Merge remote-tracking branch 'origin/master' into mergebot/pr1005

# Conflicts:
#	packages/client/ui-conversation/README.i18n.yaml

packages/client/ui-conversation/README.i18n.yaml

@@ -2,5 +2,5 @@
 # side as of the last confirmed-consistent state. Both languages carry equal authority;
 # after editing either side, bring the other along and re-record with:
 #   pnpm run verify-translation-pairing --write packages/client/ui-conversation/README.md
-README.md: 8711d41e966b521f0cca1e01002299355be59d68
-README.zh.md: bed2dba38a74a1b57cf948932b79c8b3796bdb7e
+README.md: b8ad658b0aeeb87a0bedcb78135370546ec47bd1
+README.zh.md: b252900e5bf77b2ce11208176bb54562b65c05dd

packages/client/ui-conversation/README.md

@@ -18,7 +18,7 @@ A tool call declaring the `terminal` render intent renders its command output in
 
 Tool rows are slots too — the standalone tool ring (`ToolViewRegistry`/`ctx.toolviews`/outlet) is retired. The chat entry declares the keyed `'conversation.chat.toolview'` hole (session scope; the key space is runtime-open); its render site dispatches per row via `entryKey: toolName` with `GenericToolCard` as the call-site `fallback`. The owner payload is the uniform `ToolRowOwnerProps` (`callId`/`toolName`/`block`/`openFile`) and `ToolRowProps` pre-composes it with the session standard kit. A registrant is a plain plugin: `ctx.slots.register({ name: 'conversation.chat.toolview', key: '<tool>', inject? }, Row)` with `inject: ['slots', 'conversation']` as the load-order seam (apply mounts ConversationService after the chat registration, so the service being present guarantees the slot is declared); session differentiation happens inside the component (`useSessions` reading `parentId` — the bash sample is the third-party-posture exemplar). Trajectory/waterfall toolview slots share this shape and land with their own render sites (RendersCheck rejects a declaration nobody renders).
 
-The todo surfaces are two registrations over that shape, both plain registrant plugins with `inject: ['slots', 'conversation']`. `TodoRow` takes the `'conversation.chat.toolview'` key `todo_write` and summarizes what the call attempted (`<done>/<total> 已完成 · <active item>` parsed from its args, falling back to the generic summary on malformed or wrongly-shaped model JSON, and keeping the generic dot for non-ok execution states so a cancelled call never reads as a completed update). `TodoDock` takes the `'conversation.input.dock'` list slot at `order: 10` — between Goal and Queue — and is the plan strip: it reads the host-computed `todos` projection via `useProjection` (standing plan: latest `todo/write` with no later `turn/start`) and renders `TodoPanel`, which takes the plain list, hides itself while the list is empty, and starts collapsed as a header of title plus `"<done>/<total> tasks · <n> in progress"` (status glyphs are the figma check / progress / dashed-pending set). The dock adapter owns the selection so the panel stays a pure function of its props; the standing list lives here rather than in the row so the row stays one line. Anything the input-zone composer chain hides (a `conversation.composer` takeover such as ui-question's) hides the whole dock, this strip included.
+The todo surfaces are two registrations over that shape, both plain registrant plugins with `inject: ['slots', 'conversation']`. `TodoRow` takes the `'conversation.chat.toolview'` key `todo_write` and summarizes what the call attempted (`<done>/<total> 已完成 · <active item>` parsed from its args, falling back to the generic summary on malformed or wrongly-shaped model JSON, and keeping the generic dot for non-ok execution states so a cancelled call never reads as a completed update). `TodoDock` takes the `'conversation.input.dock'` list slot at `order: -1` — above the queue rows — and is the plan strip: it reads the host-computed `todos` projection via `useProjection` (standing plan: latest `todo/write` with no later `turn/start`) and renders `TodoPanel`, which takes the plain list, hides itself while the list is empty, and starts collapsed as a header of title plus `"<done>/<total> tasks · <n> in progress"` (status glyphs are the figma check / progress / dashed-pending set). The dock adapter owns the selection so the panel stays a pure function of its props; the standing list lives here rather than in the row so the row stays one line. Anything the input-zone composer chain hides (a `conversation.composer` takeover such as ui-question's) hides the whole dock, this strip included.
 
 `QueueDock` is the terminal input-dock entry at `order: 20`. It hides while empty, renders one pending row directly, and defaults two or more rows to a collapsed `"<n> 条排队消息"` header whose button expands or collapses the complete list. The header exposes `aria-expanded` and `aria-controls`; the expanded list scrolls within a 180px height bound. An active edit or mutation keeps its rows visible, and emptying the queue restores the collapsed default for the next queue. Each visible row remains a single-line preview with its exact-occurrence edit and delete actions.
 
@@ -40,7 +40,7 @@ None; this package neither assembles nor sends a provider request.
 
 - **Stats-line durations cover the in-window flow only** — LLM and tool wall times fold the snapshot's assistant `timing` and tool call/result pairs, so nodes outside the loaded event window (older history) are not counted.
 - **Details panel is the minimal form and currently has no entry point** — selected call args/result raw display; the Input/Output/Metadata switch, Prev/Next stepping, and See-in-trajectory deep link are deferred. Tool rows stopped being details-panel click targets and nothing replaced that gesture, so `ChatViewInjected.openDetails` is implemented but uncalled and the panel (including its terminal card) is unreachable in the assembled application; its rendering stays covered by mounting it with a selection directly.
-- **Assistant per-message paging is a reserved slot** — drawn in the design, not implemented. The finalized content IconActions row (copy / branch / clock) ships under text output only; branch forks through the turn containing that message, increments the inherited title on the client, and then opens the child, while a fork or rename failure leaves the source selected.
+- **Assistant per-message paging is a reserved slot** — drawn in the design, not implemented. The finalized content IconActions row (copy / branch / clock) ships under the last content-text assistant of each turn only; mid-turn narration and Think-only nodes stay chrome-free. Branch forks through the turn containing that message, increments the inherited title on the client, and then opens the child, while a fork or rename failure leaves the source selected.
 - **The sparkle icon for the others tool row is a hand-drawn approximation** — the design glyph's vector geometry is not exportable locally; promotion into ui-primitives waits on an exact export.
 - **The approval panel's "Always allow this type" is deferred** — durable grants need a grant-storage design; only allow-once/reject answer today.
 - **TodoPanel truncates long item text to one ellipsized line** — the figma strip has no wrap or expand affordance; full text is not readable inline.

packages/client/ui-conversation/README.zh.md

@@ -18,7 +18,7 @@
 
 审批经由本包声明的链接管编辑器：`ApprovalPanel` 注册为按选择器路由的 `'conversation.composer'` 配置项（ui-question 模式），在审批等待未决期间取代 InputBar 占据编辑器（琥珀色条、理由标题、来自运行中调用参数的配对命令行、一次性的拒绝／允许）。`contract/slots.ts` 中的 `PendingApproval` 领域面在运行时 `PendingWait` 载体之上拥有 wire 编码——带审计关联的 `ApprovalResponsePayload` 值；广播的 `approval/resolved` 帧使等待落定并恢复编辑器。侧边栏通过 manager 跟踪的 `waitingApproval` 列表位（未实例化会话同样点亮）镜像该阻塞状态，其优先级高于运行中圆环，直至问题解决。未决等待完全离开消息流：问题（ui-question）与审批（ApprovalPanel）都经编辑器接管作答，不再保留只读占位卡。编辑器底行的 Access 席位挂载 `PermissionSelect`，由 host 计算的 `permissions` 投影经标准工具包 `useProjection` 供数（key 缺席即隐藏 chip）；chip 打开 Menu 原语下拉，kebab-case 预设名渲染为 Title Case 标签（与 `/permission` popup 的显示变换孪生），选中会经由输入栏注入的 `command` 回调提交 `/permission <preset>` 命令行。
 
-todo 两个面就是在该形状上的两个注册项，都是普通注册方插件，`inject: ['slots', 'conversation']`。`TodoRow` 占用 `'conversation.chat.toolview'` 的 `todo_write` key，摘要该次调用「试图写入」的内容（从其 args 解析出 `<已完成>/<总数> 已完成 · <进行中条目>`；模型 JSON 残缺或形状不对时回落到通用摘要；非 ok 执行状态保留通用状态点，使被取消的调用绝不读成一次已完成的更新）。`TodoDock` 以 `order: 10` 占用 `'conversation.input.dock'` 列表 slot（位于 Goal 和 Queue 之间），是计划条：它经 `useProjection` 读取 host 计算的 `todos` 投影（站立计划：其后没有更晚 `turn/start` 的最近一次 `todo/write`）并渲染 `TodoPanel`，后者接收纯列表，在列表为空时自我隐藏；列表非空时面板初始折叠，表头显示标题加 `"<已完成>/<总数> tasks · <n> in progress"`（状态图标为 figma 的勾选／进行中／虚线未开始一组）。选取由 dock 适配器负责，因此面板保持为其 props 的纯函数；站立列表放在此处而非行内，行才能保持单行。输入区 composer 链隐藏的一切（例如 ui-question 对 `conversation.composer` 的接管）也会隐藏整个 dock，包括这条计划条。
+todo 两个面就是在该形状上的两个注册项，都是普通注册方插件，`inject: ['slots', 'conversation']`。`TodoRow` 占用 `'conversation.chat.toolview'` 的 `todo_write` key，摘要该次调用「试图写入」的内容（从其 args 解析出 `<已完成>/<总数> 已完成 · <进行中条目>`；模型 JSON 残缺或形状不对时回落到通用摘要；非 ok 执行状态保留通用状态点，使被取消的调用绝不读成一次已完成的更新）。`TodoDock` 以 `order: -1` 占用 `'conversation.input.dock'` 列表 slot（位于队列行之上），是计划条：它经 `useProjection` 读取 host 计算的 `todos` 投影（站立计划：其后没有更晚 `turn/start` 的最近一次 `todo/write`）并渲染 `TodoPanel`，后者接收纯列表，在列表为空时自我隐藏；列表非空时面板初始折叠，表头显示标题加 `"<已完成>/<总数> tasks · <n> in progress"`（状态图标为 figma 的勾选／进行中／虚线未开始一组）。选取由 dock 适配器负责，因此面板保持为其 props 的纯函数；站立列表放在此处而非行内，行才能保持单行。输入区 composer 链隐藏的一切（例如 ui-question 对 `conversation.composer` 的接管）也会隐藏整个 dock，包括这条计划条。
 
 `QueueDock` 是 `order: 20` 的末端 input-dock 条目。队列为空时隐藏；只有一个待处理项时直接渲染该行；存在两个或更多待处理项时，默认收起为 `"<n> 条排队消息"` 表头，其按钮可展开或收起完整列表。表头暴露 `aria-expanded` 和 `aria-controls`；展开后的列表以 180px 为高度上限，并可滚动。存在进行中的编辑或变更时，列表行会保持可见；队列清空后，下一次出现队列时会恢复默认收起状态。每条可见行仍是单行预览，并提供针对精确单次入队项的编辑和删除操作。
 
@@ -40,7 +40,7 @@ todo 两个面就是在该形状上的两个注册项，都是普通注册方插
 
 - **统计行的耗时只覆盖窗口内消息流**：LLM（大语言模型）与工具墙钟时间由快照的 assistant `timing` 与工具 call/result 配对折算，落在已加载事件窗口之外的节点（更早的历史）不计入。
 - **详情面板是最小形态，且当前没有入口**：以原始形式显示已选择调用的参数／结果；Input/Output/Metadata 切换、Prev/Next 步进与 See-in-trajectory 深链接暂缓实现。工具行已不再是详情面板的点击目标，且没有任何手势接替它，因此 `ChatViewInjected.openDetails` 虽已实现却无人调用，该面板（含其终端卡片）在组装后的应用中不可达；其渲染仍由直接以选中态挂载它来覆盖。
-- **assistant 逐消息分页是预留 slot**：设计中已有图稿，尚未实现。已定稿的内容 IconActions 行（复制／分支／时钟）只挂在 text 输出下；分支会 fork 到包含该消息的轮次末尾，在 client 端递增继承标题后打开子会话，而 fork 或改名失败时源会话保持选中。
+- **assistant 逐消息分页是预留 slot**：设计中已有图稿，尚未实现。已定稿的内容 IconActions 行（复制／分支／时钟）只挂在每个轮次中最后一条带 text 内容的 assistant 下；轮次中间的叙述与纯 Think 节点不带 chrome。分支会 fork 到包含该消息的轮次末尾，在 client 端递增继承标题后打开子会话，而 fork 或改名失败时源会话保持选中。
 - **others 工具行的闪光图标是手绘近似版本**：无法在本地导出设计字形的矢量几何；等到存在精确导出后再将其提升到 ui-primitives。
 - **审批面板的「始终允许此类」暂缓**：持久授权需要授权存储设计；今天只能回答允许一次／拒绝。
 - **TodoPanel 将过长条目截成单行省略号**：figma 条没有换行或展开入口，完整文本无法在行内读完。

packages/client/ui-conversation/src/client/chat/AssistantMarkdown.tsx

@@ -4,8 +4,9 @@
 // view groups them into tool rows through its keyed toolview slot (figma
 // step-summary flow). Shared by finalized nodes and the streaming partial;
 // the turn-level loading dots live in the chat view's tail, not here.
-// Finalized content (text) nodes append IconActions once streaming ends;
-// Think / tool-head-only nodes stay chrome-free.
+// Finalized turn-tail content (text) nodes append IconActions once streaming
+// ends (`time` is omitted for mid-turn narration); Think / tool-head-only
+// nodes stay chrome-free.
 
 import { memo } from 'react'
 import type { AssistantBlock } from '@deepseek-ai/dsh-client-runtime/client'
@@ -21,7 +22,8 @@ export interface AssistantMarkdownProps {
   streaming: boolean
   /** Frozen partial of an aborted turn: rendered with a 已停止 marker. */
   interrupted?: boolean | undefined
-  /** Unix epoch ms for the finalized IconActions clock; omitted while streaming. */
+  /** Unix epoch ms for the IconActions clock; omitted while streaming or when
+   *  the parent withholds chrome (mid-turn content assistants). */
   time?: number | undefined
   /** Event sequence used as the fork boundary; omitted while streaming. */
   seq?: number | undefined

packages/client/ui-conversation/src/client/chat/ChatView.tsx

@@ -30,7 +30,7 @@ import type {
 import type { SnapshotSelectorHook } from '@deepseek-ai/dsh-client-ui-slots'
 import { IconChevronDownOutline14 } from '@deepseek-ai/dsh-client-ui-primitives'
 import type { ChatViewSlotProps } from '../contract/slots.ts'
-import { deriveChatFlow, type ChatFlowItem } from './chat-flow.ts'
+import { assistantActionsSeqs, deriveChatFlow, type ChatFlowItem } from './chat-flow.ts'
 import { AssistantMarkdown } from './AssistantMarkdown.tsx'
 import { GenericCommandCard } from './GenericCommandCard.tsx'
 import { GenericToolCard } from './GenericToolCard.tsx'
@@ -244,6 +244,9 @@ export function ChatView({ useSession, useSessions, useStore, renderSlot, sessio
   const selectedCallId = useStore(s => s.selection?.callId)
 
   const items = useMemo(() => deriveChatFlow(nodes), [nodes])
+  // Only the last content assistant of each turn owns IconActions; mid-turn
+  // text (before tools) omits `time` so AssistantMarkdown stays chrome-free.
+  const actionSeqs = useMemo(() => assistantActionsSeqs(nodes), [nodes])
 
   const listRef = useRef<HTMLDivElement | null>(null)
   const atBottomRef = useRef(true)
@@ -376,7 +379,7 @@ export function ChatView({ useSession, useSessions, useStore, renderSlot, sessio
           blocks={node.blocks}
           streaming={false}
           interrupted={node.interrupted}
-          time={node.time}
+          time={actionSeqs.has(node.seq) ? node.time : undefined}
           seq={node.seq}
           onFork={forkAt}
         />

packages/client/ui-conversation/src/client/chat/GenericCommandCard.tsx

@@ -1,6 +1,6 @@
 // GenericCommandCard: the default command row — a stripped-down
-// GenericToolCard rendering the dispatched command line and the settlement
-// text. Supplied by the chat view as the keyed commandview slot's render-site
+// GenericToolCard rendering the command name and its settlement text.
+// Supplied by the chat view as the keyed commandview slot's render-site
 // fallback (an unregistered command name lands here); registrants may compose
 // it as a base, feeding the same owner payload through.
 
@@ -20,10 +20,11 @@ export function GenericCommandCard({ node }: CommandRowOwnerProps) {
   const summary = node.outcome === null
     ? '执行中…'
     : text ?? (node.outcome.kind === 'error' ? '命令失败' : '已完成')
-  // Display line rebuilt from the structured payload (args carries its own
-  // separator whitespace verbatim); a cross-window node whose run page fell
-  // out of the window has neither.
-  const title = node.name === null ? '命令' : `/${node.name}${node.args ?? ''}`
+  // Title is the bare command name: the row already reads `name · outcome`,
+  // and the dispatched line's own `/` and arguments only restate what the
+  // settlement text says (`permission · preset workspace-write`). A
+  // cross-window node whose run page fell out of the window has no name.
+  const title = node.name ?? '命令'
   return (
     <ToolRow
       variant="others"

packages/client/ui-conversation/src/client/chat/chat-flow.ts

@@ -3,15 +3,24 @@
  * results group into consecutive-run tool groups (figma step-summary flow,
  * VERTICAL gap10) alternating with narration; everything else passes through.
  * Item identity keys are stable across snapshots so the list parent can
- * subscribe to keys only while rows subscribe to content.
+ * subscribe to keys only while rows subscribe to content. IconActions ownership
+ * (last content assistant per turn) is derived here too so ChatView and the
+ * flow share one gate.
  */
-import type { ConversationNode, ToolResultNode } from '@deepseek-ai/dsh-client-runtime/client'
+import type {
+  AssistantBlock, ConversationNode, ToolResultNode,
+} from '@deepseek-ai/dsh-client-runtime/client'
 
 /** One renderable flow item; key is the React key and the parent's identity unit. */
 export type ChatFlowItem =
   | { kind: 'node'; key: string; node: ConversationNode }
   | { kind: 'tool-group'; key: string; results: readonly ToolResultNode[] }
 
+/** True when the node has model-visible text content worth IconActions chrome. */
+function hasContentText(blocks: readonly AssistantBlock[]): boolean {
+  return blocks.some(block => block.kind === 'text' && block.text.trim() !== '')
+}
+
 /** An assistant node that renders nothing: only tool-call heads (rows render
  *  via the grouping pass) and blank text/reasoning. Skipped by the flow so it
  *  neither costs column gaps nor splits a tool-row run. Interrupted nodes
@@ -22,6 +31,21 @@ function rendersNothing(node: ConversationNode): boolean {
       || ((b.kind === 'text' || b.kind === 'reasoning') && b.text.trim() === ''))
 }
 
+/**
+ * Seq set of assistants that own IconActions: the last content-text assistant
+ * in each turn. Mid-turn narration (text before tools) stays chrome-free.
+ * @param nodes - snapshot nodes (surface order).
+ * @returns Seq values ChatView may pass as `time` into AssistantMarkdown.
+ */
+export function assistantActionsSeqs(nodes: readonly ConversationNode[]): ReadonlySet<number> {
+  const lastByTurn = new Map<number, number>()
+  for (const node of nodes) {
+    if (node.kind !== 'assistant' || !hasContentText(node.blocks)) continue
+    lastByTurn.set(node.turn, node.seq)
+  }
+  return new Set(lastByTurn.values())
+}
+
 /**
  * Group finalized nodes into the step-summary flow.
  * @param nodes - snapshot nodes (surface order).

packages/client/ui-conversation/src/client/skeleton/ApprovalPanel.module.css

@@ -19,6 +19,13 @@
   border-radius: 20px;
   background: var(--dsw-specific-input-major);
   box-shadow: var(--dsw-shadow-lv2);
+  /* Elevated surface in dark, same as the menus: `.body` inside scrolls once
+     the justification or command passes the cap, so the thumb takes the l2
+     pair. Declared on the card because the elevation belongs to the surface,
+     and the custom properties inherit down to the region that actually
+     scrolls (see ui-theme styles/scrollbar.css for the rebinding contract). */
+  --dsh-scrollbar-thumb: var(--dsw-alias-scrollbar-bg-l2);
+  --dsh-scrollbar-thumb-hover: var(--dsw-alias-scrollbar-hover-l2);
 }
 
 /* Tinted full-width header band. */
@@ -40,11 +47,22 @@
   background: var(--dsw-alias-state-warn-primary);
 }
 
+/* Scroll region: an agent's justification and its command are unbounded model
+   text (a one-line `cd` or a 40-line heredoc), and the seat sits in a
+   fixed-height column — uncapped, a long command pushed the action row past
+   the viewport and the approval could not be answered at all. The strip and
+   the action row stay outside, so the buttons are always on screen. */
 .body {
   display: flex;
   flex-direction: column;
   gap: 6px;
-  padding: 12px 16px 14px;
+  /* border-box so the cap is the region's OUTER height: the composer's draft
+     area counts its padding inside the same number, and the two seats are
+     only interchangeable if they occupy the same box. */
+  box-sizing: border-box;
+  max-height: var(--dsh-composer-text-max-height);
+  overflow-y: auto;
+  padding: 12px 16px 0;
 }
 
 /* The model's justification is the panel's message, not a footnote. */
@@ -63,11 +81,15 @@
   word-break: break-all;
 }
 
+/* Card-level row, not body content. Its padding reproduces the metrics the row
+   had inside the body: 14px above (the flex gap of 6 plus the row's 8px top
+   margin, neither of which reaches it out here) and the body's former 14px
+   bottom pad below, so the resting card is unchanged. */
 .actionRow {
   display: flex;
   justify-content: flex-end;
   gap: 8px;
-  margin-top: 8px;
+  padding: 14px 16px 14px;
 }
 
 .allow,

packages/client/ui-conversation/src/client/skeleton/ApprovalPanel.tsx

@@ -4,7 +4,11 @@
 // pending, this panel occupies the composer slot in place of the InputBar:
 // an amber "Waiting for approval" strip on the card top, the model's
 // justification as the headline, the paired command in muted code text, and
-// a right-aligned refuse/allow action row. One-shot: the buttons disable
+// a right-aligned refuse/allow action row. Justification and command are
+// unbounded model text, so they scroll inside the card at the shared composer
+// cap (`data-approval-scroll`) and the action row stays outside it — the
+// buttons must be reachable no matter how long the command is.
+// One-shot: the buttons disable
 // after a click and the panel leaves (the InputBar returns) on the broadcast
 // resolved frame. The draft's "Always allow this type" is deferred with
 // grant storage.
@@ -53,17 +57,20 @@ function ApprovalFlow({ pending, command }: { pending: PendingApproval; command?
     <div className={css.root} data-approval-key={pending.key}>
       <div className={css.card}>
         <div className={css.strip}><span className={css.dot} />等待审批</div>
-        <div className={css.body}>
+        {/* Tab stop: the region scrolls once the command passes the cap and
+            holds nothing focusable of its own, so without one a keyboard-only
+            user cannot reach the command's tail before answering. */}
+        <div className={css.body} data-approval-scroll="" tabIndex={0} role="group" aria-label="审批详情">
           <div className={css.headline}>{pending.reason ?? `工具 ${pending.toolName} 请求越权执行`}</div>
           {command !== undefined && <div className={css.command}>{command}</div>}
-          <div className={css.actionRow}>
-            <button type="button" className={css.reject} disabled={answered} onClick={() => { answer('rejected') }}>
-              拒绝
-            </button>
-            <button type="button" className={css.allow} disabled={answered} onClick={() => { answer('allowed-once') }}>
-              允许一次
-            </button>
-          </div>
+        </div>
+        <div className={css.actionRow}>
+          <button type="button" className={css.reject} disabled={answered} onClick={() => { answer('rejected') }}>
+            拒绝
+          </button>
+          <button type="button" className={css.allow} disabled={answered} onClick={() => { answer('allowed-once') }}>
+            允许一次
+          </button>
         </div>
       </div>
     </div>

packages/client/ui-conversation/src/client/skeleton/ConversationRoot.module.css

@@ -143,6 +143,14 @@
   display: flex;
   flex: none;
   flex-direction: column;
+  /* One cap for every scrolling text region a composer seat can hold: the
+     InputBar draft (figma Input 75:8208 max 14 lines × 24px line) and the
+     takeover panels' bodies top out at the same height, so electing a
+     takeover never grows the footer past the card it replaces. Declared on
+     the seat because it is the chain's only shared ancestor — fallback and
+     elected overlay are siblings — and custom properties inherit down to
+     whichever entry is mounted. */
+  --dsh-composer-text-max-height: 336px;
 }
 
 /* Active phase: header is ordinary column chrome above the scrollport (not

packages/client/ui-conversation/src/client/skeleton/InputBar.module.css

@@ -209,7 +209,9 @@
 .mirror {
   visibility: hidden;
   pointer-events: none;
-  max-height: 336px;
+  /* 14-line cap, shared with the composer takeovers (declared on
+     ConversationRoot .composerSeat). */
+  max-height: var(--dsh-composer-text-max-height);
   overflow: hidden;
 }
 

packages/client/ui-conversation/tests/chat-view.spec.tsx

@@ -16,7 +16,7 @@ import { RpcId } from '@deepseek-ai/dsh-client-connection/client'
 import type { ChatViewSlotProps, SelectionTarget } from '@deepseek-ai/dsh-client-ui-conversation/client'
 import { createChatStore } from '../src/client/stores.ts'
 import { ChatView } from '../src/client/chat/ChatView.tsx'
-import { deriveChatFlow, flowKeys } from '../src/client/chat/chat-flow.ts'
+import { assistantActionsSeqs, deriveChatFlow, flowKeys } from '../src/client/chat/chat-flow.ts'
 
 afterEach(cleanup)
 // Keyless create() persists under the bare declared key; clear between cases
@@ -61,8 +61,8 @@ const user = (seq: number, text: string): UserMessageNode => ({
   content: [{ type: 'text', text }] as never,
   source: null,
 })
-const assistant = (seq: number, text: string): AssistantMessageNode => ({
-  kind: 'assistant', seq, time: seq * 1_000, turn: 1, step: 1, blocks: [{ kind: 'text', text }],
+const assistant = (seq: number, text: string, turn = 1): AssistantMessageNode => ({
+  kind: 'assistant', seq, time: seq * 1_000, turn, step: 1, blocks: [{ kind: 'text', text }],
 })
 const toolResult = (seq: number, callId: string, name = 'bash'): ToolResultNode => ({
   kind: 'tool-result', seq, time: seq * 1_000, callId,
@@ -156,6 +156,23 @@ describe('chat-flow derivation', () => {
     expect(flowKeys(deriveChatFlow([toolResult(3, 'a'), { ...headsOnly, interrupted: true }, toolResult(5, 'b')]))).toBe('g3|n4|g5')
     expect(flowKeys(deriveChatFlow([toolResult(3, 'a'), assistant(4, 'found'), toolResult(5, 'b')]))).toBe('g3|n4|g5')
   })
+
+  it('assistantActionsSeqs keeps only the last content assistant per turn', () => {
+    const thinkOnly: AssistantMessageNode = {
+      kind: 'assistant', seq: 3, time: 3_000, turn: 1, step: 2,
+      blocks: [{ kind: 'reasoning', text: 'planning' }],
+    }
+    const seqs = assistantActionsSeqs([
+      user(1, 'hi'),
+      assistant(2, 'looking', 1),
+      thinkOnly,
+      toolResult(4, 'a'),
+      assistant(5, 'done', 1),
+      user(6, 'again'),
+      assistant(7, 'second turn', 2),
+    ])
+    expect([...seqs].sort((a, b) => a - b)).toEqual([5, 7])
+  })
 })
 
 describe('ChatView', () => {
@@ -196,6 +213,23 @@ describe('ChatView', () => {
     expect(view.getByText('run a')).toBeTruthy()
   })
 
+  it('shows assistant IconActions only on the last content message of each turn', () => {
+    const h = makeHarness({
+      nodes: [
+        user(1, 'hi'),
+        assistant(2, 'mid-turn text'),
+        toolResult(3, 'a'),
+        assistant(4, 'final answer'),
+        user(5, 'next'),
+        assistant(6, 'second turn', 2),
+      ],
+    })
+    const view = render(<h.ChatView {...h.props} />)
+    // 2 user + 2 turn-tail assistants; mid-turn text at seq 2 stays chrome-free.
+    expect(view.getAllByRole('button', { name: '复制' })).toHaveLength(4)
+    expect(view.getAllByRole('button', { name: '在新对话中分支' })).toHaveLength(4)
+  })
+
   it('forks from both user and finalized assistant message actions at their event seq', () => {
     const h = makeHarness({ nodes: [user(1, 'question'), assistant(2, 'answer')] })
     const view = render(<h.ChatView {...h.props} />)
@@ -460,10 +494,14 @@ describe('ChatView', () => {
       name: 'plan', args: '', outcome: { kind: 'success', text: '已进入 plan mode' },
       ...over,
     })
-    // Settled success: the command line is the title, the outcome text the summary.
-    const settled = makeHarness({ nodes: [user(1, 'hi'), command({})] })
+    // Settled success: the bare command name is the title, the outcome text
+    // the summary — neither the dispatched `/` nor its arguments reach the row
+    // (the settlement text already says what the command did).
+    const settled = makeHarness({ nodes: [user(1, 'hi'), command({ args: ' now' })] })
     const view = render(<settled.ChatView {...settled.props} />)
-    expect(view.getByText('/plan')).toBeTruthy()
+    expect(view.getByText('plan')).toBeTruthy()
+    expect(view.queryByText('/plan')).toBeNull()
+    expect(view.queryByText('/plan now')).toBeNull()
     expect(view.getByText('已进入 plan mode')).toBeTruthy()
 
     // Error outcome flips the row state; a text-less error gets the default copy.