Latency Tracker
Per-call and aggregated latency tracking for MEV infrastructure.
When to Use
- Implementing performance monitoring
- Debugging slow operations
- Adding instrumentation to code paths
- Tracking end-to-end latency
- Setting up alerting thresholds
Workflow
Step 1: Define Span Hierarchy
Structure spans from e2e_flow down to individual calls.
Step 2: Instrument Code
Wrap operations in tracker.span() calls.
Step 3: Set Alert Thresholds
Configure alerts for
[Description truncada. Veja o README completo no GitHub.]