Releases: 1Panel-dev/MaxKB
Releases · 1Panel-dev/MaxKB
v1.10.1-lts
New features
- Model settings: Added a large language model for the atomic capability provider of Tencent Cloud Knowledge Engine. by @wxg0103 in #2158
Enhancements
- Application: Optimize the problem of slow opening of the workflow orchestration page when there are too many nodes in the advanced orchestration application; by @shaohuzhang1 in #2250
- Application: Optimize the filtering of form data in the historical chat records sent to the AI model to avoid affecting the AI answer; by @shaohuzhang1 in #2256
- Application: Optimize the description of output thinking settings;
- Application: Optimize the advanced orchestration application, the debug preview dialog window will only be closed when the "Close" button is clicked. by @wangdan-fit2cloud in #2248
Bug Fixes
- Knowledge base: Fixed the problem that when setting the title to the association problem when uploading files, the actual association is not successful; by @wangdan-fit2cloud in #2255
- Knowledge base: Fixed the problem that the upload error occurs when the uploaded folder contains subfolders; by @wangdan-fit2cloud in #2259
- Knowledge base: Fixed the problem that the upload error occurs when the uploaded folder contains hidden files; by @wangdan-fit2cloud in #2259
- Application: Fixed the problem that the document content extraction node in the execution details displays the wrong text; by @wangdan-fit2cloud in #2223
- Application: Fixed the issue that the document content extraction node reports an error when the uploaded Excel document has a blank sheet; by @liuruibin in #2259
- Application: Fixed the issue that the occasional reply "Sorry, no relevant content was found. Please re-describe your question or provide more information."; by @shaohuzhang1 in #2256
- Application: After the application reaches the upper limit of the number of questions, the English prompt information is incorrect; by @wangdan-fit2cloud in #2223
- Application: Fixed the issue that the column name in the exported conversation log file is not displayed accurately; by @wxg0103 in #2246
- Application (X-Pack): Fixed the issue that the theme color and header font color cannot be set after clicking "Restore Default" in the display settings;
- Application (X-Pack): Fixed the issue that after clicking "Restore Default" in the display settings, the language will display null or undefined;
- System Settings (X-Pack): Fixed the issue that DingTalk users who are not in the organization can also scan the code to log in successfully;
- System Settings (X-Pack): Fixed the issue that the user's email information is not fully obtained after the DingTalk user scans the code to log in;
- System Settings (X-Pack): Fixed the issue that there should be no language setting operation in the platform settings preview.
v1.10.0-lts
New features
- Application: Support output thinking process settings; by @shaohuzhang1,@liuruibin,@wangdan-fit2cloud in #2158
- Application: Support setting thinking process labels; by @shaohuzhang1,@liuruibin,@wangdan-fit2cloud in #2158
- Application: Support automatic sending of voice input; by @liuruibin in #1971
- Application: Support automatic voice playback; by @liuruibin in #1971
- Application: Add language settings for the Q&A page in the display settings; by @shaohuzhang1,@liuruibin,@wangdan-fit2cloud,@wxg0103 in #2011
- Model settings: Support large language models, vector models, rearrangement models, speech recognition and image generation models from SILICONFLOW; by @wxg0103 in #2127
- Model settings: Support large language models and visual models from Anthropic; by @liuruibin in #1974
- Model settings: Support vector models and image understanding models from vLLM; by @liuruibin in #2038
- Model settings: support vector models of Doubao suppliers; by @liuruibin in #2006
- System management: support system language switching, the first batch supports simplified Chinese, traditional Chinese and English. by @shaohuzhang1,@liuruibin,@wangdan-fit2cloud,@wxg0103 in #2011
Enhancements
- Application: optimize the style display problem of uploaded files when users ask questions; by @wangdan-fit2cloud in #2061
- Application: you can right-click to quote or copy the AI reply content. by @liuruibin in #2005
Bug Fixes
- Knowledge base: Fixed the problem of error when uploading offline files when the file name exceeds 128 characters; by @wangdan-fit2cloud in #1962
- Application (X-Pack): Fixed the problem of incomplete reply content in WeChat for Business;
- Application: Fixed the problem of error when the opening remarks exceed 4096 characters; by @shaohuzhang1 in #2075
- Application: Fixed the problem that the uploaded document cannot read the document content after creating a new conversation on the Q&A page;
- Application: Fixed the problem that the parameters in the added application subnode are not updated in real time. by @wxg0103 in #1962
v1.9.1
功能优化
- 知识库:文档向量化时支持仅向量化非成功的分段;by @shaohuzhang1 in #1984
- 应用(X-Pack):通过第三方应用进行对话的日志,优化为按天记录每个用户的对话日志。
问题修复
- 应用:修复进行语音输入时,提示中图片加载失败的问题;by @wangdan-fit2cloud in #1966
- 应用:修复嵌入应用节点,设置变量参数后,保存不生效的问题;by @wxg0103 in #1995
- 应用:修复部分网络环境下使用图片理解模型对话结束后,提示Connection aborted的问题;by @liuruibin in #1994
- 应用:修复判断器节点的执行条件设置为any时不生效的问题;by @shaohuzhang1 in #1979
- 应用:修复添加图片理解和图片生成模型时,不能自定义输入基础模型的问题;by @liuruibin in #1983
- 应用:修复部分情况下在执行详情中会遗漏执行节点的问题。by @shaohuzhang1 in #1979
v1.9.0
新增功能
- 知识库:知识库和文档支持导出Excel文件与离线图片的ZIP包;by @shaohuzhang1 in #1869
- 知识库:上传文档选择文本文件类型时,新增XLS、XLSX、CSV、ZIP文件格式;by @shaohuzhang1 in #1869
- 知识库:上传文档选择QA问答对类型时,新增ZIP文件格式;by @shaohuzhang1 in #1869
- 应用:新增图片生成节点;by @liuruibin,@wangdan-fit2cloud in #1812
- 应用:支持上传音频文件;by @liuruibin,@wangdan-fit2cloud in #1787
- 应用:新增语音转文本节点;by @wxg0103,@wangdan-fit2cloud in #1827
- 应用:新增文本转语音节点;by @wxg0103,@wangdan-fit2cloud in #1827
- 应用:支持导出和导入应用;by @shaohuzhang1,@wangdan-fit2cloud in #1836
- 应用:工作流节点支持设置执行条件;by @shaohuzhang1,@wangdan-fit2cloud in #1888
- 应用(X-Pack):支持接入企业微信的微信客服;
- 应用(X-Pack):公众号、企业微信、微信客服支持语音提问和语音回答;
- 模型设置:创建模型时,支持设置模型参数;by @liuruibin in #1911
- 模型设置:图片理解模型新增支持Gemini供应商; by @liuruibin in #1857
- 模型设置:图片理解模型新增支持Azure OpenAI 供应商;by @shaohuzhang1 in #1857
- 模型设置:图片理解模型新增支持阿里云百炼 供应商 by @shaohuzhang1 in #1844
- 模型设置:图片理解模型新增支持Xinference、Ollama、豆包供应商; by @liuruibin in #1812
- 模型设置:图片生成模型新增支持Xinference、OpenAI、腾讯混元、通义千问、智谱AI、豆包、阿里云百炼、Azure OpenAI供应商;by @liuruibin in #1812
- 模型设置:向量模型新增支持Azure OpenAI供应商;by @liuruibin in #1868
- 模型设置:语音识别模型新增支持Azure OpenAI供应商;by @liuruibin in #1872
- 模型设置:语音合成模型新增支持Azure OpenAI供应商;by @liuruibin in #1872
功能优化
- 知识库:文档列表支持批量取消向量化和批量取消生成问题;by @shaohuzhang1 in #1896
- 应用:基本信息节点中的“用户输入“参数,添加参数时支持设置“显示默认值”选项;by @liuruibin in #1890
- 应用:表单收集节点的参数支持设置“显示默认值”选项;by @liuruibin in #1890
- 应用:对话URL支持携带“question=问题”参数,打开对话页面时自动发送问题;
- 应用:上传图片时自动生成“请解析图片内容”的问题;
- 应用:优化工作流节点的执行效率;by @shaohuzhang1 in #1913
- 模型设置:Amazon Bedrock供应商添加大语言模型时,支持ProxyURL参数。by @wxg0103 in #1901
问题修复
- 安全:修复函数库模块中远程命令执行的安全漏洞(CVE-2024-56137);by @liqiang-fit2cloud in #1856
- 应用:修复浮窗对话框中不显示“新建对话”按钮的问题;
- 应用:修复浮窗对话框右上角的icon颜色不一致的问题;by @shaohuzhang1 in #1941
- 应用:修复历史应用进行对话时会提示“缺少上下文类型”错误的问题; by @shaohuzhang1 in #1926
- 应用:修复使用低版本浏览器进行对话时报错的问题;by @shaohuzhang1 in #1870
- 应用:修复执行复杂工作流时,在部分情况下会漏执行节点的问题。by @shaohuzhang1 in #1913
v1.8.1
新增功能
- 应用:AI对话节点的历史聊天记录可设置工作流和节点的上下文;by @shaohuzhang1 in #1791
- 模型管理:支持智谱AI供应商的图片理解模型;by @liuruibin in #1785
功能优化
- 应用:语音播放时过滤表单数据;by @liuruibin in #1769
- 系统:优化部分页面的显示样式。
问题修复
- 知识库:修复文档按状态筛选结果错误的问题;by @shaohuzhang1 in #1800
- 知识库:修复拥有查看权限的成员可以导出文档的问题; by @shaohuzhang1 in #1807
- 知识库:修复拥有查看权限的成员可以修改分段启用/禁用状态的问题;
- 知识库:修复上传无目录结构的PDF文件,只能识别第一页内容的问题;
- 应用:按名称筛选结果后,不加载数据的问题;
- 应用:修复浮窗图标尺寸过大时会被引导图遮盖的问题;by @shaohuzhang1 in #1802
- 系统管理(X-Pack):修复Swagger文档中的Web站点知识库缺少参数的问题;by @wxg0103 in #1810
- 系统管理(X-Pack):修复企业微信扫码登录配置错误的问题。by @wxg0103 in #1782
v1.8.0
新增功能
- 应用:用户提问时支持上传文档和图片;by @liuruibin,@wangdan-fit2cloud in #1602
- 应用:工作流新增文档内容提取节点;by @liuruibin,@wangdan-fit2cloud in #1602
- 应用:工作流新增图片理解节点;by @liuruibin,@wangdan-fit2cloud in #1602
- 应用:工作流新增表单收集节点;by @shaohuzhang1,@wangdan-fit2cloud in #1608
- 应用:工作流支持嵌入应用节点;by @wxg0103,@wangdan-fit2cloud in #1598
- 应用:工作流画布支持收起或展开全部节点以及一键美化布局;by @shaohuzhang1 in #1644
- 应用:基础节点的用户输入参数中,支持多选框和选项卡组件;by @shaohuzhang1 in #1651
- 应用:开场白支持HTML渲染; by @shaohuzhang1 in #1608
- 应用:支持批量将对话日志添加到知识库;by @wxg0103 in #1606
- 应用:支持按创建者显示应用; by @wxg0103 in #1546
- 知识库:支持按创建者显示知识库;by @wxg0103 in #1546
- 函数库:支持按创建者显示函数;by @wxg0103 in #1546
- 模型设置:支持通义千问供应商的图片理解模型;by @liuruibin,@wangdan-fit2cloud in #1602
- 模型设置:支持OpenAI供应商的图片理解模型;by @liuruibin in #1545
- 模型设置:支持腾讯混元供应商的图片理解模型;by @liuruibin,@wangdan-fit2cloud in #1602
- 模型设置:支持阿里云百炼供应商的大语言模型;by @wxg0103 in #1642
- 系统管理(X-Pack):支持OAuth2单点登录方式;by @wxg0103
功能优化
- 知识库:优化文档状态,显示向量化和生成问题的进度详情;by @shaohuzhang1 in #1674
- 应用:基础节点的用户输入参数中,单选框组件支持标签、选项值的设置方式;by @shaohuzhang1 in #1719
- 应用:对话详情页面对AI回复内容进行标注时,保留上一次所选的知识库文档;by @wxg0103 in #1606
- 应用:优化对话日志列表的显示顺序,按最近对话时间降序排序; by @wxg0103 in #1645
- 应用:优化打开对话详情时,默认显示最新对话记录。by @wxg0103 in #1645
问题修复
- 知识库:修复PDF文档中若包含不能识别的特殊字符时,上传文档会报错的问题; by @liuruibin in #1640
- 知识库:修复部分情况下为“成功”状态的文档生成问题时,无法下发任务的问题;by @shaohuzhang1 in #1727
- 知识库:修复由于网站编码问题,导致web站点知识库无法获取文本内容的问题; by @liuruibin in #1743
- 系统管理:修复拥有知识库管理权限的成员,无法查看知识库的问题;by @wangdan-fit2cloud in #1740
- 系统管理(X-Pack):修复拥有应用查看权限的成员,能够修改应用接入配置的问题。by @wxg0103
v1.7.2
问题修复
- 应用:修复显示设置中自定义上传的用户头像、AI回复头像图片显示变形的问题;(X-Pack)
- 应用:修复工作流中知识库检索节点收起后,无法发布应用的问题;by @shaohuzhang1 in #1626
- 应用:修复AI模型的参数设置中,若删除温度或最大输出Tokens参数,无法发布应用的问题;by @shaohuzhang1 in #1629
- 应用:修复应用中关联的知识库名称过长时,偶然不显示tooltips的问题;by @wangdan-fit2cloud in #1612
- 应用:修复应用列表中查询结果显示不正确的问题。by @wangdan-fit2cloud in #1613
v1.7.1
功能优化
- 应用:高级编排应用中保存节点的收起/展开状态。by @shaohuzhang1 in #1560
问题修复
- 应用:修复应用接入认证过的公众号,提问时不回复的问题;(X-Pack)
- 应用:修复应用接入公众号配置页面的样式问题;(X-Pack)
- 应用:函数中设置了非必填参数,在应用中引用该函数时,若未传值执行会报错的问题;by @shaohuzhang1 in #1570
- 应用:修复工作流中所有节点未打开返回内容时,无法查看执行详情的问题;by @shaohuzhang1 in #1572
- 应用:修复直接回答分段内容显示问题;by @shaohuzhang1 in #1576
- 应用:修复应用发布后,提问时未使用最新发布的应用配置的问题;by @shaohuzhang1 in #1554
- 知识库:修复导入DOCX文档时,在部分情况下无法识别图片问题;by @shaohuzhang1 in #1559
- 问答页面:修复AI回答时,若返回的MarkDown表格中含有图片内容会导致显示错位的问题; by @shaohuzhang1 in #1569
- 系统设置:修复团队成员的权限设置中,搜索后点击“全选”,会选中全部而不是搜索后的内容;by @wangdan-fit2cloud in #1562
- 模型管理:修复添加Xinference向量模型时,提示 API 域名无效的问题。by @wxg0103 in #1566
v1.7.0
新增功能
- 知识库:支持知识库文档自动生成关联问题;by @liuruibin in #1339
- 应用:工作流节点支持多进多出;by @shaohuzhang1 in #1362
- 应用:支持设置语音播放的声音和语速;by @liuruibin in #1402
- 应用:支持查看和恢复历史版本;by @shaohuzhang1,@wangdan-fit2cloud in #1386
- 应用:支持设置对话日志的清除策略;by @wxg0103 in #1486
- 应用:高级编排应用支持自动保存设置,默认不开启,开启后默认5分钟保存一次;by @wangdan-fit2cloud in #1517
- 应用:工作流编排界面在添加组件时支持按名称查询组件;by @shaohuzhang1 in #1358
- 模型设置:支持自定义大语言模型和语音合成模型的参数;by @liuruibin in #1458
- 模型设置:支持按模型类型、创建人、权限查询模型;by @liuruibin in #1358
- 模型设置:模型面板中增加创建人信息;by @liuruibin in #1358
- 模型设置:支持阿里云百炼的向量模型;by @shaohuzhang1 in #1392
- 模型设置:支持 Amazon Bedrock 的向量模型;by @wxg0103 in #1399
- 模型设置:支持腾讯混元的向量模型;by @shaohuzhang1 in #1400
- 模型设置:支持讯飞星火的向量模型;by @shaohuzhang1 in #1393
- 模型设置:支持百度千帆的向量模型;by @shaohuzhang1 in #1394
- 模型设置:支持阿里云百炼的语音识别和语音合成模型; by @liuruibin in #1340
- 模型设置:支持 Xinference 的语音识别和语音合成模型; by @liuruibin in #1349
- 问答页面:支持个性化显示设置;(X-Pack)by @wxg0103,@wangdan-fit2cloud in #1428
- 问答页面:支持身份验证;(X-Pack)by @wxg0103 in #1370
- 系统设置:支持企业微信、钉钉、飞书扫码登录。(X-Pack)by @wxg0103
功能优化
- 应用:基础节点的全局变量优化为用户输入和接口传参两个参数表单;by @liuruibin in #1410
- 应用:关联知识库时支持按名称查询知识库;by @wangdan-fit2cloud in #1338
- 应用:关联知识库的名称过长时增加悬浮显示;by @wangdan-fit2cloud in #1338
- 应用:
知识库检索结果
和满足直接回答的分段内容
中包含分段标题;by @shaohuzhang1 in #1377 - 应用:导出高级编排应用时包含知识库检索节点的引用分段数据;by @wxg0103 in #1377
- 应用:在语音输入时,控制台不再输出调试信息;by @liuruibin in #1346
- 函数库:优化函数中 Python 编辑器组件,使得编写函数体验更友好;by @shaohuzhang1 in #1328
- 团队管理:设置资源权限时增加全选按钮;by @wangdan-fit2cloud in #1335
- 模型设置:创建模型时,支持按模型类型筛选供应商;by @liuruibin in #1384
- 模型设置:优化供应商列表的显示顺序和分类,将供应商划分为公有模型供应商和私有模型供应商;by @liuruibin in #1359
- 系统设置:优化修改密码的逻辑,先验证两次密码是否一致,再发邮件验证码。by @wxg0103 in #1459
问题修复
- 知识库:修复导入 License 文件后,会显示一次 License 过期的提示的问题;(X-Pack)by @shaohuzhang1 in #1429
- 知识库:修复当导出空知识库时,打开导出文件报错的问题;by @shaohuzhang1 in #1501
- 知识库:修复当知识库文档中图片不存在时,AI 回复内容显示异常的问题;by @shaohuzhang1 in #1483
- 知识库:修复在知识库上传 DOCX 文件后,部分情况下图片未正常导入的问题;by Henry-Shaw in #1482
- 知识库:修复在知识库上传 PDF 文件,选择高级分段时,设置分段长度不生效的问题;by @liuruibin in #1488
- 知识库:修复在知识库分段中插入 MarkDown 流程图,不显示文本内容的问题;
- 应用:修复当应用使用的模型为豆包智能体模型时,接入企业微信提问不回复的问题;by @wxg0103 in #1425
- 应用:修复知识来源引用分段的文件名过长时,界面显示错乱的问题;
- 应用:修复在接入的公众号中提问时,若回复内容在图片后面有文本内容, 无法返回图片的问题;(X-Pack)
- 模型管理:修复添加 Xinference 模型时,如果模型名称和 ID 不一致,无法添加模型的问题;by @wxg0103 in #1390
- 模型管理:修复添加 Xinference 模型时,没有传认证参数导致提示 API 域名无效的问题。by @dolphinlife in #1334
v1.6.1
问题修复:
- 【应用】修复使用AWS模型进行对话时报错的问题。
- 【应用】修复升级后第一次进入高级编排应用编辑页面,显示空白的问题。by @liuruibin in #1310
- 【应用】修复高级编排应用设置全局变量后,打开调试对话框,选择变量后调试对话框会被关闭的问题。 by @shaohuzhang1 in #1316
- 【应用】修复关联知识库中,知识库名称过长导致显示不全的问题。by @wangdan-fit2cloud in #1318
- 【知识库】修复分段内容超过单元格最大长度时,导出报错的问题。 by @shaohuzhang1 in #1319