Send to Datadog

Configure the OpenLIT SDK to send AI observability data to Datadog via the Datadog Agent OTLP receiver

Datadog accepts OpenTelemetry data through the Datadog Agent’s built-in OTLP receiver. Your application sends traces and metrics to the local Datadog Agent over localhost:4318, and the agent forwards them to the Datadog backend using your configured API key.

Prerequisites

This guide assumes you have a Datadog Agent already installed and configured with OTLP ingestion enabled. If you need to install the agent, refer to the Datadog Agent Installation Guide.

To enable OTLP in the Datadog Agent, add the following to your datadog.yaml:

otlp_config:
  receiver:
    protocols:
      http:
        endpoint: 0.0.0.0:4318

Get started

1. Install the OpenLIT SDK

   Python

   pip install openlit
   

   TypeScript

   npm install openlit
   
2. Configure the SDK to send to the Datadog Agent

   Point otlp_endpoint at the Datadog Agent running on localhost. No API key or authentication header is needed in the application — the agent handles authentication to the Datadog backend.

   Python

   Using function arguments:

   import openlit
   
   openlit.init(
       otlp_endpoint="http://localhost:4318"
   )
   

   Using environment variables:

   export OTEL_EXPORTER_OTLP_ENDPOINT="http://localhost:4318"
   export OTEL_SERVICE_NAME="my-ai-service"
   export OTEL_DEPLOYMENT_ENVIRONMENT="production"
   

   import openlit
   
   openlit.init()
   

   TypeScript

   Using function arguments:

   import Openlit from 'openlit';
   
   Openlit.init({
       otlpEndpoint: 'http://localhost:4318'
   });
   

   Using environment variables:

   export OTEL_EXPORTER_OTLP_ENDPOINT="http://localhost:4318"
   

   import Openlit from 'openlit';
   
   Openlit.init();
   

   The Datadog Agent handles authentication and forwarding to Datadog. You do not need to set otlp_headers or a Datadog API key in your application configuration.

Full example

import openlit
import openai

openlit.init(
    otlp_endpoint="http://localhost:4318",
    service_name="my-ai-service",
    environment="production"
)

client = openai.OpenAI()
response = client.chat.completions.create(
    model="gpt-4o",
    messages=[{"role": "user", "content": "Hello!"}]
)
print(response.choices[0].message.content)

Traces appear in Datadog APM under the service name my-ai-service.

---

Send to OpenLIT

Send to a self-hosted OpenLIT platform instance

Send to Grafana Cloud

Configure the Grafana OTLP gateway endpoint and Basic Auth headers

Integrations

60+ LLM providers, AI frameworks, and vector databases