v2.1.0

## What's Changed
## ✨ Added

**QwenPaw OS Shell**

- Open apps in movable, resizable windows with a launcher, taskbar, notifications, and saved layouts ([#6645](https://github.com/agentscope-ai/QwenPaw/pull/6645))
- Installed and marketplace apps now share one catalog across the App Center and QwenPaw OS Shell ([#6718](https://github.com/agentscope-ai/QwenPaw/pull/6718))

**Unified Files Workspace**

- Browse, preview, edit, compare, upload, and download workspace files without leaving Chat ([#6504](https://github.com/agentscope-ai/QwenPaw/pull/6504))
- Open the Files workspace directly from QwenPaw OS Shell ([#6758](https://github.com/agentscope-ai/QwenPaw/pull/6758))
- Edit Robot Framework files with syntax highlighting ([#6519](https://github.com/agentscope-ai/QwenPaw/pull/6519))

**QwenPaw Creator**

- Turn an idea or existing footage into a finished video with a team of specialized Agents ([#6284](https://github.com/agentscope-ai/QwenPaw/pull/6284))
- Select clips, subtitles, effects, assets, or text for focused Agent edits while keeping direct controls ([#6740](https://github.com/agentscope-ai/QwenPaw/pull/6740))
- Create with images, video, voice, subtitles, motion, overlays, and parallel production steps ([#6740](https://github.com/agentscope-ai/QwenPaw/pull/6740))
- Add media task recovery, rendering fixes, structured logs, and review feedback ([#6556](https://github.com/agentscope-ai/QwenPaw/pull/6556), [#6641](https://github.com/agentscope-ai/QwenPaw/pull/6641))
- Add a settings center, external Skills, asynchronous media generation, more rendering formats, and improved Windows and narrow-screen support ([#6870](https://github.com/agentscope-ai/QwenPaw/pull/6870))

**Cross-harness Integration**

- Create Agents powered by Codex or Qoder, with per-session models, permissions, Skills, and MCP services ([#6397](https://github.com/agentscope-ai/QwenPaw/pull/6397))
- ACP clients can pass temporary models, project folders, and MCP services without saving runtime credentials ([#6754](https://github.com/agentscope-ai/QwenPaw/pull/6754))
- ACP sessions now report available and current models to compatible clients ([#6531](https://github.com/agentscope-ai/QwenPaw/pull/6531))

**Browser-use**

- Browser-use remains in Beta. It can use a separate browser session or connect selected tabs from the user's signed-in Chrome. Browser-side code execution is not protected by a host-level sandbox, and action policies are not yet enforced ([#6276](https://github.com/agentscope-ai/QwenPaw/pull/6276), [#6157](https://github.com/agentscope-ai/QwenPaw/pull/6157))

**Computer-use**

- Add Windows desktop app automation. macOS packaging and signing are complete, while device-level end-to-end validation is still in progress. Operations require per-app approval and stop when the user takes over input ([#6424](https://github.com/agentscope-ai/QwenPaw/pull/6424))
- Native Computer-use adds keyboard sequences, improves Windows modal targeting and text editing, and makes macOS menus, dragging, capture, and focus more reliable ([#6891](https://github.com/agentscope-ai/QwenPaw/pull/6891), [#6913](https://github.com/agentscope-ai/QwenPaw/pull/6913), [#6977](https://github.com/agentscope-ai/QwenPaw/pull/6977))

**Context, Recovery & Long-running Work**

- Snapshot and restore conversations, memory, and selected workspace files without changing project Git history ([#6269](https://github.com/agentscope-ai/QwenPaw/pull/6269))
- Long conversations retain the active task while older exact history moves into a searchable index ([#6323](https://github.com/agentscope-ai/QwenPaw/pull/6323))
- History search returns paired requests and responses and understands date-based queries ([#6237](https://github.com/agentscope-ai/QwenPaw/pull/6237))
- Visual Compact reduces long-context input for supported multimodal models while keeping source recovery ([#6456](https://github.com/agentscope-ai/QwenPaw/pull/6456))
- Long-running tools can continue in the background with progress, extension, and cancellation controls ([#6151](https://github.com/agentscope-ai/QwenPaw/pull/6151))
- Tool workflows can reuse results, branch, loop, and run independent steps in parallel ([#5698](https://github.com/agentscope-ai/QwenPaw/pull/5698))
- ReMe Light memory search can optionally rerank candidates through an OpenAI-compatible endpoint and falls back gracefully when reranking is unavailable ([#6398](https://github.com/agentscope-ai/QwenPaw/pull/6398))
- ReMe Light adds validated, hot-updatable embedding settings, Daily Paper and service-level schedules, per-Agent runtime status, manual reindex controls, and independent search and notification options ([#6772](https://github.com/agentscope-ai/QwenPaw/pull/6772))

**Console & Models**

- Agent Statistics now shows token usage and narrows results to the current Agent ([#6503](https://github.com/agentscope-ai/QwenPaw/pull/6503), [#6402](https://github.com/agentscope-ai/QwenPaw/pull/6402), [#6862](https://github.com/agentscope-ai/QwenPaw/pull/6862))
- Project directory selection can include hidden folders when requested ([#6878](https://github.com/agentscope-ai/QwenPaw/pull/6878))
- Code blocks now share a responsive, theme-aware design with syntax highlighting, copy and download actions; LaTeX and Mermaid blocks add localized Preview and Source tabs ([#6911](https://github.com/agentscope-ai/QwenPaw/pull/6911))
- Explore memory relationships in an interactive 3D graph with focus, filtering, and reduced-motion support ([#6922](https://github.com/agentscope-ai/QwenPaw/pull/6922))

**Channels**

- Feishu, QQ, WeCom, Xiaoyi, and Yuanbao can use opt-in custom gateway endpoints for private deployments and local integration testing without changing their defaults ([#6907](https://github.com/agentscope-ai/QwenPaw/pull/6907))
- Saving an enabled built-in channel warns when the same bot identity is already used by another running Agent, while still allowing an explicit override ([#6909](https://github.com/agentscope-ai/QwenPaw/pull/6909))

## 🔄 Changed

- The App Center separates installed and available apps, expands app details, and adds a Simple Mode entry ([#6553](https://github.com/agentscope-ai/QwenPaw/pull/6553), [#6638](https://github.com/agentscope-ai/QwenPaw/pull/6638))
- Chat keeps loop modes aligned with the active Agent and lists localized Oh-My-Paw modes ([#6711](https://github.com/agentscope-ai/QwenPaw/pull/6711), [#6521](https://github.com/agentscope-ai/QwenPaw/pull/6521))
- Slash-command Skills show injected content, duplicate copies are skipped, and QwenPaw Skill links can be imported directly ([#6345](https://github.com/agentscope-ai/QwenPaw/pull/6345), [#6650](https://github.com/agentscope-ai/QwenPaw/pull/6650), [#6517](https://github.com/agentscope-ai/QwenPaw/pull/6517))
- Windows selects sandboxing based on current privileges; unsupported restrictions are reported on every platform ([#6383](https://github.com/agentscope-ai/QwenPaw/pull/6383), [#6657](https://github.com/agentscope-ai/QwenPaw/pull/6657))
- OneBot/QQ preserves text and media order and expands quoted or forwarded messages ([#6543](https://github.com/agentscope-ai/QwenPaw/pull/6543), [#6769](https://github.com/agentscope-ai/QwenPaw/pull/6769))
- QwenPaw OS Shell places the Simple/Desktop Mode switches side by side and shortens the Desktop Mode label ([#6763](https://github.com/agentscope-ai/QwenPaw/pull/6763))
- QwenPaw OS Shell now uses one browser entry, launches apps from inside the desktop, and removes obsolete Coding launch paths ([#6868](https://github.com/agentscope-ai/QwenPaw/pull/6868))
- Proactive Mode reuses the active workspace, releases its resources when disabled, and uses lightweight web tools for research ([#6712](https://github.com/agentscope-ai/QwenPaw/pull/6712))
- MiniMax models and context limits are updated for the M3 and M2.7 generations ([#6479](https://github.com/agentscope-ai/QwenPaw/pull/6479), [#6554](https://github.com/agentscope-ai/QwenPaw/pull/6554))
- Plugin installation is more tolerant of packages that declare a newer compatibility ceiling ([#6532](https://github.com/agentscope-ai/QwenPaw/pull/6532))

## 🐛 Fixed

**QwenPaw Creator**

- Failed work resumes when inputs change, duplicate video renders are avoided, and stale scenes are reviewed before final composition ([#6937](https://github.com/agentscope-ai/QwenPaw/pull/6937))

**Desktop**

- Computer-use reuses the signed macOS app identity so permissions remain across launches ([#6590](https://github.com/agentscope-ai/QwenPaw/pull/6590))
- Desktop startup handles port errors; Windows shutdown checks are bounded and stay hidden ([#6705](https://github.com/agentscope-ai/QwenPaw/pull/6705), [#6203](https://github.com/agentscope-ai/QwenPaw/pull/6203))
- Chrome reconnects after interruption, packaged browsers are found on desktop, and Windows restore locking prevents concurrent recovery ([#6669](https://github.com/agentscope-ai/QwenPaw/pull/6669))
- QwenPaw OS Shell windows keep the menu bar visible, stay inside the viewport, and allow text selection and copying ([#6801](https://github.com/agentscope-ai/QwenPaw/pull/6801), [#6861](https://github.com/agentscope-ai/QwenPaw/pull/6861))

**Runtime & Agents**

- Provider-specific tool-call fields are no longer lost during streaming, history storage, compaction, or replay ([#6759](https://github.com/agentscope-ai/QwenPaw/pull/6759))
- When a model reports a context-limit error, Scroll compacts the context and retries once; history migration also preserves session identity ([#6267](https://github.com/agentscope-ai/QwenPaw/pull/6267), [#6068](https://github.com/agentscope-ai/QwenPaw/pull/6068))
- Codex-backed Agents stay manageable when the Codex runtime is unavailable ([#6664](https://github.com/agentscope-ai/QwenPaw/pull/6664))
- Sub-agents inherit the current session approval level; an empty batch field no longer turns a single task into a batch request ([#6508](https://github.com/agentscope-ai/QwenPaw/pull/6508), [#6595](https://github.com/agentscope-ai/QwenPaw/pull/6595))
- Loop limits are written back for older configurations, and renamed per-tool entries persist ([#6530](https://github.com/agentscope-ai/QwenPaw/pull/6530), [#6682](https://github.com/agentscope-ai/QwenPaw/pull/6682))
- Shell execution preserves multiline commands and no longer hangs on inherited background pipes ([#6412](https://github.com/agentscope-ai/QwenPaw/pull/6412), [#6748](https://github.com/agentscope-ai/QwenPaw/pull/6748))
- Headless tasks construct the effective user message, and workspace commits use a local default Git identity when none is configured ([#6616](https://github.com/agentscope-ai/QwenPaw/pull/6616), [#6350](https://github.com/agentscope-ai/QwenPaw/pull/6350))
- Copying an Agent no longer creates empty Skills or scheduled-task scaffolding ([#6493](https://github.com/agentscope-ai/QwenPaw/pull/6493))
- Background sub-agent tasks wait for authoritative fork finalization and report failures without discarding worktree changes ([#6725](https://github.com/agentscope-ai/QwenPaw/pull/6725))
- Corrupted Agent configuration and invalid JSON produce a recoverable configuration error instead of breaking config loading ([#6615](https://github.com/agentscope-ai/QwenPaw/pull/6615))
- Scroll recall finds complete CJK turns for substring and multi-term searches and rejects reversed expansion ranges ([#6824](https://github.com/agentscope-ai/QwenPaw/pull/6824))
- Legacy sessions load local-path media again, including Windows UNC paths and percent-encoded non-ASCII paths ([#6873](https://github.com/agentscope-ai/QwenPaw/pull/6873))
- Memory hints no longer break conversations, and ordinary non-streaming responses can generate chat titles ([#6360](https://github.com/agentscope-ai/QwenPaw/pull/6360), [#6816](https://github.com/agentscope-ai/QwenPaw/pull/6816))
- Continuation summaries respect disabled thinking and preserve interruption or timeout behavior ([#6818](https://github.com/agentscope-ai/QwenPaw/pull/6818))
- Headless ACP sessions honor temporary model context and output limits without saving them ([#6933](https://github.com/agentscope-ai/QwenPaw/pull/6933))

**Console & UI**

- Reconnects and Agent switches no longer restore the wrong chat, stale channel, or duplicate pending messages ([#6602](https://github.com/agentscope-ai/QwenPaw/pull/6602), [#6546](https://github.com/agentscope-ai/QwenPaw/pull/6546), [#6382](https://github.com/agentscope-ai/QwenPaw/pull/6382))
- Rich messages enter the send queue, and multi-file previews wrap inside the composer ([#6798](https://github.com/agentscope-ai/QwenPaw/pull/6798), [#6662](https://github.com/agentscope-ai/QwenPaw/pull/6662))
- Large tool output uses a bounded preview, and long commands wrap inside cards ([#6637](https://github.com/agentscope-ai/QwenPaw/pull/6637), [#6677](https://github.com/agentscope-ai/QwenPaw/pull/6677))
- Tool controls stay hidden until streaming finishes ([#6749](https://github.com/agentscope-ai/QwenPaw/pull/6749))
- Settings stays visible when the sidebar is collapsed ([#6488](https://github.com/agentscope-ai/QwenPaw/pull/6488))
- Production Console builds no longer omit dependency CSS ([#6639](https://github.com/agentscope-ai/QwenPaw/pull/6639))
- Chat wheel scrolling stays stable, and the load-more spinner no longer runs while off screen ([#6904](https://github.com/agentscope-ai/QwenPaw/pull/6904), [#6896](https://github.com/agentscope-ai/QwenPaw/pull/6896))
- The header resources menu no longer shows a duplicate GitHub link ([#6903](https://github.com/agentscope-ai/QwenPaw/pull/6903))
- Workspace previews support Unicode PDF filenames and SVG media types, with corrected dark-mode styling for preview and edit surfaces ([#6915](https://github.com/agentscope-ai/QwenPaw/pull/6915))
- Long multiline assistant and tool output preserves line breaks, while wide code blocks and tables scroll within their containers ([#6890](https://github.com/agentscope-ai/QwenPaw/pull/6890))
- Chat falls back to the standard textarea input, and desktop and end-to-end verification support both textarea and contenteditable inputs ([#6934](https://github.com/agentscope-ai/QwenPaw/pull/6934))
- Image uploads no longer inflate the context-usage ring by counting Base64 data as text ([#6968](https://github.com/agentscope-ai/QwenPaw/pull/6968))
- Large screenshots can be previewed, and completed tools no longer keep a spinning status icon ([#6965](https://github.com/agentscope-ai/QwenPaw/pull/6965), [#6967](https://github.com/agentscope-ai/QwenPaw/pull/6967))
- Daily memory notes are grouped and sorted by their path date ([#6941](https://github.com/agentscope-ai/QwenPaw/pull/6941))

**Governance & Security**

- Browser debugging requires authentication unless the user explicitly opts out ([#6500](https://github.com/agentscope-ai/QwenPaw/pull/6500))
- Project import only reads approved paths and reports skipped sensitive folders ([#6487](https://github.com/agentscope-ai/QwenPaw/pull/6487), [#6713](https://github.com/agentscope-ai/QwenPaw/pull/6713))
- Sandbox cleanup handles interrupted sessions; invalid settings fall back without crashing and display a warning ([#6600](https://github.com/agentscope-ai/QwenPaw/pull/6600), [#6582](https://github.com/agentscope-ai/QwenPaw/pull/6582), [#6747](https://github.com/agentscope-ai/QwenPaw/pull/6747), [#6857](https://github.com/agentscope-ai/QwenPaw/pull/6857))
- Sandboxed child processes no longer inherit `PYTHONHOME`, preventing them from resolving the wrong Python runtime ([#6902](https://github.com/agentscope-ai/QwenPaw/pull/6902))
- Conversation command arguments are no longer written to logs ([#6692](https://github.com/agentscope-ai/QwenPaw/pull/6692))

**Memory**

- Fix memory and checkpoint saving during context migration and web-workspace recovery ([#6592](https://github.com/agentscope-ai/QwenPaw/pull/6592), [#6597](https://github.com/agentscope-ai/QwenPaw/pull/6597))
- Automatic-memory progress survives context compression and session recovery; failed submissions can be retried without polluting chat history ([#6830](https://github.com/agentscope-ai/QwenPaw/pull/6830))

**MCP**

- Terminated MCP sessions recover without dropping every tool from the active Agent ([#6894](https://github.com/agentscope-ai/QwenPaw/pull/6894))

**CLI & Channels**

- Scheduled tasks fire while the service is idle ([#6481](https://github.com/agentscope-ai/QwenPaw/pull/6481))
- Pausing and resuming a scheduled task preserves its enabled state ([#6691](https://github.com/agentscope-ai/QwenPaw/pull/6691))
- Scheduled-task results follow the configured delivery behavior, including migrated tasks ([#6182](https://github.com/agentscope-ai/QwenPaw/pull/6182), [#6511](https://github.com/agentscope-ai/QwenPaw/pull/6511))
- Pausing or resuming a missing scheduled task returns a not-found response ([#6858](https://github.com/agentscope-ai/QwenPaw/pull/6858))
- The task command reports a missing Agent configuration without printing a traceback ([#6865](https://github.com/agentscope-ai/QwenPaw/pull/6865))
- CLI project paths are accepted after global options ([#6892](https://github.com/agentscope-ai/QwenPaw/pull/6892))
- `qwenpaw update --prerelease` now discovers newer prereleases from PyPI ([#6939](https://github.com/agentscope-ai/QwenPaw/pull/6939))
- OneBot listens locally by default and requires a token when exposed to the network ([#6676](https://github.com/agentscope-ai/QwenPaw/pull/6676))
- OneBot configuration declarations stay aligned with the configuration API ([#6864](https://github.com/agentscope-ai/QwenPaw/pull/6864))
- Channel approvals preserve routing fields through the driver gate, and Xiaoyi messages hide internal Scroll headlines ([#6833](https://github.com/agentscope-ai/QwenPaw/pull/6833), [#6888](https://github.com/agentscope-ai/QwenPaw/pull/6888))
- DingTalk organization approval now fills returned credentials into channel setup ([#6709](https://github.com/agentscope-ai/QwenPaw/pull/6709))
- Matrix can connect and decrypt messages on Python 3.12 ([#6486](https://github.com/agentscope-ai/QwenPaw/pull/6486))
- Timestamps without a timezone are interpreted consistently as local process time ([#6685](https://github.com/agentscope-ai/QwenPaw/pull/6685), [#6855](https://github.com/agentscope-ai/QwenPaw/pull/6855))

**Other**

- Backup restore waits for the background operation to finish and accepts zero-byte lock files ([#6735](https://github.com/agentscope-ai/QwenPaw/pull/6735), [#6703](https://github.com/agentscope-ai/QwenPaw/pull/6703))
- OpenRouter preserves known multimodal support and retries recoverable stream or reasoning errors ([#6733](https://github.com/agentscope-ai/QwenPaw/pull/6733), [#6714](https://github.com/agentscope-ai/QwenPaw/pull/6714), [#6721](https://github.com/agentscope-ai/QwenPaw/pull/6721))
- Video input is forwarded to models that declare video support, and channel audio is transcribed again ([#6495](https://github.com/agentscope-ai/QwenPaw/pull/6495), [#6573](https://github.com/agentscope-ai/QwenPaw/pull/6573))
- Local images keep their original content across file changes, and temporary media errors no longer disable multimodal support ([#6930](https://github.com/agentscope-ai/QwenPaw/pull/6930))

## New Contributors

- @axelray-dev made their first contribution in [#6412](https://github.com/agentscope-ai/QwenPaw/pull/6412)
- @WilShi made their first contribution in [#6477](https://github.com/agentscope-ai/QwenPaw/pull/6477)
- @FittyAr made their first contribution in [#6479](https://github.com/agentscope-ai/QwenPaw/pull/6479)
- @yaodong-shen made their first contribution in [#6502](https://github.com/agentscope-ai/QwenPaw/pull/6502)
- @niuda-kok made their first contribution in [#6531](https://github.com/agentscope-ai/QwenPaw/pull/6531)
- @GMsure made their first contribution in [#6543](https://github.com/agentscope-ai/QwenPaw/pull/6543)
- @ningblue made their first contribution in [#6723](https://github.com/agentscope-ai/QwenPaw/pull/6723)
- @AntiQuality made their first contribution in [#6898](https://github.com/agentscope-ai/QwenPaw/pull/6898)
- @xingliu228 made their first contribution in [#6873](https://github.com/agentscope-ai/QwenPaw/pull/6873)

**Full Changelog**: [v2.0.1...v2.1.0](https://github.com/agentscope-ai/QwenPaw/compare/v2.0.1...v2.1.0)