Analytics Overview
Analytics in Guardrails provides deep visibility into how your AI systems behave in production.
It enables you to measure safety, detect risk, optimize performance, and prove compliance across all guardrail executions.
Why analytics matters
Without analytics, AI safety systems become opaque and unverifiable.
Guardrails analytics allows you to:
- Track every guardrail execution
- Measure pass/fail rates over time
- Identify risky prompts and failure patterns
- Monitor latency and performance impact
- Audit usage for compliance and governance
What is tracked
Each guardrail execution emits structured analytics events, including:
- Execution outcome (pass / fail)
- Execution latency
- Guardrail-level results
- Profile used
- API key and user context
- Timestamp and environment metadata
These events form the foundation for dashboards, alerts, and audits.
Analytics architecture (high-level)
Guardrails analytics follows an event-driven model:
- Runtime execution emits analytics events
- Events are ingested into the analytics pipeline
- Queries aggregate events into metrics
- Dashboards visualize trends and anomalies
This design ensures analytics is decoupled, scalable, and non-blocking.
Who should use analytics
Analytics is designed for:
- Engineering teams monitoring production AI
- Security teams auditing safety controls
- Compliance teams validating policy enforcement
- Product teams optimizing user experience
Next steps
- Learn how events are structured → Events
- Understand analytics queries → Queries
- Explore dashboards and metrics → Dashboards