Crate inference_runtime_openai 

inference-runtime-openai

OpenAI Chat Completions runtime + Azure OpenAI variant. Doc §10.3.

Implements the inference_core::ModelRunner contract over HTTP/2 (via reqwest). SSE chunks are parsed by inference-remote-core::sse and lifted into provider-typed deltas by wire.

Re-exports

pub use config::OpenAiConfig;

pub use config::OpenAiVariant;

pub use cost::OpenAiPricing;

pub use error::classify_openai_error;

pub use runner::OpenAiRunner;

Modules

config

OpenAiConfig — parameters for the OpenAI / Azure OpenAI runtime.

cost

Per-million-token pricing for the OpenAI catalog. Values are public list pricing as of the doc snapshot; operators override via inference-cli flags.

error

OpenAI-specific error refinement. Layers on top of the generic classify_http_status from inference-remote-core to recognise provider-specific shapes that should not be retried (content filter refusals, context-length-exceeded).

runner

OpenAiRunner — ModelRunner impl for OpenAI Chat Completions / Azure OpenAI.

wire

OpenAI Chat Completions wire types — request envelope + SSE chunk shape. Kept minimal: only what’s needed to round-trip the prompts / deltas / usage info that the actor system needs.