v0.1.7

What's Changed

• feat: v0.1.7 - comprehensive examples, ROLL integration, and bug fixes by @sunhailin-Leo in #2

New Contributors

• @sunhailin-Leo made their first contribution in #2

Full Changelog: v0.1.6...v0.1.7

Version 0.1.6

Version 0.1.6 - 2025/10/16

新增功能 (New Features)

• 增加 example 代码，展示使用 schedule 装饰器
• 增加 httpx 的 json monkey patch
  • 可以通过环境变量 JSON_LIB 控制使用 ujson 或 orjson
• 利用 httpx 的 hook 方式增加自定义 path 的使用
• 增加 patch_openai，用于兼容自定义 embedding 接口的返回格式

性能优化 (Performance Optimization)

• 更新 hedged_request_proportion 默认值

文档与测试 (Documentation & Testing)

• 添加单元测试代码

移除功能 (Removed Features)

• 移除 speed_up_retry，替换为 schedule
• 移除 embeddings 相关接口中 custom_url 的参数

---

Version 0.1.6 - 2025/10/16

New Features

• Added example code to demonstrate the usage of the schedule decorator
• Added json monkey patch for httpx
  • The environment variable JSON_LIB can be used to control whether to use ujson or orjson
• Utilized httpx hooks to add support for custom paths
• Added patch_openai to maintain compatibility with custom embedding interface return formats

Performance Optimization

• Updated the default value of hedged_request_proportion

Documentation & Testing

• Added unit test code

Removed Features

• Removed speed_up_retry, replaced with schedule
• Removed the custom_url parameter from embedding-related interfaces

Version 0.1.5

version 0.1.5 - 2025/09/29

新增功能 (New Features)

• 增加 OpenAIChatProxyManager 类，用于管理代理
• 新增 keepalive 配置，兼容多种 http 框架
• 新增 retry 逻辑
• RewardServiceProxy 增加 transport 配置
• 新增 openai proxy 功能
• 新增 reward service client

性能优化 (Performance Optimization)

• 优化 speed_up_retry 功能
• 引入 aiolimiter 提速任务处理（相比 0.1.0 版本提升 2 倍以上性能）
  • 信号量控制全局并发，aiolimiter 精准控制并发上下文
• 通过信号量控制并发（支持大批量和小批量任务处理）

文档与测试 (Documentation & Testing)

• 更新 reward service 接口参数、使用方式以及使用示例
• 更新 example 和项目依赖
• 补充单元测试代码（内部版本代码覆盖率：91.56%）

代码质量 (Code Quality)

• 其他代码格式化改进

---

version 0.1.5 - 2025/09/29

New Features

• Added OpenAIChatProxyManager class for proxy management
• Added keepalive configuration compatible with multiple HTTP frameworks
• Added retry logic
• Added transport configuration to RewardServiceProxy
• Added openai proxy functionality
• Added reward service client

Performance Optimization

• Optimized speed_up_retry functionality
• Introduced aiolimiter for faster task processing (2x+ performance improvement over version 0.1.0)
  • Semaphore controls global concurrency, aiolimiter precisely controls concurrent context
• Concurrency control via semaphore (supporting both large batch and small batch task processing)

Documentation & Testing

• Updated reward service interface parameters, usage methods and examples
• Updated examples and project dependencies
• Added unit test coverage (internal version code coverage: 91.56%)

Code Quality

• Other code formatting improvements