Loudness range (LRA) — EBU Tech 3342 short-term distribution, P95−P10 after gating

Tokens per second meter for pi CLI — live TPS, rolling avg, mean, p95

LLM & tool-call telemetry: storage interface, in-memory store, and pure aggregation (llmCallStats/toolCallStats with p95).

Lightweight, zero-dependency Express middleware for server latency profiling — p50/p95/p99, slow-request capture, event-loop lag, GC tracking.

Zero-dependency HTTP monitoring, APM metrics, percentile latencies (p50/p90/p95/p99), threshold alert callbacks, Prometheus format exporter, slow request profiler, and real-time embedded web dashboard for Express.js.

Lightweight, zero-dependency Express middleware for server latency profiling — p50/p95/p99, slow-request capture, event-loop lag, GC tracking.

Benchmark LLM streaming — TTFT, TPS, p50/p95/p99, cost ceiling, and regression diffing across OpenAI and Anthropic.

API latency testing with p50, p95, p99 percentile analysis for CI/CD

Generic performance harness v0.3 strict — p50/p95/p99 measurement + regression detection + baseline persistence + 3-layer harness (serial + concurrent + memory) + strict mode (iter 400 + Welch |t|>3 + p95 delta 10%) + 38 package coverage + 14-axis release

Turn HAR timing data into k6 SLO baselines — your recordings already know your p95

Quick GitLab pipeline stats: avg, p50, p95, and max per job and group. Groups are configured via presets for different release flows.

Vital signs for your MCP server: inspect capabilities, benchmark tool-call latency (p50/p95/p99), and assert health in CI. The ab/k6/pytest for MCP servers — non-interactive, scriptable, LLM-free.

An adaptive, speculative request-hedging library that automatically cuts p95 tail latency.

Calculate n-th percentile

Lightweight HTTP load benchmark — RPS, p50/p95/p99 latencies, status code breakdown, zero dependencies

MPT - Multi-channel LLM stress testing tool (RPM/TTFT/TPS/P95). Supports OpenAI-compatible APIs, SSE streaming, adaptive scheduling, and multi-model benchmarking.

Active Context Pruning — model-driven context management for OpenCode (hardened fork of DCP with 35 bug fixes)

k6 load-testing skill for Bluetemberg — scenario taxonomy, shared config/requests libs, p95 threshold discipline, and __ENV-parametrized suites wired into CI.

Repo-aware MCP tools for AI coding agents: intent search, symbol navigation, advisory call graphs, exact reads, and freshness-aware recovery.

A Node.js module for verifying email addresses