Serve
Overview

HTTP/1.1

Synchronous#

import serve
 
serve(8080, proc(req: Request): Response {.closure.} =
  ok("hello"))

One request handled at a time per connection, in-process — the simplest path, good for internal tools, CLIs and low-traffic services where async complexity isn't worth it.

Async, on the reactor#

import serve, serve/reactor
 
proc handler(req: Request): Response {.nimcall.} =
  response(200, "text/plain", "ok " & req.path & "\n")
 
serveHttpReactor(8080, handler)

For TLS:

serveHttpsReactor(8443, tlsContext, handler)

See the repo's reactor_http.nim and reactor_https.nim for complete, runnable versions.

What's included#

Keep-alive, request-size caps returning 413, streamed (non-truncated) responses, and the slowloris guard all apply identically whether you're on the sync or async path — they live in the shared http/tcp layer underneath both.

Routing#

serve doesn't impose a router, but the repo ships a routing example — reactor_router.nim — showing the router tested across all three protocols (HTTP/1.1, HTTP/2, HTTP/3) from one route table.

Powered by Docsbook