Crate prompty_anthropic 

Anthropic provider for Prompty.

This crate provides the Anthropic executor and processor implementations, registered under the key "anthropic".

Anthropic’s Messages API differs from OpenAI in several key ways:

• System messages are extracted to a top-level system field
• Messages always use content block arrays [{type: "text", text: "..."}]
• Tools use input_schema instead of parameters
• max_tokens is required (defaults to 4096)
• Auth uses x-api-key header (not Authorization: Bearer)
• Tool results are batched into one user message with tool_result blocks

Usage

prompty_anthropic::register();
// Now invoke/turn will use Anthropic for agents with provider="anthropic"

Re-exports

pub use executor::AnthropicExecutor;

pub use models::list_models;

pub use models::list_models_async;

pub use processor::AnthropicProcessor;

pub use processor::process_response;

Modules

executor

Anthropic executor — sends requests to the Anthropic Messages API.

models

Model discovery for Anthropic — calls GET /v1/models and maps the response to ModelInfo. Anthropic returns context_length and modality information directly, so no enrichment table is needed.

processor

Anthropic processor — extracts results from Anthropic Messages API responses.

wire

Wire format conversion for the Anthropic Messages API.

Functions

register

Register the Anthropic executor and processor in the global registry.