Crate stringflow 

Flow strings through language models.

Provider-agnostic LLM client supporting multiple wire formats:

• OpenAI Chat Completions (/v1/chat/completions)
• OpenAI Responses (/v1/responses)
• Anthropic Messages (/v1/messages)

A provider is just a base URL + wire format + auth config.

Module structure

• wire_formats/ — request/response types and parsing per format
• providers/ — provider configurations (llama, openai, anthropic, etc.)
• client — HTTP client (chat, streaming, health checks)

Re-exports

pub use providers::AuthConfig;

pub use providers::ProviderConfig;

Modules

providers

Provider configurations.

Structs

ChatMessage

A chat message (role + content), shared across all wire formats

HealthResponse

Health check response from /health

Enums

Error

StreamEvent

A streaming event from a chat response

WireFormat

Wire format for LLM API requests

Functions

chat

Send a blocking chat request. Retries on transient errors with exponential backoff + jitter.

chat_async

Send an async chat request. Retries on transient errors with exponential backoff + jitter.

chat_stream

Send an async streaming chat request. Returns a stream of events.

health_check

Send an async health check

health_check_blocking

Send a blocking health check