LobeHub
LobeHub
返回
更新日志

ComfyUI 集成与知识库优化 ⭐

在 LobeHub 中直接运行 ComfyUI 可视化工作流,通过瀑布流布局和自动解压更高效地组织知识,并支持将对话导出为 PDF。

ComfyUI 集成与知识库优化 ⭐

十月到十一月,LobeHub 聚焦于让工作流驱动的创作更实用。这次更新将 ComfyUI 可视化流程引入对话场景,优化了团队知识组织方式,并让多格式输出分享更便捷。

可视化工作流与文档导出

ComfyUI 工作流现已集成到 LobeHub,无需切换工具即可运行现有的可视化流程。以往使用 ComfyUI 进行图像或视频生成的团队需要维护独立环境,现在可以直接在对话上下文中引入并执行这些工作流。

文档导出能力也扩展至 PDF 格式,便于将对话归档以满足合规需求,或向需要离线副本的利益相关者分享结构化输出。

规模化知识组织

管理大型知识库的体验因两项实用改进而更加顺畅。新增的瀑布流布局让浏览海量文件资源时更高效,自动解压功能则省去了上传前手动解压的步骤 —— 直接拖拽 ZIP 文件,系统自动处理剩余工作。这些改进降低了团队知识导入和维护大型参考资料库时的摩擦。

服务商与模型覆盖

新增服务商支持包括 Cerebras 与 CometAPI,图像生成能力扩展至硅基流动与混元 Text-to-Image 3。模型库新增 Claude Haiku 4.5、GPT-5 Pro、MiniMax-M2 及 Vertex AI 的 Imagen 4。

体验优化与修复

  • 优化富文本链接渲染
  • 改进搜索体验
  • 新增禁用富文本编辑的选项
  • 新增删除与重新生成快捷键
  • 完善更新通知行为

版本详情

Features

  • v1.142.0Use env to control clerk allow origin feature.
  • v1.141.0Add PDF export functionality to share modal.
  • v1.140.0Add ComfyUI integration Phase1(RFC-128).
  • v1.139.0Support image generation for siliconcloud.
  • v1.138.0Support Group Chat, Mention, and Multi-Agent Orchestration with feature flag.
  • v1.137.0Add new setting for default image num.
  • v1.136.0Add new provider Cerebras.
  • v1.135.0Huanyuan text-to-image 3.
  • v1.134.0Support double-click to open multi agent window on the desktop.

What's improved

  • v1.142.7Update i18n.
  • v1.142.5Add MiniMax-M2 model.
  • v1.142.4Pre render ModelSwitchPanel, The error details of the connectivity check lead to a layout problem.
  • v1.142.3Adjust modal setting form styles for improved layout and responsiveness, Unzip file when uploading in knowledge base \LOB-500].
  • v1.142.2Improve provider modal height when creating custom provider.
  • v1.142.1Update i18n.
  • v1.141.9Improve local system tools render.
  • v1.141.8Improvement for Agent Team After Alpha Launch \LOB-517].
  • v1.141.7Allow removal of top_p and similar request parameters.
  • v1.141.5Change discover page from RSC to SPA to improve performance.
  • v1.141.4Fix model runtime cost calculate with CNY.
  • v1.141.3Update i18n.
  • v1.141.1Refactor context engine.
  • v1.141.0Add knowledge base mansory layout \LOB-496], improve rich text link display.
  • v1.139.3Show message author in minimap.
  • v1.139.2Solve when desktop the sider agent list too long.
  • v1.139.1Update i18n.
  • v1.138.5Refactor upload router into lambda and decide to remove it in V2.
  • v1.138.2Improve welcome message.
  • v1.137.10Add Claude Haiku 4.5 model.
  • v1.137.9Improve update notification.
  • v1.137.8Add region support for Vertex AI provider.
  • v1.137.7Use different favicon.ico in dev mode.
  • v1.137.5Add imagen model to vertex ai.
  • v1.137.1Change the user chatItem maxWidth should use flex 1.
  • v1.136.13Optimize OpenRouter modelFetch endpoint, update i18n.
  • v1.136.12Add more AWS regions, Update infini-ai models.
  • v1.136.11Add capability inference for web search, image output and video recognition in model parsing and update UI form items to support search, imageOutput and video abilities.
  • v1.136.10Improve search experience.
  • v1.136.9Add lab to support disable/enable rich text.
  • v1.136.3Add delete & regenerate hotkeys.
  • v1.136.2Update i18n.
  • v1.135.5Update i18n.
  • v1.135.4Add GPT-5 pro model.
  • v1.135.3Improve Korean translate.
  • v1.135.1Improve styles and fix tools calling condition.
  • v1.134.7Update i18n.
  • v1.134.4Add promptfoo to improve prompts quality.
  • v1.134.2Allow switching model type.
  • v1.134.1Update i18n.
  • v1.133.6Nano banana support aspect_ratio.
  • v1.133.5Optimized extendParams UI, update i18n.

What's fixed

  • v1.141.10Loadmore not work & navbar not show in pwa.
  • v1.141.0Ignore abort signal errors in TRPC client, slove when pwa user info have code cannot be viewed in full.
  • v1.139.4Pass threadId to messages in sendMessageInServer.
  • v1.138.4Fix response API tools calling issue.
  • v1.138.3Fix topic fetch not correct in custom agent.
  • v1.138.1Automatic topic creation switch does not work.
  • v1.137.8Fix duplicate tools id issue and fix link dialog issue.
  • v1.137.6Update Claude workflows to use oauth token, vertext ai create image.
  • v1.137.4Prevent Vertex AI JSON credentials from being split by comma.
  • v1.137.3Fix mcp server connect issue and refactor web search implement, fix tools calling long name length >64 issue.
  • v1.137.2Fix the Worker URL cross-origin issue.
  • v1.136.13Fix input cannot send markdown.
  • v1.136.7Disable rich text in markdown editor.
  • v1.136.4Add 'gemini-2.5-flash-image' to disabled models Thinking.
  • v1.136.0Fix standalone plugin rerender issue.
  • v1.134.3Type not preserved when model is sorted.
  • v1.133.6type not preserved when model is disabled or sorted.
  • v1.133.5Custom provider fails when client requests are enabled.
LobeHub

Super-Agents on Your Team

今天就把 Super-Agents 招进你的团队。