Analytics Queries

Analytics queries transform raw events into actionable insights.

They aggregate, filter, and compute metrics across time, profiles, and guardrails.

Query philosophy

Guardrails analytics follows three principles:

1. Queries are read-only
2. Queries are deterministic
3. Queries never affect runtime execution

This ensures analytics is safe and reliable in production systems.

Common query dimensions

Analytics queries typically group or filter by:

• Time range (last 24h, 7d, 30d)
• Profile
• Guardrail name
• API key
• Pass / fail outcome

These dimensions allow flexible analysis.

Example insights

Using analytics queries, you can answer questions like:

• How many guardrail executions occurred today?
• Which guardrails fail most frequently?
• What is the average execution latency?
• Which profiles have the highest failure rates?
• Are failures increasing over time?

Guardrail performance queries

Guardrail-level queries compute:

• Execution count per guardrail
• Failure rate
• Average execution time

These metrics help identify:

• Ineffective guardrails
• Performance bottlenecks
• Overly aggressive rules

Time-series queries

Time-based queries enable:

• Trend analysis
• Spike detection
• Regression monitoring

They are commonly used for dashboards and alerts.

Next steps

• See queries visualized in dashboards → Dashboards