ingero

概述

Ingero 追踪从 Linux 内核事件到 Python 源码的完整链路，开销小于 2%。它通过时间戳和 PID 关联 CPU 调度器事件、CUDA 运行时 API 调用和 CUDA 驱动 API 调用，生成因果链解释。MCP 服务器组件允许 Claude Code 和 Cursor 等 AI 助手通过自然语言询问 GPU 性能问题，并获取包含 Python 源码的已解析因果链，无需 SQL 或十六进制地址。

试试问 AI

装完之后，这里有 5 个你可以让 AI 做的事：

• mcp

  Expose Ingero as a Model Context Protocol server for AI-assisted GPU observability

  试试:What caused the GPU stall?试试:Explain the GPU performance issues试试:Analyze the CUDA events from the last 5 minutes
• trace

  Trace GPU workloads and capture CUDA events in real-time

  试试:Start tracing the GPU events试试:Capture CUDA API calls for PID 4821试试:Begin live event stream monitoring
• explain

  Generate an automated diagnosis of GPU performance issues

  试试:Explain the GPU issues from the last 5 minutes试试:Analyze the causal chain of GPU events试试:Provide a root cause analysis for GPU stalls
• check

  Verify system readiness for GPU tracing

  试试:Check if the system is ready for GPU tracing试试:Verify eBPF kernel support试试:Check if Ingero can trace the GPU
• query

  Query the GPU trace data for specific time periods

  试试:Query GPU events from the last hour试试:Show me high-latency CUDA operations试试:Find GPU preemption events
• demo

  Run demonstration scenarios with synthetic data

  试试:Run the incident demo scenario试试:Show a GPU performance issue demo试试:Run the demo without a GPU

可对比工具

安装

VERSION=0.16.0
curl -fsSL "https://github.com/ingero-io/ingero/releases/download/v${VERSION}/ingero_${VERSION}_linux_amd64.tar.gz" | tar xz
sudo mv ingero /usr/local/bin/

ingero mcp

{
  "mcpServers": {
    "ingero": {
      "command": "ingero",
      "args": ["mcp"]
    }
  }
}

FAQ

MCP 服务器提供了什么功能？

MCP 服务器为 AI 助手提供了通过自然语言询问 GPU 性能数据、因果链和诊断的工具。

使用 Ingero 的开销有多大？

开销小于 2%（在 GPU 硬件上测得，开启堆栈追踪时为 RTX 3090 到 H100 之间的 0.4-1.7%）。

ingero 对比