github agentscope-ai/AgentTeams v1.2.5

2 hours ago

What's Changed

  • docs: announce AgentTeams v1.2.4 in README news by @shiyiyue1102 in #1302
  • fix(qwenpaw): wait for agent startup before configuring it by @shiyiyue1102 in #1304
  • fix: include plugins in controller image push context by @shiyiyue1102 in #1301
  • chore(copaw): prepare worker package 1.0.4 by @shiyiyue1102 in #1303
  • fix(installer): validate usernames before starting MinIO by @shiyiyue1102 in #1308
  • fix(manager): honor MCP API domain scheme and port by @shiyiyue1102 in #1309
  • fix(teamharness): reject completion with non-terminal tasks by @014-code in #1307
  • fix(ci): repair the teamharness shard on main by @LUOSENGWA in #1326
  • fix(controller): approval proxy dials worker running-config under /api prefix by @LUOSENGWA in #1331
  • fix(controller): skip the openclaw.json storage write when reconcile produces no change by @LUOSENGWA in #1321
  • fix(manager): make the smoke-test SOUL check provisioning-aware by @LUOSENGWA in #1339
  • fix(matrix-channel): stop labelling silent turns as「已完成」 (#1320) by @LUOSENGWA in #1328
  • fix(manager): ensure the manager's own tree exists on the shared FS at boot by @LUOSENGWA in #1332
  • fix(matrix-channel): mask raw consume errors before room send by @LUOSENGWA in #1338
  • fix(matrix-channel): honor show_thinking for reasoning output by @LUOSENGWA in #1336
  • fix(controller): mark appservice mention dedup only after a successful wake by @LUOSENGWA in #1322
  • fix(teamharness): whitelist angle-bracket placeholders in the artifact sensitive scan by @LUOSENGWA in #1323
  • fix(teamharness): re-delegation sends a fresh notification instead of a deduplicated no-op by @LUOSENGWA in #1324
  • fix(matrix-channel): accept localpart-only @mentions in group rooms by @LUOSENGWA in #1337
  • feat(controller): keep legacy CoPaw upgrades working while blocking new CoPaw creation by @LUOSENGWA in #1327
  • fix(manager): reset idle_since on worker wake and ensure-ready by @LUOSENGWA in #1352
  • fix(mcp-setup): surface the resolved protocol for --api-domain registrations by @nillikechatchat in #1351
  • feat(install): add an LLM channel smoke check to post-install verification by @LUOSENGWA in #1347
  • ci: run controller tests for matrix-channel plugin changes by @LUOSENGWA in #1345
  • docs(design): define approval payload and routing for attention events by @LUOSENGWA in #1353
  • feat(manager): decouple the Manager runtime from the CoPaw worker package by @LUOSENGWA in #1319
  • feat(qwenpaw): add a pre-flight compatibility check for runtime upgrades by @LUOSENGWA in #1343
  • refactor: remove the CoPaw worker runtime from new installs by @LUOSENGWA in #1335
  • fix(ci): reject empty manifests in image build and release gates by @nillikechatchat in #1359
  • fix(installer): persist DASHBOARD_SESSION_SECRET across upgrades by @nillikechatchat in #1358
  • chore(higress): bump pinned Higress from 2.2.1 to 2.2.5 (all-in-one image + helm subchart) by @LUOSENGWA in #1355
  • feat(controller): add runtime-aware guards to worker-scoped proxies by @LUOSENGWA in #1325
  • feat(skills): add GET /api/v1/skills/{name}/download (zip stream) by @LUOSENGWA in #1330
  • fix(teamharness): guard single-file pushes against overwriting newer remote state by @LUOSENGWA in #1333
  • test(qwenpaw): fix two permanently failing unit tests by @RerankerGuo in #1316
  • fix(model-switch): refresh model fields when switching to an existing model by @RerankerGuo in #1315
  • docs(design): document session wakeups and the self-cross-session trigger by @LUOSENGWA in #1329
  • feat: expose Worker environment editing and gateway identity probes by @shiyiyue1102 in #1306
  • docs (translation): translate en only docs to zh-cn by @LUOSENGWA in #1354
  • fix(manager): restore worker settings after restart by @014-code in #1225
  • fix(mcp): derive scheme for schemeless --api-domain ports (follow-up to #1309, close #1284) by @RerankerGuo in #1314
  • docs(design): document the optional runtime block protocol (org.agentteams.run v1) by @LUOSENGWA in #1317
  • feat(agent): add Qwen Code and OpenCode runners to the coding-cli delegation skills by @LUOSENGWA in #1342
  • feat(agent): add coding-delegation guidance to the coding-cli skills by @LUOSENGWA in #1344
  • feat(controller): manage worker coding CLIs via container exec (runtime-agnostic)` by @LUOSENGWA in #1360
  • fix(controller): guard Manager force-leave with a membership check by @LUOSENGWA in #1367
  • feat(controller): read-only coding-run transcript visibility by @LUOSENGWA in #1365
  • docs(installer): retire leftover openclaw brand strings from comments by @RerankerGuo in #1371
  • Feat(qwenpaw) worker LLM runtime envs(hot stream timeouts + startup LLM tuning) by @LUOSENGWA in #1363
  • feat(coding-cli): wire approval-mode and native wall-time into the qwen runner by @LUOSENGWA in #1379
  • feat(qwenpaw): in-place QwenPaw runtime upgrade — two-phase orchestrator with rollback + /api/upgrade backend plugin by @LUOSENGWA in #1364
  • feat(coding-runs): enrich the read contract for viewer consumption by @LUOSENGWA in #1381
  • chore(base): bump Node 22 to 24 across base images and docs by @LUOSENGWA in #1361
  • fix(mcp): support split Console setup and QwenPaw Manager mcporter by @shiyiyue1102 in #1383
  • chore(release): prepare v1.2.5 with Dashboard v1.3.1 by @shiyiyue1102 in #1384

New Contributors

Full Changelog: v1.2.4...v1.2.5


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:v1.2.5

# Manager (lightweight; spawned by the embedded controller)
docker pull higress-registry.cn-hangzhou.cr.aliyuncs.com/agentteams/agentteams-manager:v1.2.5

# Worker
docker pull higress-registry.cn-hangzhou.cr.aliyuncs.com/agentteams/agentteams-worker:v1.2.5

# QwenPaw Worker
docker pull higress-registry.cn-hangzhou.cr.aliyuncs.com/agentteams/agentteams-qwenpaw-worker:v1.2.5

# DeepSeek Harness Worker (experimental)
docker pull higress-registry.cn-hangzhou.cr.aliyuncs.com/agentteams/agentteams-deepseek-harness-worker:v0.1.0

# Controller (used standalone in k8s; bundled inside agentteams-embedded for docker installs)
docker pull higress-registry.cn-hangzhou.cr.aliyuncs.com/agentteams/agentteams-controller:v1.2.5

Quick Start

Latest stable (v1.2.5)

macOS / Linux

bash <(curl -fsSL https://raw.githubusercontent.com/agentscope-ai/AgentTeams/v1.2.5/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/v1.2.5/install/agentteams-install.ps1')

Documentation

For more details, see the installation guide.

Don't miss a new AgentTeams release

NewReleases is sending notifications on new releases.