versun/RSSBox 2025.7.1

on GitHub

中文:
⚠️ 请注意！此次重构包含以下破坏性更新：

1. 翻译引擎只支持 DeepL 和 Openai 兼容接口，因此这次更新将只会保留 DeepL 的认证数据，并会将Moonshot AI、Together AI、OpenRouter AI、Groq 自动转换为 Openai 格式。
   简而言之：如果当前您使用的翻译引擎不在上述所列，则将会被删除。

2. 翻译缓存将不在以单词为单位，而是直接以文章为单位进行缓存

3. 不再额外处理html格式问题，将会直接让 AI 模型自行处理，因此将会增加 token 的消耗
   经测试，gpt-4.1-nano和claude等主流模型均能完美保持内容格式，
   因此，如果出现翻译后的内容乱码问题，建议更换更好的模型或者换用 DeepL，DeepL官方可自动完美处理内容格式

4. docker hub 镜像源更新为 versun/rsstranslator，请务必查看最新的 docker compose 文件以便后续更新

优化项：
优化添加流程和默认值设置，目前添加源只需输入源地址即可完成添加
优化更新频率选项
优化类别源的响应速度
显著降低内存和CPU的占用，日常使用内存在 200MB 左右
添加原始源的代理功能，可通过代理地址访问原始源
重构代码以方便他人开发贡献

English
⚠️ Please Note! This refactor includes the following breaking changes:

1. The translation engine now only supports DeepL and OpenAI-compatible interfaces. Therefore, this update will retain only DeepL authentication data, and Moonshot AI, Together AI, OpenRouter AI, and Groq will be automatically converted to OpenAI format.
   In short: If your current translation engine is not listed above, it will be removed.

2. Translation caching will no longer be done on a per-word basis but directly on a per-article basis.

3. HTML formatting will no longer be specially processed; instead, the AI models will handle it directly. This will increase token consumption.
   Tests show that mainstream models such as GPT-4.1-nano and Claude can perfectly maintain content formatting.
   Therefore, if you encounter garbled content after translation, it is recommended to switch to a better model or use DeepL, as DeepL officially handles content formatting perfectly and automatically.

4. The Docker Hub image source has been updated to versun/rsstranslator. Please be sure to check the latest docker compose file for future updates.

Improvements:

• Optimized the adding process and default value settings; now you only need to input the source URL to add a source.
• Optimized update frequency options.
• Improved response speed for category sources.
• Significantly reduced memory and CPU usage; typical memory usage is around 200MB.
• Added proxy functionality for original sources, allowing access to original sources through a proxy address.
• Refactored code to facilitate contributions and development by others.