class

Smith::LLM::Anthropic

Inherits Smith::LLM::AnthropicCaching < Smith::LLM::Provider < Reference < Object

Constants

API_VERSION = "2023-06-01"

DEFAULT_ENDPOINT = "https://api.anthropic.com/v1/messages"

DEFAULT_MAX_TOKENS = 4096

DEFAULT_MODEL = "claude-sonnet-5"

Constructors

new(api_key : String = ENV.fetch("ANTHROPIC_API_KEY", ""), default_model : String = DEFAULT_MODEL, timeouts : Timeouts = Timeouts.default, cache : Bool = true)

Source

Instance methods

api_key

Source

cache?

Caching needs a minimum prefix length (1024 tokens on Sonnet/Opus, 2048 on Haiku); below that the 1.25x write surcharge buys nothing, hence the switch.

Source

complete(request : Request) : Response

Source

complete_streaming(request : Request, &on_delta : String -> Nil) : Response

Streaming entry point. The default does no streaming at all — it runs complete() and hands the text over in one piece.

That default is what makes deltas universal: callers always see them, whether or not the provider streams, so neither the agent loop nor the renderers need a special case. Providers override this; when streaming is switched off they call super.

Source

default_model

Source

name

Source

supports_documents?

The one provider that takes a PDF as a document block and reads it itself, pages and all.

Source

supports_tool_result_media?

A tool_result's content may be an array of blocks, images among them — which is what lets read_file hand back a screenshot.

Source