middleware_dart_opentelemetry

OpenTelemetry SDK for Dart

License OpenTelemetry Specification Coverage Report

Middleware is an OpenTelemetry SDK to add standard observability to Dart applications. Middleware can be used with any OTel backend, it’s standards-compliant.

Flutter developers should use the Middleware_Flutter OpenTelemetry SDK which builds on top of Middleware Dart OTel.

Middleware.io provides an OpenTelemetry support, training, consulting and an Observability backend customized for Flutter apps, Dart backends, and any other service or process that produces OpenTelemetry data.

Dartastic.io tools and services for Dart and Flutter teams shipping to production.

Features

Middleware_Dart OTel is suitable for Dart backends, CLIs or any Dart application.

Dartastic OTel API is the API for the Dartastic OTel SDK. The dartastic_opentelemetry_api exists as a standalone library to strictly adhere to the OpenTelemetry specification which separates API and the SDK. All OpenTelemetry API classes on in dartastic_opentelemetry_api.

Middleware_Flutter OTel adds Middleware Dart OTel to Flutter apps with ease.

Middleware dart and flutter-sdk OTel are made with 💙

Getting started

Include this in your pubspec.yaml:

dependencies:
  middleware_dart_opentelemetry: ^1.0.5

Usage

The entrypoint to the SDK is the OTel class. OTel has static “factory” methods for all OTel API and SDK objects. OTel needs to be initialized first to point to an OpenTelemetry backend. Initialization does a lot of work under the hood including gathering a rich set of standard resources for any OS that Dart runs in. It prepares for the creation of the global default TracerProvider, MeterProvider, and LoggerProvider, with the serviceName and a default Tracer, Meter, and OTelLogger, all created on first use.

All configuration, including Trace, Metric and Log exporter configuration, can be made in code via OTel.initialize(). Codeless configuration can be done with standard OpenTelemetry environmental variables either through POSIX variable or -D or --define for Dart or with --dart-define for Flutter apps. See [Running with Environment Variables] below

Middleware Dart OpenTelemetry supports is working on support for all standard OpenTelemetry environment variables as defined in the OpenTelemetry Specification.

import 'package:middleware_dart_opentelemetry/middleware_dart_opentelemetry.dart';

Future<void> main() async {
  // Initialize - automatically reads environment variables.
  await OTel.initialize();

  // Get the default tracer.
  final tracer = OTel.tracer();

  // Per the OpenTelemetry spec, tracer.startSpan() does NOT activate the
  // span. Use tracer.withSpanAsync to make the span active for the
  // duration of doWork() so that any spans started inside are parented
  // to it via Context.current.
  final span = tracer.startSpan('my-operation');
  try {
    await tracer.withSpanAsync(span, doWork);
  } catch (e, stackTrace) {
    // The span has a status of SpanStatus.Ok on creation, set it to
    // Error when an error occurs in the span.
    span.recordException(e, stackTrace: stackTrace);
    span.setStatus(SpanStatusCode.Error, e.toString());
    rethrow;
  } finally {
    // Always end the span — even on error.
    span.end();
  }

  await OTel.shutdown();
}

Future<void> doWork() async {
  // Your business logic here.
}

Since dartastic_opentelemetry exports all the classes of opentelemetry_api, refer to opentelemetry_api for documentation of API classes.

See the /example folder for more examples.

OpenTelemetry Tracing API

The Tracing API is the primary signal in OpenTelemetry. A trace represents the end-to-end journey of a request through your system. Each trace is composed of spans — individual units of work with a name, timing, attributes, and parent-child relationships.

Concepts

Basic Tracing

import 'package:middleware_dart_opentelemetry/middleware_dart_opentelemetry.dart';

Future<void> main() async {
  await OTel.initialize(serviceName: 'my-service');

  // Get the default tracer
  final tracer = OTel.tracer();

  // Create a span and make it active for the duration of doWork() via
  // withSpanAsync. Per the OpenTelemetry spec, startSpan does NOT activate
  // the span — child spans started inside the closure are parented to
  // `span` via Context.current.
  // Prefer typed enum keys over raw strings — User.userId is
  // the OTel semantic-convention key. For app-specific attributes that
  // don't have a semantic convention, define your own typed enum (see
  // the Span Attributes section below).
  final span = tracer.startSpan(
    'main-operation',
    kind: SpanKind.server,
    attributes: OTel.attributesFromMap({
      User.userId.key: 'user-123',
      // app-specific key — would normally come from your own typed enum:
      'request.type': 'example',
    }),
  );

  try {
    await tracer.withSpanAsync(span, doWork);
  } catch (e, stackTrace) {
    // The span has a status of SpanStatus.Ok on creation, set it to
    // Error when an error occurs in the span.
    span.recordException(e, stackTrace: stackTrace);
    span.setStatus(SpanStatusCode.Error, e.toString());
    rethrow;
  } finally {
    span.end();
  }

  await OTel.shutdown();
}

Parent-Child Spans

Spans form a tree by linking child spans to parent spans via context:

final parentSpan = tracer.startSpan('parent-operation');
try {
  // Create a child span linked to the parent. Passing the parent's
  // SpanContext via `context:` parents this span without requiring the
  // parent to be active in Context.current.
  final childSpan = tracer.startSpan(
    'database.query',
    kind: SpanKind.client,
    context: OTel.context(spanContext: parentSpan.spanContext),
    attributes: OTel.attributesFromSemanticMap({
      Database.dbSystem: 'postgresql',
      Database.dbOperation: 'SELECT',
    }),
  );
  try {
    await queryDatabase();
  } catch (e, stackTrace) {
    // The span has a status of SpanStatus.Ok on creation, set it to
    // Error when an error occurs in the span.
    childSpan.recordException(e, stackTrace: stackTrace);
    childSpan.setStatus(SpanStatusCode.Error, e.toString());
    rethrow;
  } finally {
    childSpan.end();
  }
} catch (e, stackTrace) {
  // The span has a status of SpanStatus.Ok on creation, set it to
  // Error when an error occurs in the span.
  parentSpan.recordException(e, stackTrace: stackTrace);
  parentSpan.setStatus(SpanStatusCode.Error, e.toString());
  rethrow;
} finally {
  parentSpan.end();
}

Activating a span — OTel.withSpan / OTel.withSpanAsync

OTel.withSpan and OTel.withSpanAsync activate a span for the duration of fn (so Context.current.span returns it inside fn) and record any thrown exception with SpanStatusCode.Error before rethrowing. The caller still owns span.end() — the canonical OTel lifecycle is try / catch / finally:

final span = OTel.tracer().startSpan('compute-result');
try {
  final result = OTel.withSpan(span, () => computeExpensiveValue());
} catch (e, stackTrace) {
  // The span has a status of SpanStatus.Ok on creation, set it to
  // Error when an error occurs in the span.
  span.recordException(e, stackTrace: stackTrace);
  span.setStatus(SpanStatusCode.Error, e.toString());
  rethrow;
} finally {
  span.end();
}

// Async version
final fetchSpan = OTel.tracer().startSpan('fetch-data', kind: SpanKind.client);
try {
  final data = await OTel.withSpanAsync(
    fetchSpan,
    () => httpClient.get('/api/data'),
  );
} catch (e, stackTrace) {
  fetchSpan.recordException(e, stackTrace: stackTrace);
  fetchSpan.setStatus(SpanStatusCode.Error, e.toString());
  rethrow;
} finally {
  fetchSpan.end();
}

If you also want the span as a callback argument and want the span ended for you, use tracer.startActiveSpan / startActiveSpanAsync:

// Active span — span is in Context.current AND passed to fn,
// and ended automatically when fn returns.
OTel.tracer().startActiveSpan(
  name: 'process-request',
  fn: (span) {
    span.setStringAttribute(ExampleAttribute.requestId.key, 'abc-123');
    return processRequest();
  },
);

Span Attributes

Attributes are typed key-value pairs on spans. OTel restricts values to String, bool, int, double, and Lists of those types.

Prefer typed enum keys over raw strings. The API ships enums for every namespace in the OTel semantic conventions — Http, Url, ServerResource, Client, Database, User, Session, etc. Using them prevents typos, gives you autocomplete, and tracks the spec as it evolves. For app-specific attributes that aren’t in a convention, define your own enum implementing OTelSemantic:

// In your own app, name this for your domain (e.g. `CheckoutAttribute`).
enum ExampleAttribute implements OTelSemantic {
  requestType('request.type'),
  processingStage('processing.stage'),
  durationMs('duration_ms'),
  tags('tags'),
  cacheKey('cache.key'),
  cacheRegion('cache.region'),
  linkType('link.type'),
  authMethod('auth.method'),
  orderId('order.id'),
  requestId('request.id');

  @override
  final String key;
  @override
  String toString() => key;
  const ExampleAttribute(this.key);
}

// Type-safe individual attributes — mix API convention enums with your
// own ExampleAttribute for non-convention keys.
final span = tracer.startSpan('operation', attributes: OTel.attributes([
  OTel.attributeString(Http.requestMethod.key, 'GET'),
  OTel.attributeInt(Http.responseStatusCode.key, 200),
  OTel.attributeDouble(ExampleAttribute.durationMs.key, 123.45),
  OTel.attributeStringList(ExampleAttribute.tags.key, ['payment', 'critical']),
]));

// Or from a map (types are inferred automatically).
final span = tracer.startSpan('operation',
  attributes: OTel.attributesFromSemanticMap({
    Http.requestMethod: 'GET',
    Http.responseStatusCode: 200,
  }),
);

// `attributesOf<E>` is the single-enum form — every key is checked
// against `Http` at compile time, and Dart 3.10's static dot-shorthand
// can shorten each entry to `.requestMethod`, `.responseStatusCode`, …
span.addAttributes(OTel.attributesOf<Http>({
  Http.responseBodySize: 1024,
}));

// Mix-and-match: each typed-enum map spreads into a `Map<OTelSemantic, Object>`,
// which is exactly what `attributesFromSemanticMap` accepts. Useful when one
// span carries attributes from several namespaces.
span.addAttributes(OTel.attributesFromSemanticMap({
  ...<Http, Object>{Http.requestMethod: 'POST'},
  ...<Url, Object>{Url.urlFull: 'https://api.example.com/data'},
  ...<ExampleAttribute, Object>{ExampleAttribute.processingStage: 'complete'},
}));

Span Events

Events are time-stamped annotations on a span. Event names themselves are user-defined, but event attributes still benefit from typed enum keys:

span.addEvent(OTel.spanEventNow(
  'cache.hit',
  OTel.attributesFromSemanticMap({ExampleAttribute.cacheKey: 'user:123'}),
));

span.addEventNow('validation.passed');

Links connect spans across traces — useful for batch processing or fan-out patterns:

final link = OTel.spanLink(
  otherSpan.spanContext,
  attributes: OTel.attributesFromSemanticMap({ExampleAttribute.linkType: 'triggers'}),
);

final span = tracer.startSpan('batch-process', links: [link]);

SpanKind

Classifies the relationship between a span and its remote counterpart:

SpanKind Description Example
internal Default; internal operation with no remote side Business logic, local computation
server Server handling an incoming request HTTP server endpoint
client Client making an outgoing request HTTP client call, DB query
producer Producer enqueuing a message Kafka producer, queue publisher
consumer Consumer processing a message Kafka consumer, queue subscriber

Samplers

Samplers control which spans are recorded and exported. Configure via OTel.initialize() or per-Tracer.

Sampler Description Use Case
AlwaysOnSampler Samples every span (default) Development, debugging
AlwaysOffSampler Never samples Disable tracing without code changes
TraceIdRatioSampler Samples by trace ID ratio (consistent per trace) Production with consistent sampling
ProbabilitySampler Samples by random probability Testing, non-critical sampling
ParentBasedSampler Respects parent span’s sampling decision Distributed tracing across services
RateLimitingSampler Limits sampled traces per second (token bucket) Controlling overhead
CountingSampler Samples every Nth request Periodic sampling
CompositeSampler Combines samplers with AND/OR logic Complex sampling policies 
// Sample 10% of traces consistently
await OTel.initialize(
  serviceName: 'my-service',
  sampler: TraceIdRatioSampler(0.1),
);

// Respect parent decisions, sample 50% of new root traces
await OTel.initialize(
  serviceName: 'my-service',
  sampler: ParentBasedSampler(TraceIdRatioSampler(0.5)),
);

// Rate-limit to 100 traces/second
await OTel.initialize(
  serviceName: 'my-service',
  sampler: RateLimitingSampler(100),
);

Span Processors

Processors handle span lifecycle and export:

// SimpleSpanProcessor — exports each span immediately (good for debugging)
await OTel.initialize(
  spanProcessor: SimpleSpanProcessor(ConsoleExporter()),
);

// BatchSpanProcessor — batches spans for efficient production export
await OTel.initialize(
  spanProcessor: BatchSpanProcessor(
    OtlpGrpcSpanExporter(OtlpGrpcExporterConfig(endpoint: 'localhost:4317')),
    BatchSpanProcessorConfig(
      maxQueueSize: 2048,
      scheduleDelay: Duration(milliseconds: 5000),
      maxExportBatchSize: 512,
    ),
  ),
);

Span Exporters

Exporter Protocol Description
ConsoleExporter stdout Prints spans to console for debugging
OtlpGrpcSpanExporter gRPC Exports via OTLP/gRPC (production)
OtlpHttpSpanExporter HTTP/protobuf Exports via OTLP/HTTP (web-compatible) 
// Console (development)
final exporter = ConsoleExporter();

// OTLP gRPC (production)
final exporter = OtlpGrpcSpanExporter(OtlpGrpcExporterConfig(
  endpoint: 'otel-collector:4317',
  headers: {'api-key': 'your-key'},
  compression: true,
));

// OTLP HTTP (web-compatible)
final exporter = OtlpHttpSpanExporter(OtlpHttpExporterConfig(
  endpoint: 'https://otel-collector:4318',
  headers: {'api-key': 'your-key'},
));

Context Propagation

Propagate trace context across service boundaries using W3C Trace Context and Baggage:

final propagator = CompositePropagator<Map<String, String>, String>([
  W3CTraceContextPropagator(),
  W3CBaggagePropagator(),
]);

// Inject into outgoing HTTP headers
final headers = <String, String>{};
propagator.inject(Context.current, headers, MapTextMapSetter(headers));
// Send headers with your HTTP request...

// Extract from incoming HTTP headers
final extractedContext = propagator.extract(
  OTel.context(),
  incomingHeaders,
  MapTextMapGetter(incomingHeaders),
);

// Create a child span in the extracted context
await extractedContext.run(() async {
  final span = tracer.startSpan('handle-request');
  // This span is part of the same distributed trace
  span.end();
});

Context also propagates across Dart async gaps and Isolates:

// Across Isolates
final result = await Context.current.runIsolate(() async {
  // Context is automatically restored in the new Isolate
  final span = tracer.startSpan('isolate-work');
  try {
    return await computeInIsolate();
  } finally {
    span.end();
  }
});

OpenTelemetry Metrics API

The Metrics API in OpenTelemetry provides a way to record measurements about your application. These measurements can be exported later as metrics, allowing you to monitor and analyze the performance and behavior of your application.

Concepts

Instrument Types

Usage Pattern

Similar to the Tracing API, the metrics API follows a multi-layered factory pattern:

  1. API Layer: Defines interfaces and provides no-op implementations
  2. SDK Layer: Provides concrete implementations
  3. Flutter Layer: Adds UI-specific functionality

The API follows the pattern of using factory methods for creation rather than constructors:

// Get a meter from the meter provider
final meter = OTel.meterProvider().getMeter('component_name');

// Create a counter instrument
final counter = meter.createCounter('my_counter');

// Record measurements
counter.add(1, {'attribute_key': 'attribute_value'});

For asynchronous instruments:

// Create an observable counter
final observableCounter = meter.createObservableCounter(
  'my_observable_counter',
  () => [Measurement(10, {'attribute_key': 'attribute_value'})],
);

Understanding Metric Types and When to Use Them

Instrument Type Use Case Example
Counter Count things that only increase Request count, completed tasks
UpDownCounter Count things that can increase or decrease Active requests, queue size
Histogram Measure distributions Request durations, payload sizes
Gauge Record current value CPU usage, memory usage
ObservableCounter Count things that only increase, collected on demand Total CPU time
ObservableUpDownCounter Count things that can increase or decrease, collected on demand Memory usage
ObservableGauge Record current value, collected on demand Current temperature

OpenTelemetry Logs API

The Logs API provides structured logging that integrates with traces and metrics. Unlike traditional logging frameworks, OpenTelemetry logs are first-class telemetry signals that carry context, severity, attributes, and can be correlated with the span that was active when the log was emitted.

Concepts

Quick Start

import 'package:middleware_dart_opentelemetry/middleware_dart_opentelemetry.dart';

void main() async {
  // Initialize with logs enabled (default)
  await OTel.initialize(
    serviceName: 'my-service',
    enableLogs: true,  // Default is true
  );

  // Get a logger
  final logger = OTel.logger('my-component');

  // Emit log records
  logger.emit(
    body: 'Application started',
    severityNumber: Severity.INFO,
  );

  // Log with attributes — prefer typed enum keys.
  logger.emit(
    body: 'User logged in',
    severityNumber: Severity.INFO,
    attributes: OTel.attributesFromSemanticMap({
      User.userId: 'user123',
      User.userRole: 'admin',
    }),
  );

  // Log an error with exception.
  try {
    throw Exception('Something went wrong');
  } catch (e, stackTrace) {
    logger.emit(
      body: 'Operation failed: $e',
      severityNumber: Severity.ERROR,
      attributes: OTel.attributesFromSemanticMap({
        ExceptionResource.exceptionType: e.runtimeType.toString(),
        ExceptionResource.exceptionStacktrace: stackTrace.toString(),
      }),
    );
  }
}

Intercepting print() Calls

Dartastic OpenTelemetry can automatically capture print() calls and convert them to OpenTelemetry logs:

await OTel.initialize(
  serviceName: 'my-service',
  logPrint: true,  // Enable print interception
  logPrintLoggerName: 'dart.print',  // Optional custom logger name
);

// Use runWithPrintInterception to capture prints
OTel.runWithPrintInterception(() {
  print('This will be captured as an OTel log');
  print('So will this');
});

// For async code
await OTel.runWithPrintInterceptionAsync(() async {
  print('Async print captured');
  await someAsyncOperation();
});

Log Severity Levels

Severity Use Case
Severity.TRACE / Severity.TRACE2-4 Fine-grained debugging
Severity.DEBUG / Severity.DEBUG2-4 Debug information
Severity.INFO / Severity.INFO2-4 General information
Severity.WARN / Severity.WARN2-4 Warning conditions
Severity.ERROR / Severity.ERROR2-4 Error conditions
Severity.FATAL / Severity.FATAL2-4 Critical failures

Basic Logging

// Get a logger from the default provider
final logger = OTel.loggerProvider().getLogger('my-service');

// Emit a simple log. Prefer typed enum keys (User, ExampleAttribute,
// Http, etc.) over raw strings.
logger.emit(
  severityNumber: Severity.INFO,
  body: 'User successfully logged in.',
  attributes: OTel.attributesFromSemanticMap({
    User.userId: 'user-123',
    ExampleAttribute.authMethod: 'oauth',
  }),
);

// Warning log.
logger.emit(
  severityNumber: Severity.WARN,
  body: 'Cache miss for requested key.',
  attributes: OTel.attributesFromSemanticMap({
    ExampleAttribute.cacheKey: 'profile_42',
    ExampleAttribute.cacheRegion: 'us-east-1',
  }),
);

// Error log.
logger.emit(
  severityNumber: Severity.ERROR,
  body: 'Failed to connect to database.',
  attributes: OTel.attributesFromSemanticMap({
    Database.dbSystem: 'postgresql',
    ErrorSemantics.errorType: 'ConnectionTimeout',
  }),
);

Log-to-Trace Correlation

Logs can be linked to the active span through Context, enabling powerful correlation in your backend:

final span = tracer.startSpan('process-order');
try {
  logger.emit(
    severityNumber: Severity.INFO,
    body: 'Processing order.',
    context: Context.current, // Links this log to the active span
    attributes: OTel.attributesFromSemanticMap({ExampleAttribute.orderId: 'order-789'}),
  );
  await processOrder();
} finally {
  span.end();
}

Custom Log Exporters

// Use a custom exporter
final customExporter = OtlpHttpLogRecordExporter(
  OtlpHttpLogRecordExporterConfig(
    endpoint: 'https://my-collector:4318',
    headers: {'Authorization': 'Bearer token'},
  ),
);

await OTel.initialize(
  serviceName: 'my-service',
  logRecordExporter: customExporter,
);

Console Logging (Development)

// Use console exporter for development
await OTel.initialize(
  serviceName: 'my-service',
  logRecordProcessor: SimpleLogRecordProcessor(ConsoleLogRecordExporter()),
);

Configuration via Environment Variables

Logs can be configured via environment variables:

# Set logs exporter (otlp, console, or none)
export OTEL_LOGS_EXPORTER=otlp

# Set logs-specific endpoint
export OTEL_EXPORTER_OTLP_LOGS_ENDPOINT=https://logs-collector:4318

# Configure batch processor
export OTEL_BLRP_SCHEDULE_DELAY=5000
export OTEL_BLRP_MAX_QUEUE_SIZE=4096

# Set log record limits
export OTEL_LOGRECORD_ATTRIBUTE_COUNT_LIMIT=128

Severity Levels

OpenTelemetry defines a fine-grained 24-level severity scale, grouped into standard levels:

Level Severities Use Case
TRACE TRACE, TRACE2-4 Finest-grained debugging information
DEBUG DEBUG, DEBUG2-4 Debugging information
INFO INFO, INFO2-4 Normal operational messages
WARN WARN, WARN2-4 Warning conditions
ERROR ERROR, ERROR2-4 Error conditions
FATAL FATAL, FATAL2-4 System is unusable

Severity levels support comparison operators for filtering:

if (severity >= Severity.WARN) {
  // Handle warning or above
}

Flexible Log Bodies

The body parameter accepts diverse types — not just strings:

// String body
logger.emit(body: 'Simple message.');

// Structured body (Map)
logger.emit(body: {'event': 'batch_complete', 'items': 42});

// List body
logger.emit(body: [
  {'job': 'resize_images', 'status': 'ok'},
  {'job': 'generate_thumbnails', 'status': 'failed'},
]);

Named Events

Use eventName to categorize logs as discrete events:

logger.emit(
  eventName: 'user_signup',
  severityNumber: Severity.INFO,
  body: 'New user registered.',
  attributes: OTel.attributesFromMap({
    'user.email_domain': 'example.com',
    'signup.source': 'organic',
  }),
);

Running with Environment Variables

Dartastic OpenTelemetry supports for all standard OpenTelemetry environment variables as defined in the OpenTelemetry Specification.

Environment variables provide a convenient way to configure OpenTelemetry without hardcoding values. All environment variable names are available as strongly-typed constants in the SDK for compile-time safety and IDE autocomplete. See lib/src/environment/env_constants.dart for a complete list.

How It Works

Dart environment variables can be set in two ways:

  1. System Environment Variables (Non-web only): Traditional POSIX environment variables 
    export OTEL_SERVICE_NAME=my-service
dart run

  2. Compile-time Constants (All platforms including Flutter web): Passed during compilation/execution

    For Dart commands (dart run, dart compile, dart test):

    # Using --define (or -D shorthand)
dart run --define=OTEL_SERVICE_NAME=my-service
dart compile exe -D=OTEL_SERVICE_NAME=my-service -o myapp
dart test -DOTEL_SERVICE_NAME=my-service

    For Flutter commands:

    # Flutter uses --dart-define (note the different flag name)
flutter run --dart-define=OTEL_SERVICE_NAME=my-service
flutter build apk --dart-define=OTEL_SERVICE_NAME=my-service

Priority: Compile-time constants (--define or --dart-define) take precedence over system environment variables. Explicit parameters to OTel.initialize() override both. Thus, POSIX env vars cannot override --dart-defines and neither POSIX env vars nor --dart-defines can override code. This is sensible and reduces security vectors.

Web Support: Flutter web and Dart web only support compile-time constants (--define or --dart-define), as browser environments don’t have access to system environment variables.

Using Environment Variable Constants

All OpenTelemetry environment variable names are available as typed constants:

import 'package:middleware_dart_opentelemetry/middleware_dart_opentelemetry.dart';

void main() {
  // Reference constants instead of strings
  final serviceName = EnvironmentService.instance.getValue(otelServiceName);
  final endpoint = EnvironmentService.instance.getValue(otelExporterOtlpEndpoint);
  
  print('Service: $serviceName');
  print('Endpoint: $endpoint');
}

Constants are defined for all 74 OpenTelemetry environment variables. See lib/src/environment/env_constants.dart for the complete list.

Supported Environmental Variables

Service Configuration

Constant Environment Variable Description Example
otelServiceName OTEL_SERVICE_NAME Sets the service name my-dart-app
otelResourceAttributes OTEL_RESOURCE_ATTRIBUTES Additional resource attributes environment=prod,region=us-west
otelLogLevel OTEL_LOG_LEVEL SDK internal log level INFO, DEBUG, WARN, ERROR
otelSdkDisabled OTEL_SDK_DISABLED Global off-switch — when true, the SDK installs no span processors, metric readers, or log record processors (true no-op across all three signals, including explicit overrides) true

OTLP Exporter Configuration

Constant Environment Variable Description Default Example
otelExporterOtlpEndpoint OTEL_EXPORTER_OTLP_ENDPOINT OTLP endpoint URL http://localhost:4318 https://otel-collector:4317
otelExporterOtlpProtocol OTEL_EXPORTER_OTLP_PROTOCOL Transport protocol http/protobuf grpc, http/protobuf, http/json
otelExporterOtlpHeaders OTEL_EXPORTER_OTLP_HEADERS Headers (key=value,…) None api-key=secret,tenant=acme
otelExporterOtlpTimeout OTEL_EXPORTER_OTLP_TIMEOUT Timeout in milliseconds 10000 5000
otelExporterOtlpCompression OTEL_EXPORTER_OTLP_COMPRESSION Compression algorithm None gzip

Signal-Specific Configuration

Per the OTel spec, the default exporter for every signal is otlp (HTTP/protobuf to http://localhost:4318). Each OTEL_*_EXPORTER env var accepts otlp (default), console (prints to stdout — useful for local debugging), or none (skips processor/reader installation for that signal entirely). OTEL_SDK_DISABLED=true silences all three signals globally and overrides everything else.

Traces
Constant Environment Variable Description Default
otelTracesExporter OTEL_TRACES_EXPORTER Trace exporter type (otlp, console, none) otlp
otelExporterOtlpTracesEndpoint OTEL_EXPORTER_OTLP_TRACES_ENDPOINT Traces-specific endpoint  
otelExporterOtlpTracesProtocol OTEL_EXPORTER_OTLP_TRACES_PROTOCOL Traces-specific protocol  
otelExporterOtlpTracesHeaders OTEL_EXPORTER_OTLP_TRACES_HEADERS Traces-specific headers  
Metrics
Constant Environment Variable Description Default
otelMetricsExporter OTEL_METRICS_EXPORTER Metrics exporter type (otlp, console, none) otlp
otelExporterOtlpMetricsEndpoint OTEL_EXPORTER_OTLP_METRICS_ENDPOINT Metrics-specific endpoint  
otelExporterOtlpMetricsProtocol OTEL_EXPORTER_OTLP_METRICS_PROTOCOL Metrics-specific protocol  
otelExporterOtlpMetricsHeaders OTEL_EXPORTER_OTLP_METRICS_HEADERS Metrics-specific headers  
Logs
Constant Environment Variable Description Default
otelLogsExporter OTEL_LOGS_EXPORTER Logs exporter type (otlp, console, none) otlp
otelExporterOtlpLogsEndpoint OTEL_EXPORTER_OTLP_LOGS_ENDPOINT Logs-specific endpoint  
otelExporterOtlpLogsProtocol OTEL_EXPORTER_OTLP_LOGS_PROTOCOL Logs-specific protocol  
otelExporterOtlpLogsHeaders OTEL_EXPORTER_OTLP_LOGS_HEADERS Logs-specific headers  
Batch LogRecord Processor (BLRP)
Constant Environment Variable Default Description
otelBlrpScheduleDelay OTEL_BLRP_SCHEDULE_DELAY 1000 Delay between exports (milliseconds)
otelBlrpExportTimeout OTEL_BLRP_EXPORT_TIMEOUT 30000 Export timeout (milliseconds)
otelBlrpMaxQueueSize OTEL_BLRP_MAX_QUEUE_SIZE 2048 Maximum queue size
otelBlrpMaxExportBatchSize OTEL_BLRP_MAX_EXPORT_BATCH_SIZE 512 Maximum batch size per export
LogRecord Limits
Constant Environment Variable Default Description
otelLogrecordAttributeValueLengthLimit OTEL_LOGRECORD_ATTRIBUTE_VALUE_LENGTH_LIMIT No limit Maximum length of attribute values
otelLogrecordAttributeCountLimit OTEL_LOGRECORD_ATTRIBUTE_COUNT_LIMIT 128 Maximum number of attributes

For the complete list of all supported environment variables with full documentation, see lib/src/environment/env_constants.dart.

Environment Usage Examples

Dart Application with Environment Variables

Note the ‘,’ in OTEL_RESOURCE_ATTRIBUTES for POSIX env vars but a ‘;’ for –dart-define. This is due to a Dart quirk.

# Set environment variables
export OTEL_SERVICE_NAME=my-backend-service
export OTEL_RESOURCE_ATTRIBUTES="service.version=1.2.3,deployment.environment=prod"
export OTEL_EXPORTER_OTLP_ENDPOINT=https://otel-collector:4318
export OTEL_EXPORTER_OTLP_PROTOCOL=http/protobuf
export OTEL_EXPORTER_OTLP_HEADERS=api-key=your-key
export OTEL_LOG_LEVEL=DEBUG

# Run your application
dart run bin/my_app.dart

Flutter Application with –dart-define

flutter run \
  --dart-define=OTEL_SERVICE_NAME=my-flutter-app \
  --dart-define=OTEL_RESOURCE_ATTRIBUTES="service.version=1.2.3;deployment.environment=prod"
  --dart-define=OTEL_EXPORTER_OTLP_ENDPOINT=https://otel-collector:4317 \
  --dart-define=OTEL_EXPORTER_OTLP_PROTOCOL=grpc \
  --dart-define=OTEL_EXPORTER_OTLP_HEADERS=api-key=your-key

Flutter Web (requires –dart-define)

# Web MUST use --dart-define (environment variables don't work in browsers)
flutter run -d chrome \
  --dart-define=OTEL_SERVICE_NAME=my-web-app \
  --dart-define=OTEL_EXPORTER_OTLP_ENDPOINT=https://collector:4318

Combining Both (–dart-define wins)

# Environment variable
export OTEL_SERVICE_NAME=from-environment

# --dart-define takes precedence
dart run --dart-define=OTEL_SERVICE_NAME=from-dart-define

# Result: Uses "from-dart-define"

In Code

import 'package:middleware_dart_opentelemetry/middleware_dart_opentelemetry.dart';

void main() async {
  // OTel.initialize() automatically reads environment variables
  // when parameters are not explicitly provided
  await OTel.initialize();
  
  // Environment variables are read automatically:
  // - OTEL_SERVICE_NAME
  // - OTEL_EXPORTER_OTLP_ENDPOINT
  // - OTEL_EXPORTER_OTLP_PROTOCOL
  // - And 90+ others...
  
  // Explicit parameters override environment variables
  await OTel.initialize(
    serviceName: 'explicit-service',  // Overrides OTEL_SERVICE_NAME
    endpoint: 'https://override:4318', // Overrides OTEL_EXPORTER_OTLP_ENDPOINT
  );
  
  // You can also read environment variables directly
  final endpoint = EnvironmentService.instance.getValue(otelExporterOtlpEndpoint);
  print('Using endpoint: $endpoint');
}

Testing with Environment Variables

Integration tests can use real environment variables:

# Run tests with environment variables
OTEL_SERVICE_NAME=test-service dart test

# Run tests with --dart-define
dart test --dart-define=OTEL_SERVICE_NAME=test-service

# Run the provided integration test script
./tool/test_env_vars.sh

The SDK includes an integration test suite (test/integration/environment_variables_test.dart) and a test script (tool/test_env_vars.sh) that demonstrates proper environment variable usage.

Minimal Code Example

import 'package:middleware_dart_opentelemetry/middleware_dart_opentelemetry.dart';

void main() async {
  // Initialize - automatically reads environment variables
  await OTel.initialize();

  // Get the default tracer
  final tracer = OTel.tracer();

  // Create a span
  final span = tracer.startSpan('my-operation');
  
  try {
    // Your code here
    await doWork();
  } catch (e, stackTrace) {
    span.recordException(e, stackTrace: stackTrace);
    span.setStatus(SpanStatusCode.error, 'Operation failed');
  } finally {
    span.end();
  }
}

Since middleware_opentelemetry exports all the classes of opentelemetry_api, refer to opentelemetry_api for documentation of API classes.

See the /example folder for more examples.

🌐 HTTP Client Instrumentation

Middleware Dart OpenTelemetry includes automatic HTTP client instrumentation for out-of-the-box tracing of outbound HTTP requests. This allows you to:

Create spans for each HTTP request

Propagate W3C Trace Context headers (traceparent, tracestate)

Capture HTTP metadata (method, URL, status, timings, errors)

Automatically connect client spans to downstream services

This works for any Dart backend, CLI, or Flutter-based network implementation using dart:io.

✨ Features of OTelHttpClient

Usage

1. Import the package

   import 'package:middleware_dart_opentelemetry/middleware_dart_opentelemetry.dart';
   import 'dart:io';

2. Initialize OpenTelemetry

    await OTel.initialize(
      serviceName: 'my-dart-service',
      endpoint: '',
    );

3. Wrap the Dart HttpClient with OTelHttpClient

  final client = OTelHttpClient(HttpClient());

4. Make instrumented HTTP requests

  final request = await client.getUrl(Uri.parse('https://api.example.com/data'));
  final response = await request.close();
  
  print('Status: ${response.statusCode}');

That’s all you need — spans are now generated automatically and exported via your configured exporter.

🚀 Full Example

  void main() async {
    await OTel.initialize(serviceName: 'http-client-demo');
    
    final tracer = OTel.tracer();
    final client = OTelHttpClient(HttpClient());
    
    final span = tracer.startSpan('demo-operation');
    
    await Context.withSpan(span, () async {
    final request = await client.getUrl(Uri.parse('https://middleware.io'));
    final response = await request.close();
    print('Status: ${response.statusCode}');
    });
    span.end();
  }

This automatically creates:

OpenTelemetry Metrics API

The Metrics API in OpenTelemetry provides a way to record measurements about your application. These measurements can be exported later as metrics, allowing you to monitor and analyze the performance and behavior of your application.

Concepts

Instrument Types

Usage Pattern

Similar to the Tracing API, the metrics API follows a multi-layered factory pattern:

  1. API Layer: Defines interfaces and provides no-op implementations
  2. SDK Layer: Provides concrete implementations
  3. Flutter Layer: Adds UI-specific functionality

The API follows the pattern of using factory methods for creation rather than constructors:

// Get a meter from the meter provider
final meter = OTel.meterProvider().getMeter('component_name');

// Create a counter instrument
final counter = meter.createCounter('my_counter');

// Record measurements
counter.add(1, {'attribute_key': 'attribute_value'});

For asynchronous instruments:

// Create an observable counter
final observableCounter = meter.createObservableCounter(
  'my_observable_counter',
  () => [Measurement(10, {'attribute_key': 'attribute_value'})],
);

Understanding Metric Types and When to Use Them

Instrument Type Use Case Example
Counter Count things that only increase Request count, completed tasks
UpDownCounter Count things that can increase or decrease Active requests, queue size
Histogram Measure distributions Request durations, payload sizes
Gauge Record current value CPU usage, memory usage
ObservableCounter Count things that only increase, collected on demand Total CPU time
ObservableUpDownCounter Count things that can increase or decrease, collected on demand Memory usage
ObservableGauge Record current value, collected on demand Current temperature

Integration with Dart/Flutter

All three signal APIs (Traces, Metrics, Logs) follow the same multi-layered factory pattern:

  1. API Layer: Defines interfaces and provides no-op implementations
  2. SDK Layer: Provides concrete implementations with export and processing
  3. Flutter Layer: Adds UI-specific functionality (route observation, app lifecycle, etc.)

Middleware.io provides an OpenTelemetry Observability backend specifically built for Dart and Flutter applications. Features include:

License

Apache 2.0 - See the LICENSE file for details.

Commercial Support

Middleware.io provides an OpenTelemetry support, training, consulting, enhanced private packages and an Observability backend customized for Flutter apps, Dart backends, and any other service or process that produces OpenTelemetry data. Middleware.io is built on open standards, specifically catering to Flutter and Dart applications with the ability to show Dart source code lines and function calls from production errors and logs.

Middleware.io offers:

Additional information

This site is open source. Improve this page.