Benchmark latency, throughput, and telemetry across OpenAI-compatible inference servers.
npx skills add https://github.com/air-gapped/skills --skill aiperf
Please help me install this Agent Skill. Skill: aiperf Source: https://github.com/air-gapped/skills/tree/main/.claude/skills/aiperf Command: npx skills add https://github.com/air-gapped/skills --skill aiperf
AIPerf provides a vendor-neutral framework to benchmark latency, throughput, and telemetry for any OpenAI-compatible inference server.
Run a baseline benchmark against your endpoint to validate latency, throughput, and goodput using a representative workload.
High-intent search queries and answers about installing and using this skill.
You can benchmark latency and throughput by running a vendor-neutral framework against your OpenAI-compatible inference server using a Python environment and a target endpoint URL to generate per-request metrics.
Goodput measurement in AI inference benchmarking calculates the fraction of requests meeting all SLOs while simultaneously collecting GPU and server telemetry under realistic conditions.
Yes, you can replay production traces from Mooncake, Bailian, or BurstGPT datasets to accurately reproduce traffic patterns and tail latency for your inference server.
Yes, the benchmarking framework supports extensibility via a plugin system, allowing you to add new endpoints, datasets, telemetry exporters, or graders to customize your benchmarking workflow.
To run a baseline benchmark against your endpoint and validate latency, throughput, and goodput, you need a Python environment, a target endpoint URL, and a benchmark configuration using a representative workload.
TehminanazIntegrate the OpenAI Agents SDK with Groq's endpoint in FastAPI applications.
air-gappedBenchmark production vLLM deployments for latency, throughput, and SLO compliance.
annealing-inversionProfile an SGLang server run and generate a Chrome trace and server_args.json.
jmdueaIdentify and resolve performance bottlenecks in Rust applications.
