TraceLens
API observability — synthetic load, streaming percentiles, error budgets, tracing.
Load controls
Every control drives the live simulation.
300rps
T+0.0s · 30s window
Live service health
Across all endpoints
Throughput
—
requests / second
Error rate
—
non-2xx responses
p50 latency
—
median
p95 latency
—
95th percentile
p99 latency
—
99th percentile
Latency
Percentiles over time
Rolling-window p50 / p95 / p99, sampled every 500ms.
p50p95p99
p99 by endpoint
Tail latency comparison across the fleet.
Error budget
99.9% success SLO over a 30s rolling window.
Budget remaining100.0%
Burn rate
—
1.0× = spending at SLO pace
Budget burn-down
remaining %
Traces
Slowest recent requests — click one for the span waterfall.
Collecting samples…
Alert rules
Rules evaluate against live stats every tick.
New rule
Registered immediately and evaluated on the next tick.
30s
Active rules (2)
30s cooldown per rule prevents alert spam.
Error rate spike
error rate > 1.00% · all · 30s
Embeddings p99
p99 latency > 1200ms · embeddings · 30s
Alert feed
Newest first.
No alerts yet. Flip on Chaos: 10× errors and the error-rate rule will fire for real.