OpenTelemetry Integration | NVIDIA NeMo Guardrails Library Developer Guide

The NVIDIA NeMo Guardrails library follows OpenTelemetry best practices. The library uses only the API, and the host application configures the SDK. The following sections explain how to install and configure the OpenTelemetry SDK.

Installation

Choose one of the following options for installing the library with tracing support, the OpenTelemetry SDK, and the OpenTelemetry Protocol (OTLP) exporter.

For basic tracing support in the NeMo Guardrails library:

$ pip install nemoguardrails[tracing]

For development with the OpenTelemetry SDK:

$ pip install nemoguardrails[tracing] opentelemetry-sdk

For production with the OpenTelemetry SDK and the OpenTelemetry Protocol (OTLP) exporter:

$ pip install nemoguardrails[tracing] opentelemetry-sdk opentelemetry-exporter-otlp

Configuration Examples

The following examples show how to configure the library with the OpenTelemetry SDK for development and production use cases.

Console Output (Development)

1 from opentelemetry import trace
2 from opentelemetry.sdk.trace import TracerProvider
3 from opentelemetry.sdk.trace.export import BatchSpanProcessor, ConsoleSpanExporter
4 from opentelemetry.sdk.resources import Resource
5 
6 # Configure OpenTelemetry before NeMo Guardrails
7 resource = Resource.create({"service.name": "my-guardrails-app"})
8 tracer_provider = TracerProvider(resource=resource)
9 trace.set_tracer_provider(tracer_provider)
10 
11 console_exporter = ConsoleSpanExporter()
12 tracer_provider.add_span_processor(BatchSpanProcessor(console_exporter))
13 
14 # Configure NeMo Guardrails
15 from nemoguardrails import LLMRails, RailsConfig
16 
17 config_yaml = """
18 models:
19   - type: main
20     engine: openai
21     model: gpt-4o-mini
22 
23 tracing:
24   enabled: true
25   adapters:
26     - name: OpenTelemetry
27 """
28 
29 config = RailsConfig.from_content(yaml_content=config_yaml)
30 
31 rails = LLMRails(config)

OTLP Exporter (Production)

1 from opentelemetry import trace
2 from opentelemetry.sdk.trace import TracerProvider
3 from opentelemetry.sdk.trace.export import BatchSpanProcessor
4 from opentelemetry.exporter.otlp.proto.grpc.trace_exporter import OTLPSpanExporter
5 from opentelemetry.sdk.resources import Resource
6 
7 resource = Resource.create({"service.name": "my-guardrails-app"})
8 tracer_provider = TracerProvider(resource=resource)
9 trace.set_tracer_provider(tracer_provider)
10 
11 otlp_exporter = OTLPSpanExporter(endpoint="http://localhost:4317", insecure=True)
12 tracer_provider.add_span_processor(BatchSpanProcessor(otlp_exporter))
13 
14 # Use with NeMo Guardrails as above

OpenTelemetry Ecosystem Compatibility

The library works with the entire OpenTelemetry ecosystem, including the following components.

Component	Examples
Exporters	Jaeger, Zipkin, Prometheus, New Relic, Datadog, AWS X-Ray, and Google Cloud Trace.
Collectors	OpenTelemetry Collector and vendor-specific collectors.
Backends	Any system that accepts OpenTelemetry traces.

Refer to the OpenTelemetry Registry for the complete list.

Exporting Logs

To also forward guardrails Python log records into your OpenTelemetry backend with trace correlation, see OpenTelemetry Logs.