Configuration

Set the properties on the HTTPServer node before you call start().

Server properties

Property	Default	Description
`max_connections`	100	Maximum concurrent connections
`max_request_size`	10 MB	Maximum request body size
`idle_timeout`	30 s	Close idle connections after this time
`handshake_timeout`	10 s	TLS handshake timeout
`max_keep_alive_requests`	100	Maximum requests per keep-alive connection
`poll_interval`	0.05 s	Connection polling frequency
`bind_address`	`127.0.0.1`	IP address to bind. Use `0.0.0.0` for all interfaces
`drain_timeout`	5 s	Wait for in-flight requests during `shutdown()`

server.max_connections = 500
server.idle_timeout = 60.0
server.bind_address = "0.0.0.0"   # expose beyond localhost

Enable HTTPS

Call setup_tls() before start(). Pass a PEM certificate and a PEM key:

func _ready() -> void:
    var err := server.setup_tls("res://cert.pem", "res://key.pem")
    if err == OK:
        server.start(443)

The method returns an Error code when it fails to load the certificate or the key.

Lifecycle signals

The server emits these signals:

Signal	Emitted when
`server_started(port)`	The server starts listening
`server_stopped`	The server stops
`client_connected(ip)`	A client connects
`client_disconnected(ip)`	A client disconnects
`request_handled(request, response)`	A request completes
`websocket_connected(connection)`	A WebSocket upgrades
`websocket_disconnected(connection, code, reason)`	A WebSocket closes

Graceful shutdown

Call stop() to close all connections and stop listening. The server also calls stop() when the node leaves the scene tree:

func _exit_tree() -> void:
    server.stop()

Use shutdown() to stop accepting new connections and wait for in-flight requests to finish before closing. It waits up to drain_timeout seconds, then force-closes:

server.shutdown()        # waits drain_timeout (5 s by default)
server.shutdown(20.0)    # custom wait

Protocol defaults

The server sets a Server and Date header on every response. Requests that exceed max_request_size return 413 Payload Too Large instead of dropping the socket. Stalled requests return 408 Request Timeout. When the connection pool is full, new peers receive 503 Service Unavailable. The parser honours Expect: 100-continue and sends the interim 100 Continue once per request.

Chunked responses

Set is_chunked on the response to stream data, for example for server-sent events. The server then sends Transfer-Encoding: chunked instead of Content-Length.