Quick Start for Tracing Guardrails

The following is a minimal setup to enable tracing using the OpenTelemetry SDK.

1. Install the NeMo Guardrails library and the OpenTelemetry SDK.

   $
   pip install nemoguardrails[tracing] opentelemetry-sdk

2. Set up tracing as follows and save as trace_example.py.

   1
   # trace_example.py
   2
   from opentelemetry import trace
   3
   from opentelemetry.sdk.trace import TracerProvider
   4
   from opentelemetry.sdk.trace.export import BatchSpanProcessor, ConsoleSpanExporter
   5
   from opentelemetry.sdk.resources import Resource
   6
   from nemoguardrails import LLMRails, RailsConfig
   7
   8
   # Configure OpenTelemetry
   9
   resource = Resource.create({"service.name": "guardrails-quickstart"})
   10
   tracer_provider = TracerProvider(resource=resource)
   11
   trace.set_tracer_provider(tracer_provider)
   12
   tracer_provider.add_span_processor(BatchSpanProcessor(ConsoleSpanExporter()))
   13
   14
   # Configure guardrails with tracing
   15
   config_yaml = """
   16
   models:
   17
     - type: main
   18
       engine: openai
   19
       model: gpt-4o-mini
   20
   21
   rails:
   22
     config:
   23
       streaming: true
   24
   25
   tracing:
   26
     enabled: true
   27
     adapters:
   28
       - name: OpenTelemetry
   29
   """
   30
   31
   config = RailsConfig.from_content(yaml_content=config_yaml)
   32
   rails = LLMRails(config)
   33
   response = rails.generate(messages=[{"role": "user", "content": "Hello!"}])
   34
   print(f"Response: {response}")

3. Run the script:

   $
   python trace_example.py