AndersonBY/vector-vein v0.2.10

on GitHub

Release Notes - Version 0.2.10

New Features:

• Model Enhancements:
  • Added support for the gpt-4o-mini model to improve performance.
  • Introduced new model options including glm-4 variants and versioned Claude models.
  • Enhanced model selection options for media processing nodes.
• Workflow Improvements:
  • Added versioning for workflow and workflow run records.
• User Interface Updates:
  • Added dynamic model family management UI with translations.
  • Enhanced settings view with endpoint configurations and LLMs.
• Integration and Compatibility:
  • Integrated the python-vectorvein package for improved functionality.

Improvements:

• Performance Optimization:
  • Removed unused multimodal model checks.
  • Updated default model to gpt-4o-mini for better performance.
• Code Quality and Maintenance:
  • Refactored hashing to support key ignoring in objects.
  • Refactored Azure and Local LLM settings for better structure.
  • Removed deprecated Baichuan2-Turbo-192k option from templates.

Important Notes:

• Settings Structure Change:
  • Starting from this version, the structure of user settings will be changed. Old configuration information will be saved in the settings_v1.json file in the software directory. It is recommended that users back up their database before running the new version.

新特性:

• 模型增强:
  • 添加对 gpt-4o-mini 模型的支持以提升性能。
  • 引入新的模型选项，包括 glm-4 变体和版本化的 Claude 模型。
  • 增强媒体处理节点的模型选择选项。
• 工作流改进:
  • 为工作流和工作流运行记录添加版本控制。
• 用户界面更新:
  • 添加动态模型家族管理 UI 并进行翻译。
  • 增强设置视图，添加端点配置和 LLM 支持。
• 集成和兼容性:
  • 集成 python-vectorvein 包以精简项目代码。

改进:

• 性能优化:
  • 移除未使用的多模态模型检查。
  • 将默认模型更新为 gpt-4o-mini 以提升性能。
• 代码质量和维护:
  • 重构哈希处理以支持对象中的键忽略。
  • 重构 Azure 和本地 LLM 设置以提升结构化。
  • 从模板中移除已弃用的 Baichuan2-Turbo-192k 选项。

重要提示:

• 设置结构变更:
  • 从本版本开始，用户设置的结构将进行变更。旧的配置项信息将被保存在软件目录的 settings_v1.json 文件中。建议用户在运行新版本前先备份数据库。