Skip to content

Lynn v0.84.1

Choose a tag to compare

@MerkyorLynn MerkyorLynn released this 11 Jun 18:29
· 284 commits to main since this release

国内镜像站下载(推荐)

国内用户请优先使用以下镜像站地址;GitHub Assets 作为备用下载。

Lynn v0.84.1 Release Notes / 发布说明

发布日期: 2026-06-12 · StepFun 默认主链稳定性 + 实时语音 + Hanako 数据隔离 + GUI 空答修复 + 发版门禁补强

本次发版把 Lynn CLI 与桌面 GUI 同步到 v0.84.1。默认对话和任务执行统一回到 StepFun 3.7 Flash 一条龙主链,语音主入口统一到 Brain 托管 StepFun Realtime,优先保证“能正常对话、能完成工具任务、能自然实时说话”。本地蒸馏 A3B manager 仍作为显式 Lynn manager run 能力保留,但不会抢占默认 GUI/CLI 对话链路;考虑本地并发与稳定性,默认编排器切换继续暂缓。

中文重点

  • GUI 空答根因修复:旧会话复用过期设备签名时,客户端会为每次请求刷新签名;工具链最终答案晚到时,Brain 不再 8 秒硬关流并丢弃 late final answer。
  • StepFun 3.7 Flash 默认主链:普通 GUI/CLI 对话、Lynn -p、编码执行默认走 StepFun 3.7 Flash。Spark/A3B manager 仅在显式 Lynn manager run 或后续实验链路中使用。
  • 实时语音主入口:GUI 麦克风与 CLI 当前 chat 内 /voice / lynn voice 默认进入 Brain 托管 StepFun Realtime 连续对话;CLI 在聊天框下方显示状态与采样波形。--file/--record ASR 转写和 --speak TTS 朗读仍作为辅助工具保留。
  • reasoning-only 空答重试:Brain 在源头识别“只有思考、没有可见正文”的响应并重试,避免用户看到“思考完但不说话”。
  • 工具 turn 收口更诚实:工具完成后如果仍需等待模型最终答复,会显示事实性的工具完成状态,不再静默关闭或伪造本地总结。
  • GUI token/cost pipeline:SDK usage → WebSocket → store → 输入行 chip 打通,桌面端能长期显示会话 token/cost 状态。
  • Fleet 可发现性与验收面板:桌面端增加 Fleet 入口和 acceptance panel,方便把 Lynn 作为黑灯工厂 worker 调度。
  • Issue #72/#74 数据隔离与自愈:Lynn 默认不再读取或迁移 ~/.hanako,只在显式设置 LYNN_IMPORT_HANAKO_ON_FIRST_RUN=1 时导入;已被旧版污染的 ~/.lynn 会在启动时把已下线 MiMo/TokenPlan 模型引用修回 Brain 默认路由,同时保留用户 API key。
  • 安全与 BUG 修复附带:设备签名刷新、工具 turn 收口、reasoning-only 空答重试、Fleet 验收面板、发布门禁与文档漂移检查随本版一起落地。

安装

# 前置:Node.js 20 LTS 或 22 LTS with npm.
node -v

# 从 Lynn 镜像安装或覆盖升级 CLI。
npm install -g --force "https://download.merkyorlynn.com/downloads/cli/lynn-cli-0.84.1.tgz"

# 启动。
Lynn            # 交互式聊天 TUI;输入 /voice 或 lynn voice 进入实时语音
Lynn code       # 编码 agent TUI
Lynn --version  # 应输出 0.84.1
Lynn agents     # 给其他智能体/Fleet 的可复制命令

默认 Brain V2 路由:StepFun 3.7 Flash(256K 上下文,high 推理,48K 推理/生成预算)。语音默认走 Brain 托管 StepFun Realtime。本地 A3B / Spark 继续作为显式高端本地档与实验 manager/fallback 能力存在,不作为普通 GUI/CLI 默认主链。BYOK 仍可用。

验证

  • npm run release:preflight
  • npm run gate:startup
  • npm run gate:cli-task
  • npm run test:voice-cli
  • npm run test:release:ui
  • npm run test:release:static
  • CLI tarball 镜像站安装 smoke
  • macOS 打包签名、公证、staple、Gatekeeper 校验

Release date: 2026-06-12 · StepFun default-route stability + realtime voice + Hanako data isolation + GUI empty-answer recovery + release gates

This release unifies Lynn CLI and desktop GUI at v0.84.1. The default chat and task path is now the direct StepFun 3.7 Flash route again, and the default voice path is Brain-hosted StepFun Realtime. The priority is the baseline product contract: normal conversation works, tool tasks complete, final answers are visible, and realtime voice feels natural. The local distilled A3B manager remains available through explicit Lynn manager run, but it does not take over the default GUI/CLI path. Default local-manager routing remains deferred until local concurrency and stability are proven.

Highlights

  • GUI empty-answer recovery: stale device signatures are refreshed per request, and Brain no longer closes a tool turn after 8 seconds if the final assistant answer is still arriving.
  • StepFun 3.7 Flash default: normal GUI/CLI chat, Lynn -p, and coding execution use StepFun 3.7 Flash by default. Spark/A3B manager is explicit only.
  • Realtime voice default: GUI microphone and /voice / lynn voice inside the CLI chat use Brain-hosted StepFun Realtime for continuous conversation; CLI renders status and a live waveform below the chat. ASR transcription and TTS save commands remain auxiliary.
  • reasoning-only retry: Brain retries responses that contain reasoning but no visible answer at the source.
  • honest tool-turn close: tool completion is rendered factually while waiting for the model's final answer; no silent close and no fake local summary.
  • GUI token/cost pipeline: SDK usage now flows through WebSocket, store, and the input-row chip.
  • Fleet discoverability and acceptance panel: desktop Fleet entry points and acceptance UI are easier to find and inspect.
  • Issue #72/#74 data isolation and repair: Lynn no longer reads or migrates ~/.hanako by default; import is opt-in via LYNN_IMPORT_HANAKO_ON_FIRST_RUN=1. Existing polluted ~/.lynn installs repair retired MiMo/TokenPlan model references back to the Brain default route while preserving user API keys.
  • Safety and bug-fix rollup: device-signature refresh, honest tool-turn close, reasoning-only retry, Fleet acceptance UI, release gates, and doc-drift checks ship together in this release.

Install

# Prerequisite:Node.js 20 LTS or 22 LTS with npm.
node -v

# Install or update from the Lynn mirror.
npm install -g --force "https://download.merkyorlynn.com/downloads/cli/lynn-cli-0.84.1.tgz"

# Launch.
Lynn            # interactive chat TUI; type /voice or lynn voice for realtime voice
Lynn code       # coding-agent TUI
Lynn --version  # should print 0.84.1
Lynn agents     # copyable headless/Fleet commands

Default Brain V2 route: StepFun 3.7 Flash (256K context, high reasoning, 48K reasoning/generation budget). Voice defaults to Brain-hosted StepFun Realtime. Local A3B/Spark remains available as an explicit high-end local tier and experimental manager/fallback path, but it is not the ordinary GUI/CLI default. BYOK remains available.