Download Latest Version Kun-0.2.15-mac-arm64.zip (212.7 MB)
Email in envelope

Get an email when there's a new version of Kun

Home / v0.2.14
Name Modified Size InfoDownloads / Week
Parent folder
Kun-0.2.14-win-x64.exe 2026-06-18 165.9 MB
latest.yml 2026-06-18 342 Bytes
Kun-0.2.14-mac-arm64.dmg.blockmap 2026-06-18 203.7 kB
Kun-0.2.14-mac-x64.zip 2026-06-18 206.9 MB
Kun-0.2.14-mac-arm64.zip 2026-06-18 200.3 MB
Kun-0.2.14-win-x64.exe.blockmap 2026-06-18 160.1 kB
Kun-0.2.14-linux-x86_64.AppImage 2026-06-18 188.4 MB
Kun-0.2.14-mac-arm64.dmg 2026-06-18 194.4 MB
Kun-0.2.14-mac-x64.dmg 2026-06-18 200.9 MB
Kun-0.2.14-mac-x64.dmg.blockmap 2026-06-18 208.3 kB
latest-linux.yml 2026-06-18 387 Bytes
latest-mac.yml 2026-06-18 805 Bytes
Kun 0.2.14 source code.tar.gz 2026-06-18 25.8 MB
Kun 0.2.14 source code.zip 2026-06-18 26.3 MB
README.md 2026-06-18 4.7 kB
Totals: 15 Items   1.2 GB 0

更新摘要

v0.2.13 以来的变更:

✨ 新功能

  • write: add document character count in write toolbar (#305) (847f873)
  • add embedded terminal panel (#309) (3609ace)
  • skill-root: add skill root options and update related components (77b2cf4)
  • add @floating-ui/dom dependency (408e570)
  • enhance file and directory mention handling in composer (76a926e)
  • add computer use (host screen control) for vision models (ae12658)
  • implement background reconnection for MCP servers and enhance file mention handling (81eef6d)
  • sdd: wire PM-skill frameworks into the new-requirement workflow (1b0c8b1)
  • window: add close behavior prompt (#354) (f642074)
  • settings: move archived chats into settings (#353) (829a05d)
  • plugins: add mainstream mcp recommendations (#351) (fba07c4)
  • computer-use: proactively enroll Kun in macOS permission lists (201d25c)
  • worktree: integrate worktree pool into thread lifecycle (#365) (aa5262d)
  • claw: 飞书 / Lark 流式输出处理改进 (#362) (513d2bb)
  • workspace: add file tree preview panel (#356) (e278380)
  • goal: auto-resume an active goal after an interrupted turn (#370) (057e57f)
  • add design quality linter and context management (2be01d7)
  • compact: surface manual /compact as a progress row in the conversation (4d33100)
  • context: implement prompt token trust factor to handle provider inflation (96791c0)
  • compaction: update summary mode to model-generated summaries for improved context retention (000ff9c)
  • implement model-based compaction summary and update context compaction settings (32d5a4a)
  • ui: add cursor spotlight preference (#393) (65600d9)
  • updater: show release notes after updates (#388) (171ae8f)
  • research: add evidence-led research brief (#386) (ff32026)
  • compaction: implement effective history management and add tests (fc753d9)
  • add scoped model proxy settings (176a4e5)
  • compaction: reorder compaction summaries to render at the end of turns (325239b)

🐛 修复

  • chat: improve agent code block contrast (20ed49a)
  • chat: isolate model selection per thread (cf372a1)
  • tools: add read path recovery guidance (#348) (17dd0da)
  • renderer: show visible route fallback (#347) (68352e3)
  • shortcuts: accept common zoom-in key variants (#343) (ddd5678)
  • claw: limit phone agent model choices (#344) (b2b7274)
  • main: harden audited edge cases (#346) (f11b3ae)
  • FloatingComposer: conditionally render usage cache based on thread usage (75561b8)
  • computer-use: load node-mac-permissions as a runtime import (unblock electron-vite build) (c8b1799)
  • computer-use: surface 'Accessibility granted, restart to apply' state (8820d35)
  • memory: enable memory capability by default in GUI-managed config (#336) (425d036)
  • computer-use: emit screenshot as sibling of tool_result for Anthropic compat (d80f3ce)
  • computer-use: hide the kun-child Dock icon on macOS (no more two-Kun-icon) (0dd9e7f)
  • kun: inject read image results for vision models (#375) (6a9c240)
  • context: stop screenshots inflating the context gauge + retry transient gateway errors (c4265d3)
  • chat: keep all but the final assistant segment inside the 已处理 timeline (a4ac1c4)
  • skills: preserve disable preferences (#394) (e94f7a6)
  • loop: recover empty post-tool continuations (#387) (d89d553)
  • tray: resize tray icon across platforms (#384) (1d0f64a)
  • prevent duplicate streamdown markdown text (#381) (a0e8736)
  • providers: unblock save + add bulk model management (#397) (#399) (f4d0294)
  • harden macOS update startup recovery (c387346)
  • kun: default unknown model context to 128k (bf59a6e)
  • locales: update model quality description for clarity in settings (cd94868)

📝 文档

  • computer-use: warn that domestic multimodal models are unreliable for host control (2b088b0)
  • update GitHub issue templates (f1759cd)

📦 其他

  • feat:add big change (4b66ee8)
  • fix provider cache telemetry (#318) (71a5708)
  • show latest cache rate in usage tooltips (#325) (f5caca6)
  • 🐛 fix(write): preserve selection when AI-edit textarea is focused (1aff4de)
  • 🐛 fix(context): count compaction summary tokens, not just detail (2c9dc3f)

构建信息

  • Release version: 0.2.14
  • Release channel: stable
  • Branch: master
  • Commit: 8f20403
  • macOS: Developer ID signed and notarized
  • Platforms: macOS arm64/x64, Windows x64, Linux x64
Source: README.md, updated 2026-06-18