Releases: GCWing/BitFun
Releases · GCWing/BitFun
Nightly Build (0.2.11-nightly.20260626+07baddd8)
Automated nightly build from main branch.
Version: 0.2.11-nightly.20260626+07baddd8
Commit: 07baddd
Date: 2026-06-25T22:52:43Z
Warning: Nightly builds are untested and may be unstable.
v0.2.11
更新摘要
本次迭代核心
新增图片查看工具、内联 Skill 引用、自定义 Agent/模式管理,并大幅增强 PPT Live、Flow Chat、终端和后台子代理体验。
关键特性
- [图片查看工具] 新增 view_image 工具,Agent 可读取本地或远程 SSH Workspace 中的图片,并作为多模态输入进行分析。
- [内联 Skill 引用] 输入框支持通过 / 或 $ 插入 [$SkillName],Skill 会以标签形式展示,运行时也会作为结构化 Skill 引用传递给 Agent。
- [自定义 Agent 与自定义模式] 支持创建和管理自定义 mode / subagent,可配置提示词、工具、运行上下文策略和默认会话模式。
- [后台子代理活动面板] Flow Chat 顶部栏和会话导航可展示后台 subagent 活动,支持打开相关会话、取消单个任务或批量停止后台活动。
- [PPT Live 生成增强] PPT Live 支持更清晰的生成流程、子任务进度展示、单轮 cowork 生成、一框式提示引导,并改善长时间生成时的可见反馈。
- [终端体验升级] 支持多行粘贴确认、PowerShell 原生 Ctrl+V,以及“粘贴为单行”选项;同时在 tab、面板切换和窗口 resize 时更好地保留终端内容和光标状态。
- [flashgrep快速检索升级] 内置 flashgrep 升级并适配新协议,改进本地和远程 Workspace 搜索结果映射。
- [本地 HTML 快速打开] 文件菜单和 tab 菜单支持直接打开本地 HTML 文件,方便预览静态页面和 Agent 生成结果。
- [Computer-use 优化] 新增 describe_screen,为 text-only desktop automation 提供屏幕结构描述能力。
- [启动与会话响应优化] 减少启动 IPC、延后 tray 初始化、优化长会话打开和快速切换路径。
- [MiniApp 开发支持] 补充
miniapp-devskill、MiniApp 示例、API 参考和开发指导,提升 MiniApp 开发体验。
关键修复
- 修复 429 / TPM rate limit 场景下可能连续重复重试的问题,减少限流后的卡顿和资源浪费。
- 优化 Flow Chat 重试和异常结束时的消息展示,减少重复内容、残留状态和会话恢复后的混乱记录。
- 修复 Flow Chat 在流式输出、思考内容、文件卡片折叠和长 diff 内容场景下的滚动异常、文本截断和视口跳动问题。
- 修复打开或恢复历史会话时可能出现加载不完整、会话激活不稳定、旧的流式内容残留等问题。
- 修复终端在切换标签页、调整侧边栏/底部面板、窗口尺寸变化时可能出现空白、错位、内容丢失或重复绘制的问题。
- 修复 Windows 平台部分输入场景可能导致桌面端卡死的问题。
- 修复部分 OpenAI 响应格式解析不准确的问题,减少异常内容导致的消息展示或工具调用错误。
- 修复 Agent 流式输出中结束原因和工具参数预览可能重复或显示不准确的问题。
- 优化远程 Workspace、ControlHub 和 remote SSH 场景下的上下文提示,减少 Agent 在远程环境中误判路径、主机或执行位置的问题。
- 修正 relay server 启动脚本和文档说明不一致的问题,使默认监听地址行为更明确。
- 修复移动端会话列表初始化时 loading 状态闪烁的问题。
- 修复 PPT Live 在不同屏幕尺寸下的布局问题,并改进 editable PPTX 导出和 HTML 文本导出的稳定性。
- 修复编辑 persona 文档时可能破坏
IDENTITY.mdfrontmatter 的问题。 - 修复上下文压缩后部分提示结构未及时刷新的问题,减少后续对话指令不完整的情况。
- 优化自定义 Agent 发现和 Skill 输出格式,减少无效、重复或不一致的 Skill 信息。
What's Changed
Release Focus
This release adds image inspection, inline Skill references, custom Agent and mode management, and major improvements to PPT Live, Flow Chat, terminal behavior, and background subagent workflows.
Key Features
- [Image inspection] Added the
view_imagetool, allowing agents to read images from local paths or remote SSH Workspaces and analyze them as multimodal input. - [Inline Skill references] The chat input now supports inserting
[$SkillName]through/or$, with Skills shown as inline tags and passed to runtime as structured Skill references. - [Custom Agents and custom modes] Added support for creating and managing custom modes and subagents, with configurable prompts, tools, runtime context policies, and default chat mode behavior.
- [Background subagent activity panel] Flow Chat header and session navigation now show background subagent activity, with controls to open related sessions, cancel individual tasks, or stop background work in
bulk. - [PPT Live generation improvements] PPT Live now has a clearer generation flow, subtask progress visibility, single-turn cowork generation, one-box prompt guidance, and better feedback during long-running
generation. - [Terminal experience upgrade] Added multiline paste confirmation, native PowerShell Ctrl+V support, and a “paste as one line” option. Terminal content and cursor state are also preserved more reliably across
tabs, panels, and window resizing. - [flashgrep fast search upgrade] Updated the bundled flashgrep integration and adapted it to the latest protocol, improving local and remote Workspace search result mapping.
- [Open local HTML files] File and tab menus can now open local HTML files directly, making it easier to preview static pages and agent-generated outputs.
- [Computer-use improvements] Added
describe_screen, providing screen-structure descriptions for text-only desktop automation. - [Startup and session responsiveness] Reduced startup IPC, deferred tray initialization, and improved long-session opening and rapid session switching.
- [MiniApp development support] Added
miniapp-devskill content, MiniApp examples, API references, and development guidance to improve MiniApp development.
Key Fixes
- Fixed repeated retries under 429 / TPM rate-limit scenarios, reducing stalls and wasted requests after provider throttling.
- Improved Flow Chat message display after retries and abnormal endings, reducing duplicate content, leftover state, and confusing restored history.
- Fixed Flow Chat scrolling, text truncation, and viewport jumps during streaming output, thinking content, file-card collapse, and long diff rendering.
- Fixed issues when opening or restoring historical sessions, including incomplete loading, unstable session activation, and lingering old streaming content.
- Fixed terminal blank screens, cursor drift, missing content, and duplicate redraws when switching tabs, resizing side/bottom panels, or changing window size.
- Fixed a Windows desktop input issue that could cause the app to hang in certain scenarios.
- Fixed inaccurate parsing of some OpenAI response formats, reducing message rendering and tool-call errors caused by malformed content.
- Fixed inaccurate or duplicate display of stream ending reasons and tool-argument previews.
- Improved remote Workspace, ControlHub, and remote SSH context prompts so agents are less likely to confuse paths, hosts, or execution locations.
- Fixed mismatches between relay server startup scripts and documentation, making the default bind address behavior clearer.
- Fixed loading-state flicker when initializing the mobile session list.
- Fixed PPT Live layout issues across screen sizes and improved editable PPTX export and HTML text export stability.
- Fixed persona document editing so
IDENTITY.mdfrontmatter is preserved. - Fixed prompt scaffolding refresh after context compression, reducing cases where later instructions could be incomplete.
- Improved custom Agent discovery and Skill output formatting to reduce invalid, duplicate, or inconsistent Skill information.
v0.2.10
What's Changed
- refactor(agent-runtime): own prompt loop facts by @limityan in #1030
- refactor(agent-runtime): own event facts by @limityan in #1033
- fix(flow-chat): keep streaming code preview tail visible by @limityan in #1037
- refactor(agent-runtime): own prompt cache and registry facts by @limityan in #1036
- fix: improve cache baseline reuse across derived agent sessions by @wsp1911 in #1040
- feat(web-ui): add reveal-in-explorer action to plan viewer by @wsp1911 in #1042
- refactor(runtime): close product and tool owners by @limityan in #1041
- feat(web-ui): show tool descriptions in agent editor by @wsp1911 in #1047
- feat(acp): follow up session update UI work by @wgqqqqq in #1056
- fix(web-ui): keep toolbar window within desktop bounds by @wgqqqqq in #1054
- refactor(runtime): migrate product capability registry owners by @limityan in #1048
- perf(web-ui): reduce post-interactive workspace and session pressure by @limityan in #1032
- refactor(runtime): migrate scheduler owner decisions by @limityan in #1060
- fix(acp): isolate provider slash commands by @wgqqqqq in #1059
- refactor(runtime): migrate session store and tool IO owners by @limityan in #1062
- feat: unify scheduled job management across sessions and workspaces by @wsp1911 in #1064
- refactor(runtime): close function-agent and scheduled job owners by @limityan in #1065
- refactor: close agent runtime extension owners by @limityan in #1067
- refactor(agent-runtime): close delivery contract owners by @limityan in #1068
- refactor(runtime): move function-agent git service owner by @limityan in #1070
- refactor(runtime): close product assembly service boundary by @limityan in #1071
- fix(model-selection): make model selection per-session instead of global by @kev1n77 in #1069
- Fix ACP session recovery by @wgqqqqq in #1052
- refactor(tool-runtime): move remote tool execution helpers by @limityan in #1072
- refactor(agent-runtime): move lifecycle decisions out of core by @limityan in #1073
- refactor: move MiniApp host plans and developer tooling by @limityan in #1075
- perf(flow-chat): stabilize historical session rendering by @limityan in #1077
- refactor(core): layer crates and close product assembly facade by @limityan in #1076
- refactor(core): organize crates into six ownership layers by @limityan in #1081
- perf(web-ui): defer inactive session metadata startup by @limityan in #1082
- refactor(stream): move stream contracts out of adapters by @limityan in #1083
- fix(chat): prevent scroll compensation ratchet causing huge white space by @limityan in #1084
- fix(session-archive): fix stale expand toggle after archiving sessions add collapsible archived groups with full path by @kev1n77 in #1074
- feat: add /reload-skills slash command (CLI + Web UI) by @jarvis24young in #1057
- fix(subagent): sync subagent session context window from AI config by @kev1n77 in #1086
- perf: reduce unnecessary re-renders and serialization in nav and logging by @kev1n77 in #1087
- feat: add ExecCommand session tooling and background command controls by @wsp1911 in #1090
- fix(flow-chat): switch to multi-line input when target switcher is shown by @kev1n77 in #1092
- feat(terminal): support configurable terminal panel placement by @wsp1911 in #1094
- perf: replace full-state subscriptions with selector-based subscriptions by @kev1n77 in #1093
- refactor(core): remove project context service and stale locale keys by @wsp1911 in #1096
- refactor(core): move remote and shell execution owners by @limityan in #1097
- refactor(core): move workspace search owner by @limityan in #1098
- fix: hide Windows pipe exec console window by @wsp1911 in #1099
- feat(controlhub): enhance controlhub browser by @kev1n77 in #1100
- refactor(core): migrate remaining search runtime policies by @limityan in #1102
- fix: refine markdown file links and remote artifact delivery by @wsp1911 in #1104
- fix(web-ui): tighten chat slash-command matching by @wsp1911 in #1106
- refactor(core): move DeepReview runtime policy owner by @limityan in #1105
- perf: guard repeated scans and cache stable computations in ModernFlowChatContainer by @kev1n77 in #1107
- perf(flow-chat): reduce historical session hydration stalls by @limityan in #1108
- fix(flow-chat): preserve history window expansion scroll by @limityan in #1109
- feat(exec-command): honor configured local shell with Windows fallback by @wsp1911 in #1111
- perf: reduce RAF callback volume in VirtualMessageList event system by @kev1n77 in #1110
- refactor(core): move M10 runtime decisions into agent-runtime by @limityan in #1112
- refactor(core): migrate concrete runtime owners to services by @limityan in #1113
- refactor: refine agent runtime context handling by @wsp1911 in #1115
- refactor(core): move stable AI DTOs into core types by @limityan in #1119
- fix(flow-chat): prevent white screen on input collapse and blank viewport on streaming switch by @kev1n77 in #1095
- fix(remote-connect): prevent rustls CryptoProvider panic on reconnect by @ococl in #1101
- fix: clarify session usage timing diagnostics by @limityan in #1121
- refactor(core): migrate DeepReview capacity decisions to runtime by @limityan in #1123
- fix(web-ui): improve terminal output copy and scrolling by @wsp1911 in #1122
- fix: reserve left space for macOS traffic lights in ToolbarMode floating window by @nonoqing in #1125
- fix: stabilize latest tail anchoring for long sessions by @limityan in #1126
- refactor(prompt): restructure runtime context and add PowerShell exec guidance by @wsp1911 in #1127
- fix(flow-chat): preserve explicit session selection during restore by @limityan in #1129
- refactor(core): migrate MiniApp manager workflow facade by @limityan in #1128
- feat: add Write append mode and polish exec tool UX by @wsp1911 in #1131
- refactor(tools): simplify collapsed tool prompts and stub schema by @wsp1911 in #1135
- perf(flow-chat): improve long session readiness by @limityan in #1134
- feat(web-ui): add dedicated cards for WebFetch and GetToolSpec by @wsp1911 in #1139
- docs: add SDLC harness design docs by @limityan in #1137
- refactor(core): migrate DeepReview runtime ownership by @limityan in #1132
- docs: clarify lifecycle quality architecture by @limityan in #1140
- refactor(tool-contracts): tighten collapsed tool GetToolSpec contracts by @wsp1911 in #1142
- docs: refine SDLC harness design contracts by @limityan in #1141
- refactor: move MiniApp import and tool pipeline policies by @limityan in #1143
- fix(flow-chat): preserve session bottom anchor on resize by @limityan in #1144
- feat(mobile-web): add resume card for quick session continue by @jarvis24young in #937
- fix(flow-chat): prevent stale history projection handoff by @limityan in #1145
- docs: reframe SDLC harness as configurable governance by @limityan in #1146
- refactor: move session storage primitives to services core by @limityan in #1147
- refactor: improve web...
v0.2.9
What's Changed
- refactor(runtime): move remote response assembly to integrations by @limityan in #875
- refactor(runtime): close tool and product-domain owner slices by @limityan in #876
- feat: add create workspace action in chat dropdown by @kev1n77 in #865
- feat(mobile-web): add long-press message actions — copy, resend, delete by @jarvis24young in #863
- fix: refine multitask mode prompt by @wsp1911 in #878
- fix: @ file mention picker cannot be closed by clicking outside by @kev1n77 in #879
- test(agent-stream): add timeouts to stream fixture harness by @legenddcr in #880
- refactor(runtime): centralize service agent adapters by @limityan in #877
- [codex] Improve goal mode guidance by @bobleer in #883
- refactor(runtime): move tool context owner boundary by @limityan in #884
- refactor(core): split tool prompt reminders and rename request context to user context by @wsp1911 in #882
- feat(core): add session-scoped prompt cache with lazy restore by @wsp1911 in #886
- fix(ai-adapters): strip tool_choice when no OpenAI tools are attached by @bobleer in #889
- refactor(miniapp): move runtime decisions to product domains by @limityan in #885
- fix: avoid utf8 panic in string preview truncation by @wsp1911 in #890
- refactor(core): scope prompt cache reuse by prompt and context identity by @wsp1911 in #892
- refactor: move local filesystem owner to services-core by @limityan in #891
- feat(agentic): include written content in Write tool results by @bobleer in #897
- fix(agentic): bind Task subagent guidance to available_agents listing by @bobleer in #899
- fix(agentic): hide Task tool when no subagents are available by @bobleer in #900
- fix(ai): improve Anthropic reasoning, prefill trimming, and spotlight shortcuts by @bobleer in #901
- fix(agentic): reject tool-invocation syntax in Write content generation by @bobleer in #902
- fix(agentic): improve Write two-step content generation reliability by @bobleer in #903
- feat(core): add forked subagent execution with prompt cache reuse by @wsp1911 in #904
- refactor(core): switch context compression to full-prefix fork-like summary requests by @wsp1911 in #906
- fix(acp): drain turn updates before completion by @wgqqqqq in #896
- fix(mcp): complete stdio initialization handshake by @wgqqqqq in #894
- fix(cli): increase thread stack to 16 MiB to prevent stack overflow on Windows by @kev1n77 in #893
- fix(cli): wrap cursor navigation in selector panels by @kev1n77 in #895
- fix: improve AskUserQuestion resilience against stream truncation and duplicate cards by @harryfan1985 in #905
- fix(agentic): flush persisted tool result before returning by @harryfan1985 in #881
- fix(mcp): prevent handshake hang when local stdio server exits early by @harryfan1985 in #888
- fix(mcp): scope stdio handshake timeout by @wgqqqqq in #909
- ci(cli): build bitfun-cli per release and notify homebrew-tap by @harryfan1985 in #851
- feat(mobile-web): add session search, rename, and delete by @jarvis24young in #840
- perf(observability): add prefix-cache telemetry metrics by @wsp1911 in #911
- fix(bot): avoid duplicate IM menu items by @bobleer in #912
- fix: avoid session agent_type switches when running /init by @wsp1911 in #915
- chore(scripts): add changed-rust formatter and document fmt:rs workflow by @wsp1911 in #916
- refactor(web-ui): capsule/multi-line chat input auto layout by @GCWing in #913
- feat: add desktop crash diagnostics by @wgqqqqq in #914
- feat: support session archive and restore or delete archived sessions by @kev1n77 in #917
- Fix web UI test failures by @wgqqqqq in #918
- fix(ci): create release after desktop packaging by @wgqqqqq in #919
- refactor(runtime-ports): extract dialog and subagent contracts by @limityan in #924
- refactor(core): preserve prompt cache across coding mode switches by @wsp1911 in #925
- feat(chat): add prompt cache guard for mode switching by @wsp1911 in #927
- fix(web-ui): keep plan viewer dirty state in sync with editor save baseline by @wsp1911 in #928
- fix(compression): allow context compression for single-turn sessions and oversized summary requests by @wsp1911 in #929
- refactor: unify shared coding agent profiles by @wsp1911 in #930
- fix(agentic): anchor goal verification to initial session goal by @bobleer in #932
- feat(mobile-web): merge mobile stability PRs by @bobleer in #933
- refactor(runtime-ports): complete HR-C contract closure by @limityan in #935
- refactor(remote-connect): move runtime presentation policies by @limityan in #938
- feat(mobile-web): add connection health indicator by @jarvis24young in #922
- fix(ai-adapters): infer anthropic adaptive thinking by claude model capability by @wsp1911 in #936
- refactor(remote-connect): move service agent runtime assembly by @limityan in #941
- refactor(runtime-ports): move scheduler queue decisions by @limityan in #942
- ci: strengthen contribution governance by @limityan in #939
- docs: simplify contribution templates by @limityan in #945
- refactor(remote-connect): migrate command orchestration owner by @limityan in #946
- feat: add shared i18n contract guardrails by @limityan in #948
- perf(startup): reduce desktop first-screen work by @limityan in #934
- feat: refine i18n loading guardrails by @limityan in #950
- refactor: harden i18n audit coverage by @limityan in #952
- perf(flow-chat): smooth large history session rendering by @limityan in #953
- [chore] Modernize Rust dependency management by @limityan in #954
- fix(debug-log): prevent ingest route startup panic by @limityan in #956
- perf(flow-chat): render recent session history first by @limityan in #958
- refactor(i18n): harden locale audit guardrails by @limityan in #957
- refactor: add i18n governance report by @limityan in #961
- refactor: enforce i18n governance baseline by @limityan in #962
- perf(startup): add release timing harness by @limityan in #966
- feat(agentic): route PlaintextFollowup Write through post-write Read by @bobleer in #969
- docs(architecture): clarify runtime decomposition layers by @limityan in #964
- fix(auto-update): align NSIS publisher with installer by @kev1n77 in #965
- refactor: reuse shared i18n terms across surfaces by @limityan in #968
- refactor(i18n): tighten shared terms and dynamic key proofs by @limityan in #972
- refactor: add runtime services foundation by @limityan in #975
- feat: show working directory hint in terminal list items by @kev1n77 in #973
- refactor: add runtime owner boundaries by @limityan in #978
- refactor(i18n): harden locale debt governance by @limityan in #977
- perf(startup): defer editor and AI initialization by @limityan in #971
- feat(acp): add Oh My Pi (omp) as a built-in ACP agent by @harryfan1985 in #963
- feat(agentic): stabilize prompt cache reuse across skill and subagent updates by @wsp1911 in #985
- fix(core): handle bare markdown fences by ...
v0.2.8
What's Changed
- fix(desktop): agent companion pet resize safety and Windows UX by @bobleer in #687
- fix(cli): initialize terminal service during startup & fix user question and model name display by @kev1n77 in #686
- Refactor MCP auth and config helpers into services-integrations by @limityan in #688
- Stabilize Deep Review recovery and capacity waits by @limityan in #689
- refactor(mcp): move remote protocol helpers to integrations by @limityan in #691
- fix(agentic): guard Write against overwrite, allow loop recovery, and harden Write content generation by @bobleer in #690
- Fix web UI serialization guards for streamed params and pet activity by @limityan in #693
- Harden core decomposition service boundaries by @limityan in #694
- Refine Deep Review action bar UX by @limityan in #692
- chore(search): isolate flashgrep logs and bump bundle v0.2.7 by @wgqqqqq in #685
- Move integration runtime helpers out of core by @limityan in #695
- fix(desktop): suppress context menu on agent companion pet, reorder AI model config form fields by @bobleer in #697
- fix(core): add logs for stuck subagent task completion by @wsp1911 in #700
- Add backend diagnostic log redactor by @limityan in #698
- Refactor MCP runtime ownership into services integrations by @limityan in #699
- fix(remote): route Git commands through SSH for remote workspaces by @kev1n77 in #696
- Add backend diagnostic log redactor by @limityan in #703
- Refactor remote connect product surface contracts by @limityan in #704
- Improve Deep Review launch strategy and config handling by @limityan in #705
- Improve Deep Review completion and companion routing by @limityan in #707
- refactor(core): move tool registry contracts into agent-tools by @limityan in #708
- [fix] Keep Flow Chat file change counts session-scoped by @limityan in #709
- feat(ssh): stable connection IDs and session hydration after port changes by @bobleer in #710
- fix(config): tolerate null mode config entries by @wgqqqqq in #706
- feat: add on-demand tool spec discovery by @wsp1911 in #711
- fix(web-ui): shortcut digit lookup, overrides, editor canvas inheritance by @bobleer in #712
- refactor(product-domains): externalize guarded domain helpers by @limityan in #713
- Add desktop system fullscreen mode by @limityan in #714
- fix(web-ui): new session workspace and session AI/pet config layout by @bobleer in #716
- feat(deep-research): consolidate to pure 6-phase pipeline with citati… by @nonoqing in #718
- Fix remote workspace compatibility guards by @bobleer in #719
- feat(agents): support per-mode subagent availability management by @wsp1911 in #720
- feat(config): default agent companion pet to panda preset by @bobleer in #721
- Fix ACP tool diff fallback and search result rendering by @wgqqqqq in #715
- fix acp startup timeout by @wgqqqqq in #722
- chore(boundaries): guard latest-main runtime anchors by @limityan in #717
- fix file tool reliability by @bobleer in #724
- refactor(core): close services product owner boundaries by @limityan in #725
- test(core): add semantic regression baselines by @limityan in #726
- refactor(runtime): add remote runtime port baselines by @limityan in #727
- refactor(remote-connect): move tracker state to integrations by @limityan in #729
- feat(update): download daily app updates in background by @kev1n77 in #723
- test(remote-connect): add runtime migration snapshots by @limityan in #732
- fix(web-ui): drop noisy session model auto-migration toasts by @bobleer in #733
- bundle linux flashgrep in windows builds by @wgqqqqq in #728
- refactor(remote-connect): move runtime policy helpers by @limityan in #735
- fix(flow-chat): prevent card toggle while selecting text by @kev1n77 in #737
- feat(cli): expose ACP server and external agent controls & prevent ACP write turns from hanging by @kev1n77 in #731
- feat(miniapp): add local customization workflow by @limityan in #738
- refactor(remote-connect): move wire poll contracts by @limityan in #739
- refactor(product-domains): move pure product parsing policies by @limityan in #740
- refactor(remote-connect): move tracker registry ownership by @limityan in #741
- feat(miniapp): add built-in PR review inbox by @limityan in #742
- fix(miniapp): refine PR review inbox behavior by @limityan in #743
- perf(web): reduce startup bundle pressure for optional runtimes by @limityan in #746
- locale: fix Nursery->Assistant, Office Session->Cowork Session by @harryfan1985 in #745
- Optimize agent loop prompts and tool guidance by @bobleer in #747
- Refine agent loop guidance by @bobleer in #748
- feat(web-ui): IME-safe Escape, explore grouping, DeepSeek reasoning UI by @bobleer in #752
- feat(cli): add mode-aware skill & subagent management menu by @kev1n77 in #751
- feat(cli): redesign built-in themes and startup theme switching by @kev1n77 in #753
- Improve startup and historical session readiness by @limityan in #754
- fix(agent): Edit tool hardening and flow chat code block UI by @bobleer in #755
- Reduce startup background contention by @limityan in #756
- chore: web typography tokens and Rust formatting sweep by @bobleer in #758
- fix(web-ui): stabilize flow chat follow-output during collapses by @bobleer in #759
- perf(session): reduce startup and historical session overhead by @limityan in #760
- feat(agentic): add Multitask mode and improve subagent background delivery by @wsp1911 in #762
- chore(boundaries): refresh latest-main guardrails by @limityan in #764
- feat(miniapp): remember declined built-in updates by @limityan in #765
- refactor(product-domains): extend MiniApp runtime readiness by @limityan in #767
- Fix ACP file path resolution for completed tool cards by @wgqqqqq in #763
- fix(core): avoid session auto-save lock contention by @limityan in #769
- refactor(product-domains): extract runtime baseline helpers by @limityan in #768
- refactor(agent-tools): extract tool manifest contracts by @limityan in #770
- feat(miniapp): refine built-in PR review workflow by @limityan in #771
- refactor(agent-tools): prepare tool provider ownership by @limityan in #773
- token usage by @nonoqing in #774
- Reuse global ACP config for remote agents by @wgqqqqq in #766
- fix(ai-adapters): send flat tool schema for OpenAI Responses API by @kev1n77 in #772
- fix(desktop): default close button to minimize to system tray by @bobleer in #776
- fix: strip tool history from Write content-generation context by @bobleer in #777
- fix(agentic): allow local tool paths outside workspace root by @bobleer in #778
- refactor(product-domains): add runtime facade guardrails by @limityan in #779
- refactor(function-agents): wire commit facade path by @limityan in #781
- refactor(function-agents): route startchat through facade by @limityan in #782
- token usage by @nonoqing in https://github.com/GCWing/Bi...
v0.2.7
What's Changed
- fix(web-ui): poll cursor position for companion pet hover by @bobleer in #628
- fix(ai): preserve tool argument replay stability to avoid prompt caching misses by @wsp1911 in #627
- feat(web-ui): unify tool card status slot and processing indicator by @GCWing in #629
- feat(web-ui): refine TodoWrite card and compact header action by @GCWing in #630
- Add session
/usagereport with desktop UI and runtime metrics by @limityan in #631 - feat: enhance browser control, screenshot, and tool system by @bobleer in #632
- Fix agent runtime concurrency and cancellation handling by @bobleer in #633
- fix: subagent timeout disable button not working due to serde format mismatch by @bobleer in #634
- fix(web-ui): handle failed dialog completion in flow chat and notifications by @bobleer in #636
- feat(web-ui): flow chat polish, workspace strip, quick actions settings by @GCWing in #637
- feat: flow chat polish and quick actions (web-ui + core config) by @GCWing in #638
- fix(desktop): main window close on non-macOS and Windows process cleanup by @GCWing in #639
- fix(browser-control): accurate tab count, real browser detection, hide selector when CDP connected by @bobleer in #640
- feat(installer): add install detection and uninstaller launch support by @kev1n77 in #642
- support ACP permission options by @wgqqqqq in #641
- fix(flow-chat): prevent bottom spacer inflation by @bobleer in #644
- fix(flow-chat): align bottom scroll position by @bobleer in #647
- fix(flow-chat): stop follow output jitter by @bobleer in #648
- refactor: remove AI rules and memory modules by @wsp1911 in #646
- feat(ai-adapters,agentic): recover Write tool args truncated by max_t… by @nonoqing in #643
- feat(desktop,web-ui,installer): Tokyo Night theme, tray, and settings/usage by @GCWing in #649
- feat(web-ui): align Skills scene UI with Sparo-style layout by @GCWing in #650
- fix(agentic): make truncated Write recovery actionable end-to-end by @nonoqing in #653
- Improve session usage reporting and quick action localization by @limityan in #655
- refactor(core): finalize P1 decomposition safeguards by @limityan in #635
- fix(search): support remote flashgrep workspace search by @wgqqqqq in #652
- fix: markdown editor live refresh by @wgqqqqq in #654
- feat(skills): add suite mode tab reset and grouped visibility controls by @wsp1911 in #656
- feat(deep-review): add adaptive review runtime and controls by @limityan in #651
- feat(deep-review): add adaptive review runtime and orchestration by @limityan in #660
- feat(usage): complete session usage report by @limityan in #661
- Refactor core owner crate boundaries by @limityan in #662
- feat(deep-review): add scoped evidence and report surfaces by @limityan in #663
- fix(web-ui): open Write/Edit cards with resolved file paths by @kev1n77 in #658
- fix(web-ui): agent companion pet tap vs drag; persist turns only on quit by @GCWing in #665
- feat(agent): plaintext Write synthesis and flow-chat UI polish by @bobleer in #664
- feat(deep-review): harden queued reviewer execution and action workflows by @limityan in #666
- refactor(mcp): align tool naming with metadata-driven server resolution by @wsp1911 in #659
- fix(web-ui): stabilize review and ACP settings flows by @limityan in #668
- fix(web-ui): clarify Deep Review queue waits by @limityan in #669
- fix: align MCP tool metadata fields and remove unused tool APIs by @wsp1911 in #670
- feat(agentic,web-ui): task tool UI, explore region, and session event wiring by @GCWing in #672
- Sync local fixes: stream dedup, search remote/fallback, session rollback persistence, companion typewriter by @bobleer in #674
- feat(cli): update cli by @kev1n77 in #677
- [codex] Clarify Deep Review partial reviewer labels by @limityan in #675
- feat(agentic): add mode-scoped subagent visibility by @wsp1911 in #678
- refactor(core): move service integrations to owner crate by @limityan in #680
- fix(agent): harden Anthropic streaming, retries, and execution paths by @bobleer in #681
- Support ACP clients in remote workspaces and harden startup handling by @wgqqqqq in #679
- fix(core): stabilize task and skill prompt item ordering by @wsp1911 in #682
- Refactor core owner contracts for Remote SSH and MCP by @limityan in #684
- fix(cli): bind sessions and workspace to current folder & remove redundant preset themes by @kev1n77 in #683
- bump version to 0.2.7 by @wgqqqqq in #657
Full Changelog: v0.2.6...v0.2.7
v0.2.6
What's Changed
- fix(web-ui, git_tool, flow-chat): scrollbar visibility, git subcommand deduplication, tool card styling, and runtime hints by @limityan in #578
- Add ACP agent sessions with isolated client processes by @wgqqqqq in #579
- fix(skill-market): enable downloading skills to local user path while… by @kev1n77 in #580
- feat: increase max dialog rounds, web search summary, and bug fixes by @limityan in #583
- Improve flashgrep lifecycle and workspace activation by @wgqqqqq in #582
- feat(ui): Flow Chat UX polish, Deep Review action bar enhancements, and tool performance improvements by @limityan in #584
- feat(desktop,web-ui): Petdex agent companion import, presets, and delete by @bobleer in #587
- [codex] Add Agent companion desktop pet by @bobleer in #589
- fix(web-ui): restore inline ACP file approval actions by @wgqqqqq in #588
- [codex] Add desktop pet interaction moods by @bobleer in #590
- [codex] Refine desktop pet task bubbles by @bobleer in #591
- feat(snapshot,web-ui,desktop): snapshot plumbing, pet roster, session file UI by @bobleer in #593
- Add gated workspace search and multi-arch flashgrep resources by @wgqqqqq in #592
- feat(desktop): agent companion pet dynamic window sizing by @bobleer in #597
- fix(desktop): tighten companion pet bubble width measurement by @bobleer in #598
- fix: make tool loop detection signatures include args hash by @kev1n77 in #596
- fix(web-ui): NavPanel workspace list card layout (non-sticky) by @GCWing in #599
- fix: avoid windows startup console flicker by @wgqqqqq in #600
- fix(i18n): sync missing/extra keys across en-US, zh-CN, zh-TW locales by @limityan in #601
- feat(desktop,core): remote workspace tool context and snapshot/stream fixes by @bobleer in #602
- fix(web-ui): tighten compact tool card headers for file ops by @GCWing in #603
- fix(web-ui): flow chat compact tool card dense headers by @GCWing in #604
- feat(ai): support DeepSeek reasoning effort and preserve empty reasoning replay by @wsp1911 in #605
- fix(claw): add back placeholders in the claw mode system prompt by @wsp1911 in #606
- feat(agentic,web-ui): user steering with pending queue and agent runtime refactor by @bobleer in #610
- fix: suppress windows helper command popups by @wgqqqqq in #609
- fix: correct remote SSH command completion handling by @wsp1911 in #608
- feat(core, desktop): remote workspace identity for tools and subagents by @bobleer in #611
- feat(web-ui): appearance settings tab, nav layout, chat input styles by @GCWing in #612
- fix(core, web-ui): session finishing sync, explicit skills, companion activity by @bobleer in #614
- chore(core): bump builtin miniapp bundle versions by @bobleer in #616
- fix: flashgrep lifecycle cleanup by @wgqqqqq in #615
- chore(core, desktop, web-ui): remove Mermaid interactive editor by @GCWing in #617
- fix(web-ui): nav footer browser shortcut and insights in menu by @GCWing in #618
- feat(agentic,ai-adapters,web-ui): harden streaming tool arguments and pipeline by @bobleer in #621
- refactor(skills): isolate builtin skills and centralize mode skill resolution by @wsp1911 in #620
- Fix companion activity ordering and permission prompts by @bobleer in #623
- feat(web-ui): RAF-coalesce app state; polish flow-chat tool cards by @GCWing in #622
- feat(auto-update): add auto-update by @kev1n77 in #624
- feat(deep-research): add Pro mode 6-phase pipeline and ResearchSpecia… by @nonoqing in #613
- Fix Deep Review interruption recovery and reviewer continuation behavior by @limityan in #625
- chore: bump version to 0.2.6 by @wgqqqqq in #626
New Contributors
Full Changelog: v0.2.5...v0.2.6
v0.2.6-test.2
chore: bump test version to v0.2.6-test.2
v0.2.6-test.1
chore: bump test version to v0.2.6-test.1
v0.2.5
What's Changed
- feat(ai-adapters): add Anthropic extended thinking support in stream … by @limityan in #507
- fix: reset chat input mode after assistant sessions by @bobleer in #508
- fix(web-ui): clarify terminal tool execution phases by @wsp1911 in #509
- feat(agent): Computer Use mode, control tooling, and Feishu bot docs by @bobleer in #511
- feat(SessionControl): support session cancellation and refine cancel UX by @wsp1911 in #512
- fix(ui): remove dense dividers in todo list and fix window state sync by @limityan in #515
- chore(agentic): sync fork agent foundation from agentic_os by @wsp1911 in #516
- Yanzhn/deepreview by @limityan in #517
- Yanzhn/fix by @limityan in #521
- Add review team settings and remediation controls by @limityan in #522
- fix(flow-chat): improve deep review interruption UX by @limityan in #523
- fix(deep-review): honor reviewer strategy models by @limityan in #524
- refactor(core): align stream watchdog with configured idle timeout by @wsp1911 in #525
- Yanzhn/deepreview by @limityan in #526
- Deep Review UX improvements and flow-chat terminal card grouping by @limityan in #528
- FlowChat Interaction Improvements & Subagent Timeout Control by @limityan in #529
- feat(flow-chat): increase expanded preview max height for file operat… by @limityan in #530
- Session attention indicator, scroll fix, i18n terminology unification, and large-screen layout adjustments by @limityan in #532
- fix: skip unread marking for user-cancelled sessions by @limityan in #534
- docs: update AGENTS.md with relay-server, release-fast, Windows OpenS… by @limityan in #535
- improvements to Deep Review UX, desktop devtools, and session management fixes. by @limityan in #537
- feat(agentic): harden tool pipeline, execution, and SSE streaming by @bobleer in #538
- refactor(web-ui): flow chat terminal tool cards and UI refresh by @GCWing in #540
- feat(web-ui): file operation tool card expand and open-in-editor by @GCWing in #541
- hide empty subagent container to prevent white gap by @limityan in #539
- feat(web-ui): editor focus shortcuts, flow chat polish, session ordering by @GCWing in #542
- fix(web-ui): tool card header inset for trailing controls by @GCWing in #543
- docs(agentic): streamline builtin gstack skills and expand team mode prompt by @bobleer in #544
- Add dedicated ACP server runtime with model selection support by @wgqqqqq in #545
- fix(snapshot): handle stale dedup references and empty baselines by @bobleer in #549
- fix(web-ui): align subagent items padding with tool card header by @bobleer in #550
- Deep Review UX Improvements & Session Stability Fixes by @limityan in #547
- fix: session index completeness, ReviewTeamPage layout, and tool card refactor by @limityan in #551
- feat(chat): add session branching and make /btw child sessions transient by @wsp1911 in #552
- Expand Deep Review with Architecture and Frontend Reviewers by @limityan in #553
- feat(agents, skills): redesign AgentTeamCard, localize agent descriptions, and add skill shadowing indicators by @limityan in #554
- feat(web-ui): Polish review team page and session config layout by @GCWing in #555
- fix(deep-review): subagent ignore-timeout, session recovery, UI polish & frontend reviewer generalization by @limityan in #556
- feat(skills, agents): rename shadowed badge and localize agent descriptions by @limityan in #558
- test: normalize SSE fixture endings and gate raw SSE logging by @wsp1911 in #561
- ci: run bitfun-core tests on linux macos and windows by @wsp1911 in #562
- feat(nav-panel): add sticky absorption to workspace cards by @limityan in #563
- feat(deep-review): enhance error and interruption experience by @limityan in #564
- fix: modify display effect of deep review timeout by @limityan in #566
- Fix NavPanel layout, theme sync, and AI init error messaging by @limityan in #568
- Polish Flow Chat spacing, links, and terminal tool controls by @limityan in #570
- Add flashgrep-backed workspace search by @wgqqqqq in #571
- docs(readme): document flashgrep + ripgrep search gains by @GCWing in #573
- feat(web-ui): workspace flashgrep index modal and row layout by @GCWing in #575
- fix(search): honor workspace max file size for flashgrep by @wgqqqqq in #576
- Improve Flow Chat runtime feedback, tool cards, and Deep Review actions by @limityan in #574
- Bump version to 0.2.5 by @wgqqqqq in #577
Full Changelog: v0.2.4...v0.2.5