Crate sutegi_http 

A minimal HTTP/1.1 server built directly on std::net.

No async runtime, no hyper, no tokio. Connections are handled by a fixed thread pool. This is deliberately small: it keeps the binary tiny and the request lifecycle trivial to reason about — for a human reading the source, or an agent reasoning about the running app.

Structs

Limits

Server resource limits — the difference between “demo” and “won’t fall over”.

Request

A parsed HTTP request.

Response

An HTTP response.

SseSink

A flushing sink that formats Server-Sent Events (text/event-stream). Each call emits one event frame and flushes — exactly what LLM token streaming wants.

StreamSink

A flushing sink for raw streamed bytes. Every write is flushed so the client sees data immediately.

ThreadPool

A fixed-size pool of worker threads pulling jobs off a shared channel.

Enums

Body

Incoming

The outcome of parsing: a request, or a refusal the server turns into 413.

Method

HTTP request methods sutegi recognizes.

Functions

parse_request

Parse a single request off a buffered stream, enforcing limits. Returns Ok(None) if the peer closed before sending anything, or Ok(Some(Incoming::TooLarge)) if headers/body exceed the limits (so the server can reply 413 without allocating an attacker-chosen buffer).

serve

Bind to addr and serve requests with handler until the process exits. handler is shared across worker threads, so it must be Send + Sync.

serve_until

Like serve, but stops accepting new connections once shutdown is set, then drains in-flight requests (by dropping the pool, which joins workers) and returns. This is what makes a sutegi process safe to roll in a pod: on SIGTERM you flip the flag, stop taking traffic, and let live requests finish within the termination grace period.

status_reason

Map a status code to its canonical reason phrase.

write_response

Write a response to the stream. Always closes the connection (no keep-alive) to keep the server stateless and simple. Takes the response by value so a streaming body’s FnOnce producer can be invoked.

Type Aliases

StreamProducer

A response body: either fully buffered, or streamed incrementally.