发布日期: 2024-11-19
版本号: v0.4.0

2024年11月19日发布的0.4.0版本带来多项功能升级与优化。新增用户组功能，支持创建和管理用户组，实现基于分组的精细化权限控制，涵盖模型、知识库及工具访问权限，并新增LDAP认证提升安全性。连接管理增强OpenAI兼容性，支持Ollama多实例区分和API密钥管理，新增连接启用/禁用开关。界面方面重新设计了模型、提示词、协作工作区，新增工具自动选择、网页搜索状态提示、大文本转文件上传等功能，支持禁用引用显示和设置项搜索。修复了YouTube视频加载、共享聊天更新异常、DuckDuckGo速率限制等问题，优化引用相关性计算。调整功能模块至管理员面板，优化模型选择逻辑与全局管理流程，移除部分功能等待后续优化。新增实验性本地语音合成SpeechT5支持，提供一键重置模型和初始化向导。翻译优化涵盖乌克兰语、挪威语等语种。最后感谢开源贡献者与赞助商的支持。

更新内容 （中文）

[0.4.0] - 2024-11-19

新增功能

  • 👥 用户组管理: 支持创建和管理用户组，实现用户组织架构的无缝整合
  • 🔐 基于组的访问控制: 支持按用户组设置模型、知识库、提示词和工具的细粒度访问权限，构建可控安全环境
  • 🛠️ 基于组的用户权限: 便捷管理工作区权限，支持配置文件上传/删除/编辑、临时会话创建、以及模型/知识库/提示词/工具的创建权限
  • 🔑 LDAP认证支持: 新增LDAP认证功能，增强用户管理安全性与可扩展性
  • 🌐 增强OpenAI兼容连接: 新增前缀ID支持以避免模型ID冲突，并为不兼容’/models’端点的API提供显式模型ID支持
  • 🔐 Ollama API密钥支持: 支持为代理后的Ollama实例配置凭证，可使用前缀ID区分多个Ollama实例
  • 🔄 连接启用/禁用开关: 支持单独启用/禁用OpenAI和Ollama连接
  • 🎨 模型工作区重构: 全新设计的用户/组模型管理界面提升操作效率
  • 🎨 提示词工作区重构: 优化提示词管理界面布局，提升组织效率
  • 🧩 函数分类管理: 函数按类型(动作/过滤器/管道)自动分类管理
  • 💻 协作工作区优化: 增强多用户协同开发模型/知识库/提示词/工具的支持
  • 🔧 模型编辑器工具自动选择: 启用工具后自动勾选对应选项，减少手动操作
  • 🔔 网络搜索与工具状态指示器: 新增启用状态的视觉标识，降低误操作
  • 🔑 API密钥认证开关: 支持启用/禁用Open WebUI的API密钥认证功能
  • 🗂️ 代理检索增强: 通过对话历史智能预处理提升RAG检索精度
  • 📁 大文本转文件上传: 支持将粘贴的大段文本自动转为文件上传
  • 🗂️ 引用显示开关: 模型编辑器新增禁用引用显示功能
  • 🔍 用户设置搜索: 支持通过搜索快速定位设置项
  • 🗣️ 实验性SpeechT5 TTS: 新增本地SpeechT5文本转语音支持
  • 🔄 一键重置模型: 管理员设置中新增一键清除所有模型功能
  • 🛠️ 初始化向导优化: 首次设置时明确提示创建管理员账户
  • 🌐 多语言优化: 优化乌克兰语、挪威语、巴西葡萄牙语等翻译

问题修复

  • 🎥 YouTube视频附件: 修复YouTube视频加载和附加异常问题
  • 🔄 共享会话更新: 修复共享会话内容更新不及时的问题
  • 🔍 DuckDuckGo限流: 修复DuckDuckGo搜索集成在限流场景下的稳定性
  • 🧾 引用相关性计算: 调整RAG引用相关性百分比计算逻辑
  • 🔑 Jina搜索密钥: 新增Jina搜索API密钥输入选项

功能变更

  • 🛠️ 函数管理迁移: 函数作为高级插件移至管理员面板
  • 🛠️ Ollama连接管理: 模型管理迁移至"连接 > Ollama连接"，新增专用管理模态框
  • 📊 基础模型管理: 管理员设置中集中管理所有基础模型，默认私有需显式授权
  • 📌 模型选择记忆: 新建会话时继承前次模型选择，二次新建恢复默认
  • 🎨 全局设计优化: 全平台视觉体验统一性提升

移除功能

  • 📂 模型列表排序: 临时移除，后续用户组设置改进后回归
  • ⚙️ 默认模型设置: 临时移除用户默认模型设置功能

👏 特别致谢贡献者: @silentoplayz, @Peter-De-Ath, @michaelpoluektov, @konmeo, @matthewhand, @dannyl1u, @aleixdorca, @clang88, @DmitriyAlergant-T1A, @krishanbhasin-px, @taylorwilsdon, @antpyykk-kone, @oyve, @yeounhak, @Menghuan1918, @OriginalSimon, @sakiphan

🚀 赞助商致谢: @digitalbrainowners, @arcee-ai, @ModonHolding, @mushmax, @rndmcnlly, @taylorwilsdon, @GenieDev101, @PortlandKyGuy, @sushibait

更新内容 （原始）

[0.4.0] - 2024-11-19

Added

  • 👥 User Groups: You can now create and manage user groups, making user organization seamless.
  • 🔐 Group-Based Access Control: Set granular access to models, knowledge, prompts, and tools based on user groups, allowing for more controlled and secure environments.
  • 🛠️ Group-Based User Permissions: Easily manage workspace permissions. Grant users the ability to upload files, delete, edit, or create temporary chats, as well as define their ability to create models, knowledge, prompts, and tools.
  • 🔑 LDAP Support: Newly introduced LDAP authentication adds robust security and scalability to user management.
  • 🌐 Enhanced OpenAI-Compatible Connections: Added prefix ID support to avoid model ID clashes, with explicit model ID support for APIs lacking ‘/models’ endpoint support, ensuring smooth operation with custom setups.
  • 🔐 Ollama API Key Support: Now manage credentials for Ollama when set behind proxies, including the option to utilize prefix ID for proper distinction across multiple Ollama instances.
  • 🔄 Connection Enable/Disable Toggle: Easily enable or disable individual OpenAI and Ollama connections as needed.
  • 🎨 Redesigned Model Workspace: Freshly redesigned to improve usability for managing models across users and groups.
  • 🎨 Redesigned Prompt Workspace: A fresh UI to conveniently organize and manage prompts.
  • 🧩 Sorted Functions Workspace: Functions are now automatically categorized by type (Action, Filter, Pipe), streamlining management.
  • 💻 Redesigned Collaborative Workspace: Enhanced support for multiple users contributing to models, knowledge, prompts, or tools, improving collaboration.
  • 🔧 Auto-Selected Tools in Model Editor: Tools enabled through the model editor are now automatically selected, whereas previously it only gave users the option to enable the tool, reducing manual steps and enhancing efficiency.
  • 🔔 Web Search & Tools Indicator: A clear indication now shows when web search or tools are active, reducing confusion.
  • 🔑 Toggle API Key Auth: Tighten security by easily enabling or disabling API key authentication option for Open WebUI.
  • 🗂️ Agentic Retrieval: Improve RAG accuracy via smart pre-processing of chat history to determine the best queries before retrieval.
  • 📁 Large Text as File Option: Optionally convert large pasted text into a file upload, keeping the chat interface cleaner.
  • 🗂️ Toggle Citations for Models: Ability to disable citations has been introduced in the model editor.
  • 🔍 User Settings Search: Quickly search for settings fields, improving ease of use and navigation.
  • 🗣️ Experimental SpeechT5 TTS: Local SpeechT5 support added for improved text-to-speech capabilities.
  • 🔄 Unified Reset for Models: A one-click option has been introduced to reset and remove all models from the Admin Settings.
  • 🛠️ Initial Setup Wizard: The setup process now explicitly informs users that they are creating an admin account during the first-time setup, ensuring clarity. Previously, users encountered the login page right away without this distinction.
  • 🌐 Enhanced Translations: Several language translations, including Ukrainian, Norwegian, and Brazilian Portuguese, were refined for better localization.

Fixed

  • 🎥 YouTube Video Attachments: Fixed issues preventing proper loading and attachment of YouTube videos as files.
  • 🔄 Shared Chat Update: Corrected issues where shared chats were not updating, improving collaboration consistency.
  • 🔍 DuckDuckGo Rate Limit Fix: Addressed issues with DuckDuckGo search integration, enhancing search stability and performance when operating within rate limits.
  • 🧾 Citations Relevance Fix: Adjusted the relevance percentage calculation for citations, so that Open WebUI properly reflect the accuracy of a retrieved document in RAG, ensuring users get clearer insights into sources.
  • 🔑 Jina Search API Key Requirement: Added the option to input an API key for Jina Search, ensuring smooth functionality as keys are now mandatory.

Changed

  • 🛠️ Functions Moved to Admin Panel: As Functions operate as advanced plugins, they are now accessible from the Admin Panel instead of the workspace.
  • 🛠️ Manage Ollama Connections: The “Models” section in Admin Settings has been relocated to Admin Settings > “Connections” > Ollama Connections. You can now manage Ollama instances via a dedicated “Manage Ollama” modal from “Connections”, streamlining the setup and configuration of Ollama models.
  • 📊 Base Models in Admin Settings: Admins can now find all base models, both connections or functions, in the “Models” Admin setting. Global model accessibility can be enabled or disabled here. Models are private by default, requiring explicit permission assignment for user access.
  • 📌 Sticky Model Selection for New Chats: The model chosen from a previous chat now persists when creating a new chat. If you click “New Chat” again from the new chat page, it will revert to your default model.
  • 🎨 Design Refactoring: Overall design refinements across the platform have been made, providing a more cohesive and polished user experience.

Removed

  • 📂 Model List Reordering: Temporarily removed and will be reintroduced in upcoming user group settings improvements.
  • ⚙️ Default Model Setting: Removed the ability to set a default model for users, will be reintroduced with user group settings in the future.

👏 Massive thanks to our incredible contributors for their hard work and dedication to making this release possible: @silentoplayz, @Peter-De-Ath, @michaelpoluektov, @konmeo, @matthewhand, @dannyl1u, @aleixdorca, @clang88, @DmitriyAlergant-T1A, @krishanbhasin-px, @taylorwilsdon, @antpyykk-kone, @oyve, @yeounhak, @Menghuan1918, @OriginalSimon, @sakiphan

🚀 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!): @digitalbrainowners, @arcee-ai, @ModonHolding, @mushmax, @rndmcnlly, @taylorwilsdon, @GenieDev101, @PortlandKyGuy, @sushibait

下载链接