tt-provider-compat — the OpenAI-wire-compatible adapter machinery.
Many inference providers (OpenAI itself, plus Mistral, Groq, Together AI,
OpenRouter) speak the OpenAI POST /chat/completions + /embeddings wire
format. This crate holds the shared plumbing — request/response translation,
SSE streaming, error mapping, the HTTP client config, and the generic
[OpenAICompatibleProvider] — so each adapter crate is a thin
configuration over it.
tt-provider-openai builds its native adapter on top of this crate (adding
the OpenAI pricing catalog); the four BYOK adapters depend only on this
crate, not on the full OpenAI adapter.