GenPark Voice Latency

제공된 음성 파이프라인 타임스탬프를 기록하고 지연 시간, 병목 지점 및 최근접 순위 백분위수를 계산합니다. 로컬 인메모리 MCP이며 자동 계측은 없습니다.

GitHub
7
이 MCP 사용해 보기스폰서

문서

genpark-voice-latency

Record supplied voice pipeline timestamps in session memory and calculate latency, bottlenecks and nearest-rank percentiles. No automatic instrumentation.

Python 3.9+; standard library runtime; MIT license.

Install and run

Download genpark-voice-latency.mcpb from GitHub Releases and install with an MCPB-compatible client. Python must be installed and available as python.

Alternatively clone this repository and configure an MCP stdio server with command python and arguments containing the absolute path to mcp_server.py.

Smithery listing

Tools

  • record_pipeline_event
  • compute_turn_latency_breakdown
  • generate_sla_diagnostic_report
  • run_benchmark_telemetry_profiling

Run python -m unittest discover -s tests for regression checks. The official MCP SDK integration check uses the development dependency mcp: python tests/check_mcp.py.

Limitations

These are deterministic helpers operating on supplied structured data, not machine-learning models. Input and output remain in the local process. No hosted endpoint, automatic file access or network access is required. State lasts only for the current process. Benchmark tools run synthetic examples in isolated state; their status is not a production-quality certification.

Explicit timestamps must share a clock and unit (milliseconds); repeated stage-pair durations are summed. Default SLA is 800 ms. No persistence or telemetry collection occurs automatically.