发布日期: 2025-07-19
版本号: v0.6.17

0.6.17版本更新包含多项新增功能、问题修复和改动。新增功能包括：点击文件夹可查看包含所有聊天的专用页面，优化了文件夹创建流程，并支持直接向文件夹知识库上传文件。搜索聊天时可快速预览结果，笔记功能现支持插入图片并可选中部分内容进行编辑或提问，AI生成的回复呈现淡入动画效果。管理与设置方面，新增了活动用户指标、日志追踪ID、优化的用户组管理端点、笔记控制设置整合、管理员页面加载速度提升、聊天ID请求头转发以及增强的推理标签处理。还添加了OAuth端点认证方法支持和Redis Sentinel高可用方案，并改进了多语言本地化。 修复的问题有：恢复了混合搜索功能，在AI生成期间禁用了笔记编辑按钮，修复了数据库凭证重复字符问题，删除文件时同步清理向量数据，修复了文件上传模态框的翻译，并对不支持附件的模型禁用了拖拽上传。此外还修复了Ollama工具调用、MIME类型提示文本、笔记编辑器权限、聊天源处理、表情生成错误处理、文件夹系统提示权限执行、WebSocket锁超时类型转换等问题，并添加了PostHog依赖以解决兼容性问题。 主要改动是将底层富文本编辑器升级至v3版本，并清理了冗余或未使用的代码与字符串。

更新内容 （中文）

[0.6.17] - 2025-07-19

新增功能

  • 📂 带聊天列表的专属文件夹视图：点击文件夹现在会显示一个全新的落地页，展示该文件夹内的所有聊天列表，使导航更简单，并让团队能即时查看项目相关的对话。
  • 🆕 精简的文件夹创建模态框：创建新文件夹现在是一个无缝的统一体验，拥有一个在视觉和功能上与编辑文件夹流程相匹配的专用模态框，使所有用户的工作区组织更直观、更不易出错。
  • 🗃️ 直接上传文件至文件夹知识库：您现在可以直接将文件上传至文件夹的知识库——无需预先创建知识库，即可通过直接添加资源和文档来丰富项目空间。
  • 🔎 搜索中的聊天预览：在搜索聊天时，无需打开即可在上下文中即时预览结果——使发现、审核和回顾变得异常快捷，尤其在大型、活跃的团队中。
  • 🖼️ 笔记中的图片上传与内联插入：笔记现在支持在文本中直接插入图片，让您能以更自然、更吸引人的方式创建富视觉结构化的文档、头脑风暴或报告——图片不再仅仅是附件。
  • 📱 增强的笔记选择编辑与问答：选择笔记的任意部分，以仅编辑高亮部分或就该内容提出针对性问题——精简工作流程，提升生产力，并使审核或AI增强更具针对性。
  • 📝 将笔记复制为富文本：将整个笔记——包括所有格式、图片和结构——直接复制为富文本，以便无缝粘贴到电子邮件、报告或其他工具中，在WebUI之外保持清晰度和一致性。
  • ⚡ 渐入流式文本体验：实时生成的响应现在在AI流式输出时优雅地渐入，创造出更自然、视觉上更吸引人的阅读体验；如果您偏爱静态显示，可在界面设置中轻松关闭此功能。
  • 🔄 后续提示的设置：微调您的后续提示体验——通过新的控制选项，您可以选择保持其可见，或将其直接插入消息输入框，而非自动提交，从而为您的工作流程提供更多灵活性和控制权。
  • 🔗 提示变量文档快速链接：在提示编辑器模态框中一键访问提示变量文档——缩短学习曲线，使高级提示构建更易上手。
  • 📈 遥测的活跃与总用户指标：通过跟踪活跃和总用户的新指标，获得有关使用模式和平台参与度的宝贵洞察——增强大型组织的审计和规划能力。
  • 🏷️ 通过日志追踪和跨度ID实现可追溯性：每个日志条目现在都包含详细的追踪和跨度ID，使管理员更容易在分布式系统或复杂故障排除中定位和解决问题。
  • 👥 用户组添加/移除端点：通过新的改进端点，轻松地将用户添加到组或从组中移除——使管理员和团队负责人对协作和权限的控制更快、更清晰。
  • ⚙️ 笔记设置与控制项精简：笔记的主要“设置”现在简称为“控制项”，笔记文件现在位于专用的控制项部分，清理了导航，使查找和配置笔记相关选项更容易。
  • 🚀 更快的管理员用户页面加载：管理员的用户列表端点已进行优化，排除了大型头像，加快了大型团队的加载速度，并减少了管理任务中的等待时间。
  • 📡 聊天ID请求头转发：Ollama和OpenAI路由器请求现在在请求头中包含聊天ID，从而在AI模型集成中实现更好的请求关联和调试能力。
  • 🧠 增强的推理标签处理：改进并扩展了推理标签解析，以更健壮地处理各种标签格式，包括标准XML风格标签和自定义分隔符，确保更好的AI推理透明度和调试能力。
  • 🔐 OAuth令牌端点身份验证方法：新增可配置的OAuth令牌端点身份验证方法支持，为企业OAuth集成和身份提供商兼容性提供增强的灵活性和安全选项。
  • 🛡️ Redis哨兵高可用性支持：全面的Redis哨兵故障转移实现，具有自动主节点发现、连接故障的智能重试逻辑以及主节点宕机时的无缝操作——消除单点故障，确保生产部署中的持续服务可用性。
  • 🌐 本地化与国际化改进：优化并扩展了简体中文、繁体中文、法语、德语、韩语和波兰语的翻译，确保所有支持语言的全球用户拥有更流畅、更地道的体验。

修复内容

  • 🏷️ 恢复混合搜索功能：混合搜索现在再次无缝工作——在所有RAG驱动的工作流程中，实现更准确、相关且全面的知识发现。
  • 🚦 笔记聊天 - AI生成期间编辑按钮禁用：在与笔记聊天时，AI响应期间的编辑按钮现已禁用——防止意外编辑，并确保聊天会话期间的工作流程清晰。
  • 🧹 更干净的数据库凭据：数据库连接凭据不再重复‘@’，防止潜在的连接问题，确保更顺畅、更可靠的集成。
  • 🧑‍💻 文件删除现在会移除相关向量数据：当文件从存储中删除时，现在也会从向量数据库中清除，确保干净的数据管理并防止杂乱或过时的搜索结果。
  • 📁 修复文件模态框翻译问题：所有模态框对话字符串——包括“使用整个文档”和“使用聚焦检索”——现已完全翻译，以提供更一致和本地化的UI体验。
  • 🚫 不支持模型时禁用拖放文件上传：当使用不支持附件的模型时，已禁用通过拖放上传文件——消除混淆并防止工作流程中断。
  • 🔑 Ollama工具调用现在可靠：修复了基于Ollama的工具调用问题，确保每次聊天的AI增强和工具使用不间断。
  • 📄 MIME类型帮助字符串修正：清理了mimetype帮助文本，移除了多余字符，为文件上传配置提供了更清晰的指导。
  • 📝 笔记编辑器权限修复：从笔记聊天功能中移除了不必要的仅管理员限制，使所有授权用户都能按预期访问笔记编辑功能。
  • 📋 改进的聊天来源处理：修复了来源处理逻辑，以防止在聊天消息中重复分配来源，确保对话期间来源归属更清晰、更准确。
  • 😀 表情符号生成错误处理：改进了音频路由器的错误处理并修复了表情符号生成任务的元数据结构，防止崩溃并确保更可靠的表情生成功能。
  • 🔒 文件夹系统提示词权限强制执行：文件夹编辑模态框中的系统提示词字段现在对没有系统提示词权限的用户正确隐藏，确保所有文件夹管理界面中一致的安全策略执行。
  • 🌐 WebSocket Redis锁超时类型转换：为WebSocket Redis锁超时配置修复了正确的整数类型转换和健壮的错误处理，防止潜在的配置错误并确保稳定的WebSocket连接。
  • 📦 添加PostHog依赖：添加了PostHog 5.4.0库以解决ChromaDB兼容性问题，确保稳定的向量数据库操作，并防止部署过程中的库版本冲突。

变更内容

  • 👀 Tiptap编辑器升级至v3：底层富文本编辑器已更新以面向未来，但某些支持库仍保持在v2版本以兼容。目前，请使用 ’npm install –force’ 安装依赖项以避免安装错误。
  • 🚫 移除冗余或未使用的字符串和元素：已清理杂项未使用、重复或过时的代码和翻译，以保持精简和高性能的体验。

更新内容 （原始）

[0.6.17] - 2025-07-19

Added

  • 📂 Dedicated Folder View with Chat List: Clicking a folder now reveals a brand-new landing page showcasing a list of all chats within that folder, making navigation simpler and giving teams immediate visibility into project-specific conversations.
  • 🆕 Streamlined Folder Creation Modal: Creating a new folder is now a seamless, unified experience with a dedicated modal that visually and functionally matches the edit folder flow, making workspace organization more intuitive and error-free for all users.
  • 🗃️ Direct File Uploads to Folder Knowledge: You can now upload files straight to a folder’s knowledge—empowering you to enrich project spaces by adding resources and documents directly, without the need to pre-create knowledge bases beforehand.
  • 🔎 Chat Preview in Search: When searching chats, instantly preview results in context without having to open them—making discovery, auditing, and recall dramatically quicker, especially in large, active teams.
  • 🖼️ Image Upload and Inline Insertion in Notes: Notes now support inserting images directly among your text, letting you create rich, visually structured documentation, brainstorms, or reports in a more natural and engaging way—no more images just as attachments.
  • 📱 Enhanced Note Selection Editing and Q&A: Select any portion of your notes to either edit just the highlighted part or ask focused questions about that content—streamlining workflows, boosting productivity, and making reviews or AI-powered enhancements more targeted.
  • 📝 Copy Notes as Rich Text: Copy entire notes—including all formatting, images, and structure—directly as rich text for seamless pasting into emails, reports, or other tools, maintaining clarity and consistency outside the WebUI.
  • ⚡ Fade-In Streaming Text Experience: Live-generated responses now elegantly fade in as the AI streams them, creating a more natural and visually engaging reading experience; easily toggled off in Interface settings if you prefer static displays.
  • 🔄 Settings for Follow-Up Prompts: Fine-tune your follow-up prompt experience—with new controls, you can choose to keep them visible or have them inserted directly into the message input instead of auto-submitting, giving you more flexibility and control over your workflow.
  • 🔗 Prompt Variable Documentation Quick Link: Access documentation for prompt variables in one click from the prompt editor modal—shortening the learning curve and making advanced prompt-building more accessible.
  • 📈 Active and Total User Metrics for Telemetry: Gain valuable insights into usage patterns and platform engagement with new metrics tracking active and total users—enhancing auditability and planning for large organizations.
  • 🏷️ Traceability with Log Trace and Span IDs: Each log entry now carries detailed trace and span IDs, making it much easier for admins to pinpoint and resolve issues across distributed systems or in complex troubleshooting.
  • 👥 User Group Add/Remove Endpoints: Effortlessly add or remove users from groups with new, improved endpoints—giving admins and team leads faster, clearer control over collaboration and permissions.
  • ⚙️ Note Settings and Controls Streamlined: The main “Settings” for notes are now simply called “Controls”, and note files now reside in a dedicated controls section, decluttering navigation and making it easier to find and configure note-related options.
  • 🚀 Faster Admin User Page Loads: The user list endpoint for admins has been optimized to exclude heavy profile images, speeding up load times for large teams and reducing waiting during administrative tasks.
  • 📡 Chat ID Header Forwarding: Ollama and OpenAI router requests now include the chat ID in request headers, enabling better request correlation and debugging capabilities across AI model integrations.
  • 🧠 Enhanced Reasoning Tag Processing: Improved and expanded reasoning tag parsing to handle various tag formats more robustly, including standard XML-style tags and custom delimiters, ensuring better AI reasoning transparency and debugging capabilities.
  • 🔐 OAuth Token Endpoint Authentication Method: Added configurable OAuth token endpoint authentication method support, providing enhanced flexibility and security options for enterprise OAuth integrations and identity provider compatibility.
  • 🛡️ Redis Sentinel High Availability Support: Comprehensive Redis Sentinel failover implementation with automatic master discovery, intelligent retry logic for connection failures, and seamless operation during master node outages—eliminating single points of failure and ensuring continuous service availability in production deployments.
  • 🌐 Localization & Internationalization Improvements: Refined and expanded translations for Simplified Chinese, Traditional Chinese, French, German, Korean, and Polish, ensuring a more fluent and native experience for global users across all supported languages.

Fixed

  • 🏷️ Hybrid Search Functionality Restored: Hybrid search now works seamlessly again—enabling more accurate, relevant, and comprehensive knowledge discovery across all RAG-powered workflows.
  • 🚦 Note Chat - Edit Button Disabled During AI Generation: The edit button when chatting with a note is now disabled while the AI is responding—preventing accidental edits and ensuring workflow clarity during chat sessions.
  • 🧹 Cleaner Database Credentials: Database connection no longer duplicates ‘@’ in credentials, preventing potential connection issues and ensuring smoother, more reliable integrations.
  • 🧑‍💻 File Deletion Now Removes Related Vector Data: When files are deleted from storage, they are now purged from the vector database as well, ensuring clean data management and preventing clutter or stale search results.
  • 📁 Files Modal Translation Issues Fixed: All modal dialog strings—including “Using Entire Document” and “Using Focused Retrieval”—are now fully translated for a more consistent and localized UI experience.
  • 🚫 Drag-and-Drop File Upload Disabled for Unsupported Models: File upload by drag-and-drop is disabled when using models that do not support attachments—removing confusion and preventing workflow interruptions.
  • 🔑 Ollama Tool Calls Now Reliable: Fixed issues with Ollama-based tool calls, ensuring uninterrupted AI augmentation and tool use for every chat.
  • 📄 MIME Type Help String Correction: Cleaned up mimetype help text by removing extraneous characters, providing clearer guidance for file upload configurations.
  • 📝 Note Editor Permission Fix: Removed unnecessary admin-only restriction from note chat functionality, allowing all authorized users to access note editing features as intended.
  • 📋 Chat Sources Handling Improved: Fixed sources handling logic to prevent duplicate source assignments in chat messages, ensuring cleaner and more accurate source attribution during conversations.
  • 😀 Emoji Generation Error Handling: Improved error handling in audio router and fixed metadata structure for emoji generation tasks, preventing crashes and ensuring more reliable emoji generation functionality.
  • 🔒 Folder System Prompt Permission Enforcement: System prompt fields in folder edit modal are now properly hidden for users without system prompt permissions, ensuring consistent security policy enforcement across all folder management interfaces.
  • 🌐 WebSocket Redis Lock Timeout Type Conversion: Fixed proper integer type conversion for WebSocket Redis lock timeout configuration with robust error handling, preventing potential configuration errors and ensuring stable WebSocket connections.
  • 📦 PostHog Dependency Added: Added PostHog 5.4.0 library to resolve ChromaDB compatibility issues, ensuring stable vector database operations and preventing library version conflicts during deployment.

Changed

  • 👀 Tiptap Editor Upgraded to v3: The underlying rich text editor has been updated for future-proofing, though some supporting libraries remain on v2 for compatibility. For now, please install dependencies using ’npm install –force’ to avoid installation errors.
  • 🚫 Removed Redundant or Unused Strings and Elements: Miscellaneous unused, duplicate, or obsolete code and translations have been cleaned up to maintain a streamlined and high-performance experience.

下载链接