Reference

Reference: site file

aliases

Further domains served by the same site.

cache

Response cache for the site; absent: no caching.

cache.bypass_cookies

Cookie-name prefixes that mark a personalised request (added to the profile’s).

cache.bypass_paths

Path prefixes that are never cached (added to the profile’s).

cache.bypass_query

Query parameters that disable caching (added to the profile’s).

cache.default_ttl

Freshness for 200 responses that carry no caching headers at all. Default: none (only explicitly cacheable responses are stored).

cache.enabled

Cache this application’s responses.

cache.stale_while_revalidate

Serve stale entries this long while refreshing in the background.

domain

The site’s main domain.

htaccess

Read Apache .htaccess files below root (ADR-0031).

https_redirect

Redirect plain HTTP to HTTPS (the www. redirect follows it).

name

Application name; default derived from domain.

node

Node.js: the command as a string or list, or the Node.js settings.

node.command

Command started by scalws (no shell), e.g. [node, server.js]; the application listens on PORT.

node.env

Environment variables for the process.

node.health_path

Optional readiness path; any non-5xx answer counts as ready.

node.listen

unix: a socket under runtime_dir passed as PORT; tcp: a free loopback port.

node.response_timeout

Time to wait for the response headers of the application.

node.startup_timeout

Time the process gets to become ready.

node.stop_timeout

Time the process gets to exit after SIGTERM before it is killed.

node.upstream

Proxy to a Node.js process managed elsewhere instead of starting one.

node.workers

Number of processes (min); requests are balanced across ready ones.

node.workers.autoscale

off, recommend (default) or auto (ADR-0014).

node.workers.max

Upper bound when the pool is scaled.

node.workers.min

Processes started at the beginning and kept at least.

node.workers.target_concurrency

Concurrent requests one worker is expected to handle well (utilisation basis).

php

PHP: true (a managed PHP-FPM pool), the address of an existing pool (unix:/run/php/php8.3-fpm.sock) or the PHP settings.

php.fastcgi

External pool: unix:/path.sock or tcp:host:port. Mutually exclusive with fpm.

php.fpm

Managed PHP-FPM pool settings (workers, ini values); not used with fastcgi.

php.fpm.binary

PHP-FPM binary; default: php-fpm on PATH, else the newest versioned one (php-fpm8.4, Alpine’s php-fpm84).

php.fpm.env

Environment passed to PHP (env[...]).

php.fpm.ini

php_admin_value[...] settings, e.g. memory_limit: 256M.

php.fpm.max_requests

pm.max_requests; 0 disables recycling.

php.fpm.workers

min = pm.start_servers/pm.min_spare_servers, max = pm.max_children.

php.fpm.workers.autoscale

off, recommend (default) or auto (ADR-0014).

php.fpm.workers.max

Upper bound when the pool is scaled.

php.fpm.workers.min

Processes started at the beginning and kept at least.

php.fpm.workers.target_concurrency

Concurrent requests one worker is expected to handle well (utilisation basis).

php.front_controller

Script that handles requests for paths that do not exist, e.g. /index.php.

php.index

Index files for directory requests, tried in order.

php.response_timeout

Time to wait for PHP-FPM’s response headers.

php.root

Document root; defaults to the application root.

php.startup_timeout

Time the managed PHP-FPM pool gets to become ready.

php.stop_timeout

Time the managed pool gets to exit before it is killed.

profile

Application profile, e.g. wordpress (deny rules, cache bypass, defaults).

proxy

Reverse proxy: the upstream URL or the proxy settings.

proxy.backends

Weighted split per request across pools (alternative to upstream).

proxy.backends[].pool

Name of a pool in upstreams; alternative to status.

proxy.backends[].status

Answers this share of the requests with this status (500-599) instead of sending them to a pool.

proxy.backends[].weight

Share of requests (0..=1000); 0 sends none.

proxy.client

HTTP/1 client used towards this upstream; unset: server.upstream_client.

proxy.connect_timeout

Time to establish a connection to the upstream.

proxy.preserve_host

Forward the client’s Host header (default) instead of the upstream authority.

proxy.response_timeout

Time allowed for the upstream to return response headers.

proxy.tls_ca_file

Extra CA bundle (PEM) trusted for https:// upstreams, in addition to the system roots.

proxy.upstream

http://host:port, https://host:port or unix:/absolute/path.sock (one address, no retries), or {pool: name}: a pool from upstreams.

proxy.upstream.pool

Name of a pool in upstreams.

proxy.websocket

Allow WebSocket upgrades to be tunnelled to this upstream.

python

Python: module:app (ASGI through uvicorn) or the Python settings.

python.app

module:callable, e.g. app:app or mysite.wsgi:application.

python.command

Explicit argv. Bind arguments are appended for uvicorn/hypercorn/gunicorn unless the argv already contains {socket} or {port}.

python.env

Environment variables for the processes.

python.health_path

Path requested to decide that a worker is ready and healthy.

python.listen

unix: a socket under runtime_dir; tcp: a free loopback port.

python.mode

asgi (uvicorn) or wsgi (gunicorn).

python.process_pool

scalws runs workers.min..max single-worker server processes itself instead of one server with --workers, so the pool can be scaled and restarted live (ADR-0014). Trades the server’s copy-on-write sharing for scaling.

python.response_timeout

Time to wait for the response headers of the application.

python.server

Default: uvicorn for ASGI, gunicorn for WSGI.

python.startup_timeout

Time the processes get to become ready.

python.stop_timeout

Time the processes get to exit after SIGTERM before they are killed.

python.upstream

Proxy to a Python server managed elsewhere instead of starting one.

python.virtualenv

Virtual environment whose server binary and packages are used.

python.workers

Worker processes of the application server (min).

python.workers.autoscale

off, recommend (default) or auto (ADR-0014).

python.workers.max

Upper bound when the pool is scaled.

python.workers.min

Processes started at the beginning and kept at least.

python.workers.target_concurrency

Concurrent requests one worker is expected to handle well (utilisation basis).

redirect

Redirect: the target origin or the redirect settings.

redirect.host

Replaces the request’s host; unset: taken from the request.

redirect.keep_path

Append the request’s path and query to to.

redirect.path

Replaces the request’s path; unset: taken from the request.

redirect.port

Replaces the request’s port; unset: the well-known port of a configured scheme, else taken from the request.

redirect.prefix

Replaces the matched prefix of the request’s path; unset: taken from the request.

redirect.scheme

Replaces the request’s scheme; unset: taken from the request.

redirect.status

301, 302, 303, 307 or 308.

redirect.to

Target origin, e.g. https://example.com: scheme and host (and port), no path when keep_path is on. Alternative to the fields scheme, host, port, path and prefix.

resources

Tenant resource limits; set them in one site file per tenant.

resources.burst

Token-bucket burst for requests_per_second (default: equal to the rate).

resources.cpu

cpu.max, e.g. "400%" = four CPUs.

resources.io_bandwidth

io.max: read and write bytes per second each, on every disk holding the tenant’s application roots, e.g. "20MiB".

resources.io_weight

io.weight (1-10000, default 100), best effort: needs a weight-capable I/O scheduler.

resources.max_concurrent_requests

Requests of this tenant processed concurrently; excess gets 503.

resources.memory

memory.max.

resources.pids

pids.max.

resources.requests_per_second

Sustained request rate; excess gets 429.

resources.swap

memory.swap.max when memory is set; default 0 (no swap).

root

Document root: static files, PHP scripts or the application’s directory.

root_symlinks

Symbolic links in the path of root: follow (default) or refuse (the root must be canonical and is opened without following links; for hosting panels).

routes

Path-prefix routes with their own runtime.

routes[].allow

Client addresses and CIDR blocks allowed; if set, only these clients (403 else).

routes[].basic_auth

HTTP basic authentication against an htpasswd file.

routes[].basic_auth.realm

Realm shown by the browser (default Restricted).

routes[].basic_auth.users_file

Absolute path of an htpasswd file (bcrypt, SHA-256/512 crypt or Apache MD5 entries).

routes[].cors

Cross-origin resource sharing; preflight requests are answered by ScalWS.

routes[].cors.allow_credentials

Allow credentials (cookies, authorization); not with allow_origins: ["*"].

routes[].cors.allow_headers

Request headers allowed in preflight answers (ingress-nginx defaults).

routes[].cors.allow_methods

Methods allowed in preflight answers (default: GET, PUT, POST, DELETE, PATCH, OPTIONS).

routes[].cors.allow_origins

Allowed origins, e.g. https://app.example.com, https://*.example.com, or ["*"].

routes[].cors.expose_headers

Response headers exposed to the page.

routes[].cors.max_age

How long a preflight answer may be cached, at most 365 days (default 20 days).

routes[].deny

Client addresses and CIDR blocks refused with 403; checked after allow.

routes[].headers

Request headers that must match (all of them).

routes[].headers[].name

Header name (case-insensitive) or query parameter name.

routes[].headers[].regex

Regular expression the value must match (searched anywhere unless anchored).

routes[].headers[].value

Exact, case-sensitive value.

routes[].https_redirect

Answer plain-HTTP requests that select this route with a 301 to HTTPS; unset: the application’s https_redirect.

routes[].max_body

Largest request body; overrides limits.max_request_body_bytes; 413 above.

routes[].methods

Request methods this route matches, e.g. [GET, POST]; empty: any.

routes[].path

Path of the route; how it is compared depends on path_type.

routes[].path_type

prefix (default, whole segments, longest first), exact or regex (anchored at the start, in configuration order).

routes[].query

Query parameters that must match (all of them, decoded, first occurrence).

routes[].query[].name

Header name (case-insensitive) or query parameter name.

routes[].query[].regex

Regular expression the value must match (searched anywhere unless anchored).

routes[].query[].value

Exact, case-sensitive value.

routes[].request_headers

Changes to the request headers (after the built-in proxy headers).

routes[].request_headers.add

Headers added next to any present.

routes[].request_headers.remove

Headers removed.

routes[].request_headers.set

Headers set to this value, replacing any present.

routes[].request_timeout

Time from the route match until the response head; 504 when exceeded.

routes[].response_headers

Changes to every response of the route, including errors generated by ScalWS.

routes[].response_headers.add

Headers added next to any present.

routes[].response_headers.remove

Headers removed.

routes[].response_headers.set

Headers set to this value, replacing any present.

routes[].rewrite

Path and Host rewrite before the runtime.

routes[].rewrite.host

Host sent upstream (implies preserve_host: false), host or host:port.

routes[].rewrite.path

Full path; $1..$9 from a regex route, $0 the match, $$ a literal $; ?... in it sets the query (else the query is kept).

routes[].rewrite.prefix

Replaces the matched prefix (prefix routes only).

routes[].runtime

Runtime that handles requests under the path.

routes[].runtime<static>.index

File names served for a directory request, tried in order.

routes[].runtime<static>.precompressed

Serve <file>.br / <file>.gz siblings to clients that accept them.

routes[].runtime<static>.root

Document root. Defaults to the application’s root.

routes[].runtime<proxy>.backends

Weighted split per request across pools (alternative to upstream).

routes[].runtime<proxy>.backends[].pool

Name of a pool in upstreams; alternative to status.

routes[].runtime<proxy>.backends[].status

Answers this share of the requests with this status (500-599) instead of sending them to a pool.

routes[].runtime<proxy>.backends[].weight

Share of requests (0..=1000); 0 sends none.

routes[].runtime<proxy>.client

HTTP/1 client used towards this upstream; unset: server.upstream_client.

routes[].runtime<proxy>.connect_timeout

Time to establish a connection to the upstream.

routes[].runtime<proxy>.preserve_host

Forward the client’s Host header (default) instead of the upstream authority.

routes[].runtime<proxy>.response_timeout

Time allowed for the upstream to return response headers.

routes[].runtime<proxy>.tls_ca_file

Extra CA bundle (PEM) trusted for https:// upstreams, in addition to the system roots.

routes[].runtime<proxy>.upstream

http://host:port, https://host:port or unix:/absolute/path.sock (one address, no retries), or {pool: name}: a pool from upstreams.

routes[].runtime<proxy>.upstream.pool

Name of a pool in upstreams.

routes[].runtime<proxy>.websocket

Allow WebSocket upgrades to be tunnelled to this upstream.

routes[].runtime<respond>.body

Response body (at most 64 KiB).

routes[].runtime<respond>.content_type

Content-Type of the response; default text/plain; charset=utf-8.

routes[].runtime<respond>.status

HTTP status of the fixed response (200-599).

routes[].runtime<redirect>.host

Replaces the request’s host; unset: taken from the request.

routes[].runtime<redirect>.keep_path

Append the request’s path and query to to.

routes[].runtime<redirect>.path

Replaces the request’s path; unset: taken from the request.

routes[].runtime<redirect>.port

Replaces the request’s port; unset: the well-known port of a configured scheme, else taken from the request.

routes[].runtime<redirect>.prefix

Replaces the matched prefix of the request’s path; unset: taken from the request.

routes[].runtime<redirect>.scheme

Replaces the request’s scheme; unset: taken from the request.

routes[].runtime<redirect>.status

301, 302, 303, 307 or 308.

routes[].runtime<redirect>.to

Target origin, e.g. https://example.com: scheme and host (and port), no path when keep_path is on. Alternative to the fields scheme, host, port, path and prefix.

routes[].runtime<php>.fastcgi

External pool: unix:/path.sock or tcp:host:port. Mutually exclusive with fpm.

routes[].runtime<php>.fpm

Managed PHP-FPM pool settings (workers, ini values); not used with fastcgi.

routes[].runtime<php>.fpm.binary

PHP-FPM binary; default: php-fpm on PATH, else the newest versioned one (php-fpm8.4, Alpine’s php-fpm84).

routes[].runtime<php>.fpm.env

Environment passed to PHP (env[...]).

routes[].runtime<php>.fpm.ini

php_admin_value[...] settings, e.g. memory_limit: 256M.

routes[].runtime<php>.fpm.max_requests

pm.max_requests; 0 disables recycling.

routes[].runtime<php>.fpm.workers

min = pm.start_servers/pm.min_spare_servers, max = pm.max_children.

routes[].runtime<php>.fpm.workers.autoscale

off, recommend (default) or auto (ADR-0014).

routes[].runtime<php>.fpm.workers.max

Upper bound when the pool is scaled.

routes[].runtime<php>.fpm.workers.min

Processes started at the beginning and kept at least.

routes[].runtime<php>.fpm.workers.target_concurrency

Concurrent requests one worker is expected to handle well (utilisation basis).

routes[].runtime<php>.front_controller

Script that handles requests for paths that do not exist, e.g. /index.php.

routes[].runtime<php>.index

Index files for directory requests, tried in order.

routes[].runtime<php>.response_timeout

Time to wait for PHP-FPM’s response headers.

routes[].runtime<php>.root

Document root; defaults to the application root.

routes[].runtime<php>.startup_timeout

Time the managed PHP-FPM pool gets to become ready.

routes[].runtime<php>.stop_timeout

Time the managed pool gets to exit before it is killed.

routes[].runtime<node>.command

Command started by scalws (no shell), e.g. [node, server.js]; the application listens on PORT.

routes[].runtime<node>.env

Environment variables for the process.

routes[].runtime<node>.health_path

Optional readiness path; any non-5xx answer counts as ready.

routes[].runtime<node>.listen

unix: a socket under runtime_dir passed as PORT; tcp: a free loopback port.

routes[].runtime<node>.response_timeout

Time to wait for the response headers of the application.

routes[].runtime<node>.startup_timeout

Time the process gets to become ready.

routes[].runtime<node>.stop_timeout

Time the process gets to exit after SIGTERM before it is killed.

routes[].runtime<node>.upstream

Proxy to a Node.js process managed elsewhere instead of starting one.

routes[].runtime<node>.workers

Number of processes (min); requests are balanced across ready ones.

routes[].runtime<node>.workers.autoscale

off, recommend (default) or auto (ADR-0014).

routes[].runtime<node>.workers.max

Upper bound when the pool is scaled.

routes[].runtime<node>.workers.min

Processes started at the beginning and kept at least.

routes[].runtime<node>.workers.target_concurrency

Concurrent requests one worker is expected to handle well (utilisation basis).

routes[].runtime<python>.app

module:callable, e.g. app:app or mysite.wsgi:application.

routes[].runtime<python>.command

Explicit argv. Bind arguments are appended for uvicorn/hypercorn/gunicorn unless the argv already contains {socket} or {port}.

routes[].runtime<python>.env

Environment variables for the processes.

routes[].runtime<python>.health_path

Path requested to decide that a worker is ready and healthy.

routes[].runtime<python>.listen

unix: a socket under runtime_dir; tcp: a free loopback port.

routes[].runtime<python>.mode

asgi (uvicorn) or wsgi (gunicorn).

routes[].runtime<python>.process_pool

scalws runs workers.min..max single-worker server processes itself instead of one server with --workers, so the pool can be scaled and restarted live (ADR-0014). Trades the server’s copy-on-write sharing for scaling.

routes[].runtime<python>.response_timeout

Time to wait for the response headers of the application.

routes[].runtime<python>.server

Default: uvicorn for ASGI, gunicorn for WSGI.

routes[].runtime<python>.startup_timeout

Time the processes get to become ready.

routes[].runtime<python>.stop_timeout

Time the processes get to exit after SIGTERM before they are killed.

routes[].runtime<python>.upstream

Proxy to a Python server managed elsewhere instead of starting one.

routes[].runtime<python>.virtualenv

Virtual environment whose server binary and packages are used.

routes[].runtime<python>.workers

Worker processes of the application server (min).

routes[].runtime<python>.workers.autoscale

off, recommend (default) or auto (ADR-0014).

routes[].runtime<python>.workers.max

Upper bound when the pool is scaled.

routes[].runtime<python>.workers.min

Processes started at the beginning and kept at least.

routes[].runtime<python>.workers.target_concurrency

Concurrent requests one worker is expected to handle well (utilisation basis).

routes[].runtime<container>.command

Arguments after the image; {port} is replaced by the port scalws chose.

routes[].runtime<container>.drop_capabilities

Drop all Linux capabilities.

routes[].runtime<container>.engine

Container engine: podman or docker.

routes[].runtime<container>.engine_path

Engine binary; default podman / docker from PATH.

routes[].runtime<container>.env

Environment variables for the containers.

routes[].runtime<container>.health_path

Path requested to decide that a container is ready and healthy.

routes[].runtime<container>.image

Image reference, e.g. docker.io/library/nginx:1.27.

routes[].runtime<container>.mounts

Host directories mounted into the containers.

routes[].runtime<container>.mounts[].read_only

Mount read-only.

routes[].runtime<container>.mounts[].source

Host path (absolute).

routes[].runtime<container>.mounts[].target

Path inside the container (absolute).

routes[].runtime<container>.network

bridge: publish port on a loopback port; host: share the host network.

routes[].runtime<container>.port

Port the application listens on inside the container (bridge networking).

routes[].runtime<container>.pull

When the image is pulled: missing, always or never.

routes[].runtime<container>.read_only

Read-only root filesystem (a tmpfs is mounted at /tmp).

routes[].runtime<container>.response_timeout

Time to wait for the response headers of the application.

routes[].runtime<container>.startup_timeout

Time a container gets to become ready (includes pulling the image).

routes[].runtime<container>.stop_timeout

Time a container gets to stop before it is killed.

routes[].runtime<container>.user

User the container runs as (--user).

routes[].runtime<container>.workers

Number of containers (min at start; scaled between min and max).

routes[].runtime<container>.workers.autoscale

off, recommend (default) or auto (ADR-0014).

routes[].runtime<container>.workers.max

Upper bound when the pool is scaled.

routes[].runtime<container>.workers.min

Processes started at the beginning and kept at least.

routes[].runtime<container>.workers.target_concurrency

Concurrent requests one worker is expected to handle well (utilisation basis).

runtime

The full runtime syntax of scalws.yaml, instead of a shorthand above.

runtime<static>.index

File names served for a directory request, tried in order.

runtime<static>.precompressed

Serve <file>.br / <file>.gz siblings to clients that accept them.

runtime<static>.root

Document root. Defaults to the application’s root.

runtime<proxy>.backends

Weighted split per request across pools (alternative to upstream).

runtime<proxy>.backends[].pool

Name of a pool in upstreams; alternative to status.

runtime<proxy>.backends[].status

Answers this share of the requests with this status (500-599) instead of sending them to a pool.

runtime<proxy>.backends[].weight

Share of requests (0..=1000); 0 sends none.

runtime<proxy>.client

HTTP/1 client used towards this upstream; unset: server.upstream_client.

runtime<proxy>.connect_timeout

Time to establish a connection to the upstream.

runtime<proxy>.preserve_host

Forward the client’s Host header (default) instead of the upstream authority.

runtime<proxy>.response_timeout

Time allowed for the upstream to return response headers.

runtime<proxy>.tls_ca_file

Extra CA bundle (PEM) trusted for https:// upstreams, in addition to the system roots.

runtime<proxy>.upstream

http://host:port, https://host:port or unix:/absolute/path.sock (one address, no retries), or {pool: name}: a pool from upstreams.

runtime<proxy>.upstream.pool

Name of a pool in upstreams.

runtime<proxy>.websocket

Allow WebSocket upgrades to be tunnelled to this upstream.

runtime<respond>.body

Response body (at most 64 KiB).

runtime<respond>.content_type

Content-Type of the response; default text/plain; charset=utf-8.

runtime<respond>.status

HTTP status of the fixed response (200-599).

runtime<redirect>.host

Replaces the request’s host; unset: taken from the request.

runtime<redirect>.keep_path

Append the request’s path and query to to.

runtime<redirect>.path

Replaces the request’s path; unset: taken from the request.

runtime<redirect>.port

Replaces the request’s port; unset: the well-known port of a configured scheme, else taken from the request.

runtime<redirect>.prefix

Replaces the matched prefix of the request’s path; unset: taken from the request.

runtime<redirect>.scheme

Replaces the request’s scheme; unset: taken from the request.

runtime<redirect>.status

301, 302, 303, 307 or 308.

runtime<redirect>.to

Target origin, e.g. https://example.com: scheme and host (and port), no path when keep_path is on. Alternative to the fields scheme, host, port, path and prefix.

runtime<php>.fastcgi

External pool: unix:/path.sock or tcp:host:port. Mutually exclusive with fpm.

runtime<php>.fpm

Managed PHP-FPM pool settings (workers, ini values); not used with fastcgi.

runtime<php>.fpm.binary

PHP-FPM binary; default: php-fpm on PATH, else the newest versioned one (php-fpm8.4, Alpine’s php-fpm84).

runtime<php>.fpm.env

Environment passed to PHP (env[...]).

runtime<php>.fpm.ini

php_admin_value[...] settings, e.g. memory_limit: 256M.

runtime<php>.fpm.max_requests

pm.max_requests; 0 disables recycling.

runtime<php>.fpm.workers

min = pm.start_servers/pm.min_spare_servers, max = pm.max_children.

runtime<php>.fpm.workers.autoscale

off, recommend (default) or auto (ADR-0014).

runtime<php>.fpm.workers.max

Upper bound when the pool is scaled.

runtime<php>.fpm.workers.min

Processes started at the beginning and kept at least.

runtime<php>.fpm.workers.target_concurrency

Concurrent requests one worker is expected to handle well (utilisation basis).

runtime<php>.front_controller

Script that handles requests for paths that do not exist, e.g. /index.php.

runtime<php>.index

Index files for directory requests, tried in order.

runtime<php>.response_timeout

Time to wait for PHP-FPM’s response headers.

runtime<php>.root

Document root; defaults to the application root.

runtime<php>.startup_timeout

Time the managed PHP-FPM pool gets to become ready.

runtime<php>.stop_timeout

Time the managed pool gets to exit before it is killed.

runtime<node>.command

Command started by scalws (no shell), e.g. [node, server.js]; the application listens on PORT.

runtime<node>.env

Environment variables for the process.

runtime<node>.health_path

Optional readiness path; any non-5xx answer counts as ready.

runtime<node>.listen

unix: a socket under runtime_dir passed as PORT; tcp: a free loopback port.

runtime<node>.response_timeout

Time to wait for the response headers of the application.

runtime<node>.startup_timeout

Time the process gets to become ready.

runtime<node>.stop_timeout

Time the process gets to exit after SIGTERM before it is killed.

runtime<node>.upstream

Proxy to a Node.js process managed elsewhere instead of starting one.

runtime<node>.workers

Number of processes (min); requests are balanced across ready ones.

runtime<node>.workers.autoscale

off, recommend (default) or auto (ADR-0014).

runtime<node>.workers.max

Upper bound when the pool is scaled.

runtime<node>.workers.min

Processes started at the beginning and kept at least.

runtime<node>.workers.target_concurrency

Concurrent requests one worker is expected to handle well (utilisation basis).

runtime<python>.app

module:callable, e.g. app:app or mysite.wsgi:application.

runtime<python>.command

Explicit argv. Bind arguments are appended for uvicorn/hypercorn/gunicorn unless the argv already contains {socket} or {port}.

runtime<python>.env

Environment variables for the processes.

runtime<python>.health_path

Path requested to decide that a worker is ready and healthy.

runtime<python>.listen

unix: a socket under runtime_dir; tcp: a free loopback port.

runtime<python>.mode

asgi (uvicorn) or wsgi (gunicorn).

runtime<python>.process_pool

scalws runs workers.min..max single-worker server processes itself instead of one server with --workers, so the pool can be scaled and restarted live (ADR-0014). Trades the server’s copy-on-write sharing for scaling.

runtime<python>.response_timeout

Time to wait for the response headers of the application.

runtime<python>.server

Default: uvicorn for ASGI, gunicorn for WSGI.

runtime<python>.startup_timeout

Time the processes get to become ready.

runtime<python>.stop_timeout

Time the processes get to exit after SIGTERM before they are killed.

runtime<python>.upstream

Proxy to a Python server managed elsewhere instead of starting one.

runtime<python>.virtualenv

Virtual environment whose server binary and packages are used.

runtime<python>.workers

Worker processes of the application server (min).

runtime<python>.workers.autoscale

off, recommend (default) or auto (ADR-0014).