Principle:Protectai Llm guard Observability Instrumentation
| Knowledge Sources | |
|---|---|
| Domains | Observability, DevOps, Monitoring |
| Last Updated | 2026-02-14 12:00 GMT |
Overview
An instrumentation pattern that integrates OpenTelemetry tracing and metrics into a FastAPI application to provide distributed tracing, latency monitoring, and scanner validity counters.
Description
Observability instrumentation adds three capabilities to the LLM Guard API:
- Distributed tracing: Automatic span creation for each HTTP request via FastAPIInstrumentor, with support for OTLP HTTP, console, and AWS X-Ray exporters.
- Metrics collection: Scanner validity counters tracking pass/fail rates per scanner, exportable via Prometheus, OTLP HTTP, or console.
- Health endpoints: /healthz and /readyz endpoints excluded from tracing to avoid noise.
Usage
Use this principle when deploying LLM Guard in production environments that require monitoring. Configure via the tracing and metrics sections of the YAML configuration file.
Theoretical Basis
# Pseudocode for observability setup
resource = Resource(service_name=app_name, service_version=version)
configure_tracer(resource, tracing_config) # TracerProvider + exporter
configure_metrics(resource, metrics_config) # MeterProvider + exporter
instrument_fastapi(app, excluded_urls=["healthz", "readyz", "metrics"])
Related Pages
Implemented By
Page Connections
Double-click a node to navigate. Hold to expand connections.
Principle
Implementation
Heuristic
Environment