Skip to content

Releases: ModelEngine-Group/nexent

v1.7.6

21 Nov 12:15
@SimengBian SimengBian
v1.7.6
e27d265
This commit was created on GitHub.com and signed with GitHub’s verified signature.
GPG key ID: B5690EEEBB952194
Verified
Learn about vigilant mode.

Choose a tag to compare

View all tags
v1.7.6 Latest
Latest

🚀 Nexent：开源智能体平台 / Nexent: Open Source Intelligent Agent Platform

我们很高兴地宣布Nexent v1.7.6正式发布！🎉
Nexent 是一个开源智能体平台，能够将流程的自然语言转化为完整的多模态智能体 —— 无需编排，无需复杂拖拉拽。基于 MCP 工具生态，Nexent 提供强大的模型集成、数据处理、知识库管理、零代码智能体开发能力。我们的目标很简单：将数据、模型和工具整合到一个智能中心中，使日常工作流程更智能、更互联。

We are excited to announce that Nexent v1.7.6 is released! 🎉
Nexent is an open-source agent platform that turns process-level natural language into complete multimodal agents — no diagrams, no wiring. Built on the MCP tool ecosystem, Nexent provides model integration, data processing, knowledge-base management, and zero-code agent development. Our goal is simple: to bring data, models, and tools together in one smart hub, making daily workflows smarter and more connected.

🔔 核心功能更新 / Key Enhancements

1️⃣ 新增全局导航栏 / Global Navigation Bar

我们在页面左侧新增了清晰易用的导航栏，优化了用户的操作路径，支持快速跳转至各个核心模块，显著提升使用效率与操作便捷性。

We have introduced a user-friendly global navigation bar on the left side of the page, streamlining user workflows and enabling quick access to key modules, significantly improving overall usability and navigation efficiency.

2️⃣ 智能体空间上线 / Agent Space Launch

全新推出“智能体空间”页面，用户可在此以直观的卡片形式浏览所有智能体，查看详细配置信息，并执行删除、导出等管理操作，帮助用户更好地组织与管理智能体资源。

We are excited to launch the new "Agent Space" page, where users can browse all agents in an intuitive card-based layout, view detailed configurations, and perform management actions such as deletion and export, enabling better organization and control over agent resources.

3️⃣ 视频播放支持 / Video Playback Support

前端现已支持视频内容的嵌入与播放，用户可对视频进行播放、暂停等交互操作，进一步丰富了内容展示形式与用户体验。

The frontend now supports video embedding and playback, allowing users to interact with video content through play and pause controls, enriching content presentation and enhancing the overall user experience.

4️⃣ 后端容器架构优化 / Backend Container Architecture Optimization

我们将后端容器拆分为编辑态与运行态两种模式，有效提升系统资源利用率和任务处理能力，显著增强多任务并发性能。

We have split the backend container into separate editing and runtime modes, optimizing system resource utilization and task processing capabilities, thereby significantly improving performance under concurrent workloads.

5️⃣ 知识库切块预览 / Knowledge Base Chunk Preview

新增知识库内容切块预览功能，用户可直观查看知识库中被分段处理后的详细内容，便于理解和校验知识的结构与完整性。

We've added a knowledge base chunk preview feature, allowing users to visually inspect the segmented content within their knowledge bases, making it easier to understand and verify the structure and integrity of stored knowledge.

6️⃣ BUG 修复 / Bug Fixes

本次更新还修复了其他9个已知问题，持续提升系统稳定性与用户体验。

This release also addresses 9 other known issues, further enhancing system stability and user experience.

🧭 特性预览 / Feature Preview

  • 智能体市场即将上线 / Agent Marketplace Coming Soon

我们即将推出智能体市场，为用户提供丰富多样的实用与趣味智能体，支持一键下载与部署，拓展智能体应用场景。

We will soon launch the Agent Marketplace, offering a variety of practical and entertaining agents for users to download and deploy with one click, expanding the possibilities of agent applications.

  • 多模态 SDK 开发 / Multimodal SDK in Development

我们正在开发全新的多模态 SDK，以支持图像、音频等多模态数据在智能体间的无缝、高效流转，赋能更复杂的跨模态交互场景。

A new Multimodal SDK is currently in development, designed to facilitate seamless and efficient transmission of multimodal data (such as images and audio) between agents, enabling more sophisticated cross-modal interaction scenarios.

What's Changed

New Contributors

Full Changelog: v1.7.5.3...v1.7.6

Contributors

lvxiaohai99, Jasonxia007, and 15 other contributors
Assets 2
Loading

v1.7.5.3-hotfix

17 Nov 07:02
@WMC001 WMC001
v1.7.5.3-hotfix
d666c6c
This commit was created on GitHub.com and signed with GitHub’s verified signature.
GPG key ID: B5690EEEBB952194
Verified
Learn about vigilant mode.

Choose a tag to compare

View all tags
v1.7.5.3-hotfix

🔔 热修复 / Hotfix

1️⃣ Ray 进程初始化失败修复 / Ray Process Initialization Failure Fix

修复了本地环境中 Ray 进程启动时因配置加载异常导致的初始化失败问题

Fixed the initialization failure of Ray processes in the local environment caused by abnormal configuration loading.

Loading

v1.7.5.3

15 Nov 07:11
@SimengBian SimengBian
v1.7.5.3
9c165e6
This commit was created on GitHub.com and signed with GitHub’s verified signature.
GPG key ID: B5690EEEBB952194
Verified
Learn about vigilant mode.

Choose a tag to compare

View all tags
v1.7.5.3

🚀 Nexent：开源智能体平台 / Nexent: Open Source Intelligent Agent Platform

我们很高兴地宣布Nexent v1.7.5.3正式发布！🎉
Nexent 是一个开源智能体平台，能够将流程的自然语言转化为完整的多模态智能体 —— 无需编排，无需复杂拖拉拽。基于 MCP 工具生态，Nexent 提供强大的模型集成、数据处理、知识库管理、零代码智能体开发能力。我们的目标很简单：将数据、模型和工具整合到一个智能中心中，使日常工作流程更智能、更互联。

We are excited to announce that Nexent v1.7.5.3 is released! 🎉
Nexent is an open-source agent platform that turns process-level natural language into complete multimodal agents — no diagrams, no wiring. Built on the MCP tool ecosystem, Nexent provides model integration, data processing, knowledge-base management, and zero-code agent development. Our goal is simple: to bring data, models, and tools together in one smart hub, making daily workflows smarter and more connected.

🔔 本周核心BUG修复 / This Week's Key Bug Fixes

1️⃣ 知识库总结失败：修复了知识库总结失败的问题

Knowledge Base Summary Failure: Fixed issue where knowledge base summary failed

2️⃣ 聊天框状态切换问题：修复了聊天框状态应切换为"已读"的问题

Chat Box Status Switch Issue: Fixed issue where Dialog status should switch to "read"

3️⃣ 页面显示和跳转逻辑：修复了轻量版中页面显示逻辑和完整版中页面跳转逻辑

Page Display/Jump Logic: Fixed page display logic in speed mode and page jump logic in full mode

4️⃣ 登录窗口关闭问题：修复了登录窗口无法关闭的问题

Login Window Cannot Be Closed: Fixed issue where login window cannot be closed

5️⃣ 其他BUG修复：修复了其他5个BUG，持续提升系统稳定性和用户体验

Other bug fixes: Fixed 5 other bugs, continuously improving system stability and user experience

What's Changed

  • 🐛 Fixed an issue where batch modification of the maximum token count was not synchronized in a single model setting. #1575 by @YehongPan in #1590
  • 🐛 When a user swipes and taps the chat box, the Dialog status should switch to "read". #1388 by @YehongPan in #1591
  • 🐛 Knowledge base summary failed. #1619 by @porkpink in #1620
  • 🐛 Page display logic fix in speed mode and page jump logic fix in full mode by @WMC001 in #1627
  • 🐛 In the MCP configuration, connectivity is verified, and after a tool is deleted or added (or its status changes), the tool list and agent list are refreshed synchronously. #1345 by @YehongPan in #1633
  • ✨[Request] Abstract Storage SDK by @Zhi-a in #1628
  • Add user feedback on automatic agent creation feature by @JQTXZS2 in #1645
  • Revert "✨[Request] Abstract Storage SDK" by @Phinease in #1668
  • ✨ Now user can save or discard changes in agent config page by @Jasonxia007 in #1654
  • 🐛 Even when there is no historical login token, the login prompt needs to pop up #1476 by @WMC001 in #1648
  • Add tip section for new open source contributors by @0501-maomao in #1655
  • Add tip from jeery1 about Huawei ICT event by @Jeery1 in #1656
  • ✨[Request] Abstract Storage SDK #1219 by @Zhi-a in #1651
  • 🐛 Bugfix: login window cannot be closed #1476 by @WMC001 in #1679
  • 🐛 Agents with abnormal status should also be allowed to operate #1683 by @WMC001 in #1687
  • 🐛 Bugfix: generate button in Agent config page returns error: Empty agent ID in creation mode by @Jasonxia007 in #1689
  • ♻️ Imporvement: nexent-data-process now takes less memory, elasticsearch takes more available memory by @Jasonxia007 in #1691

New Contributors

Full Changelog: v1.7.5.2...v1.7.5.3

Contributors

Jasonxia007, Phinease, and 22 other contributors
Loading

v1.7.5.2

11 Nov 01:36
@SimengBian SimengBian
v1.7.5.2
747761f
This commit was created on GitHub.com and signed with GitHub’s verified signature.
GPG key ID: B5690EEEBB952194
Verified
Learn about vigilant mode.

Choose a tag to compare

View all tags
v1.7.5.2

🚀 Nexent：开源智能体平台 / Nexent: Open Source Intelligent Agent Platform

我们很高兴地宣布Nexent v1.7.5.2正式发布！🎉
Nexent 是一个开源智能体平台，能够将流程的自然语言转化为完整的多模态智能体 —— 无需编排，无需复杂拖拉拽。基于 MCP 工具生态，Nexent 提供强大的模型集成、数据处理、知识库管理、零代码智能体开发能力。我们的目标很简单：将数据、模型和工具整合到一个智能中心中，使日常工作流程更智能、更互联。

We are excited to announce that Nexent v1.7.5.2 is released! 🎉
Nexent is an open-source agent platform that turns process-level natural language into complete multimodal agents — no diagrams, no wiring. Built on the MCP tool ecosystem, Nexent provides model integration, data processing, knowledge-base management, and zero-code agent development. Our goal is simple: to bring data, models, and tools together in one smart hub, making daily workflows smarter and more connected.

🔔 本周核心BUG修复 / This Week's Key Bug Fixes

1️⃣ 批量添加模型失败：修复了因缺少分块大小参数导致批量添加模型失败的问题

Batch Model Adding Failure: Fixed issue where batch adding model failed due to missing expected chunk size parameter

2️⃣ Markdown意外换行：修复了最终答案中Markdown意外换行的问题

Unexpected Line Break in Markdown: Fixed issue where unexpected Markdown line break appears in final answer

3️⃣ 模型连接失败时的提示错误：修复了在中文环境下模型连接失败时显示英文错误信息的问题

Wrong Error Message When Model Connection Failed: Fixed issue where English error message shows up under Chinese environment when model connection fails

4️⃣ 同名Agent状态修复：修复了同名Agent均可运行的问题；现在当存在同名Agent时，系统会将其标记为不可用并提醒用户修改名称。

Duplicate Agent Name Status Repair: Fixed issue where multiple agents with the same name could all run; now duplicate-named agents will be marked as unusable and guide the user to modify the name.

5️⃣ 其他BUG修复：修复了其他9个BUG，持续提升系统稳定性和用户体验

Other bug fixes: Fixed 9 other bugs, continuously improving system stability and user experience

What's Changed

  • 🐛 [Bug] Fixed the issue where the Agent prompt text description overwrote the button. by @Zhi-a in #1500
  • 🐛 [Improvement] When creating an Agent, the information items of the created agent should be clear by @Zhi-a in #1508
  • 🐛 [Bug] English error message shows up under chinese language when model is not connected #1502 by @Mermaid97 in #1543
  • ✨ Now user can adjust the chunking size when adding or editing the embedding model by @Jasonxia007 in #1512
  • 🎨 Tool test run results formatting #1399 by @porkpink in #1556
  • 🐛 [Bug] Translation unavailable when generating agent without business description by @Zhi-a in #1557
  • 🐛 [Bug] An agent with the same name should be displayed as unusable. by @Zhi-a in #1560
  • 🐛 Fix unexpected line break on markdown display of final answer #1475 by @YehongPan in #1563
  • 🐛 [Bug] Remove unnecessary floating tooltip. by @Zhi-a in #1564
  • 🐛 Button text in agent tools modal are not centered. #1506 by @porkpink in #1565
  • 🐛 Fix the issue where the frontend still throws an exception after successfully batch modifying the model configuration #1439 by @YehongPan in #1567
  • 🐛 Repair English error message shows up under chinese language when model is not connected by @Mermaid97 in #1513
  • 🎨 [Improvement] When uploading the same document multiple times, the knowledge base summary will classify them into multiple categories. by @Mermaid97 in #1551
  • 🐛 Generating title error when using tokenpony deepseek-r1 #1339 by @porkpink in #1570
  • 🐛 When not logged in, access to the backend interface is not allowed #1476 by @WMC001 in #1571
  • 🐛 Bugfix: bacth create model failed due to missing expected chunk size by @Jasonxia007 in #1573
  • 🐛 Bugfix: bacth create model failed due to missing expected chunk size by @Jasonxia007 in #1574
  • [release] 📝 Update APP version by @WMC001 in #1584
  • Release/v1.7.5.2 by @WMC001 in #1613

New Contributors

Full Changelog: v1.7.5.1...v1.7.5.2

Contributors

Jasonxia007, YuhangZho, and 25 other contributors
Loading

v1.7.5.1

31 Oct 09:20
@SimengBian SimengBian
v1.7.5.1
2f62c2a
This commit was created on GitHub.com and signed with GitHub’s verified signature.
GPG key ID: B5690EEEBB952194
Verified
Learn about vigilant mode.

Choose a tag to compare

View all tags
v1.7.5.1

🚀 Nexent：开源智能体平台 / Nexent: Open Source Intelligent Agent Platform

我们很高兴地宣布Nexent v1.7.5.1正式发布！🎉
Nexent 是一个开源智能体平台，能够将流程的自然语言转化为完整的多模态智能体 —— 无需编排，无需复杂拖拉拽。基于 MCP 工具生态，Nexent 提供强大的模型集成、数据处理、知识库管理、零代码智能体开发能力。我们的目标很简单：将数据、模型和工具整合到一个智能中心中，使日常工作流程更智能、更互联。

We are excited to announce that Nexent v1.7.5.1 is released! 🎉
Nexent is an open-source agent platform that turns process-level natural language into complete multimodal agents — no diagrams, no wiring. Built on the MCP tool ecosystem, Nexent provides model integration, data processing, knowledge-base management, and zero-code agent development. Our goal is simple: to bring data, models, and tools together in one smart hub, making daily workflows smarter and more connected.

🔔 本周核心BUG修复 / This Week's Key Bug Fixes

1️⃣ 知识库PDF处理失败：修复了上传PDF文件时知识库处理失败的问题

Knowledge Base PDF Processing Failure: Fixed issue where knowledge base failed to process when uploading .pdf files

2️⃣ 应用图标重新上传和保存问题：修复了无法重新上传或保存应用图标的问题

App Icon Re-upload/Save Issue: Fixed issue where cannot re-upload or save app icon

3️⃣ 导入Agent配置时模型信息丢失：修复了导入Agent配置时大语言模型信息未写入数据库的问题

Model Info Loss When Importing Agent Config: Fixed issue where LLM information wasn't written to database when importing agent configuration

4️⃣ 手动输入模式参数值不生效：修复了MCP工具测试手动输入模式下更改参数值不生效的问题

Manual Input Parameter Not Effective: Fixed issue where in MCP tool test run changing parameter value in manual input mode doesn't take effect

5️⃣ 其他BUG修复：修复了其他8个BUG，持续提升系统稳定性和用户体验

Other bug fixes: Fixed 8 other bugs, continuously improving system stability and user experience

What's Changed

  • 🐛 Modify the knowledge base front-end rendering #1458#1456 by @Mermaid97 in #1459
  • 🐛 Bugfix: knowledgebase fail to process when upload .pdf files by @Jasonxia007 in #1474
  • [Improvement] The suffixes of LLM and embedding models are not consistent during configuration #1402 by @Zhi-a in #1472
  • Fixed the default filling of email account and password when adding models by @Zhi-a in #1481
  • ✨ Allow to edit unavailable agents #1312 by @WMC001 in #1486
  • 🐛 When debugging the agent is allowed, it is possible to have multi-t… #1340 by @Mermaid97 in #1488
  • 🐛 Bugfix: cannot re-upload or save app icon by @Jasonxia007 in #1487
  • #1475 🐛 Modify markdown page rendering by @Mermaid97 in #1489
  • #1450 🐛 Friendly error messages when adding a model by @Mermaid97 in #1491
  • 🐛 [Bugfix] Fixed the issue where large language model information was not written to the database when importing agent configuration. by @Zhi-a in #1480
  • 🐛 Changing the parameter value in manual input mode does not take effect. #1471 by @porkpink in #1492
  • 🐛 [Request] Administrator can go to any configuration page #1346 by @Mermaid97 in #1493
  • 🐛 update the app version. by @porkpink in #1499
  • Release/v1.7.5.1 by @WMC001 in #1514

New Contributors

Full Changelog: v1.7.5...v1.7.5.1

Contributors

Jasonxia007, WMC001, and 23 other contributors
Loading

v1.7.5

25 Oct 07:35
@WMC001 WMC001
v1.7.5
bdc0e70
This commit was created on GitHub.com and signed with GitHub’s verified signature.
GPG key ID: B5690EEEBB952194
Verified
Learn about vigilant mode.

Choose a tag to compare

View all tags
v1.7.5

🔔 核心功能更新 / Key Enhancements

1️⃣ Mermaid图表支持 / Mermaid Diagram Support

支持展示Mermaid图表，在与智能体对话时可正确显示流程图、时序图、甘特图等，并支持放大/缩小、下载和查看源码等操作。

Now supports displaying Mermaid diagrams, correctly rendering flowcharts, sequence diagrams, Gantt charts, etc., during conversations with agents, with support for zooming in/out, downloading, and viewing source code.

2️⃣ 默认模型配置优化 / Default Model Configuration Optimization

修改模型配置逻辑，在创建智能体的所有环节中，若无特殊指定，则默认使用系统大语言模型，简化操作流程，实现一次配置即可全流程通用。

Optimized the model configuration logic - when creating agents, the system's large language model is now used by default in all steps unless otherwise specified, simplifying operations and enabling one-time configuration for seamless end-to-end usage.

3️⃣ MCP工具测试增强 / MCP Tool Testing Enhancement

更新MCP工具测试逻辑，支持以JSON格式输入原始测试参数，也可单独填写每个参数的值，提供更灵活的测试方式。

Enhanced MCP tool testing logic, now supporting raw test parameter input in JSON format or individual parameter value entry, offering greater testing flexibility.

4️⃣ 知识库总结算法升级 / Knowledge Base Summary Algorithm Upgrade

优化知识库内容总结功能，引入分层聚类算法，能够准确提炼多主题复杂知识库的关键内容，并以分点形式清晰陈列。

Upgraded the knowledge base summarization feature with a hierarchical clustering algorithm, enabling accurate extraction of key content from complex multi-topic knowledge bases and presenting it in a bulleted format.

5️⃣ 问题修复 / Bug Fixes

累计修复5个关键缺陷，持续提升系统稳定性和用户体验。

Fixed 5 bugs, further improving system stability and user experience.

🧭 v1.7.6特性预览 / v1.7.6 Feature Preview

  • 多模态模型拓展支持 / Enhanced Multimodal Model Support

即将支持更丰富的多模态模型，提供更强大的视觉-语言理解与生成能力。

Soon you'll be able to leverage more comprehensive multimodal models for enhanced visual-language understanding and generation.

  • 多模态MCP工具接入 / Multimodal MCP Tool Integration

将支持接入多模态MCP工具，扩展智能体的多模态交互能力。

Future updates will enable integration of multimodal MCP tools, expanding agents' multimodal interaction capabilities.

  • 前端多模态数据支持 / Frontend Multimodal Data Support

前端将支持图片、视频等多模态数据的展示、保存及相关操作。

The frontend will support display, storage, and operations for multimodal data including images and videos.

Loading

v1.7.4.1

17 Oct 11:06
@SimengBian SimengBian
v1.7.4.1
210b54c
This commit was created on GitHub.com and signed with GitHub’s verified signature.
GPG key ID: B5690EEEBB952194
Verified
Learn about vigilant mode.

Choose a tag to compare

View all tags
v1.7.4.1

🚀 Nexent：开源智能体平台 / Nexent: Open Source Intelligent Agent Platform

我们很高兴地宣布Nexent v1.7.4.1正式发布！🎉
Nexent 是一个开源智能体平台，能够将流程的自然语言转化为完整的多模态智能体 —— 无需编排，无需复杂拖拉拽。基于 MCP 工具生态，Nexent 提供强大的模型集成、数据处理、知识库管理、零代码智能体开发能力。我们的目标很简单：将数据、模型和工具整合到一个智能中心中，使日常工作流程更智能、更互联。

We are excited to announce that Nexent v1.7.4.1 is released! 🎉
Nexent is an open-source agent platform that turns process-level natural language into complete multimodal agents — no diagrams, no wiring. Built on the MCP tool ecosystem, Nexent provides model integration, data processing, knowledge-base management, and zero-code agent development. Our goal is simple: to bring data, models, and tools together in one smart hub, making daily workflows smarter and more connected.

🔔 本周核心更新 / This Week's Key Updates

1️⃣ LaTeX渲染异常：修复了LaTeX内容渲染异常的问题，现在可正常展示LaTeX数学公式

Fix LaTeX content rendering exception, and now nexent can present LaTeX math formula correctly

2️⃣ 文件上传重复问题：修复了上传同名文件时的重复问题，现在会自动添加后缀

Upload file with the same names will now add suffix automatically and avoid duplication

3️⃣ 工具分组及测试运行：对本地工具进行了分组并添加了工具测试运行功能

Group local tools for effciently selection, and add tool test run function to ensure tool healthy before selection

4️⃣ 文档更新：修复了文档中表述不当内容，提升可读性

Repair improper expression in documents to improve readability

5️⃣ 其他BUG修复：修复了其他5个BUG，持续提升系统稳定性和用户体验。
Other bug fixes: Fixed 5 other bugs, continuously improving system stability and user experience.

What's Changed

New Contributors

Full Changelog: v1.7.4...v1.7.4.1

Contributors

xuyaqist, Jasonxia007, and 16 other contributors
Loading

v1.7.4

30 Sep 07:47
@SimengBian SimengBian
v1.7.4
a59b528
This commit was created on GitHub.com and signed with GitHub’s verified signature.
GPG key ID: B5690EEEBB952194
Verified
Learn about vigilant mode.

Choose a tag to compare

View all tags
v1.7.4

🚀 Nexent：开源智能体平台 / Nexent: Open Source Intelligent Agent Platform

我们很高兴地宣布Nexent v1.7.4正式发布！🎉 我们称该版本内部代号为 Jingyuan，以感谢 @ljy65535 为 Nexent 从零到一所做出的卓越贡献。✨
Nexent 是一个开源智能体平台，能够将流程的自然语言转化为完整的多模态智能体 —— 无需编排，无需复杂拖拉拽。基于 MCP 工具生态，Nexent 提供强大的模型集成、数据处理、知识库管理、零代码智能体开发能力。我们的目标很简单：将数据、模型和工具整合到一个智能中心中，使日常工作流程更智能、更互联。

We are excited to announce that Nexent v1.7.4 is released! 🎉 We internally codenamed this release Jingyuan, in honor of @ljy65535's excellent contributions to Nexent from zero to one. ✨
Nexent is an open-source agent platform that turns process-level natural language into complete multimodal agents — no diagrams, no wiring. Built on the MCP tool ecosystem, Nexent provides model integration, data processing, knowledge-base management, and zero-code agent development. Our goal is simple: to bring data, models, and tools together in one smart hub, making daily workflows smarter and more connected.

🔔 核心功能更新 / Key Enhancements

1️⃣ 统一LLM配置 / Unified LLM Configuration

我们取消了原有的主副模型设定，现在您只需配置一个系统大语言模型即可。后续所有使用大语言模型的场景将默认采用此模型，简化配置流程，提升操作效率。

We have streamlined the model configuration by removing the primary/secondary model setup. Now you only need to configure one system large language model, which will be used by default across all LLM-related functions, simplifying setup and improving operational efficiency.

2️⃣ 灵活模型选择 / Flexible Model Selection

在开发智能体时，您可以为每个智能体配置不同的大语言模型。针对任务简单明确的场景，可选用轻量模型以提升响应速度；对于复杂规划任务，则可选用能力更强的模型以提高准确率，实现灵活适配。在为知识库生成总结，或为智能体生成提示词时，您也可以根据需求选择不同的大语言模型。

When developing agents, you can configure different large language models for each agent. Use lighter models for simple tasks to improve speed, or more capable models for complex planning tasks to enhance accuracy – achieving optimal flexibility. When generating summary for knowledge bases or generating prompts for agents, you can also select different LLMs as needed.

3️⃣ 版本号展示 / Version Number Display

在首页左下角新增版本号展示，并在每个容器部署日志中输出当前版本号，便于用户快速定位和诊断问题。

The version number is now displayed in the lower-left corner of the homepage and printed in each container's deployment logs, helping users quickly identify and troubleshoot issues.

4️⃣ 模型供应商图标标识 / Model Vendor Icons

在模型配置页面，我们为每个模型添加了供应商图标（如硅基流动、DeepSeek等），方便您直观识别模型来源。

Added vendor icons (such as Silicon Flow, DeepSeek, etc.) for each model in the model configuration page, making it easier to visually identify model sources.

5️⃣ 向量化模型更换风险提示 / Embedding Model Change Risk Warning

在更换向量化模型时，系统会明确提示此操作可能存在风险，可能导致部分知识库不可用，引导用户重新选择知识库，确保操作安全。

When changing embedding models, the system now displays a clear warning about potential risks, including possible unavailability of some knowledge bases, guiding users to reselect knowledge bases for safer operations.

6️⃣ 账户注销 / Account Deletion

普通用户现在可以自主注销自己的账户，进一步保障用户对个人账户的管控权利。

Regular users can now delete their own accounts, providing greater control over personal account management.

7️⃣ 问题修复 / Bug Fixes

累计修复12个关键缺陷，持续提升系统稳定性和用户体验。

Fixed 12 bugs, further improving system stability and user experience.

🧭 v1.7.5特性预览 / v1.7.5 Feature Preview

  • 智能体图标自定义 / Agent Icon Customization

即将支持为智能体选择预设图标或上传自定义图标，让智能体更具个性化。

Soon you'll be able to select preset icons or upload custom ones for your agents, enhancing personalization.

  • Mermaid流程图与LaTeX公式支持 / Mermaid Flowchart & LaTeX Formula Support

将支持Mermaid流程图和LaTeX数学公式的生成与渲染展示。

Future updates will bring support for generating and rendering Mermaid flowcharts and LaTeX mathematical formulas.

  • 知识库内容总结优化 / Knowledge Base Content Summary Optimization

我们将优化知识库内容总结算法，支持对多主题复杂知识库进行关键内容提炼。

We'll optimize the knowledge base content summarization algorithm to support extracting key information from complex, multi-topic knowledge bases.

🔍 在线体验 / Online Experience

立即访问：http://nexent.tech/
欢迎您随时提交反馈或建议（通过GitHub Issues）。

Visit Now: http://nexent.tech/
Feel free to submit your feedback or suggestions (GitHub Issues).

🎁 早期贡献者福利 / Early Contributor Benefits

我们永远不会忘记Nexent的早期支持者！从专属徽章、纪念品到未来奖励，我们将全力回馈所有助力Nexent成长的伙伴。

We will never forget our early supporters!From exclusive badges and souvenirs to future rewards, we are committed to thanking all the partners who have helped Nexent grow.

📢 助力传播 / Help Us Spread the Word

最重要的是让更多人了解Nexent：为仓库点星⭐️、关注我们、分享给朋友——每一次点击都是对开源的有力支持！

The most important thing is to get Nexent in front of more people: Star the repository, follow us, and share it with your friends — every click is a powerful support for open-source!

What's Changed

  • ✨ add model/llm_list interface by @ljy65535 in #1236
  • ✨ use model_id when generate prompt by @ljy65535 in #1245
  • 🐛 a better way to distinguish between executable and generated code #832 by @Phinease in #1249
  • 📝 Added terminal tool user guide in Chinese and English #1234 by @WMC001 in #1244
  • Update opensource-memorial-wall.md by @wcq23 in #1256
  • ♻️ Refresh the tool list when backend service is initialized #1250 by @WMC001 in #1252
  • ✨ Can specify the model when configuring the agent. by @ljy65535 in #1259
  • Update opensource-memorial-wall.md by @Johnny-zbb in #1260
  • ♻️ Embedding reinforcement PART 1 by @Jasonxia007 in #1255
  • ✨ Remove all sub-model calls, provide users with the main models available to choose from in the knowledge base summary. by @Mermaid97 in #1241
  • ✨ use specify the model when running agent;long-text model use main model default by @ljy65535 in #1261
  • ✨ remove field of llmSecondary by @ljy65535 in #1266
  • Update opensource-memorial-wall.md by @fy-create in #1271
  • 🐛 When integrating the deep thinking model, the content of the think label in the final answer should be removed. #1254 by @porkpink in #1268
  • ✨ Support select model in multi cases by @ljy65535 in #1267
  • 🐛 The conversation will be stopped and the reason will be prompted when uploaded too large files #1246 #1247 by @WMC001 in #1275
  • ✨ Add performance monitor module by @Phinease in #1281
  • ♻️ Improvement: Warn the user trying to proceed when embedding model is not set by @Jasonxia007 in #1272
  • ♻️ Improvement: Now ModelEngine interface won't wait till timeout while related environment variables are not set by @Jasonxia007 in #1276
  • 🐛 Page text display optimization by @Mermaid97 in #1277
  • 🐛 fix the problems of the last week commit by @Mermaid97 in #1279
  • 🔨 use arm host for unitest runner and update monitor auto disable by @Phinease in #1286
  • 🐛 Fix tooltip display in agent pool and tool pool by @ljy65535 in #1280
  • 🎨 Incorrect rendering for examples in generated promt & chat interface. #688 by @porkpink in #1287
  • 🎨 [Improvement] Style improvement and minor bug fix by @Jasonxia007 in #1290
  • Update opensource-memorial-wall.md by @wuling786 in #1291
  • ♻️ refactor voice_app #1177 by @WMC001 in #1288
  • Update opensource-memorial-wall.md by @Blxh-4 in #1292
  • 🐛 [Improvement] Description on top of the chat box should be updated #826 by @Mermaid97 in #1289
  • ♻️ Refactor: unify common components Part 1 #1029 by @WMC001 in #1294
  • 🔨 improve test output and quality by @Phinease in #1296
  • Revert " ♻️ Refactor: unify common components Part 1 #1029" by @WMC001 in #1299
  • 🐛 Contribution wall is required when registering an administrator account by @ljy65535 in #1300
  • ✨ Now supports regular user account deletion by @Jasonxia007 in #1298
  • 🐛 Optimize Elasticsearch Startup Parameters and Monitoring Parameters by @Phinease in #1303
  • ♻️ Refactor: frontend internationalization component optimization #1107 by @WMC001 in #1302
  • Update opensource-memorial-wall.md by @xufanchao6657 in #1305
  • ✨ Version number display #1231 by @porkpink in #1304
  • ♻️ Refactor: remove unnecessary url constant #1073 by @WMC001 in #1306
  • 🐛 Record the settings of tools by @ljy65535 in #1308
  • 🐛 Fix ID verification exception when updating the model #1311 by @WMC001 in #1313
  • [release] 🐛 Fix ID verification exception when updating the model #...
Read more

Contributors

fy-create, SimengBian, and 11 other contributors
Loading

v1.7.3.4

19 Sep 09:49
@SimengBian SimengBian
v1.7.3.4
c07f15a
This commit was created on GitHub.com and signed with GitHub’s verified signature.
GPG key ID: B5690EEEBB952194
Verified
Learn about vigilant mode.

Choose a tag to compare

View all tags
v1.7.3.4

🚀 Nexent：开源智能体平台 / Nexent: Open Source Intelligent Agent Platform

我们很高兴地宣布Nexent v1.7.3.4正式发布！🎉
Nexent 是一个开源智能体平台，能够将流程的自然语言转化为完整的多模态智能体 —— 无需编排，无需复杂拖拉拽。基于 MCP 工具生态，Nexent 提供强大的模型集成、数据处理、知识库管理、零代码智能体开发能力。我们的目标很简单：将数据、模型和工具整合到一个智能中心中，使日常工作流程更智能、更互联。

We are excited to announce that Nexent v1.7.3.4 is released! 🎉
Nexent is an open-source agent platform that turns process-level natural language into complete multimodal agents — no diagrams, no wiring. Built on the MCP tool ecosystem, Nexent provides model integration, data processing, knowledge-base management, and zero-code agent development. Our goal is simple: to bring data, models, and tools together in one smart hub, making daily workflows smarter and more connected.

🔔 本周核心BUG修复 / This Week's Key Bug Fixes

1️⃣ 登出与会话获取问题：修复了用户登出异常和无法获取会话的问题
Logout and Session Retrieval Issue: Fixed problems with user logout and failure to retrieve sessions

2️⃣ 工具识别错误：修复了LLM错误将no_think识别为工具/助手的问题
Tool Identification Error: Fixed issue where LLM incorrectly regarded no_think as a tool/assistant

3️⃣ 北向导入与API错误：修复了北向导入错误及北向API无法存储和获取会话历史列表的问题
Northbound Import and API Error: Fixed northbound import error and API failure in storing/retrieving conversation history lists

4️⃣ 终端工具调试模式启动问题：修复了终端工具在调试模式下无法启动的问题
Terminal Tool Debug Mode Startup Issue: Fixed problem where terminal tool failed to start in debug mode

5️⃣ 其他BUG修复：修复了其他8个BUG，持续提升系统稳定性和用户体验。
Other bug fixes: Fixed 8 other bugs, continuously improving system stability and user experience.

What's Changed

  • 🐛 Bug fix: logout problem and failed to get session by @Jasonxia007 in #1187
  • ♻️ Refactor: Partial constant extraction in the backend by @Bavichi in #1171
  • 🐛 Fix terminal tool working directory permission #1183 by @WMC001 in #1184
  • ♻️ Refactor: cleanup console log. by @porkpink in #1191
  • 🐛 Bug fix: northbound import error by @Jasonxia007 in #1192
  • 🐛 Bug fix: re-include make/ folder in .dockerignore by @Jasonxia007 in #1195
  • Update opensource-memorial-wall.md by @CSJSAJA in #1198
  • ♻️ Add cursor rules of frontend by @WMC001 in #1197
  • Update opensource-memorial-wall.md by @EXUAN0312 in #1204
  • 🐛 Terminal tool does not start in debug mode #1157 by @WMC001 in #1215
  • 🐛 llm incorrectly regard no_think as a tool/assistant by @SimengBian in #1223
  • 🐛 Bug fix: northbound api cannot store and get conversation history list by @Jasonxia007 in #1203
  • 🔨 Rewrite voice APPID/TOKEN when deploying in service by @ljy65535 in #1207
  • 🐛 Better agent running log printing by @ljy65535 in #1221
  • Update opensource-memorial-wall.md by @TypeABC in #1226
  • 🐛 Refreshing the knowledge base or Agent configuration page will return to the model configuration page. by @porkpink in #1225
  • 🐛 Use conversation_id and user_id as the primary key of the agent run thread pool by @ljy65535 in #1224
  • 📝 add cursor rule for test file by @ljy65535 in #1227
  • ✨ The terminal tool supports password login #1052 by @WMC001 in #1228
  • 🐛 When generating prompts using deep thinking models, it is necessary to remove the deep thinking content. #1229 by @porkpink in #1240
  • 🐛 a better way to distinguish between executable and generated code #832 by @Phinease in #1248
  • [release] ♻️ Refresh the tool list when backend service is initialized #1250 by @WMC001 in #1253
  • Release/v1.7.3.4 by @ljy65535 in #1263

New Contributors

Full Changelog: v1.7.3.3...v1.7.3.4

Contributors

SimengBian, Jasonxia007, and 8 other contributors
Loading

v1.7.3.3

12 Sep 10:12
@SimengBian SimengBian
v1.7.3.3
6e44c28
This commit was created on GitHub.com and signed with GitHub’s verified signature.
GPG key ID: B5690EEEBB952194
Verified
Learn about vigilant mode.

Choose a tag to compare

View all tags
v1.7.3.3

🚀 Nexent：开源智能体平台 / Nexent: Open Source Intelligent Agent Platform

我们很高兴地宣布Nexent v1.7.3.3正式发布！🎉
Nexent 是一个开源智能体平台，能够将流程的自然语言转化为完整的多模态智能体 —— 无需编排，无需复杂拖拉拽。基于 MCP 工具生态，Nexent 提供强大的模型集成、数据处理、知识库管理、零代码智能体开发能力。我们的目标很简单：将数据、模型和工具整合到一个智能中心中，使日常工作流程更智能、更互联。

We are excited to announce that Nexent v1.7.3.3 is released! 🎉
Nexent is an open-source agent platform that turns process-level natural language into complete multimodal agents — no diagrams, no wiring. Built on the MCP tool ecosystem, Nexent provides model integration, data processing, knowledge-base management, and zero-code agent development. Our goal is simple: to bring data, models, and tools together in one smart hub, making daily workflows smarter and more connected.

🔔 本周核心BUG修复 / This Week's Key Bug Fixes

1️⃣ 嵌入模型维度偶尔为0：修复了嵌入模型偶尔出现维度为0的问题
Embedding Model Occasionally Has Dimension of 0: Fixed issue where embedding model occasionally had a dimension of 0

2️⃣ 部署时环境变量默认值错误：修复了基础设施模式下部署时ENV默认值错误设置为Y的问题
Incorrect ENV Default Value in Deployment: Fixed issue where ENV default was incorrectly set to Y when deploying in Infrastructure mode

3️⃣ 登出和会话获取问题：修复了登出问题和获取会话失败的问题
Logout and Session Retrieval Issues: Fixed logout problem and failed to get session

4️⃣ 终端工具工作目录权限问题：修复了终端工具工作目录权限问题
Terminal Tool Working Directory Permission: Fixed terminal tool working directory permission issue

5️⃣ 其他BUG修复：修复了其他7个BUG，持续提升系统稳定性和用户体验
Other bug fixes: Fixed 7 other bugs, continuously improving system stability and user experience.

What's Changed

  • 🔨 Add Doc Build Check. by @porkpink in #1137
  • 🐛 Avoid recording sensitive information. by @ljy65535 in #1146
  • ♻️ Refactor: Front-end agent configuration, knowledge base related constants and types are placed in specific files by @Bavichi in #1141
  • ♻️ Refactor: The Agent information bar uses AntD style #1133 by @WMC001 in #1144
  • ♻️ Front-end model configuration, knowledge base related constants and types are placed in specific files by @Bavichi in #1149
  • ♻️ refactor user_management_xxx.py by @ljy65535 in #1148
  • ♻️ refactor image module. by @porkpink in #1150
  • 🐛 Bug fix: users can not start new conversation when memory is still adding. by @Jasonxia007 in #1134
  • 🔨 Update mdc cursor rules and claude.md for backend by @Jasonxia007 in #1152
  • 📝 modify sdk readme file by @ljy65535 in #1159
  • ♻️ Front-end model configuration, chat page related constants and types are placed in specific files by @Bavichi in #1153
  • 🐛 When loading the agent list on the /chat page, the drop-down box flashes in the upper right corner. by @ljy65535 in #1161
  • ♻️ refactor me model module. by @porkpink in #1154
  • ✨ Truncation information of files that are too large is displayed to the user #1119 by @WMC001 in #1158
  • 🔨 Add auto web type check. by @porkpink in #1163
  • ♻️ Front-end services, hooks, chat page related constants and types are placed in specific files by @Bavichi in #1162
  • 🐛 The embedding model occasionally has a dimension of 0 by @Bavichi in #1165
  • ♻️ Translate error messages and Chinese comments in backend code into English #1131 by @Mermaid97 in #1156
  • ♻️ Refactor: Sort backend imports (utils) #1037 by @WMC001 in #1168
  • ♻️ Front-end services BUG FIXS by @Bavichi in #1169
  • 🐛 The default change of ENV is Y when deploying in Infrastructure mode. by @porkpink in #1167
  • ♻️ Trans exception from format JSONResponse to format HTTPException in user_managerment_app.py by @ljy65535 in #1170
  • 🐛 Delete the cache of agent information on the agent configuration page. by @ljy65535 in #1166
  • Update README.md by @Davina-jcx in #1176
  • ♻️ Refactor: cleanup data_process_app.py by @porkpink in #1174
  • ♻️ Complete the front-end Chinese comments. by @Mermaid97 in #1175
  • ♻️ Refactor: cleanup model_managment_app.py, memory_config_app.py by @Jasonxia007 in #1172
  • ♻️ Refactor: Partial constant extraction in the backend #1171 by @Bavichi in #1182
  • [release] 🐛 Bug fix: logout problem and failed to get session by @Jasonxia007 in #1188
  • [release] 🐛 Fix terminal tool working directory permission #1183 by @WMC001 in #1185
  • [release] 🐛 Bug fix: northbound import error by @Jasonxia007 in #1193
  • [release] 🐛 Bug fix: re-include make/ folder in .dockerignore by @Jasonxia007 in #1196
  • Release/v1.7.4 by @ljy65535 in #1199

Full Changelog: v1.7.3.2...v1.7.3.3

Contributors

Jasonxia007, ljy65535, and 5 other contributors
Loading