Rikkahub Revised v2.4.8-revised.4
·
12 commits
to main
since this release
更新内容
- 优化千问、DeepSeek、豆包及兼容供应商的模型能力、上下文窗口与推理参数识别。
- 规范 GPT Image 2 的尺寸、比例与质量参数,完善 Gemini、Grok 和 Seedream 生图适配。
- 升级图库预览、详情、文件夹、图片导入、元数据备份与批量添加、移动和删除。
- 改进自动重连,覆盖网络切换、限流、连接中断、聊天续答及图片生成与编辑。
- 提升完整对话与文件的上下文统计准确性,增强上下文窗口识别和自动压缩可靠性。
- 修复 DOC、DOCX、XLSX 解析,优化大型多媒体附件与复杂大体积 MIDI 的处理。
- 汉化错误提示,新增按条目保存七天的完整诊断日志及查看入口。
- 完善 MCP OAuth 授权、连接状态与重连反馈。
What's Changed
- Improved model capability, context-window, and reasoning-parameter detection for Qwen, DeepSeek, Doubao, and compatible providers.
- Standardized GPT Image 2 size, aspect-ratio, and quality parameters, with improved Gemini, Grok, and Seedream image generation support.
- Upgraded gallery previews, details, folders, image import, metadata backup, and batch add, move, and delete operations.
- Improved automatic reconnection for network switching, rate limits, connection interruptions, response continuation, and image generation/editing.
- Improved full conversation and attachment context accounting, context-window detection, and automatic compression reliability.
- Fixed DOC, DOCX, and XLSX parsing and optimized large multimedia attachments and complex large MIDI processing.
- Localized error messages and added complete per-entry diagnostic logs retained for seven days with a dedicated viewer.
- Improved MCP OAuth authorization, connection status, and reconnection feedback.