samples

page_type	sample
languages	typescript
products	azure-monitor
urlFragment	microsoft-opentelemetry-typescript

Microsoft OpenTelemetry distribution samples for TypeScript

These sample programs show how to use the @microsoft/opentelemetry distribution in common scenarios.

File Name	Description
basicConnection.ts	Demonstrates how to configure Microsoft OpenTelemetry using a connection string.
cloudRole.ts	Demonstrates how to set Cloud Role Name and Cloud Role Instance using OpenTelemetry Resource attributes.
customMetric.ts	Demonstrates how to generate custom metrics that will be sent to Azure Monitor.
customTrace.ts	Demonstrates how to generate custom traces that will be sent to Azure Monitor.
liveMetrics.ts	Demonstrates how to enable or disable Live Metrics for real-time monitoring.
offlineStorage.ts	Demonstrates how to configure offline storage and automatic retries for telemetry.
otlpExporter.ts	Demonstrates how to enable the OTLP exporter alongside Azure Monitor to send telemetry to two locations.
redactQueryStrings.ts	Demonstrates how to redact URL query strings from telemetry to protect sensitive information.
sampling.ts	Demonstrates how to enable sampling to reduce data ingestion volume and control costs.
langchainInstrumentation.ts	Demonstrates how to enable LangChain instrumentation to trace GenAI operations, including the enableSensitiveData toggle for capturing message content.
aks-langchain	Deploys a loader-based LangChain auto-instrumentation reproduction to AKS with a Dockerfile, Kubernetes manifest, and Azure Monitor walkthrough.
openaiInstrumentation.ts	Demonstrates how to enable OpenAI Agents SDK instrumentation to trace GenAI operations.
a365Export.ts	Demonstrates A365 observability export: token resolver setup, dual export with Azure Monitor, and span routing by tenant/agent.
a365ManualScopes.ts	Traces a full agent turn with manual scopes (InvokeAgent → Inference → ExecuteTool → Inference → Output) and cross-service context propagation.
a365HostingMiddleware.ts	Demonstrates A365 hosting middleware (BaggageMiddleware, OutputLoggingMiddleware, ObservabilityHostingManager, ScopeUtils).

Prerequisites

• Node.js LTS (>= 20.0.0)
• TypeScript (install via npm install -g typescript)
• An Azure subscription with an Application Insights resource

Setup

1. Install dependencies:

npm install

2. Compile the samples:

npm run build

3. Copy sample.env to .env and fill in the variables needed by the samples you want to run:

cp sample.env .env

Variable	Used by	Description
APPLICATIONINSIGHTS_CONNECTION_STRING	Most samples	Connection string from your Application Insights resource.
OTEL_SERVICE_NAME	cloudRole.ts	Service name mapped to Cloud Role Name.
OTEL_SERVICE_NAMESPACE	cloudRole.ts	Service namespace prepended to Cloud Role Name.
OTEL_SERVICE_INSTANCE_ID	cloudRole.ts	Service instance mapped to Cloud Role Instance.
AZURE_OPENAI_API_KEY	langchainInstrumentation.ts, openaiInstrumentation.ts	API key for your Azure OpenAI resource.
AZURE_OPENAI_INSTANCE_NAME	langchainInstrumentation.ts, openaiInstrumentation.ts	Azure OpenAI resource name (e.g. contoso).
AZURE_OPENAI_DEPLOYMENT_NAME	langchainInstrumentation.ts, openaiInstrumentation.ts	Model deployment name (e.g. gpt-4o).
AZURE_OPENAI_API_VERSION	langchainInstrumentation.ts, openaiInstrumentation.ts	Azure OpenAI API version (default 2024-06-01).
A365_BEARER_TOKEN	a365Export.ts, a365ManualScopes.ts	Auth token for the Agent365 observability API. In production you would acquire this via MSAL; the samples read it from this env var as a shortcut.

4. Run a sample:

node dist/basicConnection.js

Or pass the connection string directly:

APPLICATIONINSIGHTS_CONNECTION_STRING="<your connection string>" node dist/basicConnection.js

Capturing GenAI message content (LangChain)

By default, LangChain instrumentation hides sensitive GenAI message content — prompts, completions, tool call arguments/results, and system instructions — so telemetry carries only non-sensitive metadata (model, token counts, latency, tool names, etc.).

langchainInstrumentation.ts demonstrates the enableSensitiveData toggle. Set the top-level option to record message content:

useMicrosoftOpenTelemetry({
  // Hidden by default. Only enable in trusted, non-production environments
  // where capturing message content is intentional.
  enableSensitiveData: true,
  azureMonitor: {
    /* ... */
  },
  instrumentationOptions: {
    langchain: { enabled: true },
  },
});

Note: enableSensitiveData defaults to false. Only enable it in trusted, non-production environments where capturing message content is intentional. This setting currently applies to LangChain instrumentation.