Repository navigation
LLM API Concurrent Test v1.0 Release Notes
🎉 首次正式发布!/ First Official Release!
这是 LLM API Concurrent Test 的首次正式发布版本,为您提供专业的 LLM API 并发性能测试解决方案。
This is the first official release of LLM API Concurrent Test, providing you with a professional concurrent performance testing solution for LLM APIs.
✨ 核心功能 / Core Features
🚀 并发负载测试 / Concurrent Load Testing
- 支持 1-20 个并发连接同时测试
- Support for 1-20 concurrent connections simultaneous testing
- 智能超时控制,确保测试数据准确性
- Intelligent timeout control ensuring test data accuracy
📊 精确性能指标 / Precise Performance Metrics
- 首次响应时间测量: 精确测量到第一个内容 token 的时间
- First Token Latency Measurement: Precisely measures time to first content token
- Token 输出速度: 使用官方
tiktoken库进行精确计算 - Token Output Speed: Uses official
tiktokenlibrary for precise calculation - 总计性能统计: 包含成功率、平均响应时间、总输出速度
- Comprehensive Statistics: Including success rate, average response time, total output speed
🌐 多语言界面 / Multilingual Interface
- 支持中文 (
cn) 和英文 (en) 界面 - Support for Chinese (
cn) and English (en) interfaces - 配置文件和程序输出均支持双语
- Both configuration files and program output support bilingual
⚙️ 灵活配置 / Flexible Configuration
- YAML 配置文件,支持详细参数设置
- YAML configuration file with detailed parameter settings
- 可配置并发数、超时时间、API 端点等
- Configurable concurrent count, timeout, API endpoints, etc.
🛠️ 技术特性 / Technical Features
- .NET 8: 基于最新 .NET 8 框架构建
- 自包含部署: 无需安装 .NET 运行时
- Self-contained Deployment: No need to install .NET runtime
- 跨平台支持: 支持 Windows、Linux、macOS 多架构
- Cross-platform Support: Supports Windows, Linux, macOS multi-architecture
- MIT 开源协议: 最宽松的开源许可证
- MIT License: Most permissive open source license
📦 支持的平台 / Supported Platforms
| 平台/Platform | 架构/Architecture | 文件名/Filename |
|---|---|---|
| Windows | x64 | LLMApiConcurrentTest-v1.0-win-x64.zip |
| Windows | x86 | LLMApiConcurrentTest-v1.0-win-x86.zip |
| Windows | ARM64 | LLMApiConcurrentTest-v1.0-win-arm64.zip |
| Linux | x64 | LLMApiConcurrentTest-v1.0-linux-x64.zip |
| Linux | ARM64 | LLMApiConcurrentTest-v1.0-linux-arm64.zip |
| macOS | x64 (Intel) | LLMApiConcurrentTest-v1.0-osx-x64.zip |
| macOS | ARM64 (Apple Silicon) | LLMApiConcurrentTest-v1.0-osx-arm64.zip |
🚀 快速开始 / Quick Start
- 下载对应平台的版本 / Download the version for your platform
- 解压到任意目录 / Extract to any directory
- 复制
config.yaml.template为config.yaml/ Copyconfig.yaml.templatetoconfig.yaml - 配置你的 API 设置 / Configure your API settings
- 运行程序开始测试 / Run the program to start testing
Windows:
LLMApiConcurrentTest.exeLinux/macOS:
./LLMApiConcurrentTest📖 配置示例 / Configuration Example
# 并发请求数量 / Number of concurrent requests
concurrent_num: 3
# API 基础 URL / API Base URL
baseurl: "https://api.openai.com/v1"
# API 密钥 / API Key
api_key: "your-api-key-here"
# 模型名称 / Model Name
model_name: "gpt-3.5-turbo"
# 超时时间(确保并发测试准确性)/ Timeout (ensuring concurrent test accuracy)
timeoutsec: 10
# 界面语言 / Interface Language
language: "cn" # or "en"📊 性能优势 / Performance Benefits
- ⚡ 高精度: 使用官方 tiktoken 库,token 计算误差 < 10%
- ⚡ High Precision: Uses official tiktoken library, token calculation error < 10%
- 🎯 并发准确性: 智能超时机制确保真实并发测试
- 🎯 Concurrent Accuracy: Intelligent timeout ensures genuine concurrent testing
- 📈 详细指标: 首次响应时间、输出速度、成功率等完整指标
- 📈 Detailed Metrics: Complete metrics including first response time, output speed, success rate
🔗 项目链接 / Project Links
- GitHub: https://github.com/gesong2077/LLMApiConcurrentTest
- Issues: https://github.com/gesong2077/LLMApiConcurrentTest/issues
- Documentation: 详见项目 README.md / See project README.md
🙏 致谢 / Acknowledgments
感谢所有使用的开源组件:
- Newtonsoft.Json - MIT License
- YamlDotNet - MIT License
- Tiktoken - MIT License
开始你的 LLM API 性能测试之旅吧!/ Start your LLM API performance testing journey! 🚀