github agentscope-ai/QwenPaw v2.0.0-beta.1

pre-release5 hours ago

⚠️ This is an early beta release for QwenPaw 2.0.0 (v2.0.0-beta.1), which is currently undergoing active development — it may contain breaking changes and instability. Intended for developers and early adopters only; not recommended for production use.

What's Changed

  • refactor: migrate agentscope from version 1.x to 2.0.0 by @qbc2016 in #4846
  • Add uninstall hooks and expose skill provider API by @wangfei010313 in #5008
  • fix: propagate ModelInfo.max_input_length to AgentScope context_size for auto-compaction by @qbc2016 in #5018
  • feat(tui): bundle the terminal chat UI into qwenpaw (Phase 1) by @ekzhu in #5032
  • feat(runtime): Runtime 2.0 modular architecture with enhanced tool-call coordination by @XiuShenAl in #5078
  • feat(driver): introduce Agent OS Driver — unified abstraction for external capabilities (MCP/A2A/ACP/...) by @xiaoming-qxm in #5067
  • feat: initial governance & sandbox interface disscussion by @weidankong in #5088
  • bump: version to 2.0.0a1 by @rayrayraykk in #5281
  • feat(context): migrate context management from LightContextManager to AgentScope 2.0 native compression by @qbc2016 in #5309
  • feat(memory): migrate QwenPaw memory runtime to ReMe4 by @jinliyl in #5349
  • Merge/2.0 into main by @rayrayraykk in #5412
  • docs(roadmap): update roadmap for QwenPaw 2.0.0 by @cuiyuebing in #5461
  • fix: macos sandbox missing close bracket by @weidankong in #5454
  • refactor(memory): restructure auto-memory lifecycle, enhance /compact and add /system_prompt command by @jinliyl in #5450
  • revert(upload_file): replace manual file size check with utility func… by @zhijianma in #5467
  • perf: parallelize shutdown to reduce close time by @rayrayraykk in #5469
  • feat(coding): expand chat panel to 50% and shrink editor area by @zhaozhuang521 in #5473
  • Revert/relative path file preiview by @zhijianma in #5478
  • feat(ci): end-to-end UI verification for desktop releases by @yutai78786 in #5428
  • fix(cron): restore cron session compatibility and sanitize malformed tool_call history by @qbc2016 in #5475
  • fix(cron): allow editing and deleting enabled cron jobs by @lalaliat in #5483
  • fix(dashscope): honour extra_body generate_kwargs and avoid sending default enable_thinking by @qbc2016 in #5491
  • fix(token_usage): restore chat token/context usage ring & popover under agentscope 2.0 by @yuanxs21 in #5493
  • feat(dingtalk): support custom API endpoint for private deployments by @hongxicheng in #5111
  • feat(discord): download attachments locally and add media_dir config by @hongxicheng in #5468
  • fix(channel): restore streaming path and split multi-segment reply into separate streaming boxes by @hongxicheng in #5487
  • feat(memory): Enhance memory search functionality and simplify metadata display by @jinliyl in #5482
  • Feat/channel page dual layout by @zhaozhuang521 in #5504
  • style(chat): remove max-width constraints in wide mode styles by @zhijianma in #5502
  • feat(tools): redesign Tools page with enabled/available split layout by @zhaozhuang521 in #5509
  • fix(WeCom): WeCom QR code fetch fails due to greedy regex in settings parsing by @hongxicheng in #5460
  • fix(cron): cron session visibility, memory isolation and hot-reload stability by @lalaliat in #5494
  • test(integration): stabilise flaky integration tests + remove dead /api/agent code by @yutai78786 in #5507
  • fix(runtime): align envelope event translation with v1 streaming protocol by @XiuShenAl in #5495
  • docs(locales): update "tools" translations for consistency across mul… by @zhijianma in #5516
  • feat(console): mobile adaptation for sidebar, header, and page header by @yaozy2020 in #5444
  • fix(chat): prevent ModelSelector dropdown overflow and marquee long model names on mobile by @yaozy2020 in #5445
  • fix(chat): mobile header actions, session dropdown, and unified right drawers by @yaozy2020 in #5446
  • feat(console): mobile adaptation for Debug page by @yaozy2020 in #5449
  • feat(console): mobile adaptation for Security page by @yaozy2020 in #5451
  • feat(console): mobile adaptation for Agent Skills page by @yaozy2020 in #5458
  • feat(console): mobile adaptation for Skill Market page by @yaozy2020 in #5459
  • feat(console): make TokenUsage tables horizontally scrollable on mobile by @yaozy2020 in #5463
  • feat(console): mobile adaptation for Skill Pool page by @yaozy2020 in #5464
  • feat(console): mobile adaptation for Voice Transcription settings page by @yaozy2020 in #5470
  • fix(pack): repair desktop packaging builds by @jinglinpeng in #5518
  • feat(skills): split Skills page into enabled/disabled sections with d… by @zhaozhuang521 in #5521
  • fix: override format_tools for gemini by @qbc2016 in #5517
  • fix: add DashScopeChatModel to ChatModelName literal type by @qbc2016 in #5513
  • fix(console): improve MCP access policy layout by @xiaoming-qxm in #5213
  • fix: inline $ref/$defs in tool schemas for GLM model compatibility by @qbc2016 in #5496
  • feat: add OpenAI Response API provider by @qbc2016 in #5519
  • fix: fix tool input json decode by @qbc2016 in #5486
  • fix: refine mcp tool name to pase openai api by @qbc2016 in #5485
  • fix(matrix): use nio client.download for encrypted media by @Morxi in #5059
  • feat(inbox): add source type filter for push messages by @lalaliat in #5522
  • fix(skillpool): restore SkillPool styles after PR #5521 by @zhaozhuang521 in #5532
  • refactor(readme): add trending badge by @lalaliat in #5534
  • fix(tui): restore ACP commands and inline approvals by @ekzhu in #5443
  • Document Langfuse Docker deployment by @totoyang in #5380
  • fix: add missing emptyState and emptyIcon styles for TokenUsage page by @zhaozhuang521 in #5544
  • fix(chat): preserve assistant markdown newlines by @wananing in #5538
  • fix: cap the file size of send_file_to_user by @weidankong in #5457
  • fix(skill): zip metadata error handling by @Leirunlin in #5481
  • feat(memory): refactor auto memory system with turn-based tracking by @jinliyl in #5540
  • feat(tui): support project-scoped code sessions by @ekzhu in #5448
  • fix(models): count only configured providers online by @wananing in #5537
  • fix(skill): remove enable all during init by @Leirunlin in #5477
  • test(e2e): adapt for agentscope 2.0 — drop Plan Mode, fix selectors and fixtures by @yutai78786 in #5542
  • feat(desktop): add tauri auto updater by @jinglinpeng in #4669
  • test(integration): parallelize with pytest-xdist and harden coverage pipeline by @yutai78786 in #5531
  • fix: standalone "type": "null" in functionDeclaration schema breaks third-party proxies by @NotFoundOvO in #5545
  • fix(hooks): inject user_id and channel into ContextVarsSetupHook by @XiuShenAl in #5552
  • feat(channel): add Slack channel with full multimodal & streaming support by @celestialhorse51D in #5193
  • fix(channel): use event content as primary text source in streaming end by @hongxicheng in #5553
  • Feat/session switch performance by @zhaozhuang521 in #5559

New Contributors

Full Changelog: v1.1.12.post2...v2.0.0-beta.1

Don't miss a new QwenPaw release

NewReleases is sending notifications on new releases.