Module llm 

LLM clients for the openai and anthropic wire protocols.

Re-exports

pub use crate::retry::RetryPolicy;

pub use crate::retry::Retryable;

pub use crate::retry::with_retry;

Structs

AnthropicClient

Client for the anthropic protocol (/messages).

Completion

A parsed completion.

CompletionRequest

A single-turn completion request. The model is bound to the client.

OpenAiClient

Client for the openai protocol (/chat/completions, /embeddings).

Reasoning

Reasoning extracted from a response.

Usage

Token counts reported by the provider.

Enums

LlmError

Errors from an LLM provider.

ProtocolClient

A client for one of the two wire protocols.

SetupError

Errors while building a client from the configuration.

Traits

LlmClient

A chat model bound to one provider and one model ID.

Functions

connect

Builds the client for role: resolves the provider’s API key (literal or Vault reference) and uses pipeline.request_timeout_secs as request timeout. A provider without api_key gets no auth header, which suits local servers.