Skip to content

v1.2.6

Latest
Latest

Choose a tag to compare

View all tags
@github-actions github-actions released this 08 Nov 11:48
· 17 commits to main since this release
v1.2.6
9301e53
This commit was created on GitHub.com and signed with GitHub’s verified signature.
GPG key ID: B5690EEEBB952194
Verified
Learn about vigilant mode.

AgentTeams 1.2.6 Release Notes

AgentTeams 1.2.6 重点增强多 Agent 协作的过程可见性、任务控制和多用户权限管理，默认搭配 Dashboard v1.4.0。新版工作台加入语音输入与执行追踪，并完善运行时配置、文件存储和升级流程；实验性 DeepSeek Harness Worker 镜像更新至 v0.1.1。

新增与增强

Dashboard v1.4.0：更集中的工作台与语音输入

  • 将常用功能归为「助手、工作台、资产中心、成员与团队」四个主菜单，改善多团队会话切换与任务查看体验。
  • 聊天新增语音转文字，默认使用浏览器本地免费识别；识别结果先进入输入框，可编辑后手动发送。管理员也可配置百炼语音识别。
  • 支持中英文语音输入，改进中文转写结果处理；本地模型首次使用需要加载，之后会缓存。
  • 优化消息时间与正文的排版、已读状态持久化，并新增中性主题及可选的日夜切换。

原生会话追踪与执行诊断

  • 新增可选的 Worker 原生会话追踪，支持关联 QwenPaw、Hermes 的 turn、模型请求和工具调用。
  • Dashboard 增加按用户权限过滤的会话列表与执行瀑布图，便于查看耗时分布、调用细节及失败步骤。
  • 补充 Tracing 后端配置与 Manager 配置指引，并修正追踪记录的团队身份归属。
  • Matrix 增加静默丢弃消息的工作区记录，改善「消息已发出但 Agent 没有响应」时的排查线索。

任务控制、人工介入与协作记录

  • TeamHarness 支持暂停、恢复任务派发与 Loop 推进，并记录 Loop 预算和控制操作审计。
  • 新增团队级人工事项收件箱，提供项目 attention 查询、结构化人工答复，以及逐接收人的通知结果和投递缺口。
  • 为任务状态冲突提供结构化错误；项目、任务元数据支持版本条件读取和更新，降低并发操作覆盖较新状态的风险。
  • 完善交付产物的 @mention 摘要通知、报告长度限制和评审上下文；创建项目时可通过 --team 邀请整个团队，并查看实际入群结果。

多用户权限与模型授权

  • 明确资源管理边界：Manager 和平台管理员负责资源管理；Team Leader 保留本团队任务协调、状态读取和自身就绪上报，不能直接管理其他 Worker 的配置与生命周期。
  • 新增按用户、角色配置的模型授权及共享授权表，对 Worker 主模型、子 Agent 模型和 Coding CLI 模型的配置变更执行服务端校验。
  • 新增工作区访问授权与限制，完善团队人类成员变更后的权限同步及已有团队房间的管理员授权。
  • 人类用户列表不再返回初始密码，改进 Worker 凭据原子写入和人类用户并发更新时的凭据保留。

运行时配置、模块更新与 DSH 改进

  • 更新 QwenPaw、Hermes 和 OpenClaw 运行时；补充 OpenClaw、Hermes 的配置读取、更新与回滚能力。
  • 新增 Worker、Manager 重启接口，并在支持的运行时路径中热应用主模型变更。
  • 补充 Matrix、TeamHarness 的运行时适配包，以及团队、Worker、Manager 的模块更新接口；插件发布包携带版本清单和校验和。
  • 改善 Coding CLI 安装的持久化、Qwen Code 参数兼容和 QwenPaw × Qwen Code 的委派、唤醒与审批映射。
  • 实验性 DeepSeek Harness Worker 镜像更新至 v0.1.1，修复工具调用 ID 的流式记录与会话重放、重复文件投递、模型切换及 Matrix 历史消息补取问题。

用户文件区与存储同步

  • Controller 新增用户独立文件区、回收站、配额和文件同步授权，提供文件清单、预签名上传下载及增量同步接口。
  • 文件读取支持 ETag、304、HEAD、Range 与 If-Range，便于客户端执行缓存校验和断点续传。
  • 新增外部 S3 兼容存储后端；改进对象访问、目录修改时间保留和同步状态查询，并明确返回存储不可用错误。
  • 文件同步新增阻塞式 flush 屏障与逐文件回执；上传清单保护尚未推送的本地写入，减少并发同步中的数据覆盖。

修复与稳定性

  • Matrix 消息协作：修复临时发送错误无法正常重试、同步超时配置过小导致频繁轮询、无关事件误触发重试，以及 Agent 之间的拒绝访问回复循环；改进定时任务最终回复投递和睡眠中 Leader 的唤醒。
  • 权限与安全：收紧 OpenClaw elevated exec 的默认 Matrix 来源范围，迁移历史通配符配置并保留自定义授权；修复 Dashboard Worker token 保护及多项权限边界问题。
  • 任务与产物：修复 Markdown 包裹的占位模板被误判为敏感凭据的问题，完善非法状态转换、完成报告和通知结果记录。
  • 模型与运行时：修复自定义 LLM IP 地址在不同初始化路径下的接入，改进运行时版本一致性检查和 DSH 依赖兼容性。
  • 部署与升级：升级流程增加预检、备份和分层执行；修复端口解析、WASM 检查、Dashboard 新配置卷写权限，以及外部 Higress 接入和 Dashboard 配置持久化问题。
  • 构建与依赖：完善 Node 24 LTS 版本约束、Controller 依赖更新和插件打包；发布构建改用原生架构 runner，解决 ARM64 模拟编译挂起及 Hermes 发布缺失构建上下文的问题。

升级与兼容说明

  • AgentTeams 核心镜像使用 v1.2.6，Dashboard 使用独立版本 v1.4.0，实验性 DeepSeek Harness Worker 镜像使用 v0.1.1。DSH Worker 镜像版本与其内置上游软件包版本分别管理。
  • 升级前备份配置和持久化数据，并按部署模式更新相应镜像。安装脚本已对齐以上组件版本。
  • Team Leader 的资源管理权限已收紧；依赖 Leader 直接创建、修改或唤醒其他 Worker 的自定义技能，应调整为交由 Manager 或平台管理员处理。
  • 人类用户修改 Worker 或 Coding CLI 的模型配置受有效模型授权约束；现有模型保持运行与授权变更是不同操作，升级后应检查所需授权。
  • 原生会话追踪需要配置后端并显式启用；QwenPaw、Hermes 的通用追踪覆盖不等同于 OpenClaw、DeepSeek Harness 或 OpenHuman。不同运行时的配置、热应用和插件能力以对应支持范围为准。
  • 本地语音识别默认免费，首次使用需要加载模型；百炼语音识别、Tracing 后端和外部存储需单独配置。
  • 更广的历史场景仍独立跟踪：#1284（部分 MCP Console 更新与服务注册行为）、#575（旧版 CoPaw 配置恢复）。

完整变更范围：Core v1.2.5…v1.2.6；配套界面版本：Dashboard v1.4.0。

English Release Notes

AgentTeams 1.2.6 improves visibility into multi-agent execution, task control, and multi-user permissions. It pairs with Dashboard v1.4.0, bringing voice input and execution tracing alongside stronger runtime configuration, file storage, and upgrade support. The experimental DeepSeek Harness Worker image moves to v0.1.1.

New features and improvements

Dashboard v1.4.0: a focused workspace and voice input

  • Organizes navigation into Assistant, Workbench, Asset Center, and Members & Teams, with improved multi-team conversation switching and task access.
  • Adds speech-to-text input with free, browser-local recognition by default. Transcripts remain editable in the composer and require a manual send. Administrators can optionally configure Alibaba Cloud Bailian recognition.
  • Supports Chinese and English input and improves Chinese transcript normalization. Local model assets load on first use and are cached afterward.
  • Improves message timestamps, persisted read markers, neutral themes, and optional day/night switching.

Native session tracing and execution diagnostics

  • Adds optional native Worker session tracing for QwenPaw and Hermes, correlating turns, model requests, and tool calls.
  • Adds permission-scoped session lists and waterfall views in Dashboard to inspect timing, calls, and failures.
  • Improves tracing backend setup guidance and corrects team identity attribution.
  • Records silently dropped inbound Matrix messages in a workspace ledger to improve diagnosis when an agent does not respond.

Task control, human intervention, and collaboration records

  • Adds pause/resume controls for task dispatch and Loop progression, with Loop budget records and control-action audits.
  • Adds a team attention inbox, project attention queries, structured human answers, and per-recipient notification outcomes and delivery gaps.
  • Returns structured task-transition errors and supports version-conditional project/task metadata reads and writes to reduce concurrent state overwrites.
  • Improves artifact summary mentions, report limits, and review context. Project creation can invite the complete roster through --team and report actual join outcomes.

Multi-user permissions and model grants

  • Reserves resource management for the Manager and platform administrators. Team Leaders retain scoped coordination, reads, and self-ready reporting without configuration or lifecycle authority over other Workers.
  • Adds per-human, per-role model grants and shared grant tables, with server-side checks on Worker primary/subagent model changes and Coding CLI model settings.
  • Adds workspace access grants and restrictions, improves human-membership synchronization, and grants administrator privileges in existing team rooms.
  • Removes initial passwords from human lists and improves atomic Worker credential writes and credential preservation during concurrent human updates.

Runtime configuration, module updates, and DSH improvements

  • Updates the QwenPaw, Hermes, and OpenClaw runtimes and adds native configuration read, update, and rollback support for OpenClaw and Hermes.
  • Adds Worker/Manager restart endpoints and hot application of primary-model changes on supported runtime paths.
  • Adds runtime adapter bundles for Matrix and TeamHarness and module-update endpoints for teams, Workers, and Managers. Plugin packages include manifests and checksums.
  • Improves Coding CLI installation persistence, Qwen Code argument compatibility, and QwenPaw × Qwen Code delegation, wakeup, and approval mapping.
  • Updates the experimental DeepSeek Harness Worker image to v0.1.1, fixing streamed tool-call IDs and session replay, duplicate file delivery, model changes, and limited Matrix timeline backfill.

Human file areas and storage synchronization

  • Adds per-human file areas, trash, quotas, file-sync grants, listing, presigned transfers, and incremental synchronization endpoints.
  • Supports ETag, 304, HEAD, Range, and If-Range on file reads for cache validation and resumable downloads.
  • Adds an external S3-compatible storage backend and improves object access, modification-time preservation, sync-status queries, and storage-unavailable errors.
  • Adds a blocking flush barrier and per-file receipts, with manifest protection for unpushed local writes.

Fixes and stability

  • Matrix collaboration: fixes retry classification, excessively short sync timeouts, unrelated-event retriggers, and access-denied reply loops. Improves final cron replies and sleeping-Leader wakeups.
  • Permissions and security: restricts default OpenClaw elevated-exec Matrix senders, migrates historical wildcards while preserving custom lists, and fixes Dashboard Worker-token protection and permission boundaries.
  • Tasks and artifacts: fixes false positives for Markdown-wrapped credential placeholders and improves transition, completion-report, and notification records.
  • Models and runtimes: repairs custom LLM IP endpoints across bootstrap paths and improves runtime-version checks and DSH dependency compatibility.
  • Deployment and upgrades: adds layered preflight/backup handling and fixes port parsing, WASM checks, fresh Dashboard volume permissions, external Higress integration, and Dashboard configuration persistence.
  • Builds and dependencies: improves Node 24 LTS constraints, Controller dependencies, and plugin packaging. Native architecture runners avoid ARM64 emulation hangs, and Hermes publication receives its required build context.

Upgrade and compatibility notes

  • Core images use v1.2.6, Dashboard uses v1.4.0, and the experimental DeepSeek Harness Worker image uses v0.1.1. The DSH Worker image and its bundled upstream package have separate versioning.
  • Back up configuration and persistent data before upgrading and update the images for your deployment mode. Installer defaults are aligned with these component versions.
  • Custom Leader skills that directly create, modify, or wake other Workers must delegate resource management to the Manager or a platform administrator.
  • Human-initiated Worker and Coding CLI model changes are checked against effective model grants. Continuing to run a current model and changing model authorization are separate operations; review grants after upgrading.
  • Native tracing requires backend configuration and explicit enablement. QwenPaw/Hermes tracing coverage does not imply equivalent coverage for OpenClaw, DeepSeek Harness, or OpenHuman. Configuration, hot application, and plugin capabilities vary by runtime.
  • Local voice recognition is free by default and loads model assets on first use. Cloud speech recognition, tracing backends, and external storage require separate configuration.
  • Broader historical cases remain tracked independently: #1284 covers additional MCP Console update and registration behavior; #575 covers legacy CoPaw configuration recovery.

Full change range: Core v1.2.5…v1.2.6. Companion UI release: Dashboard v1.4.0.

完整 PR 记录与贡献者 / Pull requests and contributors

What's Changed

  • fix(matrix): send-gate retrigger carries the placeholder root flag by @LUOSENGWA in #1393
  • feat(teamharness): map loop_id across project state and task meta by @LUOSENGWA in #1401
  • feat(events): signal emission — shared contract, Go five endpoints, Python task/loop emitters, list_events by @LUOSENGWA in #1405
  • feat(server): expose projection freshness and strong-consistency reads by @LUOSENGWA in #1407
  • fix(matrix): send-gate retrigger only fires on relevant room events by @LUOSENGWA in #1408
  • feat(teamharness): structured illegal-transition errors + SIG-MISMATCH emission by @LUOSENGWA in #1404
  • feat(storage): external S3-compatible storage backend (provider=s3) by @LUOSENGWA in #1387
  • deps(controller): bump x/net and grpc for reachable advisories by @Portll in #1234
  • docs(readme): add v1.2.5 news, sync ja-JP news, drop broken zh-CN image by @max-wc in #1399
  • fix(manager-bridge): activate runtime ACL flags for the allowlist policy (#1388) by @LUOSENGWA in #1419
  • fix(project-management): surface worker auto-join outcome (#1390) by @LUOSENGWA in #1420
  • fix(openclaw): restrict elevated.allowFrom to trusted Matrix IDs by @nillikechatchat in #1425
  • fix(upgrade): block upgrade when explicit --check-deps dry-run is unrunnable by @LUOSENGWA in #1398
  • docs: bump pinned AGENTTEAMS_VERSION in all three READMEs to v1.2.3 by @RerankerGuo in #1370
  • fix(matrix-channel): look up retryable errcodes via the live class name by @RerankerGuo in #1368
  • fix(teamharness): whitelist markdown-wrapped placeholder templates in the artifact scan by @RerankerGuo in #1372
  • chore(migrate): drop dead CMD_SOURCES map from skill analyze.sh by @RerankerGuo in #1374
  • test(teamharness): fill the mcp_server fixture stubs needed by the full lifecycle by @RerankerGuo in #1373
  • fix(matrix-channel): clamp sync_timeout_ms to a sane minimum by @RerankerGuo in #1369
  • refactor(teamharness): split mcp/server.py into modules (no behaviour change) by @LUOSENGWA in #1452
  • docs(coding-cli): taskflow visibility by @LUOSENGWA in #1410
  • feat(controller): worker-level read-only MCP proxy (GET /api/v1/workers/{name}/mcp) by @LUOSENGWA in #1453
  • docs(usage): single- and multi-user deployments share one deployment model by @LUOSENGWA in #1416
  • feat(controller): add a worker-level read-only plugin registry proxy by @LUOSENGWA in #1454
  • feat(plugins): release any manifest-carrying plugin from one workflow by @LUOSENGWA in #1455
  • feat(teamharness): record structured answer payload on request_attention by @LUOSENGWA in #1449
  • fix(teamharness): publish artifacts with an @ summary event (m.mentions) by @LUOSENGWA in #1435
  • fix(matrix): propagate display flags from CRD to worker config by @LUOSENGWA in #1450
  • feat: dual ledger — loop budget (Python) + control-action audit with kind filter by @LUOSENGWA in #1406
  • feat(teamharness): attention record fields + notification outcome log (S0+S4a) (#1394) by @LUOSENGWA in #1418
  • feat(plugins): add runtime adapter bundle skeletons by @LUOSENGWA in #1427
  • fix(coding-cli): persist tab installs in the runtime's persisted home (QwenPaw working dir; OpenClaw/Hermes home) by @LUOSENGWA in #1409
  • refactor(install): extract a safe, layered upgrade orchestrator (P1: preflight/backup/controller) by @LUOSENGWA in #1430
  • fix(sync): push manifest protects unpushed writes (#1391) by @LUOSENGWA in #1417
  • fix(coding-cli): adapt wrapper to qwen-code 0.25.3 --json-file semantics drift by @LUOSENGWA in #1397
  • fix(controller): wake a sleeping team leader on mention by @RerankerGuo in #1462
  • feat(install): support external Higress adapter and dashboard persistence by @nillikechatchat in #1396
  • feat(plugins): add OpenClaw and Hermes adapters to the matrix bundle by @LUOSENGWA in #1458
  • build(node): use latest Node 24 LTS (floor 24.22.1) by @LUOSENGWA in #1422
  • build(openclaw): pin upstream v2026.10.6 || build(openclaw): pin upstream v2026.10.6 by @LUOSENGWA in #1460
  • fix(matrix): finish final-mode cron reply delivery by @shiyiyue1102 in #1466
  • test(qwenpaw): fix stale source tests and wire CI by @shiyiyue1102 in #1484
  • fix(controller): sync Human grants to Worker allowlists by @benjamin-zhao-075 in #1456
  • feat(server): read endpoint for signal events (#1405 follow-up) by @LUOSENGWA in #1433
  • feat(plugins): implement OpenClaw and Hermes adapters for the teamharness bundle by @LUOSENGWA in #1459
  • feat(controller): read-only project attention endpoint by @LUOSENGWA in #1434
  • fix(build): derive the OpenClaw base tag from the pin, and assert the gateway version in the probe and at manager startup by @LUOSENGWA in #1479
  • test(legacy-copaw): gate seeding on readiness and keep failure evidence by @LUOSENGWA in #1476
  • feat(teamharness): pause and resume task dispatch and loop progression by @LUOSENGWA in #1402
  • docs(zh-cn): add Chinese translation of worker-environment-and-gateway-probes by @RerankerGuo in #1378
  • fix(controller): persist team human membership updates by @shiyiyue1102 in #1487
  • fix(install): separate WASM plugin and runtime image registries by @shiyiyue1102 in #1488
  • fix(gateway): repair custom LLM IP endpoints across bootstrap paths by @shiyiyue1102 in #1489
  • feat(matrix): runtime-neutral core and Python adapters for the AgentTeams Matrix channel (#1468) by @LUOSENGWA in #1469
  • feat(human): add per-human file area, trash, and quota by @LUOSENGWA in #1426
  • fix(matrix): prevent access-denied reply loops between agents by @shiyiyue1102 in #1491
  • fix(controller): grant admin power in existing team rooms by @shiyiyue1102 in #1490
  • feat(teamharness): report per-recipient delivery gaps on attention notifications by @LUOSENGWA in #1495
  • fix(ci): make the human-files job wait for RustFS, and stop publishing empty image tars by @LUOSENGWA in #1504
  • feat(project-management): create-project.sh --team invites the whole Team by @dchaudhari7177 in #1429
  • feat(teamharness): protocol guards for decision classes, attention stalls and review gates (#1451) by @LUOSENGWA in #1467
  • fix(controller): atomically persist worker credential files by @shiyiyue1102 in #1517
  • feat(runtimeconfig): per-runtime config surface types and registry by @LUOSENGWA in #1492
  • feat(matrix): record silent inbound drops in a per-workspace ledger by @LUOSENGWA in #1497
  • fix(controller): register the worker plugin registry route by @LUOSENGWA in #1525
  • feat(runtimeconfig): openclaw config surface, member config routes, reconcile & rollback (P8a) by @LUOSENGWA in #1493
  • feat(hermes): upgrade the Worker runtime to hermes-agent v2026.9.24 by @LUOSENGWA in #1472
  • feat(workers): restart endpoints for workers and managers, and hot-apply the primary model by @LUOSENGWA in #1473
  • fix(controller): worker list — single indexed team lookup + opt-in fields projection by @LUOSENGWA in #1477
  • feat(controller): ETag/304 for stable spec projections (#1475) by @LUOSENGWA in #1480
  • test(plugins): bundle L3/L4 lifecycle conformance (P-b / P-c) by @LUOSENGWA in #1530
  • feat(controller): team-level attention inbox with lazy overdue events by @LUOSENGWA in #1503
  • feat(human): file_sync capability + manifest endpoints (#1414 P3) by @LUOSENGWA in #1499
  • chore(teamharness): harden loader smoke, contract guards and agent guardrails by @LUOSENGWA in #1523
  • fix(dsh): backfill limited Matrix timelines by @bingdongni in #1528
  • feat(controller): per-human model grants — Human.spec.modelAccess + ModelAccessTable by @LUOSENGWA in #1513
  • feat(controller): presigned download URLs for human file area (?presign=1) by @LUOSENGWA in #1501
  • feat(qwenpaw): upgrade the Worker and QwenPaw Manager runtimes to 2.2.2.post1 by @LUOSENGWA in #1471
  • feat(sync): blocking flush barrier with per-file receipts (#1391) by @LUOSENGWA in #1500
  • Add OpenClaw channel plugin bridging AgentTeams Matrix via the sidecar by @LUOSENGWA in #1470
  • feat(human-files): presign uploads with server-side commit accounting by @LUOSENGWA in #1519
  • feat(human-files): list the file area with size/etag and a recursive key cursor by @LUOSENGWA in #1518
  • feat(runtimeconfig): hermes config surface, delegation projection, adapter & restart window (P8b) by @LUOSENGWA in #1494
  • feat(teamharness): commit task metadata through the controller's conditional endpoint by @LUOSENGWA in #1498
  • feat(controller): add module update endpoints for index, teams, workers and managers by @LUOSENGWA in #1526
  • feat(matrix): add the bundle manifest and QwenPaw adapter package by @LUOSENGWA in #1524
  • feat(human): workspace access grants and restrictions (P2) by @LUOSENGWA in #1502
  • feat(plugin): QwenPaw × QwenCode bridge — dispatch, wake, worker-proxy approval mapping by @LUOSENGWA in #1411
  • ci: verify downloaded image archives and retry once by @LUOSENGWA in #1534
  • build(plugins): publish QwenPaw adapter zips with checksums and manifests by @LUOSENGWA in #1527
  • fix(controller): reuse the shared team index in the MCP catalog by @LUOSENGWA in #1511
  • feat(teamharness): accountable completion, room invites, report caps and review context by @LUOSENGWA in #1521
  • deps(controller): move to Go 1.26 and bump modules for reachable advisories by @Portll in #1529
  • feat(controller): team and human list summaries + keep initialPassword out of lists (#1475) by @LUOSENGWA in #1481
  • feat(matrix): steer capability, optimistic config writes, isolation bundle and setup playbook by @LUOSENGWA in #1522
  • fix(auth): reserve resource control for Manager and Admin by @shiyiyue1102 in #1536
  • feat(observability): add optional native Worker session tracing by @shiyiyue1102 in #1532
  • fix(oss): list directory entries through the SDK so modification times survive by @LUOSENGWA in #1539
  • feat(files): add change-log based ?since= delta endpoint by @LUOSENGWA in #1541
  • feat(meta): add GET project/task meta with ETag (read half of the CAS pair) by @LUOSENGWA in #1540
  • perf(oss): serve the object hot path from the S3 SDK by @LUOSENGWA in #1537
  • fix(oss): map store-availability faults to 503 storage_unavailable by @LUOSENGWA in #1538
  • feat(controller): enforce per-human model grants on worker and coding-CLI writes by @LUOSENGWA in #1514
  • test: repair 1.2.6 shell regression fixtures by @shiyiyue1102 in #1542
  • fix(cli): make default Human creation use a valid permission level by @shiyiyue1102 in #1543
  • fix(manager): guide users through Dashboard tracing setup by @shiyiyue1102 in #1546
  • fix(tracing): derive team identity from projected runtime by @shiyiyue1102 in #1547
  • fix(deepseek-harness): pin compatible Cordis service dependencies by @shiyiyue1102 in #1548
  • fix(deepseek-harness): preserve streamed tool IDs for session replay by @shiyiyue1102 in #1549
  • fix(deepseek-harness): avoid duplicate outbox file delivery by @shiyiyue1102 in #1550
  • fix(deepseek-harness): apply projected model changes per turn by @shiyiyue1102 in #1551
  • feat(files): add 304/Range/HEAD/If-Range to human file GET and manifest GET by @LUOSENGWA in #1544
  • fix(humans): preserve initial credentials during concurrent updates by @shiyiyue1102 in #1554
  • perf(sync): answer sync/status from manifest user metadata by @LUOSENGWA in #1545
  • fix(install): harden controller upgrade port and wasm checks by @LUOSENGWA in #1553
  • docs(modules): add the module update & distribution guide and refresh plugin docs by @LUOSENGWA in #1555
  • fix(build): provide Hermes trace context for image publishing by @shiyiyue1102 in #1556
  • fix(build): avoid emulated OpenClaw base compilation hangs by @shiyiyue1102 in #1558
  • chore: prepare v1.2.6 component release pins by @shiyiyue1102 in #1552

New Contributors

Full Changelog: v1.2.5...v1.2.6

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:v1.2.6

# Manager (lightweight; spawned by the embedded controller)
docker pull higress-registry.cn-hangzhou.cr.aliyuncs.com/agentteams/agentteams-manager:v1.2.6

# Worker
docker pull higress-registry.cn-hangzhou.cr.aliyuncs.com/agentteams/agentteams-worker:v1.2.6

# QwenPaw Worker
docker pull higress-registry.cn-hangzhou.cr.aliyuncs.com/agentteams/agentteams-qwenpaw-worker:v1.2.6

# DeepSeek Harness Worker (experimental)
docker pull higress-registry.cn-hangzhou.cr.aliyuncs.com/agentteams/agentteams-deepseek-harness-worker:v0.1.1

# Controller (used standalone in k8s; bundled inside agentteams-embedded for docker installs)
docker pull higress-registry.cn-hangzhou.cr.aliyuncs.com/agentteams/agentteams-controller:v1.2.6

Quick Start

Latest stable (v1.2.6)

macOS / Linux

bash <(curl -fsSL https://raw.githubusercontent.com/agentscope-ai/AgentTeams/v1.2.6/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/v1.2.6/install/agentteams-install.ps1')

Documentation

For more details, see the installation guide.

Contributors

Portll, shiyiyue1102, and 7 other contributors
Assets 4
Loading