github agentscope-ai/AgentTeams v1.2.7
AgentTeams v1.2.7
on GitHub

one hour ago

AgentTeams 1.2.7 Release Notes

AgentTeams 1.2.7 重点完善一键安装与保留数据升级，统一六种 Harness 的镜像命名和版本，并将 Tracing 扩展为跨 Agent、团队、会话和任务的通用排障能力。默认搭配 Dashboard v1.4.1。

新增与增强

一键启动与升级

  • 简化本地安装流程，支持先启动工作区，再通过 Dashboard 配置 Manager 模型。
  • 完善保留现有数据和配置的升级流程，修复镜像选择、版本检测和 Bash / PowerShell 默认值不一致的问题。
  • 默认推荐 Core v1.2.7 与独立版本 Dashboard v1.4.1；保留显式版本、自定义镜像以及不安装 Dashboard 的选择。

六种共享 Harness

  • OpenClaw、QwenPaw、Hermes、DeepSeek、OpenHuman、Codex 统一使用 agentteams-harness-xxx:v1.2.7 镜像，提供 linux/amd64 和 linux/arm64。
  • Manager 与 Worker 复用 Harness 镜像，不再发布专用 Manager 镜像。本版本 Manager 支持 OpenClaw 和 QwenPaw，其余四种为 Worker 运行时。
  • 补齐 Codex 默认镜像、OpenHuman Worker / Leader 原生工具注册，并修复管理员身份传递、Worker 凭据保留和运行时升级问题。

通用 Tracing 与诊断

  • Controller 集中管理可观测配置和查询，Dashboard 左侧工作台增加 Tracing 入口；未配置时提供配置指引。
  • agt tracing 支持按 Agent / Worker、团队、Session、Task 等维度查询，不再要求必须有正式任务记录。
  • 支持时间范围、结果数量、JSON / 表格输出，以及 trace 详情查询，便于排查模型请求、工具调用和失败步骤。

Dashboard v1.4.1

  • 团队、Manager 和未归属 Worker 统一列表管理；团队拓扑和 Agent 详情在页面内展示，统一 Provider → Model 选择流程。
  • 优化工作区会话、任务看板、人工待办、个人文件共享和首次登录引导。
  • 同一轮思考和工具调用集中展示，执行中提供有界实时预览，结束后自动折叠；调整消息操作栏、间距、字号和导航收起位置。
  • 修复中文输入法确认候选词时意外发送、会话深链、行内代码及成员聊天权限等问题；移除自动日夜主题切换。
  • 修复模型、路由、MCP、Skill 和文件管理的回归，改善登录与认证服务暂时不可用时的提示。

修复与可靠性

  • 强化任务派发、团队状态同步及项目完成条件，要求任务完成并登记交付产物后才能完成项目。
  • 修复 Matrix 同步回调阻塞、空白输出的 NO_REPLY 判断和消息去重缓存增长问题。
  • 完善团队管理员的实时权限解析、L1 Human 与网关管理授权，以及系统管理员的个人文件空间。
  • 改善文件列表缓存、压缩与回收站查询效率；Consumer 创建支持自动生成 API Key，并提供安全的列表查询。

升级说明

  • 建议使用下方固定版本安装脚本升级并保留原数据；有自定义镜像或固定版本的部署请核对覆盖项。
  • 旧的 agentteams-manager、agentteams-worker 等名称不再作为本版本默认镜像；新部署使用统一 Harness 名称。
  • Tracing 为可选能力，需要配置 Collector 和可观测后端；Dashboard v1.4.1 的通用查询与配置管理建议搭配 Core v1.2.7。
完整 PR 变更清单

What's Changed

  • fix(matrix-channel): make the NO_REPLY final-line detector total on whitespace-only output by @RerankerGuo in #1557
  • fix(controller): bound the appservice mention dedup map with a TTL by @RerankerGuo in #1559
  • feat(teamharness): require completed and artifact-registered tasks for complete_project by @LUOSENGWA in #1512
  • ci: reduce redundant release builds and model regressions by @shiyiyue1102 in #1560
  • fix(install): refresh token plan glm model by @zcxGGmu in #1085
  • fix(worker): preserve live Matrix token during config pulls by @shiyiyue1102 in #1561
  • fix(manager): use runtime-specific model-switch instructions by @shiyiyue1102 in #1562
  • ci: trim redundant setup and scope PR regression coverage by @shiyiyue1102 in #1563
  • fix(matrix): prevent inbound callbacks from stalling sync by @014-code in #1222
  • fix(teams): resolve Team Admin access live and confirm administrator by @shiyiyue1102 in #1568
  • fix(runtime): keep runtime provider id independent of gateway provider by @LUOSENGWA in #1566
  • docs(runtime): warn against using QwenPaw built-in provider ids by @LUOSENGWA in #1565
  • fix(humans): provide a personal Human workspace for the system admin by @shiyiyue1102 in #1574
  • feat(auth): authorize L1 Humans and gateway management in Controller by @shiyiyue1102 in #1573
  • fix(worker): allow clearing model provider through update API by @shiyiyue1102 in #1571
  • refactor(hermes): move runtime to group harness and upgrade to v0.23.0 by @shiyiyue1102 in #1575
  • feat(files): ETag/304 and gzip for human file listings by @LUOSENGWA in #1578
  • perf(files): index trash listings to avoid per-entry meta reads by @LUOSENGWA in #1577
  • feat: register independently packaged Codex member runtime by @shiyiyue1102 in #1576
  • refactor: migrate DeepSeek harness to agentteams-group by @shiyiyue1102 in #1582
  • refactor: migrate QwenPaw harness to group repository by @shiyiyue1102 in #1586
  • feat(controller): support Manager model setup after installation by @shiyiyue1102 in #1584
  • build: consume OpenClaw Worker from group harness submodule by @shiyiyue1102 in #1581
  • refactor: run Manager on shared Harness runtimes by @shiyiyue1102 in #1580
  • fix: provide a default Codex Worker image by @shiyiyue1102 in #1588
  • feat(install): simplify local startup and data-preserving upgrades by @shiyiyue1102 in #1585
  • docs: align homepage installation with Dashboard quickstart by @shiyiyue1102 in #1589
  • fix(release): resolve 1.2.7 Harness and upgrade regressions by @shiyiyue1102 in #1595
  • feat(tracing): centralize configuration and task diagnosis in controller by @shiyiyue1102 in #1597
  • fix(release): use shared Harness images for Managers by @shiyiyue1102 in #1598
  • fix(tests): resolve the canonical Manager collaboration room by @shiyiyue1102 in #1600
  • fix(worker): propagate the configured administrator identity by @shiyiyue1102 in #1601
  • fix(openhuman): register native tools for Worker and Leader roles by @shiyiyue1102 in #1602
  • feat(tracing): add general Agent team session and task queries by @shiyiyue1102 in #1603
  • fix(ci): use runner networking for embedded dependency installation by @shiyiyue1102 in #1606
  • fix: generate optional Consumer API keys and support safe listing by @shiyiyue1102 in #1604
  • ci: publish six shared Harness images with unified versions by @shiyiyue1102 in #1607
  • fix: prevent OpenClaw policy sync during startup admission by @shiyiyue1102 in #1609
  • fix: harden task delegation and team state synchronization by @shiyiyue1102 in #1605
  • fix(release): verify Docker manifest source labels on both architectures by @shiyiyue1102 in #1610
  • release: unify six Harness image versions at v1.2.7 by @shiyiyue1102 in #1608
  • chore(release): pin companion Dashboard v1.4.1 by @shiyiyue1102 in #1593

New Contributors

Full Changelog: v1.2.6...v1.2.7

Docker Images

Multi-architecture images (amd64 + arm64): 

# Embedded all-in-one (infra + controller; pulled by the installer)
docker pull higress-registry.cn-hangzhou.cr.aliyuncs.com/agentteams/agentteams-embedded:v1.2.7

# OpenClaw Harness (Manager, Leader and Worker)
docker pull higress-registry.cn-hangzhou.cr.aliyuncs.com/agentteams/agentteams-harness-openclaw:v1.2.7

# QwenPaw Harness (Manager, Leader and Worker)
docker pull higress-registry.cn-hangzhou.cr.aliyuncs.com/agentteams/agentteams-harness-qwenpaw:v1.2.7

# DeepSeek Harness Worker (experimental)
docker pull higress-registry.cn-hangzhou.cr.aliyuncs.com/agentteams/agentteams-harness-deepseek:v1.2.7

# Controller (used standalone in k8s; bundled inside agentteams-embedded for docker installs)
docker pull higress-registry.cn-hangzhou.cr.aliyuncs.com/agentteams/agentteams-controller:v1.2.7

Quick Start

Latest stable (v1.2.7)

macOS / Linux

bash <(curl -fsSL https://raw.githubusercontent.com/agentscope-ai/AgentTeams/v1.2.7/install/agentteams-install.sh)

Windows (PowerShell 7+)

Set-ExecutionPolicy Bypass -Scope Process -Force; $wc=New-Object Net.WebClient; $wc.Encoding=[Text.Encoding]::UTF8; iex $wc.DownloadString('https://raw.githubusercontent.com/agentscope-ai/AgentTeams/v1.2.7/install/agentteams-install.ps1')

Documentation

For more details, see the installation guide.

Check out latest releases or
releases around agentscope-ai/AgentTeams v1.2.7

Don't miss a new AgentTeams release

NewReleases is sending notifications on new releases.

Get notifications