Observability Reference

Reference for key metrics, filters, and rule controls

Metrics Reference

Metric	Meaning	Typical Use
Total Requests	Count of inferences in range	Volume trend checks
Success Rate	Share of successful requests	Reliability monitoring
Avg / p95 / p99 Latency	End-to-end response time	Performance regression detection
TTFT	Time to first token	Streaming responsiveness
Input / Output Tokens	Token consumption totals	Usage and cost analysis
Requests per Second	Throughput over time	Capacity planning

Common Request Filters

Filter	Purpose
Date Range	Bound investigation window
Project / Deployment	Scope to workload owner
Endpoint Type	Separate chat/completion patterns
Success Status	Focus on failures or successes
Latency Bounds	Isolate slow requests
Token Bounds	Identify heavy prompts or outputs

Rule Operations

Operation	Result
Create	Adds new enforcement rule
Edit	Updates criteria, scope, or status
Disable	Stops enforcement but keeps history
Delete	Removes rule from active inventory