Skip to main content
Version: 3.79.0

Configuration

Platformatic Runtime is configured with a configuration file. It supports the use of environment variables as setting values with environment variable placeholders.

Configuration Files​

The Platformatic CLI automatically detects and loads configuration files found in the current working directory with the file names listed here.

Alternatively, you can use the --config option to specify a configuration file path for most wattpm CLI commands. The configuration examples in this reference use the JSON format.

Supported File Formats​

For detailed information on supported file formats and extensions, please visit our Supported File Formats and Extensions page.

Settings​

Configuration settings containing sensitive data should be set using environment variable placeholders.

info

The autoload and applications settings can be used together, but at least one of them must be provided. When the configuration file is parsed, autoload configuration is translated into applications configuration.

autoload​

The autoload configuration is intended to be used with monorepo applications. autoload is an object with the following settings:

  • path (required, string) - The path to a directory containing the applications to load. In a traditional monorepo application, this directory is typically named packages.
  • exclude (array of strings) - Child directories inside path that should not be processed.
  • mappings (object) - Each applicaiton is given an ID and is expected to have a Platformatic configuration file. By default, the ID is the application's directory name, and the configuration file is expected to be a well-known Platformatic configuration file. mappings can be used to override these default values. Supported properties are the same of entries in application, except path, url, and gitBranch.

preload​

The preload configuration is intended to be used to register Application Performance Monitoring (APM) agents. preload should contain a path or a list of paths pointing to a CommonJS or ES module that is loaded at the start of the app worker thread.

extensions​

While preload injects code in each application worker thread, extensions loads custom code in the runtime main thread. Extensions can observe and control the whole runtime: they receive the Runtime object and an ITC (Inter-Thread Communication) facade which allows them to register custom commands that applications can invoke from their worker threads.

extensions can be a path, an object with path, options, and build properties, or an array of either:

{
"extensions": [
{
"path": "./runtime-extension.js",
"options": {
"bucket": "{S3_BUCKET}"
}
}
]
}

Each file must export a setup function, which is invoked during the runtime initialization, before any application is started. TypeScript files are supported out of the box via Node.js type stripping.

export default async function setup ({ runtime, itc, sharedContext, logger, options, root, metrics, health }) {
// React to runtime events
runtime.on('application:worker:started', payload => {
logger.info({ payload }, 'worker started')
})

// Publish state to every application worker
await sharedContext.update({ reconciler: { ready: true } })

// Register a custom command that applications can invoke via
// getITC().send('acme:hello', payload)
itc.handle('acme:hello', async payload => {
return { hello: payload.name }
})

// Register main-thread metrics with the shared Runtime metrics pipeline
const { client, registry } = metrics
const jobs = new client.Gauge({
name: 'acme_extension_jobs',
help: 'Number of jobs tracked by the extension',
registers: [registry]
})
jobs.set(0)

// Contribute readiness/liveness checks and diagnostic routes on the health probes server
health.registerReadinessCheck('dispatchable', async () => {
return { status: true }
})

health.registerLivenessCheck('control-plane', async () => true)

health.registerRoutes(async app => {
app.get('/inventory', async () => ({ ok: true }))
})

return {
async start () {
// Invoked after applications are prepared, before they accept traffic.
// Can add/start dynamic applications; Runtime will not start them again.
},
async stop () {
// Invoked during shutdown after the entrypoint stops and before remaining
// applications stop. Useful for control-plane handoff.
},
async close () {
// Invoked when the runtime is closed, after all applications have stopped.
}
}
}

How the setup function is resolved​

The setup function can be exported either as the default export or as a named setup export, so both of the following are valid:

// Default export
export default async function setup (context) {}
// Named export
export async function setup (context) {}

Extensions are also commonly shipped as CommonJS or as "faux ESM" modules, in which a transpiler or a bundler puts the real exports on module.exports together with an __esModule marker. In those modules Node.js exposes the whole module.exports object as the ESM default export, so the setup function ends up one level deeper. Runtime transparently unwraps that extra level, which means all of these work as well:

// CommonJS, transpiled default export: module.exports.default is the setup function
exports.__esModule = true
exports.default = async function setup (context) {}
// CommonJS, named setup export: module.exports.setup is the setup function
exports.setup = async function setup (context) {}

The resolution order is: the default export, the named setup export, then the same two properties of the default export. The first one that is a function wins, so a module exporting both a default function and a named setup function uses the default export. If none of them is a function, the runtime fails to start with PLT_RUNTIME_INVALID_EXTENSION.

The setup function receives a context object with the following properties:

  • runtime - The Runtime instance. It is an EventEmitter, so extensions can subscribe to all runtime events and invoke any public method, including application and worker introspection (getApplications(), getWorkers()), application configuration and environment (getApplicationConfig(id), getApplicationEnv(id)), and metrics (getMetrics()).

  • itc - A facade over the runtime ITC:

    • handle(name, handler) - Registers a custom command invocable from any application via the ITC API returned by getITC() from @platformatic/globals. The name must not clash with the commands reserved by the runtime or with a command registered by another extension.
    • send(target, name, payload) - Sends a request to a worker and awaits its response. target is an application ID (a worker is chosen in round-robin) or application:worker-index for a specific worker.
    • notify(target, name, payload) - Sends a fire-and-forget notification. When target is an application ID, all its running workers are notified; use application:worker-index to target a specific worker. Workers receive notifications via getITC().on(name, handler).
  • sharedContext - The shared context API used by application workers. get() synchronously returns a snapshot of the current context in main-thread extensions. update(update, options?) merges update into the context and broadcasts the result to every running worker; pass { overwrite: true } to replace the context instead. Newly started workers receive the latest context. Use update() rather than mutating the object returned by get().

  • logger - A child of the runtime logger.

  • options - The options object specified in the configuration, if any.

  • root - The runtime project root directory.

  • metrics - A per-extension Prometheus client and registry:

    • client - The @platformatic/prom-client module (same client used by application workers).
    • registry - A dedicated Registry for this extension. Metrics registered here appear once in Runtime.getMetrics(), the management metrics API, and the existing /metrics endpoint. They are not duplicated per application worker.

    Label behavior: extension metrics are main-thread metrics. Runtime never invents a workerId or application ID label for them. Only static labels from the runtime metrics.labels configuration are applied (the configured application label name is omitted, same as process-level metrics). Extensions that need application- or worker-specific labels must set them explicitly when creating metrics.

    Metric family names must be unique across extensions and must not collide with runtime process metrics, restart metrics, or application worker metrics. Collisions fail with PLT_RUNTIME_METRIC_FAMILY_COLLISION, identifying the extension and metric family. The registry is cleared when the extension closes (including partial startup failures).

  • health - API for contributing readiness/liveness checks and diagnostic routes on Watt's health probes server (the metrics server when probes are shared with it, or the dedicated health probes server when configured separately):

    • registerReadinessCheck(name, check) - Registers a named check that participates in /ready. Check names must be unique across extensions. The check may return a boolean or { status, statusCode?, body? }. Rejected, timed-out, or malformed checks fail closed. Timeouts use the configured health checks timeout. Readiness-only failures do not fail /status (liveness), so a temporary control-plane condition can stop traffic without triggering a pod restart loop. Returns an unregister function.
    • registerLivenessCheck(name, check) - Same contract as readiness, but the check participates in /status. Returns an unregister function.
    • registerRoutes(plugin) - Registers a Fastify plugin on the health probes server before it starts listening. Works for shared and separate probe servers. Route collisions fail startup with a coded error identifying the extension and route. Returns an unregister function that disables the routes. Closing the extension removes its checks and routes.

The setup function can optionally return an object with awaited lifecycle hooks:

  • preBuild(context) - Called before an application capability is built. context contains applicationId and the absolute applicationPath.
  • onBuild(context, build) - Wraps the application capability build. Call and await build() to continue the build. The hook may perform work immediately before and after it, or omit the call to replace the capability build. Its return value becomes the build result.
  • postBuild(context, result) - Called after a successful application build with the final result returned by the onBuild hook chain.
  • start() - Called once applications have been prepared and registered, and before the originally configured applications start accepting traffic. Runtime awaits every extension start hook. An extension may add and start dynamic applications here; those applications are excluded from the normal startup pass so they are not started twice. Runtime does not report started until extension and application startup complete. If a later extension or application fails, Runtime stops and closes already-started extensions during cleanup.
  • stop() - Called during shutdown after the entrypoint has been stopped and before the remaining applications stop. Runtime awaits every extension stop hook so control-plane extensions can settle work and hand off state.
  • close() - Called when the runtime is being closed, after applications have stopped. The extension metrics registry is cleared after close.

When multiple extensions are configured, they are set up and started in registration order. preBuild runs in registration order. onBuild hooks are nested middleware with the first registered extension outermost. postBuild, stop, and close run in reverse registration order. Each of stop and close is invoked at most once per Runtime life, including repeated stop/close calls and failed-start cleanup paths. Existing extensions that only return close() keep their previous behavior. Health checks and routes registered by an extension are cleaned up with it.

Listening to Runtime EventEmitter events is not a substitute for these hooks: emit() does not await asynchronous listeners. Lifecycle ordering is enforced by Runtime itself so it covers signal handling, internal failure cleanup, and future lifecycle entry points.

Extensions are not loaded when building applications unless their object configuration explicitly sets "build": true. Build-enabled extensions are set up before application workers are created, receive the build hooks above for every application, and are closed when the build Runtime closes. Their start and stop hooks are not called during a build.

{
"extensions": [
{
"path": "./build-extension.js",
"build": true
}
]
}

For a complete worked example — enabling continuous profiling on every worker when it starts (or is restarted) and shipping the captured profiles — see the Capture Flamegraphs on Health Events guide.

applications​

applications is an array of objects that defines the applications managed by the runtime. Each application object supports the following settings:

  • id (required, string) - A unique identifier for the application.

  • enabled (boolean, string, or object) - If false, the application is disabled and will not be loaded by the runtime. Boolean strings and environment variable placeholders are supported. It can also be an object where each key is an environment name and each value is a boolean. If the current environment does not match any key, the application is enabled. Default: true.

  • path (required, string) - The path to the directory containing the application. It can be omitted if url is provided.

  • url (required, string) - The URL of the application remote GIT repository, if it is a remote application. It can be omitted if path is provided. You can specify a branch using the URL fragment syntax: https://github.com/user/repo.git#branch-name.

  • gitBranch (string) - The branch of the application to resolve. Takes precedence over the branch specified in the URL fragment.

  • config (string) - The configuration file used to start the application.

  • useHttp (boolean) - The application will be started on a random HTTP port on 127.0.0.1, and exposed to the other applications via that port, on default it is set to false. Set it to true if you are using @fastify/express.

  • websocket (boolean) - The application will be started on a random HTTP port on 127.0.0.1 so that the gateway can proxy WebSocket connections to it, but, unlike useHttp, HTTP traffic between applications keeps using the in-memory mesh network when the capability supports in-thread dispatching (for example the service family); for the other capabilities it flows through the bound TCP port, as under useHttp. Set it to true when a non-entrypoint application behind the gateway needs to accept WebSocket connections. Default: false.

  • reuseTcpPorts: Enable the use of the reusePort option whenever any TCP server starts listening on a port. The default is true. The values specified here overrides the values specified in the runtime.

  • workers - The number of workers to start for this application. If the application is the entrypoint or if the runtime is running in development mode this value is ignored and hardcoded to 1. This can be specified as:

    • number - A fixed number of workers
    • object - Advanced worker configuration with the following properties:
      • static (number) - A fixed number of workers
      • dynamic (boolean) - Enable dynamic worker scaling. This is only meaningful when set to false to disable dynamic scaling for this application.
      • minimum (number) - Minimum number of workers when using dynamic scaling
      • maximum (number) - Maximum number of workers when using dynamic scaling
  • health (object): Configures the health check and low-level resource defaults for each worker of the application. It supports all the properties also supported in the runtime health property. The values specified here override the values specified in the runtime.

  • arguments (array of strings) - The arguments to pass to the application. They will be available in process.argv.

  • envfile (string) - The path to an .env file to load for the application. By default, the .env file is loaded from the application directory.

  • env (object) - An object containing environment variables to set for the application. Values set here takes precedence over values set in the envfile.

  • sourceMaps (boolean) - If true, source maps are enabled for the application. Default: false.

  • packageManager (string) - The package manager to use when using the install-dependencies or the resolve commands of wattpm-utils. Default is to autodetect it, unless it is specified via command line.

  • preload (string or array of strings): A file or a list of files to load before the application code.

  • nodeOptions (string): The NODE_OPTIONS to apply to the application. These options are appended to any existing option.

  • execArgv (array of strings): Additional arguments to pass to application worker threads via the execArgv option. These arguments are passed to the Node.js executable when creating worker threads. See Node.js Worker Threads documentation for more information. Note that execArgv options are automatically inherited by any child worker threads created by the application. For a complete example of early per-worker tracer initialization, see Datadog tracing with multiple applications and workers.

  • permissions (object): Configure application-level security permissions to restrict file system access. Supported properties are:

    • fs:
      • read (array of strings): Array of file system paths the application is permitted to read from. Uses the same syntax as Node.js --allow-fs-read.
      • write (array of strings): Array of file system paths the application is permitted to write to. Uses the same syntax as Node.js --allow-fs-write.

    When filesystem permissions are enabled, certain paths are automatically added to maintain application functionality:

    • The current Watt project's node_modules directory
    • The application's own node_modules directory
    • Any node_modules directories found in parent directories of the runtime path

    The security permissions are based on Node.js permission model and therefore the application will have restricted access to native modules, child processes, worker threads, the inspector protocol, and WASI. See the Node.js Permission Model Constraints for complete details.

  • dependencies (array of strings): A list of applications that must be started before attempting to start the current application. Note that the runtime will not perform any attempt to detect or solve dependencies cycles.

  • management (boolean or object): Grants the application access to runtime management operations via the ITC (Inter-Thread Communication) channel. See the management section for details.

  • telemetry (object): containing an instrumentations array to optionally configure additional open telemetry intrumentations per application, e.g.:

"applications": [
{
"id": "api",
"path": "./services/api",
"telemetry": {
"instrumentations": ["@opentelemetry/instrumentation-express"]
}
}
]

It's possible to specify the name of the export of the instrumentation and/or the options:

"applications": [
{
"id": "api",
"path": "./services/api",
"telemetry": {
"instrumentations": [{
"package": "@opentelemetry/instrumentation-express",
"exportName": "ExpressInstrumentation",
"options": {}
}]
}
}
]

An alias for applications. If both are present, their content will be merged.

It's also possible to disable the instrumentation by setting the enabled value property to false (env variables are also supported):

"applications": [
{
"id": "api",
"path": "./services/api",
"telemetry": {
"enabled": "false",
"instrumentations": [{
"package": "@opentelemetry/instrumentation-express",
}]
}
}
]

env​

An object containing environment variables to set for all applications in the runtime. Any environment variables set in the env object will be merged with the environment variables set in the envfile and env properties of each application, with application-level environment variables taking precedence.

envfile​

The path to an .env file to load for the runtime. By default, the .env file is loaded from the application directory.

strictEnv​

Controls what happens when a {PLT_*} environment variable placeholder references a variable which is not set:

  • false (the default): the placeholder is silently replaced with an empty string.
  • true: loading the configuration fails at startup with an error listing all the missing variables.
  • "warn": a warning listing the missing variables is logged, but the placeholders are still replaced with an empty string.

Not every unset variable is reported as missing. When loading the configuration of an application, the runtime resolves any variable whose name ends in _URL to the internal URL of that application, so such a variable gets a value even when it is not set. When strictEnv is enabled, these variables are listed in a separate warning. They are never turned into an error, not even when strictEnv is true, because they do resolve to a value and failing on them would change which configurations are able to boot.

The value is also applied when loading the configuration files of the applications in the runtime.

{
"strictEnv": true
}

sourceMaps​

If true, source maps are enabled for all applications. Default: false. This setting can be overridden at the application level.

resolvedServicesBasePath​

The base path, relative to the configuration file to store resolved applications. Each application will be saved in {resolvedServicesBasePath}/{id}. Default: external.

entrypoint​

The Platformatic Runtime's entrypoint is an application that is exposed publicly. This optional value must be the ID of an application defined via the autoload or applications configuration.

If entrypoint is omitted, the runtime automatically selects one when there is a single application or exactly one Gateway application. If it cannot select a single entrypoint, the runtime starts without a public entrypoint; applications remain reachable through their internal .plt.local URLs and APIs such as runtime.inject().

workers​

Configures the default number of workers to start per each application. Some values can be overridden at the application level.

This can be specified as:

  • number - A fixed number of workers (minimum 1)
  • object - Advanced worker configuration with the following properties:
    • static (number) - A fixed number of workers
    • dynamic (boolean) - Enable dynamic worker scaling (default: false). The dynamic worker scaler automatically adjusts the number of workers for each application based on Event Loop Utilization (ELU) and available system memory. It can be overridden at the application level.
    • minimum (number) - The minimum number of workers that can be used for each application. Default: 1.
    • maximum (number) - The maximum number of workers that can be used for each application. Default: global total value.
    • total (number) - The maximum number of workers that can be used for all applications. Default: os.availableParallelism() (typically the number of CPU cores).
    • maxMemory (number) - The maximum total memory in bytes that can be used by all workers. Default: 90% of the system's total memory.
    • cooldown (number) - The amount of milliseconds the scaling algorithm will wait after making a change before scaling up or down again. This prevents rapid oscillations. Default: 20000.
    • gracePeriod (number) - The amount of milliseconds after a worker is started before the scaling algorithm will start collecting metrics for it. This allows workers to stabilize after startup. Default: 30000.
    • scaleUpELU (number) - The Event Loop Utilization (ELU) threshold an application's average ELU must exceed before the scaler adds a worker. Must be between 0 and 1. It can be overridden at the application level. Default: 0.8.
    • scaleDownELU (number) - The ELU threshold an application's average ELU must fall below before the scaler removes a worker. Must be between 0 and 1. It can be overridden at the application level. Default: 0.2.

This value is hardcoded to 1 if the runtime is running in development mode or when applying it to the entrypoint.

workersRestartDelay​

Configures the amount of milliseconds to wait before replacing another worker of an application during a restart.

gracefulShutdown​

Configures the amount of milliseconds to wait before forcefully killing an application or the runtime.

The object supports the following settings:

  • application (number) - The graceful shutdown timeout for an application. Default: 10000 (ten seconds).
  • runtime (number) - The graceful shutdown timeout for the entire runtime. Default: 30000 (thirty seconds).

watch​

An optional boolean, set to default false, indicating if hot reloading should be enabled for the runtime. If this value is set to false, it will disable hot reloading for any applications managed by the runtime. If this value is true, then hot reloading for individual applications is managed by the configuration of that application.

Note that watch should be enabled for each individual application in the runtime.

warning

While hot reloading is useful for development, it is not recommended for use in production.

startTimeout​

The number of milliseconds to wait before considering an application as failed to start. Default: 30000.

restartOnError​

The number of milliseconds to wait before attempting to restart an application that unexpectedly exit.

If not specified or set to true, the default value is 5000, set to 0 or false to disable.

Any value smaller than 10 will cause immediate restart of the application.

This setting is ignored in production, where applications are always restarted immediately.

exitOnUnhandledErrors​

When enabled (default), Platformatic automatically installs error handlers for uncaughtException and unhandledRejection events on each worker process. These handlers will automatically restart the affected worker when such errors occur.

If application code installs its own listeners for these events, Platformatic tracks them, removes them from process, and invokes them before terminating the worker. This keeps Platformatic in control of the shutdown while still allowing error reporting tools to observe fatal errors.

Set this to true to terminate the worker after 100 milliseconds. Set this to a positive number to use that number of milliseconds instead.

Setting this to false, 0, or a negative number disables the automatic error handling, making you responsible for implementing proper error handling in your application code.

health​

Configures per-worker health checks and low-level worker resource defaults. Health checks are enabled only if restartOnError is greater than zero; bufferPoolSize and defaultHighWaterMark are applied during worker startup regardless.

The object supports the following settings:

  • enabled (boolean): If to enable the health check. Default: true.
  • interval (number): The interval between checks in milliseconds. Default: 30000.
  • gracePeriod (number): How long after the application started before starting to perform health checks. Default: 30000.
  • maxUnhealthyChecks (number): The number of consecutive failed checks before killing the worker. Default: 10.
  • maxELU (number): The maximum allowed Event Loop Utilization. The value must be a percentage between 0 and 1. Default: 0.99.
  • maxELULogLevel (string): The log level for messages reporting that maxELU was exceeded. Accepts trace, debug, info, warn, error, fatal, or silent to suppress the message. Default: error.
  • maxEventLoopDelay (number): The maximum allowed event loop delay in milliseconds. When set, each worker samples its own event loop delay (via perf_hooks.monitorEventLoopDelay) and reports it once per second as an eventLoopDelay health signal; a worker whose maximum delay over a check window exceeds this value counts as unhealthy. This catches long individual stalls at low average utilization (a 150ms synchronous stall each second is an ELU of just 0.15), which are invisible to maxELU. Disabled by default.
  • maxEventLoopDelayP99 (number): The maximum allowed p99 event loop delay in milliseconds. Same mechanism as maxEventLoopDelay, but evaluated against the worst per-second p99 reported over the check window instead of the absolute maximum, so single outlier stalls are smoothed out. Either option activates the sampler; they can be combined. Disabled by default.
  • maxHeapUsed (number): The maximum allowed memory utilization. The value must be a percentage between 0 and 1. Default: 0.99.
  • maxHeapUsedLogLevel (string): The log level for messages reporting that maxHeapUsed was exceeded. Accepts trace, debug, info, warn, error, fatal, or silent to suppress the message. Default: error.
  • maxHeapTotal (number or string): The maximum allowed memory allocatable by the process. The value must be an amount in bytes, in bytes or in memory units. Default: 4GB.
  • maxYoungGeneration(number or string): The maximum amount of memory that can be used by the young generation. The value must be an amount in bytes, in bytes or in memory units. Default: 128MB
  • codeRangeSize (number or string): The maximum amount of memory that can be used for code range (compiled code). The value must be an amount in bytes or in memory units. Default: 268435456 (256MB).
  • bufferPoolSize (number or string): Sets Node.js Buffer.poolSize for each application worker. The value must be an amount in bytes or in memory units. Default: 262144 (256KB).
  • defaultHighWaterMark (number or string): Sets the default high water mark for byte streams via Node.js stream.setDefaultHighWaterMark(false, value). The value must be an amount in bytes or in memory units. Default: 262144 (256KB).

bufferPoolSize is intentionally larger than the Node.js default and should be treated as an aggressive server-oriented default. Node.js pools Buffer.allocUnsafe() allocations only when size < (Buffer.poolSize >>> 1), so the 256KB default serves allocations smaller than 128KB from the pool. This covers common server allocation sizes such as HTTP parser buffers, stream chunks, and small file reads in the 4KB-64KB range, reducing allocator work and Worker-thread contention. The trade-off is one larger pool per worker/realm, and RSS can stay higher while pooled slices are retained; lower this value for memory-constrained deployments.

maxELULogLevel and maxHeapUsedLogLevel can also be overridden in each application's health configuration. They only control threshold messages: health checks, worker replacement, and replacement error logs remain unchanged, including when either level is silent.

healthProbes​

Enables the Kubernetes readiness and liveness probe endpoints. It can be a boolean, a string, or an object. Default: true.

If healthProbes is true, unset, or a string, the probe endpoints are installed on the Prometheus server. Individual probe endpoints can still be configured with metrics.readiness and metrics.liveness.

Set this to false to disable both probe endpoints globally.

Use an object to configure the health probes server. Health probes are exposed on a standalone server only when the resolved hostname and port differ from the Prometheus server. Otherwise, they are installed on the Prometheus server.

  • enabled (boolean or string). Enables the health probe endpoints. Default: true.
  • hostname (string). The hostname where the health probes server will be listening. Default: 0.0.0.0.
  • port (number or string). The port where the health probes server will be listening. Default: 9090.
  • readiness (object or boolean). Optional readiness endpoint configuration. If omitted, metrics.readiness is used when present.
  • liveness (object or boolean). Optional liveness endpoint configuration. If omitted, metrics.liveness is used when present.
Example health probes on a standalone server
{
"metrics": {
"hostname": "0.0.0.0",
"port": 9090
},
"healthProbes": {
"hostname": "0.0.0.0",
"port": 9091,
"readiness": {
"endpoint": "/health"
},
"liveness": {
"endpoint": "/live"
}
}
}

telemetry​

Open Telemetry is optionally supported with these settings:

  • applicationName (required, string) — Name of the application as will be reported in open telemetry. In the runtime case, the name of the applications as reported in traces is ${applicationName}-${applicationId}, where applicationId is the id of the application in the runtime.
  • version (string) — Optional version (free form)
  • skip (array). Optional list of operations to skip when exporting telemetry defined object with properties:
    • method: GET, POST, PUT, DELETE, PATCH, HEAD, OPTIONS, TRACE
    • path. e.g.: /documentation/json
  • exporter (object or array) — Exporter configuration. If not defined, the exporter defaults to console. If an array of objects is configured, every object must be a valid exporter object. The exporter object has the following properties:
    • type (string) — Exporter type. Supported values are console, otlp,