Download Latest Version v2.7.0 source code.zip (87.0 MB) Google Add to Preferred Sources
Home / v2.7.0
Name Modified Size InfoDownloads / Week
Parent folder
README.md 2026-09-29 29.3 kB
0
v2.7.0 source code.tar.gz 2026-09-29 85.2 MB
0
v2.7.0 source code.zip 2026-09-29 87.0 MB
0
Totals: 3 Items   172.3 MB 0

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

我们很高兴地宣布Nexent v2.7.0 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 v2.7.0 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.

新功能 / New Features

  • 推出智能体工作台（Agent Workbench）：启动后直达工作台界面，支持以自然语言创建智能体，并提供 Skill 与知识库选择、创建历史、失败重试与恢复、会话续接等完整创建流程。 / Introduced the Agent Workbench, which lands directly on the workbench after launch and lets users create agents from natural language, with Skill and knowledge-base pickers, creation history, retry and recovery, and conversation resume.
  • 内置官方智能体资产：随仓库提供政务（招标分析、公文写作、报告识别、安全报告、社媒洞察）与医疗（血检、病例生成、造影剂量、肺 CT 诊断/质控、医保审核）共 11 个预置智能体，含配套知识库与技能，并提供一键部署脚本。 / Bundled official agent assets: 11 pre-built agents across government (bid analysis, official document writing, report identification, security reporting, social media insight) and healthcare (blood test, case generation, contrast dosage, lung CT diagnosis/QC, insurance review), with companion knowledge bases, Skills, and a one-click deployment script.
  • 支持安全审计日志：在 API Key、CAS、用户组、邀请、北向接口、OAuth、租户与用户管理等关键操作上记录审计流水。 / Added security audit logging across API key, CAS, group, invitation, northbound, OAuth, tenant, and user-management operations.
  • 完善租户资源规格限额：对单租户的智能体、知识库、MCP、Skill、评测集与会话等资源设置上限，配套统一错误码、429 响应与中英文提示。 / Enforced tenant resource quotas for agents, knowledge bases, MCP services, Skills, evaluation sets, and conversations, with unified error codes, 429 responses, and localized messages.
  • 重构模型配置页：以默认模型槽位网格替代原有弹窗，槽位带优先级徽标、连通性状态与一键校验，模型库保留表格、筛选与增删改流程。 / Redesigned the model configuration page with a default-model slot grid (priority badges, per-slot connectivity hints, one-click verification) while keeping the model library table, filters, and full CRUD flows.
  • 支持模型调用日志分类与路由能力，便于按类别追踪与排查模型调用。 / Added classification and routing for model invocation logs to make model calls easier to track and troubleshoot.
  • 支持将用户上下文传递至 MCP 工具与 A2A 智能体，实现工具侧鉴权。 / Passed user context through to MCP tools and A2A agents for tool-side authorization.
  • 优化知识库管理：知识库界面全新改版，并支持 AIDP 知识文件的删除与下载。 / Revamped the knowledge-base interface and enabled deletion and download of AIDP knowledge files.
  • 新增智能体发布与使用指引，帮助用户快速上手与发布智能体。 / Added an agent publishing and usage guide to help users get started and publish agents quickly.
  • 支持模型思考挡位（reasoning effort）配置，并提供 MCP 地址与配置校验工具。 / Added model reasoning-effort configuration plus MCP URL and config validation utilities.
  • 优化资源管理体验：统一资源页与资源卡片、优化超管管理页与 Skill 列表操作入口。 / Unified the resource page and resource cards, and polished the super-admin management page and Skill list actions.
  • 优化智能体仓库与列表：支持分页列表中的仓库状态展示、卡片标签布局与名称溢出处理。 / Improved the agent repository and lists with repository status in paged views, card tag layout, and name-overflow handling.
  • 前端构建提速：升级 Next.js 以启用更快的 Turbopack 构建。 / Upgraded Next.js for faster Turbopack builds.

Bug 修复 / Bug Fixes

  • 修复租户管理员无法管理本租户模型的问题：/model/manage/* 改为「SU 任意租户、ADMIN 仅限本租户」的角色加租户作用域校验，恢复资源管理页模型列表与增删改。 / Fixed tenant admins being unable to manage their own tenant's models: /model/manage/* now uses a role-plus-tenant-scope check (SU for any tenant, ADMIN for its own), restoring the model list and CRUD on the resource page.
  • 修复模型权限隔离：模型管理仅对管理员开放，DEV 角色不再看到模型页，同时保留只读权限供智能体编辑选用模型。 / Fixed model permission isolation: model management is restricted to admins and the model page is hidden from DEV, while read access remains for agent editors.
  • 修复技能标签数据异常导致智能体页面崩溃的问题，并对标签做前后端双重规范化。 / Fixed an agents-page crash caused by malformed skill tags, with tag normalization on both frontend and backend.
  • 修复配额分页与平台配额面板：知识库配额明细分页、面板加载态与公平分摊显示修正。 / Fixed quota breakdown pagination and the platform quota panel (KB pagination, loading states, and fair-share display).
  • 修复监控异常上报，完善监控身份标识与指标归属。 / Fixed abnormal monitoring reporting and improved monitoring identity and metric attribution.
  • 修复评估相关缺陷：任务列表多智能体过滤、AIDP 知识库用例生成支持、移除用例生成的损坏文件上传，并修复评估删除对租户管理员的回退逻辑。 / Fixed evaluation issues: task-list filtering for multiple agents, AIDP knowledge-base case generation, removal of the broken file-upload path, and tenant-admin fallback for evaluation deletion.
  • 修复智能体配置更新特定场景下无限循环的问题。 / Fixed an infinite loop when agent configuration updates failed in certain scenarios.
  • 修复 Agent 输出协议的重试与流式显示（develop 分支）。 / Fixed output-protocol retry and streaming display on the develop branch.
  • 修复对话资源限制未本地化的问题，并拒绝超过 10 MB 的附件上传。 / Localized conversation resource limits and rejected attachments larger than 10 MB.
  • 修复 AIDP 相关缺陷：折叠高级选项后权限丢失、文档列表历史合并、处理顺序与创建时间。 / Fixed AIDP issues: permissions lost when collapsing advanced options, and document-list history merge, processing order, and creation time.
  • 修复知识库预览在轮询期间被刷新打断的问题。 / Fixed knowledge previews being refreshed during polling.
  • 修复工作台创建智能体后的重定向与智能体配置页展示不全问题。 / Fixed redirection after creating an agent in the Workbench and incomplete rendering on the agent configuration page.
  • 移除模型自动重命名行为，并加固配额、标签、模型与 API Key 等接口契约。 / Removed automatic model renaming and hardened quota, tag, model, and API-key contracts.
  • 修复 OpenAI 兼容 provider 发现被 SSRF 防护误拦的问题，支持私网地址模型服务发现。 / Fixed the SSRF guard blocking OpenAI-compatible provider discovery, enabling privately addressed model services.
  • 修复定时任务展示：改为显示下次执行时间并在任务结束后正确展示结果。 / Fixed scheduled-task display to show the next execution time and surface results after completion.

文档与部署变化 / Docs & Deployment Changes

  • 合并 SQL 迁移为 v2.7.0_merged_migrations.sql，统一承载自 v2.6.1 以来的迁移内容（含智能体图标字段重命名等）。 / Consolidated SQL migrations into v2.7.0_merged_migrations.sql, carrying all migrations since v2.6.1 (including the agent icon column rename).
  • 新增官方智能体部署脚本 deploy/deploy-official-agents.sh，支持一键导入预置智能体资产。 / Added deploy/deploy-official-agents.sh for one-click import of bundled official agent assets.
  • 更新部署配置与环境变量模板：K8s Helm values/configmap、Docker Compose、构建脚本，新增租户资源限额配置项。 / Updated deployment configs and env templates (K8s Helm values/configmap, Docker Compose, build scripts) with new tenant resource-limit settings.
  • 完善项目规范与测试资产：新增测试资产 skill 及配套参考文档。 / Expanded engineering standards and test assets with a new test-assets skill and its reference docs.

What's Changed

New Contributors

Full Changelog: https://github.com/ModelEngine-Group/nexent/compare/v2.5.1...v2.7.0

Source: README.md, updated 2026-09-29