发布日期: 2025-05-05
版本号: v0.6.6

Open WebUI 0.6.6 版本（2025年5月5日）带来了多项重要更新。新增了AI增强笔记功能，支持音频转写和会议录音导入；增强了协作与集成能力，包括Markdown笔记导入、按用户组控制笔记权限、OneDrive/SharePoint集成；优化了管理功能，如分页用户列表、细粒度聊天分享权限、用户角色更改确认对话框及登录失败审计日志。系统还提升了搜索性能，支持全知识库并行搜索与Firecrawl、Yacy等新搜索引擎集成；改进了文档处理能力，支持可配置的Docling OCR、Sentence Transformers和Pinecone向量数据库集成。此外，新增了工具依赖自动安装、令牌过期自动登出、YouTube自动嵌入、多语言支持更新及S3文件标签等功能。修复方面重点加强了HTML安全、Ollama代理模型处理、视频上传、WebSocket稳定性、动作模块执行、通知推送、知识库重索引等，并完善了OpenAI o系列模型支持。本版本还更新了自定义许可证及贡献者协议，升级了CUDA Docker镜像至12.8，并进行了后端重构以提高可靠性。最后，衷心感谢所有贡献者和赞助商的支持，并提示企业版客户可联系销售团队获取增强功能。

更新内容 （中文）

[0.6.6] - 2025-05-05

新增

  • 📝 AI 增强笔记（含音频转录）：轻松创建笔记，附加会议或语音音频，AI 将利用音频转录即时增强、总结或优化您的笔记，以最少的努力使您的文档更智能、更简洁、更具洞察力。
  • 🔊 会议音频录制与导入：无缝录制会议音频或捕获屏幕音频，并将其附加到您的笔记中——使回顾、注释和提取重要讨论的见解更加容易。
  • 🁢 轻松导入 Markdown 笔记：通过导入 Markdown 笔记将您现有的知识库带入 Open WebUI，以便立即利用所有高级笔记管理和 AI 功能。
  • 👥 按用户组划分的笔记权限：根据用户角色或组别精细调整笔记的访问和编辑权限，以便您可根据需要委派撰写任务或限制敏感信息。
  • ☁️ OneDrive 与 SharePoint 集成：通过将笔记和文件直接与 OneDrive 或 SharePoint 连接来保持内容同步——解锁快速的企业级导入/导出功能，并与您现有的工作流程实现无缝协作。
  • 🗂️ 管理面板中的分页用户列表：通过全新的分页用户列表，轻松管理和搜索大型团队——节省时间，简化大型组织中的用户管理。
  • 🕹️ 精细的聊天共享与导出权限：享受对谁可以共享或导出聊天的增强控制，在团队和企业环境中实现更严格的治理和隐私保护。
  • 🛑 用户角色更改确认对话框：通过在更新用户角色前设置必需的确认步骤来减少意外的权限更改——提高安全性，防止团队管理中代价高昂的错误。
  • 🚨 登录失败尝试审计日志：通过审计跟踪中详细的认证失败日志，快速检测未授权访问尝试或排查用户登录问题。
  • 💡 聊天专用“生成标题”按钮：快速组织每一次对话——点击新按钮，让 AI 为您的所有聊天创建相关、清晰的标题，节省时间并减少杂乱。
  • 💬 通知声音常开选项：通过将声音警报设置为始终播放来掌控您的通知——帮助您在繁忙环境中及时了解重要更新。
  • 🆔 S3 文件标签支持：上传到 S3 的文件现在包含标签，以便更好地组织、搜索，并与您的文件管理策略集成。
  • 🛡️ OAuth 阻止组支持：通过明确阻止指定的 OAuth 组来获得对基于组的访问的更多控制——非常适合复杂的身份或安全要求。
  • 🚀 优化后的更快网络搜索与多线程查询：通过改进的多线程搜索，享受显著更快的网络搜索和 RAG（检索增强生成）——在更短的时间内获得更丰富、更准确的结果。
  • 🔍 全知识库并行搜索：即使在非混合模式下，现在也可以并行搜索整个知识库，加快响应速度，并为每个问题提高知识准确性。
  • 🌐 新增 Firecrawl 与 Yacy 网络搜索集成：通过两个新的高级搜索引擎扩展您的信息世界——Firecrawl 用于更深入的网页洞察，Yacy 用于去中心化、注重隐私的搜索能力。
  • 🧠 可配置的 Docling OCR 引擎与语言：使用环境变量微调 Docling OCR 引擎和支持的语言，以实现更智能、更定制化的文档提取和 RAG 工作流程。
  • 🗝️ 增强的 Sentence Transformers 配置：新增环境变量，便于设置和高级自定义 Sentence Transformers——确保最适合您的嵌入需求。
  • 🌲 Pinecone 向量数据库集成：通过完全原生支持 Pinecone 作为您的向量数据库，实现企业级规模的知识索引、搜索和管理——轻松处理即使最大的文档集。
  • 🔄 工具与功能的自动依赖项安装：再也不必担心重启时丢失依赖项——外部函数和工具依赖项现在会在启动时自动安装，确保工具始终“开箱即用”。
  • 🔒 令牌过期时自动登出：安全更智能——如果用户认证令牌过期，现在会自动注销，保护敏感内容并确保合规性，且不受干扰。
  • 🎬 YouTube 嵌入自动检测：粘贴 YouTube 链接即可在聊天中即时嵌入视频——无需再手动嵌入，使知识共享和媒体消费对每个团队来说都更加容易。
  • 🔄 扩展的语言与区域支持：丹麦语、法语、俄语、繁体中文、简体中文、泰语、加泰罗尼亚语、德语和韩语的翻译已升级，为整个平台提供更流畅、更自然的用户体验。

修复

  • 🔒 更严格的 HTML 令牌安全：HTML 渲染现在仅限于管理员上传的令牌，降低了任何 XSS 风险，保护您的数据安全。
  • 🔐 改进的 HTML 安全性与令牌处理：进一步加固了 HTML 令牌和内容的处理方式，保证更强的抵御安全漏洞和攻击的能力。
  • 🔏 修正 Ollama 代理前缀的模型使用：增强了模型引用处理，使得使用自定义前缀时，Ollama 中的代理模型也能始终正确下载和运行。
  • 📥 视频文件上传处理：防止视频文件被错误分类为文本，修复了上传错误，确保媒体文件按预期工作。
  • 🔄 消除依赖性 WebSocket 顺序延迟：简化了 WebSocket 操作以防止延迟并保持快速的实时协作，尤其是在多用户环境中。
  • 🛠️ 更稳健的模块动作执行：模块中的多个动作现在会按设计触发，增加了自动化和脚本灵活性。
  • 📧 通知 Webhook：确保即使用户当前不活跃，也会始终为用户事件发送通知 Webhook。
  • 🗂️ 更智能的知识库重建索引：即使遇到损坏或缺失的集合，知识重建索引也会继续，使您的搜索功能保持可靠运行。
  • 🏷️ 带个人资料图片的用户导入：导入用户时，他们的个人资料图片现在也会随之带入——使入职和协作从第一天起就更清晰明了。
  • 💬 OpenAI o 系列通用支持：所有 OpenAI o 系列模型现在都能被无缝识别和支持，为每个工作流程解锁更强大的能力和更多模型选择。

变更

  • 📜 自定义许可证更新与贡献者协议：Open WebUI 现在在默认需要贡献者许可证协议的自定义许可证下运行——详情请见 https://docs.openwebui.com/license/，确保社区的可持续开放创新。
  • 🔨 CUDA Docker 镜像更新至 12.8：升级了 CUDA 镜像支持，以实现更快、更兼容的模型推理和面向未来的 GPU 性能，服务于您的 AI 基础设施。
  • 🧱 为可靠性进行的通用后端重构：持续的稳定性改进精简了后端逻辑，减少了错误，并为下一波功能发布奠定了更坚实的基础——所有这些都在底层进行，以实现更可靠的 WebUI。

[!TIP]
寻找 企业计划？ – 立即联系我们的销售团队！

获取增强功能，包括自定义主题和品牌、服务水平协议 (SLA) 支持、长期支持 (LTS) 版本等！

👏 向我们了不起的贡献者们致以最诚挚的感谢，感谢他们的辛勤工作和奉献，使本次版本得以发布： @agrimmAilio, @PeterDaveHello, @TESTPERSONAL, @izumi0004, @aleixdorca, @salbahra, @tth37, @therealmichaelberna, @Mister-Hope, @yak1ex, @shcodenick, @schulti, @oleguldberg, @panda44312, @stephen304, @spammenotinoz, @tarmst, @hurxxxx, @ayan4m1, @athoik, @AnapatChaiwongse, @jmtatsch, @NoMoreFood, @flefevre, @omahs, @TiancongLx, @SadmL

赞助商 🙌

🚀 我们要向我们了不起的赞助商们表示衷心的感谢，感谢他们的慷慨支持（注：我们已将私人赞助商从此列表中排除。如果您希望在此列出，请随时联系我们！）

祖母绿级

  • n8n n8n

@ModonHolding, @agency-agency, @roosi-gmbh, @Unsupervisedcom, @feddersen-group, @SRKConsulting, @mushmax, @rndmcnlly, @icanbwell, @blogic-cz, @Krambu-Travis, @GenieDev101, @PortlandKyGuy, @CubicleGangster, @GitDatamind, @TOWaB, @TheZigGroup, @Classic298, @comet-ml, @chAyubTahir

更新内容 （原始）

[0.6.6] - 2025-05-05

Added

  • 📝 AI-Enhanced Notes (With Audio Transcription): Effortlessly create notes, attach meeting or voice audio, and let the AI instantly enhance, summarize, or refine your notes using audio transcriptions—making your documentation smarter, cleaner, and more insightful with minimal effort.
  • 🔊 Meeting Audio Recording & Import: Seamlessly record audio from your meetings or capture screen audio and attach it to your notes—making it easier to revisit, annotate, and extract insights from important discussions.
  • 📁 Import Markdown Notes Effortlessly: Bring your existing knowledge library into Open WebUI by importing your Markdown notes, so you can leverage all advanced note management and AI features right away.
  • 👥 Notes Permissions by User Group: Fine-tune access and editing rights for notes based on user roles or groups, so you can delegate writing or restrict sensitive information as needed.
  • ☁️ OneDrive & SharePoint Integration: Keep your content in sync by connecting notes and files directly with OneDrive or SharePoint—unlocking fast enterprise import/export and seamless collaboration with your existing workflows.
  • 🗂️ Paginated User List in Admin Panel: Effortlessly manage and search through large teams via the new paginated user list—saving time and streamlining user administration in big organizations.
  • 🕹️ Granular Chat Share & Export Permissions: Enjoy enhanced control over who can share or export chats, enabling tighter governance and privacy in team and enterprise settings.
  • 🛑 User Role Change Confirmation Dialog: Reduce accidental privilege changes with a required confirmation step before updating user roles—improving security and preventing costly mistakes in team management.
  • 🚨 Audit Log for Failed Login Attempts: Quickly detect unauthorized access attempts or troubleshoot user login problems with detailed logs of failed authentication right in the audit trail.
  • 💡 Dedicated ‘Generate Title’ Button for Chats: Swiftly organize every conversation—tap the new button to let AI create relevant, clear titles for all your chats, saving time and reducing clutter.
  • 💬 Notification Sound Always-On Option: Take control of your notifications by setting sound alerts to always play—helping you stay on top of important updates in busy environments.
  • 🆔 S3 File Tagging Support: Uploaded files to S3 now include tags for better organization, searching, and integration with your file management policies.
  • 🛡️ OAuth Blocked Groups Support: Gain more control over group-based access by explicitly blocking specified OAuth groups—ideal for complex identity or security requirements.
  • 🚀 Optimized Faster Web Search & Multi-Threaded Queries: Enjoy dramatically faster web search and RAG (retrieval augmented generation) with revamped multi-threaded search—get richer, more accurate results in less time.
  • 🔍 All-Knowledge Parallel Search: Searches across your entire knowledge base now happen in parallel even in non-hybrid mode, speeding up responses and improving knowledge accuracy for every question.
  • 🌐 New Firecrawl & Yacy Web Search Integrations: Expand your world of information with two new advanced search engines—Firecrawl for deeper web insight and Yacy for decentralized, privacy-friendly search capabilities.
  • 🧠 Configurable Docling OCR Engine & Language: Use environment variables to fine-tune Docling OCR engine and supported languages for smarter, more tailored document extraction and RAG workflows.
  • 🗝️ Enhanced Sentence Transformers Configuration: Added new environment variables for easier set up and advanced customization of Sentence Transformers—ensuring best fit for your embedding needs.
  • 🌲 Pinecone Vector Database Integration: Index, search, and manage knowledge at enterprise scale with full native support for Pinecone as your vector database—effortlessly handle even the biggest document sets.
  • 🔄 Automatic Requirements Installation for Tools & Functions: Never worry about lost dependencies on restart—external function and tool requirements are now auto-installed at boot, ensuring tools always “just work.”
  • 🔒 Automatic Sign-Out on Token Expiry: Security is smarter—users are now automatically logged out if their authentication token expires, protecting sensitive content and ensuring compliance without disruption.
  • 🎬 Automatic YouTube Embed Detection: Paste YouTube links and see instant in-chat video embeds—no more manual embedding, making knowledge sharing and media consumption even easier for every team.
  • 🔄 Expanded Language & Locale Support: Translations for Danish, French, Russian, Traditional Chinese, Simplified Chinese, Thai, Catalan, German, and Korean have been upgraded, offering smoother, more natural user experiences across the platform.

Fixed

  • 🔒 Tighter HTML Token Security: HTML rendering is now restricted to admin-uploaded tokens only, reducing any risk of XSS and keeping your data safe.
  • 🔐 Refined HTML Security and Token Handling: Further hardened how HTML tokens and content are handled, guaranteeing even stronger resistance to security vulnerabilities and attacks.
  • 🔏 Correct Model Usage with Ollama Proxy Prefixes: Enhanced model reference handling so proxied models in Ollama always download and run correctly—even when using custom prefixes.
  • 📥 Video File Upload Handling: Prevented video files from being misclassified as text, fixing bugs with uploads and ensuring media files work as expected.
  • 🔄 No More Dependent WebSocket Sequential Delays: Streamlined WebSocket operation to prevent delays and maintain snappy real-time collaboration, especially in multi-user environments.
  • 🛠️ More Robust Action Module Execution: Multiple actions in a module now trigger as designed, increasing automation and scripting flexibility.
  • 📧 Notification Webhooks: Ensured that notification webhooks are always sent for user events, even when the user isn’t currently active.
  • 🗂️ Smarter Knowledge Base Reindexing: Knowledge reindexing continues even when corrupt or missing collections are encountered, keeping your search features running reliably.
  • 🏷️ User Import with Profile Images: When importing users, their profile images now come along—making onboarding and collaboration visually clearer from day one.
  • 💬 OpenAI o-Series Universal Support: All OpenAI o-series models are now seamlessly recognized and supported, unlocking more advanced capabilities and model choices for every workflow.

Changed

  • 📜 Custom License Update & Contributor Agreement: Open WebUI now operates under a custom license with Contributor License Agreement required by default—see https://docs.openwebui.com/license/ for details, ensuring sustainable open innovation for the community.
  • 🔨 CUDA Docker Images Updated to 12.8: Upgraded CUDA image support for faster, more compatible model inference and futureproof GPU performance in your AI infrastructure.
  • 🧱 General Backend Refactoring for Reliability: Continuous stability improvements streamline backend logic, reduce errors, and lay a stronger foundation for the next wave of feature releases—all under the hood for a more dependable WebUI.

[!TIP]
Looking for an Enterprise Plan? – Speak with Our Sales Team Today!

Get enhanced capabilities, including custom theming and branding, Service Level Agreement (SLA) support, Long-Term Support (LTS) versions, and more!

👏 Massive thanks to our incredible contributors for their hard work and dedication to making this release possible: @agrimmAilio, @PeterDaveHello, @TESTPERSONAL, @izumi0004, @aleixdorca, @salbahra, @tth37, @therealmichaelberna, @Mister-Hope, @yak1ex, @shcodenick, @schulti, @oleguldberg, @panda44312, @stephen304, @spammenotinoz, @tarmst, @hurxxxx, @ayan4m1, @athoik, @AnapatChaiwongse, @jmtatsch, @NoMoreFood, @flefevre, @omahs, @TiancongLx, @SadmL

Sponsors 🙌

🚀 We’d like to extend a heartfelt thank you to our amazing sponsors for their generous support (Note: We’ve excluded private sponsors from this list. If you’d like to get featured here, feel free to reach out to us!)

Emerald

  • n8n n8n

@ModonHolding, @agency-agency, @roosi-gmbh, @Unsupervisedcom, @feddersen-group, @SRKConsulting, @mushmax, @rndmcnlly, @icanbwell, @blogic-cz, @Krambu-Travis, @GenieDev101, @PortlandKyGuy, @CubicleGangster, @GitDatamind, @TOWaB, @TheZigGroup, @Classic298, @comet-ml, @chAyubTahir

下载链接