)

Pricing:

  • Cache writes (5-min): 1.25x base input token price
  • Cache writes (1-hour): 2x base input token price
  • Cache reads: 0.1x base input token price

Constructors

ephemeral

Create ephemeral cache control (5-minute TTL, default)

Standard caching with 5-minute duration. Cache writes cost 1.25x base price.

Source
new(pull : JSON::PullParser)
Source
new(type : String = "ephemeral", ttl : Int32 | Nil = nil)
Source
new(*, __pull_for_json_serializable pull : JSON::PullParser)
Source
one_hour

Create extended cache control (1-hour TTL)

Extended caching for longer context retention. Cache writes cost 2x base price. Requires the beta header: extended-cache-ttl-2025-04-11

# Use via beta.messages.create with the beta header
client.beta.messages.create(
  betas: [Anthropic::EXTENDED_CACHE_TTL_BETA],
  model: ...,
  messages: [...]
)
Source
with_ttl(seconds : Int32) : self

Create cache control with custom TTL in seconds

Note: Custom TTLs may require specific beta headers.

Source

Instance methods

type
Source