v2.0.0

# What's Changed

## ✨ Added

**Core Architecture**

- **Runtime 2.0**: Refactor the kernel based on AgentScope 2.0 ([#5078](https://github.com/agentscope-ai/QwenPaw/pull/5078), [#4846](https://github.com/agentscope-ai/QwenPaw/pull/4846), [#5018](https://github.com/agentscope-ai/QwenPaw/pull/5018), [#5552](https://github.com/agentscope-ai/QwenPaw/pull/5552))
- **Agent OS Driver**: Unified management of MCP/A2A/ACP connections, encrypted credential storage, and per-call approval policy ([#5067](https://github.com/agentscope-ai/QwenPaw/pull/5067))
- **Loop Engineering**: New Loop gate design with three Loop templates: Normal Mode / Mission Mode / Goal Mode ([#5665](https://github.com/agentscope-ai/QwenPaw/pull/5665), [#5727](https://github.com/agentscope-ai/QwenPaw/pull/5727))
- **Governance & Sandbox**: Policy engine with allow/deny/ask rules, ToolGuard detectors merged into a single policy evaluation path, and a generalized governance pattern for plugins ([#5088](https://github.com/agentscope-ai/QwenPaw/pull/5088), [#5546](https://github.com/agentscope-ai/QwenPaw/pull/5546), [#5301](https://github.com/agentscope-ai/QwenPaw/pull/5301))

**Context Management**

- **Scroll Context**: SQLite-backed working history (history.db) with FTS5 full-text search, eviction index, turn-level persistence, and a sandboxed `recall_history_python` REPL for on-demand recall of older turns. 30-day default retention, auto-imports existing sessions on startup ([#5321](https://github.com/agentscope-ai/QwenPaw/pull/5321))
- **Native Compression**: Refactor using AgentScope 2.0's built-in context compression strategy ([#5309](https://github.com/agentscope-ai/QwenPaw/pull/5309))

**Long-term Memory**

- **ReMe v0.4.0 backend**: Replace the in-house memory runtime with ReMe v0.4.0 — vector store, auto-tracking, and retrieval are now handled by the ReMe v0.4.0 library ([#5349](https://github.com/agentscope-ai/QwenPaw/pull/5349))
- **Turn-based tracking**: Memory writes now happen per-turn via middleware, no longer tied to session close ([#5540](https://github.com/agentscope-ai/QwenPaw/pull/5540))
- **Usage-aware search**: Memory search now considers recency and frequency; backend-specific embedding model selection added to config UI ([#5820](https://github.com/agentscope-ai/QwenPaw/pull/5820))
- **ADBPG REST mode**: ADBPG memory backend switched to REST-only, with auto search ([#5296](https://github.com/agentscope-ai/QwenPaw/pull/5296))
- **Search UI simplification**: Cleaner memory search display, metadata fields reduced to essentials ([#5482](https://github.com/agentscope-ai/QwenPaw/pull/5482))
- **'none' backend**: Config option to fully disable the memory system for lightweight deployments ([#5732](https://github.com/agentscope-ai/QwenPaw/pull/5732))

**Terminal UI (TUI)**

- **Full-screen TUI**: A textual-based terminal chat app (`qwenpaw tui`) — streaming output, tool-call rendering, status bar, and session management all in the terminal ([#5032](https://github.com/agentscope-ai/QwenPaw/pull/5032))
- **Project-scoped sessions**: TUI can launch a Coding Mode session scoped to a specific project directory ([#5448](https://github.com/agentscope-ai/QwenPaw/pull/5448))
- **Context-usage bar**: Live token/context usage indicator in the TUI status bar ([#5673](https://github.com/agentscope-ai/QwenPaw/pull/5673))
- **ACP commands in TUI**: `/approve`, `/deny` inline approval commands + transcript panel improvements ([#5443](https://github.com/agentscope-ai/QwenPaw/pull/5443), [#5714](https://github.com/agentscope-ai/QwenPaw/pull/5714))

**Console & UI**

- **Approval card**: Larger action buttons, clearer diff preview, one-click approve/deny ([#5855](https://github.com/agentscope-ai/QwenPaw/pull/5855))
- **Long text upload**: Paste or upload large text blocks directly into the chat input ([#5854](https://github.com/agentscope-ai/QwenPaw/pull/5854))
- **Session-level tool approval**: Override global tool policy per session ([#5685](https://github.com/agentscope-ai/QwenPaw/pull/5685), [#5832](https://github.com/agentscope-ai/QwenPaw/pull/5832))
- **Non-owner banner**: Show "view-only" banner when viewing someone else's session tab ([#5664](https://github.com/agentscope-ai/QwenPaw/pull/5664))
- **Mobile layouts**: Responsive CSS + drawer-based navigation for all console pages ([#5444](https://github.com/agentscope-ai/QwenPaw/pull/5444)–[#5470](https://github.com/agentscope-ai/QwenPaw/pull/5470))
- **Tools page split view**: Enabled vs. available tools shown in separate columns; per-tool config-aware toggle ([#5509](https://github.com/agentscope-ai/QwenPaw/pull/5509), [#5604](https://github.com/agentscope-ai/QwenPaw/pull/5604))
- **Skill market improvements**: Star/download counts, enabled/disabled split, auto-sync from CDN ([#5521](https://github.com/agentscope-ai/QwenPaw/pull/5521), [#5706](https://github.com/agentscope-ai/QwenPaw/pull/5706), [#5639](https://github.com/agentscope-ai/QwenPaw/pull/5639))
- **Channels page dual layout**: Enabled channels on top, available below ([#5504](https://github.com/agentscope-ai/QwenPaw/pull/5504))
- **Coding Mode 50/50 split**: Chat panel takes half the viewport ([#5473](https://github.com/agentscope-ai/QwenPaw/pull/5473))
- **Hidden folders in project browser**: `.` prefixed directories now selectable ([#5828](https://github.com/agentscope-ai/QwenPaw/pull/5828))
- **Session drawer**: Date-grouped sessions with inline search ([#5728](https://github.com/agentscope-ai/QwenPaw/pull/5728))
- **Token usage popover**: Ring chart showing token/context usage per turn, restored for Runtime 2.0 ([#5493](https://github.com/agentscope-ai/QwenPaw/pull/5493), [#5544](https://github.com/agentscope-ai/QwenPaw/pull/5544))
- **Inbox filters**: Filter push notifications by source type ([#5522](https://github.com/agentscope-ai/QwenPaw/pull/5522))
- **Message anchors**: Jump-to-message navigator for user turns ([#5515](https://github.com/agentscope-ai/QwenPaw/pull/5515))
- **Request timeout**: Configurable HTTP timeout + AbortSignal propagation for backend calls ([#5764](https://github.com/agentscope-ai/QwenPaw/pull/5764))

**Model Providers**

- **OpenAI Response API**: New provider for the OpenAI Response API format ([#5519](https://github.com/agentscope-ai/QwenPaw/pull/5519))
- **DashScope updates**: Qwen3.7 + DeepSeek V4 added to built-in model list; `extra_body` passthrough for `generate_kwargs`; per-model thinking relay toggle ([#5646](https://github.com/agentscope-ai/QwenPaw/pull/5646), [#5491](https://github.com/agentscope-ai/QwenPaw/pull/5491), [#5687](https://github.com/agentscope-ai/QwenPaw/pull/5687), [#5730](https://github.com/agentscope-ai/QwenPaw/pull/5730))
- **GitHub Models**: Migrate to new API endpoint; support fine-grained PATs ([#5735](https://github.com/agentscope-ai/QwenPaw/pull/5735))
- **Telegram custom base URL**: `api_base_url` config for self-hosted Telegram Bot API servers ([#5651](https://github.com/agentscope-ai/QwenPaw/pull/5651))

**Channels**

- **Matrix streaming**: Edit-in-place streaming, no longer sends a new message per chunk ([#5585](https://github.com/agentscope-ai/QwenPaw/pull/5585))
- **Plugin-registered channels**: Plugins can register custom channel types with a JSON schema that auto-generates config UI ([#4693](https://github.com/agentscope-ai/QwenPaw/pull/4693))
- **Per-channel debounce toggle**: `no_text_debounce` flag so fast channels don't batch messages ([#5617](https://github.com/agentscope-ai/QwenPaw/pull/5617))
- **Slack**: Full channel implementation with multimodal attachments, thread replies, and streaming ([#5193](https://github.com/agentscope-ai/QwenPaw/pull/5193))
- **DingTalk private deploy**: `api_endpoint` config for on-prem DingTalk instances ([#5111](https://github.com/agentscope-ai/QwenPaw/pull/5111))
- **Azure Bot channel**: Azure Bot channel added as a plugin, with i18n/custom-icon/doc_url support for plugin channels ([#5849](https://github.com/agentscope-ai/QwenPaw/pull/5849))
- **Discord attachments**: Download to local `media_dir`, serve via static file route ([#5468](https://github.com/agentscope-ai/QwenPaw/pull/5468))

**Desktop**

- **Tauri migration**: Desktop build moved from Electron to Tauri — smaller binary, native webview, NSIS installer on Windows ([#5734](https://github.com/agentscope-ai/QwenPaw/pull/5734), [#4669](https://github.com/agentscope-ai/QwenPaw/pull/4669), [#4041](https://github.com/agentscope-ai/QwenPaw/pull/4041))
- **Bundled ACP runtime**: Node.js runtime + ACP modules ship inside the desktop package for out-of-box agent communication ([#5814](https://github.com/agentscope-ai/QwenPaw/pull/5814), [#5794](https://github.com/agentscope-ai/QwenPaw/pull/5794))
- **DevTools gesture**: 8× logo click opens Chromium DevTools for debugging ([#5805](https://github.com/agentscope-ai/QwenPaw/pull/5805))

**Security & Sandbox**

- **Windows AppContainer sandbox**: Code execution isolated via Windows native AppContainer API ([#5525](https://github.com/agentscope-ai/QwenPaw/pull/5525))
- **Linux bubblewrap sandbox**: Mount namespace isolation via bwrap for Linux code execution ([#5310](https://github.com/agentscope-ai/QwenPaw/pull/5310))
- **Cron tool-safety toggle**: Per-job flag to require/skip tool approval for scheduled tasks ([#5847](https://github.com/agentscope-ai/QwenPaw/pull/5847))
- **Dangerous command detection**: ToolGuard detects `find -delete` and similar destructive shell patterns ([#5843](https://github.com/agentscope-ai/QwenPaw/pull/5843))
- **Access policy**: Declarative allow/deny/ask rules evaluated per capability call

**Plugins**

- **AgentScope middleware hooks**: Plugins can register AgentScope-compatible middleware with semver versioning ([#5221](https://github.com/agentscope-ai/QwenPaw/pull/5221))
- **Version-gated catalog**: Plugin market and CDN only show plugins compatible with the running QwenPaw version ([#5661](https://github.com/agentscope-ai/QwenPaw/pull/5661))
- **Lifecycle hooks**: `on_uninstall` hook, validator fixes, and public skill-provider API for plugin authors ([#5008](https://github.com/agentscope-ai/QwenPaw/pull/5008))

**Tools**

- **`web_search` and `web_fetch`**: Built-in web search and fetch tools with product identity ([#5890](https://github.com/agentscope-ai/QwenPaw/pull/5890))
- **`grep_search` show_file option**: Add `show_file` flag to display matched file content ([#5840](https://github.com/agentscope-ai/QwenPaw/pull/5840))
- **Tool error passthrough**: Tool result error state forwarded to frontend for better UX ([#5912](https://github.com/agentscope-ai/QwenPaw/pull/5912))

**CLI**

- **`cron update`**: Edit schedule/prompt/config of existing cron jobs without delete+recreate ([#5210](https://github.com/agentscope-ai/QwenPaw/pull/5210))

**Observability**

- **Langfuse 2.0**: Trace grouping restored via hook integration; each agent loop is a Langfuse generation span ([#5511](https://github.com/agentscope-ai/QwenPaw/pull/5511))

## 🔄 Changed

- **relay_reasoning**: `preserve_thinking` renamed to `relay_reasoning`; defaults to off for DashScope to avoid duplicate reasoning in output ([#5876](https://github.com/agentscope-ai/QwenPaw/pull/5876))
- **Rate limiting**: Multi-dimensional (IP + user + endpoint) rate limiting with configurable windows ([#5738](https://github.com/agentscope-ai/QwenPaw/pull/5738))
- **Memory lifecycle**: `/compact` command reworked; new `/system_prompt` command; search state simplified to avoid stale embeddings ([#5450](https://github.com/agentscope-ai/QwenPaw/pull/5450), [#5815](https://github.com/agentscope-ai/QwenPaw/pull/5815))
- **SessionItem component**: All session renderers consolidated into one `SessionItem` component ([#5754](https://github.com/agentscope-ai/QwenPaw/pull/5754))
- **Skill UI**: SkillPool/SkillMarket components refactored; removed dead CSS ([#5753](https://github.com/agentscope-ai/QwenPaw/pull/5753), [#5532](https://github.com/agentscope-ai/QwenPaw/pull/5532))
- **Plan Mode removed**: Agent config no longer includes plan_mode; UI references cleaned up ([#5605](https://github.com/agentscope-ai/QwenPaw/pull/5605), [#5618](https://github.com/agentscope-ai/QwenPaw/pull/5618))
- **Plugin version retention**: Old plugin versions kept on disk after update; failed-load state cleaned up on next boot ([#5695](https://github.com/agentscope-ai/QwenPaw/pull/5695))

## 🐛 Fixed

**Context & Memory**

- Eviction no longer removes the active turn; graduated pressure relief prevents OOM on large tool outputs ([#5765](https://github.com/agentscope-ai/QwenPaw/pull/5765))
- Live turn anchored with a seam banner so the model knows where eviction ends ([#5871](https://github.com/agentscope-ai/QwenPaw/pull/5871))
- Evicted spans without headlines now get auto-generated labels in the index ([#5848](https://github.com/agentscope-ai/QwenPaw/pull/5848))
- Strip headline markers from HTTP history path ([#5672](https://github.com/agentscope-ai/QwenPaw/pull/5672))
- Adjust system prompt wording to stop the model from writing every observation to memory ([#5629](https://github.com/agentscope-ai/QwenPaw/pull/5629))
- Truncate oversized tool results before context insertion to prevent token overflow ([#5510](https://github.com/agentscope-ai/QwenPaw/pull/5510))

**Providers & Tool Schemas**

- OpenAI Responses API formatting and reasoning_content compatibility fixed ([#5920](https://github.com/agentscope-ai/QwenPaw/pull/5920))
- Responses API function_call names preserved correctly ([#5913](https://github.com/agentscope-ai/QwenPaw/pull/5913))
- Fix reasoning_content alignment by improving base formatter drop prediction ([#5865](https://github.com/agentscope-ai/QwenPaw/pull/5865))
- Fix provider-aware context threshold display ([#5902](https://github.com/agentscope-ai/QwenPaw/pull/5902))
- Retry on OpenAI streaming errors that include a body status code ([#5799](https://github.com/agentscope-ai/QwenPaw/pull/5799))
- Restore OpenRouter OAuth route accidentally dropped during Runtime 2.0 migration ([#5806](https://github.com/agentscope-ai/QwenPaw/pull/5806))
- `reasoning_content` field errors in streaming no longer crash the response ([#5582](https://github.com/agentscope-ai/QwenPaw/pull/5582))
- Tool schema sanitization: nullable fields, Gemini `const`, standalone `type:null` all handled ([#5549](https://github.com/agentscope-ai/QwenPaw/pull/5549), [#5827](https://github.com/agentscope-ai/QwenPaw/pull/5827), [#5545](https://github.com/agentscope-ai/QwenPaw/pull/5545))
- Inline `$ref`/`$defs` for GLM models that don't support JSON Schema references ([#5496](https://github.com/agentscope-ai/QwenPaw/pull/5496))
- Normalize MCP tool names for OpenAI API compliance; gracefully decode malformed tool input JSON ([#5485](https://github.com/agentscope-ai/QwenPaw/pull/5485), [#5486](https://github.com/agentscope-ai/QwenPaw/pull/5486))
- Empty `api_domain` or missing proto descriptors no longer cause hard crashes on Yuanbao ([#5804](https://github.com/agentscope-ai/QwenPaw/pull/5804))
- Only actually configured providers count toward the "online" indicator ([#5537](https://github.com/agentscope-ai/QwenPaw/pull/5537))
- Gemini `format_tools` override to match their expected schema format ([#5517](https://github.com/agentscope-ai/QwenPaw/pull/5517))

**Runtime & Agents**

- Interrupted turns now persist their partial state; empty cancel messages suppressed ([#5838](https://github.com/agentscope-ai/QwenPaw/pull/5838))
- `cancel_envelope` properly yielded on task cancellation ([#5674](https://github.com/agentscope-ai/QwenPaw/pull/5674))
- Workspace prompt files (`.qwenpaw.md`) and skill injection restored in system prompt assembly ([#5396](https://github.com/agentscope-ai/QwenPaw/pull/5396), [#5680](https://github.com/agentscope-ai/QwenPaw/pull/5680))
- Remove stale `language` arg from memory section call ([#5803](https://github.com/agentscope-ai/QwenPaw/pull/5803))
- `spawn_subagent` works again under Runtime 2.0 ([#5660](https://github.com/agentscope-ai/QwenPaw/pull/5660))
- Using `project_dir` without `coding_mode` enabled no longer silently proceeds ([#5640](https://github.com/agentscope-ai/QwenPaw/pull/5640))
- Self-paired tool messages (call + result from same turn) no longer dropped during message sanitation ([#5792](https://github.com/agentscope-ai/QwenPaw/pull/5792))
- SSE envelope events aligned with v1 protocol format for backward compat ([#5495](https://github.com/agentscope-ai/QwenPaw/pull/5495))
- Windows backslash paths converted to `file://` URLs; reverse parsing fixed ([#5635](https://github.com/agentscope-ai/QwenPaw/pull/5635))
- Handle `TextBlock` objects in `offload_tool_result` ([#5831](https://github.com/agentscope-ai/QwenPaw/pull/5831))
- Leading whitespace before JSON objects no longer breaks parsing ([#5766](https://github.com/agentscope-ai/QwenPaw/pull/5766), [#5841](https://github.com/agentscope-ai/QwenPaw/pull/5841))
- Content-safety image rejections no longer treated as "model can't do images" ([#5533](https://github.com/agentscope-ai/QwenPaw/pull/5533))
- Loop gate state reset on new user turn; peer-reset on continuation ([#5916](https://github.com/agentscope-ai/QwenPaw/pull/5916))
- Error envelope uses structured Error object for frontend SDK compatibility ([#5905](https://github.com/agentscope-ai/QwenPaw/pull/5905))
- `/stop` isolated by user_id to prevent cross-user task cancellation ([#5883](https://github.com/agentscope-ai/QwenPaw/pull/5883))

**Desktop**

- Python runtime download pinned to exact release URL ([#5875](https://github.com/agentscope-ai/QwenPaw/pull/5875))
- Plugin dep install no longer cascades into a storm of concurrent pip processes ([#5570](https://github.com/agentscope-ai/QwenPaw/pull/5570))
- Linux browser detection handles `env`-wrapped Exec lines in .desktop files ([#5526](https://github.com/agentscope-ai/QwenPaw/pull/5526))
- PyInstaller spec and Tauri config fixed for release packaging ([#5518](https://github.com/agentscope-ai/QwenPaw/pull/5518))
- Desktop screenshot tool now allowed in workspace governance ([#5641](https://github.com/agentscope-ai/QwenPaw/pull/5641))

**TUI**

- Fix CJK/IME composition on macOS/Linux ([#5671](https://github.com/agentscope-ai/QwenPaw/pull/5671))
- Improve approvals and warmup sessions ([#5892](https://github.com/agentscope-ai/QwenPaw/pull/5892))

**Console & Chat**

- Fix `rootSessionId` assignment in approval request routing ([#5886](https://github.com/agentscope-ai/QwenPaw/pull/5886))
- Mobile chat history drawer renders sessions correctly ([#5744](https://github.com/agentscope-ai/QwenPaw/pull/5744))
- Remove `white-space: pre-wrap` from message bubbles (caused double line breaks) ([#5596](https://github.com/agentscope-ai/QwenPaw/pull/5596))
- Preserve Markdown newlines in assistant messages ([#5538](https://github.com/agentscope-ai/QwenPaw/pull/5538))
- Route plugin market detail links through the link guard ([#5750](https://github.com/agentscope-ai/QwenPaw/pull/5750))
- Show stream completion timestamp ([#5742](https://github.com/agentscope-ai/QwenPaw/pull/5742))
- Timestamp always visible; fix double `/api` prefix in fetch URLs ([#5802](https://github.com/agentscope-ai/QwenPaw/pull/5802))
- Fix MCP access policy page layout ([#5213](https://github.com/agentscope-ai/QwenPaw/pull/5213))
- Push tool-guard approval notifications to IM channels ([#5601](https://github.com/agentscope-ai/QwenPaw/pull/5601))
- Plugin market compatibility labels now display correctly ([#5926](https://github.com/agentscope-ai/QwenPaw/pull/5926))
- Auto-enable preview mode for image files in coding session ([#5878](https://github.com/agentscope-ai/QwenPaw/pull/5878))

**MCP**

- Fix runtime approval level not properly propagated ([#5864](https://github.com/agentscope-ai/QwenPaw/pull/5864))
- Invalid MCP server configs in `agent.json` no longer crash agent startup ([#5755](https://github.com/agentscope-ai/QwenPaw/pull/5755))

**Channels**

- Split multi-segment streaming replies into separate message boxes ([#5487](https://github.com/agentscope-ai/QwenPaw/pull/5487))
- Streaming end event uses event content as text ([#5553](https://github.com/agentscope-ai/QwenPaw/pull/5553))
- Channel instance passed directly to notification path
- Matrix encrypted media downloaded via nio library ([#5059](https://github.com/agentscope-ai/QwenPaw/pull/5059))
- WeCom QR code URL extraction fixed ([#5460](https://github.com/agentscope-ai/QwenPaw/pull/5460))
- DingTalk delivery failures now surfaced to the user ([#5654](https://github.com/agentscope-ai/QwenPaw/pull/5654))
- Matrix token login failure due to duplicate auth parameters ([#5873](https://github.com/agentscope-ai/QwenPaw/pull/5873))
- Telegram typing indicator fires before consume ([#5700](https://github.com/agentscope-ai/QwenPaw/pull/5700))

**Cron**

- Run timestamps stored in the job's configured timezone ([#5783](https://github.com/agentscope-ai/QwenPaw/pull/5783))
- Cron sessions properly isolated ([#5494](https://github.com/agentscope-ai/QwenPaw/pull/5494))
- Sanitize malformed `tool_call` entries in cron history on load ([#5475](https://github.com/agentscope-ai/QwenPaw/pull/5475))
- Heartbeat timeout now configurable ([#5557](https://github.com/agentscope-ai/QwenPaw/pull/5557))
- Enabled cron jobs can be edited/deleted from the UI ([#5483](https://github.com/agentscope-ai/QwenPaw/pull/5483))

**Security & Governance**

- Fix OFF mode incorrectly triggering approval popups ([#5623](https://github.com/agentscope-ai/QwenPaw/pull/5623))
- Strict mode properly overrides ALLOW rules ([#5682](https://github.com/agentscope-ai/QwenPaw/pull/5682))
- External runner hard-block policy strengthened ([#5642](https://github.com/agentscope-ai/QwenPaw/pull/5642))
- Pet plugin approval hook was missing the `scope` param ([#5830](https://github.com/agentscope-ai/QwenPaw/pull/5830))
- Split `rm` detection/extraction to prevent `${HOME}` bypass ([#5866](https://github.com/agentscope-ai/QwenPaw/pull/5866))
- macOS sandbox config had a missing closing bracket ([#5454](https://github.com/agentscope-ai/QwenPaw/pull/5454))

**Memory**

- Heartbeat/cron requests skip auto-memory and search ([#5592](https://github.com/agentscope-ai/QwenPaw/pull/5592))

**Workspace & Tools**

- Workspace detectors cache key updated ([#5500](https://github.com/agentscope-ai/QwenPaw/pull/5500))
- Fix Playwright memory leak ([#5686](https://github.com/agentscope-ai/QwenPaw/pull/5686))
- Skill market install queue cleared between installs ([#5650](https://github.com/agentscope-ai/QwenPaw/pull/5650))

**Packaging & CI**

- Add `project.urls` to pyproject.toml for GitHub dependency graph ([#5839](https://github.com/agentscope-ai/QwenPaw/pull/5839))
- ACP version pinned to 0.3.x ([#5798](https://github.com/agentscope-ai/QwenPaw/pull/5798))
- Empty `extra_flags` expansion guarded for bash 3.2 on macOS ([#5743](https://github.com/agentscope-ai/QwenPaw/pull/5743))
- Dead bootstrap step removed from Tauri CI ([#5578](https://github.com/agentscope-ai/QwenPaw/pull/5578))
- supervisord restart limit added ([#5613](https://github.com/agentscope-ai/QwenPaw/pull/5613))

## ⚡ Performance

- `SidebarSessionList` virtualized (react-window); polling deduplicated; date formatting cached ([#5643](https://github.com/agentscope-ai/QwenPaw/pull/5643))
- Session switching no longer flashes blank; queue alert flicker resolved ([#5559](https://github.com/agentscope-ai/QwenPaw/pull/5559), [#5683](https://github.com/agentscope-ai/QwenPaw/pull/5683))
- Shutdown tasks parallelized — close time cut by ~40% ([#5469](https://github.com/agentscope-ai/QwenPaw/pull/5469))
- `send_file_to_user` caps file size to prevent memory spikes on large attachments ([#5457](https://github.com/agentscope-ai/QwenPaw/pull/5457))

## 🏗️ Infrastructure

- Desktop verification extracted into reusable composite GitHub Actions ([#5681](https://github.com/agentscope-ai/QwenPaw/pull/5681))
- Docker `init: true` added to prevent zombie accumulation ([#5649](https://github.com/agentscope-ai/QwenPaw/pull/5649))
- QwenPaw review bot auto-comments on PRs with lint/style findings ([#5736](https://github.com/agentscope-ai/QwenPaw/pull/5736))
- Windows nightly CI HTTP timeout raised; per-test hang safeguard added ([#5627](https://github.com/agentscope-ai/QwenPaw/pull/5627))

## 🧪 Testing

- Crons module: 51 unit tests covering scheduling, execution, history ([#5423](https://github.com/agentscope-ai/QwenPaw/pull/5423))
- Chats module: 38 unit tests for session CRUD and message handling ([#5422](https://github.com/agentscope-ai/QwenPaw/pull/5422))
- Frontend M2: Stores + Hooks + Control page component tests ([#5409](https://github.com/agentscope-ai/QwenPaw/pull/5409))
- Frontend M3-A: Agent hooks + Settings page tests ([#5434](https://github.com/agentscope-ai/QwenPaw/pull/5434))
- Frontend M3-B: Inbox + API module tests ([#5438](https://github.com/agentscope-ai/QwenPaw/pull/5438))
- E2E adapted for AgentScope 2.0 ([#5542](https://github.com/agentscope-ai/QwenPaw/pull/5542))
- Desktop e2e verification added to release workflow ([#5428](https://github.com/agentscope-ai/QwenPaw/pull/5428))
- Integration tests parallelized via pytest-xdist; coverage gate hardened ([#5531](https://github.com/agentscope-ai/QwenPaw/pull/5531))
- Channels module unit tests ([#5812](https://github.com/agentscope-ai/QwenPaw/pull/5812))
- Approvals module unit tests ([#5811](https://github.com/agentscope-ai/QwenPaw/pull/5811))
- Inbox module unit tests ([#5809](https://github.com/agentscope-ai/QwenPaw/pull/5809))
- Console contract-guard tests for 12 API modules ([#5807](https://github.com/agentscope-ai/QwenPaw/pull/5807))
- Console hooks + stores unit tests ([#5808](https://github.com/agentscope-ai/QwenPaw/pull/5808))
- Large session regression tests ([#5810](https://github.com/agentscope-ai/QwenPaw/pull/5810))
- Integration tests: Sprint 4.1 covering `/api/tool-calls/*` + `/api/console/chat/task` ([#5895](https://github.com/agentscope-ai/QwenPaw/pull/5895))
- Flaky test stabilization: retry decorators, dead `/api/agent` endpoint removed ([#5507](https://github.com/agentscope-ai/QwenPaw/pull/5507))

## 📖 Docs

- Architecture page added (en + zh) explaining Runtime, Driver, Workspace layers ([#5653](https://github.com/agentscope-ai/QwenPaw/pull/5653))
- TUI documentation page ([#5644](https://github.com/agentscope-ai/QwenPaw/pull/5644))
- Security docs: Sandbox and Access Policy sections ([#5621](https://github.com/agentscope-ai/QwenPaw/pull/5621), [#5678](https://github.com/agentscope-ai/QwenPaw/pull/5678))
- Context management rewritten around scroll + episodic memory model ([#5614](https://github.com/agentscope-ai/QwenPaw/pull/5614), [#5631](https://github.com/agentscope-ai/QwenPaw/pull/5631))
- v1→v2 migration guide for plugin authors ([#5752](https://github.com/agentscope-ai/QwenPaw/pull/5752))
- Roadmap and general docs updated for 2.0 ([#5610](https://github.com/agentscope-ai/QwenPaw/pull/5610), [#5461](https://github.com/agentscope-ai/QwenPaw/pull/5461))
- HiClaw renamed to AgentTeams in all practice guides ([#5636](https://github.com/agentscope-ai/QwenPaw/pull/5636))
- Blog section + SEO improvements on the website ([#5697](https://github.com/agentscope-ai/QwenPaw/pull/5697))
- Langfuse docker deployment docs updated ([#5380](https://github.com/agentscope-ai/QwenPaw/pull/5380))
- Whisper installation notes added to console docs ([#5924](https://github.com/agentscope-ai/QwenPaw/pull/5924))
- Multi-language "tools" translations aligned ([#5516](https://github.com/agentscope-ai/QwenPaw/pull/5516))

## New Contributors

- @ustc-mkh made their first contribution in [#5525](https://github.com/agentscope-ai/QwenPaw/pull/5525)
- @wfeng007 made their first contribution in [#4041](https://github.com/agentscope-ai/QwenPaw/pull/4041)
- @biaobiaobiao108 made their first contribution in [#5533](https://github.com/agentscope-ai/QwenPaw/pull/5533)
- @WujunhaoGPU made their first contribution in [#5651](https://github.com/agentscope-ai/QwenPaw/pull/5651)
- @y1y5 made their first contribution in [#5752](https://github.com/agentscope-ai/QwenPaw/pull/5752)
- @VectorPeak made their first contribution in [#5766](https://github.com/agentscope-ai/QwenPaw/pull/5766)
- @Osamaali313 made their first contribution in [#5792](https://github.com/agentscope-ai/QwenPaw/pull/5792)
- @hehuang139 made their first contribution in [#5799](https://github.com/agentscope-ai/QwenPaw/pull/5799)
- @manjieqi made their first contribution in [#5210](https://github.com/agentscope-ai/QwenPaw/pull/5210)
- @RerankerGuo made their first contribution in [#5924](https://github.com/agentscope-ai/QwenPaw/pull/5924)

**Full Changelog**: [v1.1.12.post3...v2.0.0](https://github.com/agentscope-ai/QwenPaw/compare/v1.1.12.post3...v2.0.0)