Wide range of new features, upgrades, and bugfixes

[garylin2099]

Main changes

Numerical statistics

• Huge number of commits: increased from ~1300 to ~2000
• Significantly improved test coverage: from ~30% to 80%+

Ease of use

• Example optimization: Now only 7 lines of code are needed to implement a general election debate. Simplified debate and knowledge base related examples
• Pure asynchronous: LLM API calls are changed to pure asynchronous, removing all synchronization code
• Installation process optimization: now you can use pip install -e .[dev] / pip install -e .[test] to install the package corresponding to the scenario

Framework update

• Integrated code to support Huggingface and AgentStore
• Added serialization and deserialization of important objects
• Support breakpoint recovery
• All types are modified based on pydantic v2. In order to keep it simple enough, all private variables are changed to public

LLM provider update

• OpenAI package upgraded to v1.6.0
• Support and optimize LLM related codes: Gemini, ZhipuAI, Ollama, OpenLLM etc.

Other update

• Removed OpenInterpreter and StableDiffusion part of the code
• Add ToT strategy
• BUGFIX: Fixed a lot of bugs

主要变动

数值统计

• 巨量的commits数：从~1300提升到了~2000
• 大幅提升测试覆盖率：从~30%提升到了80%+

易用性

• 例子优化：现在只需要7行代码就可以实现一个大选辩论。简化了debate和知识库相关示例
• 纯粹异步：LLM API调用改为纯异步，去除了所有的同步代码
• 安装流程优化：现在可以pip install -e .[dev] / pip install -e .[test] 来安装对应场景的包

框架更新

• 集成了用于支撑Huggingface和AgentStore的代码
• 增加各重要对象的序列化反序列化
• 支持断点恢复
• 全类型基于pydantic v2进行改造，为保持足够简洁，将所有私有变量改为公有

LLM provider更新

• OpenAI包升级至v1.6.0
• 支持与优化LLM相关代码：Gemini, ZhipuAI, Ollama, OpenLLM etc.

其他

• 移除OpenInterpreter和StableDiffusion部分代码
• 增加ToT策略
• BUGFIX：修复了大量bug

[geekan]

LGTM