fix(web): address reference review defects

.agents/notes/implemented/feature/2026-07-21-cross-session-references.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 .agents/notes/implemented/feature/2026-07-21-cross-session-references.md
-2026-07-21-cross-session-references.md: 78d18da23ef682df18653aea73467e281266de9c
-2026-07-21-cross-session-references.zh.md: 4fe53e21f85fab950b8480bb28869df6254eed0a
+2026-07-21-cross-session-references.md: 48a3241871f0b65270aaf74e42c9d02c45eaf027
+2026-07-21-cross-session-references.zh.md: ef723115f0de4312c0f4b1d215303b4a6a31be47

.agents/notes/implemented/feature/2026-07-21-cross-session-references.md

@@ -34,7 +34,7 @@ Reference preparation is not a new delivery protocol and does not create a turn
 
 The unified Web `@` source combines session candidates with Host-backed file discovery. Session candidate lookup matches case-insensitive substrings of the session id, cwd, or latest folded title, displays that title, and falls back to the session id when a title observation is absent or fails. Lookup follows the request's cancellation signal, and session id, cwd, and mention labels escape external control characters while the canonical URI retains the original id.
 
-Web exposes file and session discovery through generated Remote methods on their owning services, as detailed in [Web file and session references](2026-07-27-web-file-and-session-references.md). Session picks are atomic chips backed by the Host-produced canonical mention. Ordinary `session.prompt` delivery carries that mention without a reference-specific API Proxy route. Replay associates the separate session-reference context with its neighboring direct message and renders a compact source summary instead of exposing the snapshot JSON.
+Web exposes file and session discovery through generated Remote methods on their owning services, as detailed in [Web file and session references](2026-07-27-web-file-and-session-references.md). Session picks are atomic chips backed by the Host-produced canonical mention. Ordinary `session.prompt` delivery carries that mention without a reference-specific API Proxy route. Replay associates the separate session-reference context with the direct message immediately before it and renders a compact source summary instead of exposing the snapshot JSON.
 
 The [automation-only ACP transport](../simplification/2026-07-23-acp-automation-only-protocol.md) deliberately does not mount session-query or session-reference services.
 
@@ -55,7 +55,7 @@ Each of at most three references is independently capped at 65,536 UTF-8 bytes b
 
 ## Verification
 
-Unit and integration coverage pins URI round-trips and text-boundary punctuation, explicit malformed references, id/cwd/title candidate matching and ranking, failed title-observation fallback, candidate cancellation, control-character escaping, projection exclusions, non-recursive snapshot projection, backend-independent compact checkpoints, tag-safe framing, deduplication, self-reference, count limits, all-or-nothing reads, cancellation against a non-settling storage read, independent per-source byte retention, frozen message ownership, pre-step parsing and insertion, downstream rejection, node-owned replay association, title isolation, and the generated Remote discovery faces. A keyless Web snapshot pins the assembled reference selection path.
+Unit and integration coverage pins URI round-trips and text-boundary punctuation, explicit malformed references, id/cwd/title candidate matching and ranking, failed title-observation fallback, candidate cancellation, control-character escaping, projection exclusions, non-recursive snapshot projection, backend-independent compact checkpoints, tag-safe framing, deduplication, self-reference, count limits, all-or-nothing reads, cancellation against a non-settling storage read, independent per-source byte retention, frozen message ownership, pre-step parsing and insertion, downstream rejection, Chat-projected following-recall association, title isolation, and the generated Remote discovery faces. A keyless Web snapshot pins the assembled reference selection path.
 
 ## Consequences
 

.agents/notes/implemented/feature/2026-07-21-cross-session-references.zh.md

@@ -34,7 +34,7 @@ Web 用户需要把另一场对话中的相关工作带入一条新消息，但
 
 统一的 Web `@` source 把会话候选与 Host 支持的文件发现组合在一起。会话候选查询会对 session id、cwd 或最新折叠后的标题执行不区分大小写的子串匹配，显示该标题，并在没有标题观察结果或标题观察失败时回退到 session id。查询遵循请求的取消信号；session id、cwd 和提及标签中的外部控制字符会被转义，但规范 URI 仍保留原始 id。
 
-Web 通过所属服务上的生成 Remote 方法提供文件与会话发现，详见 [Web 文件与会话引用](2026-07-27-web-file-and-session-references.md)。session 选择项是由 Host 生成的规范 mention 支撑的原子 chip。普通 `session.prompt` 投递会携带该 mention，无需引用专用 API Proxy 路由。回放会把独立的 session-reference 上下文与相邻直接消息关联起来，并渲染精简来源摘要，而不暴露快照 JSON。
+Web 通过所属服务上的生成 Remote 方法提供文件与会话发现，详见 [Web 文件与会话引用](2026-07-27-web-file-and-session-references.md)。session 选择项是由 Host 生成的规范 mention 支撑的原子 chip。普通 `session.prompt` 投递会携带该 mention，无需引用专用 API Proxy 路由。回放会把独立的 session-reference 上下文与紧邻其前的直接消息关联起来，并渲染精简来源摘要，而不暴露快照 JSON。
 
 [仅面向自动化的 ACP（Agent Client Protocol）传输层](../simplification/2026-07-23-acp-automation-only-protocol.md)有意不挂载会话查询或会话引用服务。
 
@@ -55,7 +55,7 @@ Web 通过所属服务上的生成 Remote 方法提供文件与会话发现，
 
 ## 验证
 
-单元与集成测试覆盖 URI 无损往返与文本边界标点、显式格式错误的引用、按 id／cwd／标题进行候选匹配与排序、标题观察失败时的回退、候选查询取消、控制字符转义、投影排除规则、快照的非递归投影、与后端无关的压缩检查点、标签安全封套、去重、自引用、数量限制、读取的全有或全无、存储读取不结束时的取消、逐源独立字节保留、冻结消息所有权、pre-step 解析和插入、下游拒绝、节点负责的回放关联、标题隔离，以及生成的 Remote 发现接口。一个无密钥 Web 快照会固定组装后的引用选择路径。
+单元与集成测试覆盖 URI 无损往返与文本边界标点、显式格式错误的引用、按 id／cwd／标题进行候选匹配与排序、标题观察失败时的回退、候选查询取消、控制字符转义、投影排除规则、快照的非递归投影、与后端无关的压缩检查点、标签安全封套、去重、自引用、数量限制、读取的全有或全无、存储读取不结束时的取消、逐源独立字节保留、冻结消息所有权、pre-step 解析和插入、下游拒绝、Chat 投影的后继召回关联、标题隔离，以及生成的 Remote 发现接口。一个无密钥 Web 快照会固定组装后的引用选择路径。
 
 ## 后果
 

.agents/notes/implemented/feature/2026-07-27-web-file-and-session-references.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 .agents/notes/implemented/feature/2026-07-27-web-file-and-session-references.md
-2026-07-27-web-file-and-session-references.md: 09619837b6eda304106276064ed5423b10053119
-2026-07-27-web-file-and-session-references.zh.md: 52cba697c76635ab60eb862475bb47c1a27f99a4
+2026-07-27-web-file-and-session-references.md: 51447d0c22c2e5beb4eae03e7f955239525fc16d
+2026-07-27-web-file-and-session-references.zh.md: f2f87aabaa818fe5d104b0ee1488658c792f59b9

.agents/notes/implemented/feature/2026-07-27-web-file-and-session-references.md

@@ -16,7 +16,7 @@ The file capability follows the three-package seam: `@deepseek-ai/dsh-file-refer
 
 A session pick is a structured composer reference. Its visible form uses a chat-bubble glyph and business-color session title without a capsule, while its clipboard and model form is the canonical `@[label](dsh-session:…)` mention produced by the Host. The complete `@label` display text remains in the transparent textarea, and the same-size backdrop colors that range and replaces its leading marker with the domain glyph. Native glyph metrics therefore determine width, wrapping, selection, and caret placement without truncation. The occurrence range retains reference identity for serialization; Backspace or Delete at its boundary removes it whole, and editing inside it turns the remaining characters into ordinary text. Ordinary `session.prompt` delivery carries the canonical mention unchanged. The session-reference service parses accepted direct user messages at `agent/pre-step`, captures every source, replaces the canonical mention with readable text while preserving the direct message id, and inserts the frozen snapshot immediately after that message. The recalled-context row uses the same chat glyph while other context keeps the document glyph. The API Proxy contains no reference-specific route, dependency, or error code.
 
-The input machine keeps ordinary draft text and atomic references until the default sink reports Host acceptance. Serialization or prompt transport failure returns the same draft to editing. After acceptance, reference preparation belongs to the agent turn; a malformed mention, failed source read, cancellation, or budget failure terminates that turn. The logged prompt remains the replay authority. The chat renders the durable direct-message-then-recall order, decorates recognized file and session mentions as icon-and-text references, and keeps snapshot JSON behind the collapsed recall row.
+The input machine keeps ordinary draft text and atomic references until the default sink reports Host acceptance. Its session-store mirror persists each occurrence's canonical clipboard projection, so remounting without the occurrence table retains a parseable reference instead of a display-only label. Serialization or prompt transport failure returns the same draft to editing. After acceptance, reference preparation belongs to the agent turn; a malformed mention, failed source read, cancellation, or budget failure terminates that turn. The logged prompt remains the replay authority. The chat renders the durable direct-message-then-recall order and associates exact session labels only from the immediately following sourced recall, which preserves multi-word titles and keeps consecutive references independent. It decorates recognized file and session mentions as icon-and-text references, treats unquoted `@path` tokens including extensionless basenames as files, leaves sentence punctuation outside the reference range, and keeps snapshot JSON behind the collapsed recall row.
 
 ## Reference transaction
 
@@ -42,7 +42,7 @@ File lookup is advisory and cancellable; selection itself performs no read. Sess
 
 ## Verification
 
-Package tests pin shared file grammar and ranking, cache invalidation and lifecycle cleanup, parallel Web lookup, quoted paths, independent candidate failure, cancellation, source-title suppression through pending and ready states, grouped headings that do not alter option indexes, file/directory continuation, structured file and session references, complete inline labels, domain glyphs, adjacent-reference and adjacent-text reference projection, codec round-trip, generated Remote type inference, direct-before-recall pre-step preparation, downstream rejection, and durable chat order. The keyless assembled Web snapshot renders the available reference sections without the raw source title, selects a file, then selects a session reference through the real client composition.
+Package tests pin shared file grammar and ranking, cache invalidation and lifecycle cleanup, parallel Web lookup, quoted paths, independent candidate failure, cancellation, source-title suppression through pending and ready states, grouped headings that do not alter option indexes, file/directory continuation, structured file and session references, complete inline labels, domain glyphs, disabled-layer ownership, canonical draft persistence across remount, adjacent-reference and adjacent-text projection, extensionless file and sentence-punctuation rendering, codec round-trip, generated Remote type inference, direct-before-recall pre-step preparation, downstream rejection, and following-recall association for multi-word and consecutive labels. The keyless assembled Web snapshot renders the available reference sections without the raw source title, selects a file, then selects a session reference through the real client composition, and replays a multi-word session label in direct-before-recall order.
 
 ## Consequences
 

.agents/notes/implemented/feature/2026-07-27-web-file-and-session-references.zh.md

@@ -16,7 +16,7 @@ Web 通过 `@deepseek-ai/dsh-client-ui-reference` 暴露一个合并的 `@file`
 
 选择会话会创建一个结构化输入框引用。可见形式使用聊天气泡图标与业务色会话标题，不使用胶囊容器；剪贴板和模型形式则是宿主生成的规范 `@[label](dsh-session:…)` mention。完整的 `@label` 展示文本会保留在透明 textarea 中，同尺寸 backdrop 会为这段范围着色，并把开头的 marker 替换为对应领域图标。因此宽度、换行、选择区与光标位置都由原生字形度量决定，不会截断。occurrence 范围会保留引用身份以供序列化；在边界按 Backspace 或 Delete 会整段删除引用，在范围内部编辑则会把剩余字符转为普通文本。普通 `session.prompt` 投递会原样携带规范 mention。session-reference 服务会在 `agent/pre-step` 解析已接受的直接用户消息，捕获每个源，在保留直接消息 id 的同时把规范 mention 替换为可读文本，并把冻结快照插入到该消息紧后。召回上下文行使用同一个聊天图标，其他上下文保留文档图标。API Proxy 不包含引用专用路由、依赖或错误码。
 
-输入状态机在默认 sink 报告宿主已接受前，会保留普通草稿文本和原子引用。序列化或提示词传输失败后，同一草稿会回到可编辑状态。接受后，引用准备属于 agent 轮次；格式错误的 mention、源读取失败、取消或预算失败会终止该轮次。已记录的提示词仍是回放权威。聊天界面按照持久的直接消息后接召回行顺序渲染，将识别到的文件与会话 mention 装饰成图标加文字的引用，并把快照 JSON 保留在默认收起的召回行中。
+输入状态机在默认 sink 报告宿主已接受前，会保留普通草稿文本和原子引用。它写入会话 store 的镜像会持久化每个 occurrence 的规范剪贴板投影，因此在 occurrence 表缺失的情况下重新挂载时，仍会保留可解析的引用，而不是仅供显示的标签。序列化或提示词传输失败后，同一草稿会回到可编辑状态。接受后，引用准备属于 agent 轮次；格式错误的 mention、源读取失败、取消或预算失败会终止该轮次。已记录的提示词仍是回放权威。聊天界面按照持久的直接消息后接召回行顺序渲染，并且只从紧随其后的带来源召回中关联准确的会话标签，因此既能保留多词标题，也能让连续引用彼此独立。它会把识别到的文件与会话 mention 装饰成图标加文字的引用，把包括无扩展名 basename 在内的未加引号 `@path` token 视为文件，将句末标点留在引用范围之外，并把快照 JSON 保留在默认收起的召回行中。
 
 ## 引用事务
 
@@ -42,7 +42,7 @@ type @ → parallel file/session Remote calls → pick folder text or atomic fil
 
 ## 验证
 
-包（package）测试固定共享文件语法和排序、缓存失效及生命周期清理、Web 并行查询、带引号的路径、候选项独立失败、取消、在 pending 与 ready 状态下隐藏 source 标题、不改变候选项索引的分组标题、文件／目录继续补全、结构化文件与会话引用、完整行内标签、领域图标、相邻引用及相邻文本条件下的引用投影、codec 无损往返、生成的 Remote 类型推断、pre-step 中直接消息先于召回的准备、下游拒绝，以及持久聊天顺序。无密钥的装配 Web 快照会在不显示原始 source 标题的情况下渲染可用的引用分组，并通过真实客户端组合依次选择文件和会话引用。
+包（package）测试固定共享文件语法和排序、缓存失效及生命周期清理、Web 并行查询、带引号的路径、候选项独立失败、取消、在 pending 与 ready 状态下隐藏 source 标题、不改变候选项索引的分组标题、文件／目录继续补全、结构化文件与会话引用、完整行内标签、领域图标、禁用状态下的层级归属、跨重新挂载的规范草稿持久化、相邻引用及相邻文本条件下的引用投影、无扩展名文件与句末标点渲染、codec 无损往返、生成的 Remote 类型推断、pre-step 中直接消息先于召回的准备、下游拒绝，以及多词与连续标签的后继召回关联。无密钥的装配 Web 快照会在不显示原始 source 标题的情况下渲染可用的引用分组，通过真实客户端组合依次选择文件和会话引用，并按直接消息先于召回的顺序回放多词会话标签。
 
 ## 后果
 

apps/web/tests/reference-composer.e2e.ts

@@ -68,7 +68,7 @@ function targetSessionFixture(): string {
   const session = Session.create(SessionId(TARGET_SESSION_ID))
   session.append('turn/start', { turn: 1 })
   const user = session.append('user/message', createUserMessage({
-    content: [{ type: 'text', text: '@Research what changed?' }],
+    content: [{ type: 'text', text: '@Research notes what changed?' }],
     source: { kind: 'user' },
   }), { surfaceOp: 'append' })
   session.append('user/message', createUserMessage({
@@ -79,7 +79,7 @@ function targetSessionFixture(): string {
       version: 1,
       references: [{
         sessionId: SOURCE_SESSION_ID,
-        label: 'Research',
+        label: 'Research notes',
         capturedThroughSeq: 4,
         compacted: false,
         originalMessages: 2,
@@ -176,12 +176,12 @@ describe.skipIf(MODE === 'record')('web e2e: file and session references through
     const target = page.getByRole('treeitem').filter({ hasText: /^dsh-web-e2e-ws-/ }).first()
     await target.waitFor({ timeout: 15_000 })
     await target.click()
-    await page.getByRole('button', { name: /^Session recall\s*Research$/ }).waitFor({ timeout: 15_000 })
+    await page.getByRole('button', { name: /^Session recall\s*Research notes$/ }).waitFor({ timeout: 15_000 })
 
     const snapshot = (await captureStableAria(page, '[class*="centerCol"]', scaffold.workspaceCwd))
       .split(TARGET_SESSION_ID).join('{{targetId}}')
     await compareOrRefreshGolden(ORDER_EXPECTED, snapshot, MODE)
-    expect(snapshot.indexOf('Research what changed?')).toBeLessThan(snapshot.indexOf('Session recall Research'))
+    expect(snapshot.indexOf('Research notes what changed?')).toBeLessThan(snapshot.indexOf('Session recall Research notes'))
     expect(tripwire.pageErrors).toEqual([])
     expect(tripwire.warnings).toEqual([])
     await assertFixtureInventory(SNAPSHOT_DIR, ['menu.expected.md', 'order.expected.md'])

apps/web/tests/snapshots/reference-composer/order.expected.md

@@ -7,12 +7,12 @@
   - tablist:
     - tab "Chat" [selected]
     - tab "Trajectory"
-- text: Research what changed? {{clock}}
+- text: Research notes what changed? Referenced session · Research notes {{clock}}
 - button "Copy":
   - img
-- button "Session recall Research":
+- button "Session recall Research notes":
   - img
-  - text: Session recall Research
+  - text: Session recall Research notes
 - textbox "Message the agent"
 - button "Commands":
   - img

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: f1a4e5de4056712feab7cc1627589909ef70f526
-README.zh.md: a942c994d39652b3811c88ce99bc149cd47b2118
+README.md: cbd57ba05a96da36ac9198817e78594c1a2cefff
+README.zh.md: cbe13efd43abda020edae41a0adee09d991314fe

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

@@ -167,7 +167,11 @@ function projectUserText(text: string, sessionLabels: readonly string[]): ReactN
   let m: RegExpExecArray | null
   while ((m = re.exec(text)) !== null) {
     const tokenStart = m.index + (m[1]?.length ?? 0)
-    const label = m[2] ?? ''
+    const rawLabel = m[2] ?? ''
+    const label = rawLabel.startsWith('@"')
+      ? rawLabel
+      : rawLabel.replace(/[.,;:!?，。；：！？]+$/gu, '')
+    if (label.length <= 1) continue
     ranges.push({ start: tokenStart, end: tokenStart + label.length, label, kind: 'plain' })
   }
   ranges.sort((a, b) => a.start - b.start
@@ -181,7 +185,7 @@ function projectUserText(text: string, sessionLabels: readonly string[]): ReactN
     const referenceKind = kind === 'session'
       ? 'session'
       : label.startsWith('@')
-        ? label.endsWith('/') ? 'folder' : /[./\\]/u.test(label.slice(1)) ? 'file' : 'session'
+        ? label.endsWith('/') ? 'folder' : 'file'
         : undefined
     const displayLabel = referenceKind === undefined
       ? label

packages/client/ui-conversation/src/client/conversation-nodes/chat-snapshot-builder.ts

@@ -5,6 +5,7 @@ import type {
   ConversationViewBuilder, ConversationViewDefinition, LegacyConversationSlice,
   PartialAssistant, RunningToolCall,
 } from '@deepseek-ai/dsh-client-runtime/client'
+import { sessionRecallLabels } from '@deepseek-ai/dsh-client-runtime/client'
 import type { ChatNode } from '../contract/chat-nodes.ts'
 import { isRunningTool } from '../contract/chat-nodes.ts'
 
@@ -138,6 +139,99 @@ function orderedVisible(nodes: readonly ChatConversationViewNode[]): ChatConvers
     .sort((left, right) => left.anchorSeq - right.anchorSeq || left.key.localeCompare(right.key))
 }
 
+function referenceMessageSeq(node: ChatConversationViewNode): number | undefined {
+  const candidate = node as ChatNode
+  return candidate.kind === 'user' || candidate.kind === 'steering'
+    ? candidate.data.seq
+    : undefined
+}
+
+function followingRecall(node: ChatConversationViewNode): {
+  readonly messageSeq: number
+  readonly labels: readonly string[]
+} | undefined {
+  const candidate = node as ChatNode
+  if (candidate.kind !== 'context') return undefined
+  return {
+    messageSeq: candidate.data.seq - 1,
+    labels: sessionRecallLabels(candidate.data.source),
+  }
+}
+
+function withReferenceLabels(
+  node: ChatConversationViewNode,
+  labels: readonly string[],
+): ChatConversationViewNode {
+  const candidate = node as ChatNode
+  if (candidate.kind !== 'user' && candidate.kind !== 'steering') return node
+  const current = candidate.data.referenceLabels ?? EMPTY_KEYS
+  const hasLabels = Object.hasOwn(candidate.data, 'referenceLabels')
+  if (sameReferences(current, labels) && hasLabels === (labels.length > 0)) return node
+  const data: Record<string, unknown> = { ...candidate.data }
+  if (labels.length === 0) delete data.referenceLabels
+  else data.referenceLabels = labels
+  return { ...candidate, data }
+}
+
+/** Associates a direct message with the sourced recall event that immediately follows it. */
+class ReferenceLabelProjector {
+  private readonly messagesBySeq = new Map<number, string>()
+  private readonly labelsByMessageSeq = new Map<number, readonly string[]>()
+
+  replace(nodes: readonly ChatConversationViewNode[]): readonly ChatConversationViewNode[] {
+    this.messagesBySeq.clear()
+    this.labelsByMessageSeq.clear()
+    for (const node of nodes) {
+      const messageSeq = referenceMessageSeq(node)
+      if (messageSeq !== undefined) this.messagesBySeq.set(messageSeq, node.key)
+      const recall = followingRecall(node)
+      if (recall !== undefined && recall.labels.length > 0) {
+        this.labelsByMessageSeq.set(recall.messageSeq, recall.labels)
+      }
+    }
+    return nodes.map((node) => {
+      const messageSeq = referenceMessageSeq(node)
+      return messageSeq === undefined
+        ? node
+        : withReferenceLabels(node, this.labelsByMessageSeq.get(messageSeq) ?? EMPTY_KEYS)
+    })
+  }
+
+  apply(
+    upserts: readonly ChatConversationViewNode[],
+    store: ChatNodeStore,
+  ): readonly ChatConversationViewNode[] {
+    const byKey = new Map(upserts.map(node => [node.key, node]))
+    const affected = new Set<number>()
+    for (const node of upserts) {
+      const messageSeq = referenceMessageSeq(node)
+      if (messageSeq !== undefined) {
+        this.messagesBySeq.set(messageSeq, node.key)
+        affected.add(messageSeq)
+      }
+      const recall = followingRecall(node)
+      if (recall === undefined) continue
+      const current = this.labelsByMessageSeq.get(recall.messageSeq)
+      if (recall.labels.length === 0) this.labelsByMessageSeq.delete(recall.messageSeq)
+      else {
+        this.labelsByMessageSeq.set(
+          recall.messageSeq,
+          current !== undefined && sameReferences(current, recall.labels) ? current : recall.labels,
+        )
+      }
+      affected.add(recall.messageSeq)
+    }
+    for (const messageSeq of affected) {
+      const key = this.messagesBySeq.get(messageSeq)
+      if (key === undefined) continue
+      const node = byKey.get(key) ?? store.get(key)
+      if (node === undefined) continue
+      byKey.set(key, withReferenceLabels(node, this.labelsByMessageSeq.get(messageSeq) ?? EMPTY_KEYS))
+    }
+    return [...byKey.values()]
+  }
+}
+
 interface LegacyContribution {
   readonly anchorSeq: number
   readonly nodes: readonly ConversationNode[]
@@ -384,6 +478,7 @@ export class ChatSnapshotBuilder implements ConversationViewBuilder<ChatConversa
   private readonly store = new MutableChatNodeStore()
   private readonly locations = new MutableChatLocationIndex()
   private readonly legacy = new LegacySliceBuilder()
+  private readonly referenceLabels = new ReferenceLabelProjector()
   private order: readonly string[] = EMPTY_KEYS
   readonly empty: ChatSnapshot
 
@@ -395,19 +490,21 @@ export class ChatSnapshotBuilder implements ConversationViewBuilder<ChatConversa
     readonly nodes: readonly ChatConversationViewNode[]
     readonly timeline: ConversationTimelineSnapshot
   }): ChatSnapshot {
-    this.store.replace(input.nodes)
-    this.order = orderedVisible(input.nodes).map(node => node.key)
+    const nodes = this.referenceLabels.replace(input.nodes)
+    this.store.replace(nodes)
+    this.order = orderedVisible(nodes).map(node => node.key)
     this.locations.rebuild(this.order, this.store)
-    return this.snapshot(input.timeline, this.legacy.replace(input.nodes, input.timeline))
+    return this.snapshot(input.timeline, this.legacy.replace(nodes, input.timeline))
   }
 
   apply(input: {
     readonly upserts: readonly ChatConversationViewNode[]
     readonly timeline: ConversationTimelineSnapshot
   }): ChatSnapshot {
+    const upserts = this.referenceLabels.apply(input.upserts, this.store)
     let structural = false
     const contentOnly: ChatConversationViewNode[] = []
-    for (const node of input.upserts) {
+    for (const node of upserts) {
       const previous = this.store.get(node.key)
       const nodeStructural = previous === undefined
         || previous.anchorSeq !== node.anchorSeq
@@ -416,14 +513,14 @@ export class ChatSnapshotBuilder implements ConversationViewBuilder<ChatConversa
       structural ||= nodeStructural
       if (!nodeStructural) contentOnly.push(node)
     }
-    this.store.upsert(input.upserts)
+    this.store.upsert(upserts)
     if (structural) {
       const next = orderedVisible(this.store.values()).map(node => node.key)
       this.order = sameReferences(this.order, next) ? this.order : next
       this.locations.rebuild(this.order, this.store)
     }
     this.locations.touch(contentOnly)
-    return this.snapshot(input.timeline, this.legacy.apply(input.upserts, input.timeline))
+    return this.snapshot(input.timeline, this.legacy.apply(upserts, input.timeline))
   }
 
   private snapshot(

packages/client/ui-conversation/src/client/conversation-nodes/message.ts

@@ -3,18 +3,18 @@ import type {
   ContextMessageNode, ConversationNodeDefinition, SteeringMessageNode, UserMessageNode,
 } from '@deepseek-ai/dsh-client-runtime/client'
 import {
-  contextForm, contextProvenance, isAppendSurfaceEvent, isReplacementSurfaceEvent, sessionRecallLabels,
+  contextForm, contextProvenance, isAppendSurfaceEvent, isReplacementSurfaceEvent,
 } from '@deepseek-ai/dsh-client-runtime/client'
 import type { InboxState } from './inbox.ts'
 import { chatNode } from './common.ts'
 
 interface ReferencedUserMessageNode extends UserMessageNode {
-  /** Labels cited by the immediately preceding session-reference context. */
+  /** Labels cited by the immediately following session-reference context. */
   readonly referenceLabels?: readonly string[]
 }
 
 interface ReferencedSteeringMessageNode extends SteeringMessageNode {
-  /** Labels cited by the immediately preceding session-reference context. */
+  /** Labels cited by the immediately following session-reference context. */
   readonly referenceLabels?: readonly string[]
 }
 
@@ -61,11 +61,6 @@ export const messageDefinition: ConversationNodeDefinition<MessageNode> = {
       }
     }
     const claimed = reader.previous<InboxState>('inbox-next-step')?.state.claimed.has(String(event.data.id)) === true
-    const previous = reader.previous<MessageNode>('input-message')
-    const labels = previous?.state.kind === 'context' && previous.state.seq + 1 === event.seq
-      ? sessionRecallLabels(previous.state.source)
-      : []
-    const referenceLabels = labels.length === 0 ? {} : { referenceLabels: labels }
     return claimed
       ? {
         kind: 'steering',
@@ -74,7 +69,6 @@ export const messageDefinition: ConversationNodeDefinition<MessageNode> = {
         time: event.time,
         content: event.data.content,
         source: event.data.source,
-        ...referenceLabels,
       }
       : {
         kind: 'user',
@@ -82,7 +76,6 @@ export const messageDefinition: ConversationNodeDefinition<MessageNode> = {
         time: event.time,
         content: event.data.content,
         source: event.data.source,
-        ...referenceLabels,
       }
   },
   update: context => context.state,

packages/client/ui-conversation/src/client/input/facade.ts

@@ -17,7 +17,7 @@ import type {
   PasteComponent, QueuedMessage, SessionInput, SubmitAttempt,
 } from './contract.ts'
 import type { InputSubmitMode } from '../contract/composer-submission.ts'
-import { InputMachine } from './machine.ts'
+import { InputMachine, projectClipboard } from './machine.ts'
 
 /** Popup face the shell needs (dismissal only; typed structurally to avoid a value import). */
 export interface PopupDismissFace {
@@ -98,7 +98,7 @@ export class SessionInputShell implements SessionInput {
   // production (the machine's no-clock default is a constant for pure tests).
   private readonly core = new InputMachine({ now: () => Date.now() })
   private noticeSeq = 0
-  private lastDraft = ''
+  private lastMirroredDraft = ''
   private imageIds: readonly DraftAttachmentId[] = []
   /** One image-only send at a time: Enter during the Host round-trip is a no-op. */
   private imageSendInFlight = false
@@ -596,9 +596,10 @@ export class SessionInputShell implements SessionInput {
   private publish(): void {
     const next = this.compose()
     this.state.set(next)
-    if (next.draft !== this.lastDraft) {
-      this.lastDraft = next.draft
-      this.mirrorFn?.(next.draft)
+    const mirroredDraft = projectClipboard(next)
+    if (mirroredDraft !== this.lastMirroredDraft) {
+      this.lastMirroredDraft = mirroredDraft
+      this.mirrorFn?.(mirroredDraft)
     }
   }
 }

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

@@ -147,6 +147,11 @@
   pointer-events: none;
 }
 
+.backdropDisabled,
+.backdropDisabled :is(.hlToken, .hint, .textRef, .chip, .chipInvalid) {
+  color: var(--dsw-alias-label-tertiary);
+}
+
 .hlToken {
   background-color: transparent;
   color: var(--dsw-alias-state-warn-label);
@@ -191,6 +196,7 @@
   outline: none;
   background: transparent;
   color: transparent;
+  -webkit-text-fill-color: transparent;
   /* Business blue, not brand-primary: that token resolves to ink in this sheet. */
   caret-color: var(--dsw-alias-state-business-primary);
 }
@@ -230,12 +236,15 @@
 /* figma 34:10434: #ADB2B8 light / #81858C dark — the caption pair exactly. */
 .input::placeholder {
   color: var(--dsw-alias-label-caption);
+  -webkit-text-fill-color: var(--dsw-alias-label-caption);
   user-select: none;
 }
 
-/* Running lock: grayed but the draft stays visible; the turn ending re-enables. */
+/* The backdrop owns disabled draft color; the textarea remains caret-only so
+   its marker glyphs cannot cover the reference icons beneath it. */
 .input:disabled {
-  color: var(--dsw-alias-label-tertiary);
+  color: transparent;
+  -webkit-text-fill-color: transparent;
   cursor: not-allowed;
 }
 

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

@@ -287,7 +287,7 @@ export function InputBar({
     }
     // Absent machine without a Workspace recovery action stays disabled; the
     // guard narrows the faces for the paths below.
-    if (keyboard === undefined || inputActions === undefined) return
+    if (input === undefined || keyboard === undefined || inputActions === undefined) return
     // Shift+Enter is the native newline UNCONDITIONALLY — decided before the
     // IME guard so a composition-closing Shift+Enter still breaks the line.
     if (e.key === 'Enter' && e.shiftKey) return
@@ -651,7 +651,14 @@ export function InputBar({
             which a compositor-driven gesture outruns and leaves the words trailing the caret. */}
         <div ref={scrollRef} className={css.scroll} data-input-scroll>
           <div className={css.grow}>
-            <div aria-hidden className={css.backdrop} data-input-backdrop>{backdrop}</div>
+            <div
+              aria-hidden
+              className={clsx(css.backdrop, textareaDisabled && css.backdropDisabled)}
+              data-input-backdrop
+              data-disabled={textareaDisabled || undefined}
+            >
+              {backdrop}
+            </div>
             <textarea
               ref={inputRef}
               className={css.input}

packages/client/ui-conversation/tests/chat-branch-tails.client.spec.tsx

@@ -107,6 +107,41 @@ describe('MessageItem arms', () => {
     expect(view.getByText('引用会话 · 你好')).toBeTruthy()
   })
 
+  it('renders the complete metadata-confirmed multi-word session label', () => {
+    const view = render(
+      <MessageItem
+        t={t}
+        referenceLabels={['Research notes']}
+        node={{
+          kind: 'user',
+          seq: 1,
+          time: 1_000,
+          content: [{ type: 'text', text: '@Research notes what changed?' }] as never,
+          source: null,
+        }}
+      />,
+    )
+    expect(view.container.querySelector('[data-ref-chip="session"]')?.textContent).toBe('Research notes')
+    expect(view.getByText('what changed?')).toBeTruthy()
+    expect(view.getByText('引用会话 · Research notes')).toBeTruthy()
+  })
+
+  it('renders no-extension paths as files and leaves sentence punctuation outside the reference', () => {
+    const view = render(
+      <MessageItem t={t} node={{
+        kind: 'user',
+        seq: 1,
+        time: 1_000,
+        content: [{ type: 'text', text: 'Read @Dockerfile and @src/README.md, please.' }] as never,
+        source: null,
+      }} />,
+    )
+    const files = [...view.container.querySelectorAll('[data-ref-chip="file"]')]
+    expect(files.map(file => file.textContent)).toEqual(['Dockerfile', 'README.md'])
+    expect(files.every(file => file.querySelector('svg') !== null)).toBe(true)
+    expect(view.container.textContent).toContain('README.md, please.')
+  })
+
   it('user bubbles expose clock / copy and neither branch nor edit; copy writes the text', () => {
     const writeText = vi.fn().mockResolvedValue(undefined)
     Object.defineProperty(navigator, 'clipboard', {

packages/client/ui-conversation/tests/conversation-node-definitions.client.spec.ts

@@ -528,49 +528,98 @@ describe('built-in conversation node Definitions', () => {
     })
   })
 
-  it('associates session-reference labels inside the adjacent direct-message node', () => {
-    const referenceSource = {
-      kind: 'session-reference',
-      form: 'recall',
-      version: 1,
-      references: [
-        { sessionId: 'source-a', label: 'Research' },
-        { sessionId: 'source-b', label: 'Review' },
-      ],
-    }
+  it('associates each direct message with its immediately following session recall', () => {
     const value = assembler([
-      at(1, 'user/message', {
-        ...textMessage('reference-context', 'snapshot'),
-        source: referenceSource,
+      at(1, 'user/message', textMessage('citing-research', '@Research notes what changed?'), { surfaceOp: 'append' }),
+      at(2, 'user/message', {
+        ...textMessage('research-context', 'snapshot'),
+        source: {
+          kind: 'session-reference',
+          form: 'recall',
+          version: 1,
+          references: [{ sessionId: 'source-a', label: 'Research notes' }],
+        },
+      }, { surfaceOp: 'append' }),
+      at(3, 'user/message', textMessage('citing-review', '@Review next'), { surfaceOp: 'append' }),
+      at(4, 'user/message', {
+        ...textMessage('review-context', 'snapshot'),
+        source: {
+          kind: 'session-reference',
+          form: 'recall',
+          version: 1,
+          references: [{ sessionId: 'source-b', label: 'Review' }],
+        },
       }, { surfaceOp: 'append' }),
-      at(2, 'user/message', textMessage('citing-user', '@Research and @Review'), { surfaceOp: 'append' }),
-      at(4, 'user/message', textMessage('later-user', 'unrelated'), { surfaceOp: 'append' }),
+      at(6, 'user/message', textMessage('later-user', 'unrelated'), { surfaceOp: 'append' }),
     ])
 
     const current = snapshot(value)
+    const messages = [...current.nodes.values()]
+      .filter(candidate => candidate.kind === 'user' || candidate.kind === 'context')
     const users = [...current.nodes.values()].filter(candidate => candidate.kind === 'user')
-    expect(users[0]?.data).toMatchObject({ referenceLabels: ['Research', 'Review'] })
-    expect(users[1]?.data).not.toHaveProperty('referenceLabels')
+    expect(messages.map(candidate => candidate.kind)).toEqual(['user', 'context', 'user', 'context', 'user'])
+    expect(users[0]?.data).toMatchObject({ referenceLabels: ['Research notes'] })
+    expect(users[1]?.data).toMatchObject({ referenceLabels: ['Review'] })
+    expect(users[2]?.data).not.toHaveProperty('referenceLabels')
   })
 
-  it('keeps a direct message before its following session-reference context', () => {
+  it('updates an already published direct node when its following recall arrives', () => {
     const value = assembler([
-      at(1, 'user/message', textMessage('citing-user', '@Research what changed?'), { surfaceOp: 'append' }),
-      at(2, 'user/message', {
-        ...textMessage('reference-context', 'snapshot'),
+      at(1, 'user/message', textMessage('citing-user', '@Research notes what changed?'), { surfaceOp: 'append' }),
+    ])
+    const before = node(snapshot(value), 'user')
+    expect(before?.data).not.toHaveProperty('referenceLabels')
+
+    value.append(at(2, 'user/message', {
+      ...textMessage('reference-context', 'snapshot'),
+      source: {
+        kind: 'session-reference',
+        form: 'recall',
+        version: 1,
+        references: [{ sessionId: 'source-a', label: 'Research notes' }],
+      },
+    }, { surfaceOp: 'append' }))
+    value.flush()
+
+    const current = snapshot(value)
+    const nodes = [...current.nodes.values()]
+      .filter(candidate => candidate.kind === 'user' || candidate.kind === 'context')
+    expect(nodes.map(candidate => candidate.kind)).toEqual(['user', 'context'])
+    expect(nodes[0]?.key).toBe(before?.key)
+    expect(nodes[0]?.data).toMatchObject({ referenceLabels: ['Research notes'] })
+    expect(current.legacy.nodes[0]).toMatchObject({ referenceLabels: ['Research notes'] })
+  })
+
+  it('associates a claimed steering message with its following recall', () => {
+    const steering = textMessage('steering-reference', '@Research notes continue')
+    const value = assembler([
+      at(1, 'agent/inbox/spliced', {
+        target: 'next-step',
+        start: 0,
+        inserted: [steering],
+      }),
+      at(2, 'agent/inbox/spliced', {
+        target: 'next-step',
+        start: 0,
+        removedCount: 1,
+        inserted: [],
+      }),
+      at(3, 'user/message', steering, { surfaceOp: 'append' }),
+      at(4, 'user/message', {
+        ...textMessage('steering-reference-context', 'snapshot'),
         source: {
           kind: 'session-reference',
           form: 'recall',
           version: 1,
-          references: [{ sessionId: 'source-a', label: 'Research' }],
+          references: [{ sessionId: 'source-a', label: 'Research notes' }],
         },
       }, { surfaceOp: 'append' }),
     ])
 
-    const nodes = [...snapshot(value).nodes.values()]
-      .filter(candidate => candidate.kind === 'user' || candidate.kind === 'context')
-    expect(nodes.map(candidate => candidate.kind)).toEqual(['user', 'context'])
-    expect(nodes[0]?.data).not.toHaveProperty('referenceLabels')
+    expect(node(snapshot(value), 'steering')?.data).toMatchObject({
+      messageId: 'steering-reference',
+      referenceLabels: ['Research notes'],
+    })
   })
 
   it('keeps replacement copies out of Chat business nodes', () => {

packages/client/ui-conversation/tests/input-bar.client.spec.tsx

@@ -1182,6 +1182,21 @@ describe('decorations', () => {
     expect(shell.snapshot.occurrences[0]).toMatchObject({ offset: 3, length: 4 })
   })
 
+  it('keeps the textarea glyph layer transparent when a structured reference becomes disabled', () => {
+    const { view, shell, session, textarea } = bench()
+    act(() => {
+      shell.setDraft('@w1')
+      shell.insertReference({
+        source: 'reference', ref: 'w1', label: '会话一', appearance: 'session', clipboardText: '@w1',
+      }, { start: 0, end: 3, draftRev: shell.snapshot.draftRev })
+      session.set(snapshotOf({ removed: true }))
+    })
+    const backdrop = view.container.querySelector('[data-input-backdrop]')
+    expect(textarea.disabled).toBe(true)
+    expect(backdrop?.getAttribute('data-disabled')).toBe('true')
+    expect(backdrop?.querySelector('[data-decoration="chip"] svg')).not.toBeNull()
+  })
+
   it('Backspace and Delete remove a reference as one range at its boundaries', () => {
     const reference = {
       source: 'reference', ref: 'w1', label: '会话一', appearance: 'session' as const, clipboardText: '@w1',

packages/client/ui-conversation/tests/input-reference-submit.client.spec.ts

@@ -10,6 +10,7 @@ import { SessionInputShell } from '../src/client/input/facade.ts'
 import type { DraftAttachmentId } from '../src/client/input/contract.ts'
 
 const mention = '@[Research](dsh-session:InNvdXJjZSI)'
+const spacedMention = '@[Research notes](dsh-session:InNvdXJjZSI)'
 const commandImages = {
   serialize: () => Promise.resolve([]),
   release: () => {},
@@ -32,6 +33,42 @@ function chip(shell: SessionInputShell): void {
 }
 
 describe('reference submission', () => {
+  it('mirrors canonical reference text so a persisted draft remains resolvable after remount', async () => {
+    const mirror = vi.fn()
+    const first = new SessionInputShell({
+      actx: {} as ClientContext,
+      defaultSink: vi.fn(),
+      commandImages,
+    })
+    first.bindMirror(mirror)
+    first.setDraft('@res')
+    expect(first.insertReference({
+      source: 'reference',
+      ref: spacedMention,
+      label: 'Research notes',
+      appearance: 'session',
+      clipboardText: spacedMention,
+    }, {
+      start: 0,
+      end: 4,
+      draftRev: first.snapshot.draftRev,
+    })).toBe(true)
+    expect(first.snapshot.draft).toBe('@Research notes ')
+    expect(mirror).toHaveBeenLastCalledWith(`${spacedMention} `)
+
+    const sink = vi.fn(() => Promise.resolve<SubmitOutcome>({ kind: 'success' }))
+    const restored = new SessionInputShell({
+      actx: {} as ClientContext,
+      defaultSink: sink,
+      commandImages,
+    })
+    restored.setDraft(mirror.mock.calls.at(-1)?.[0] as string)
+    restored.submit()
+    await vi.waitFor(() => {
+      expect(sink).toHaveBeenCalledWith(spacedMention, [], 'queue', expect.any(AbortSignal))
+    })
+  })
+
   it('retains the chip on Host failure and clears it only after a later accepted retry', async () => {
     const serializeReference = vi.fn(() => Promise.resolve(mention))
     const sink = vi.fn<(

packages/context/session-reference/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/context/session-reference/README.md
-README.md: 71ed6891a3a0bee480b2064d844df8758349ee38
-README.zh.md: 5a761115105914a47c1f3d2673df4b2f3e3e76da
+README.md: 1ce5ab0a3b0cb5342c1087ddd644933a92b2a596
+README.zh.md: 2fc30fa01d0e428a2e7ef0be1680a1937f06660c

packages/context/session-reference/README.md

@@ -32,7 +32,7 @@ Retention applies `maxReferenceBytes` independently to each source, keeps compac
 
 #### What the model sees
 
-The model sees two consecutive user-role messages: the current message with its readable `@label`, then the `## Referenced sessions` untrusted snapshot. The warning forbids following instructions, permission claims, or tool requests from the snapshot unless the citing user message requests their use. Labels, cwd values, ids, and conversation text are serialized as JSON inside `<referenced-sessions>` tags; every data `<` is emitted as the lossless JSON escape `\u003c`, so source text cannot spell a framing tag.
+The model sees two consecutive user-role messages: the current message with its readable `@label`, then the `## Referenced sessions` untrusted snapshot. The warning forbids following instructions, permission claims, or tool requests from the snapshot unless the current user explicitly repeats them. Labels, cwd values, ids, and conversation text are serialized as JSON inside `<referenced-sessions>` tags; every data `<` is emitted as the lossless JSON escape `\u003c`, so source text cannot spell a framing tag.
 
 #### Token effect
 

packages/context/session-reference/README.zh.md

@@ -32,7 +32,7 @@
 
 #### 模型看到的内容
 
-模型会看到两条连续的 user 角色消息：先是带可读 `@label` 的当前消息，再是 `## Referenced sessions` 不受信任快照。警告禁止遵循快照中的指令、权限声明或工具请求，除非引用它的 user 消息要求使用这些内容。标签、cwd 值、id 与会话文本会作为 JSON 在 `<referenced-sessions>` 标签中序列化；数据中的每个 `<` 都会以无损 JSON 转义 `\u003c` 的形式发出，因此源文本无法拼出定界标签。
+模型会看到两条连续的 user 角色消息：先是带可读 `@label` 的当前消息，再是 `## Referenced sessions` 不受信任快照。警告禁止遵循快照中的指令、权限声明或工具请求，除非当前用户明确重复这些内容。标签、cwd 值、id 与会话文本会作为 JSON 在 `<referenced-sessions>` 标签中序列化；数据中的每个 `<` 都会以无损 JSON 转义 `\u003c` 的形式发出，因此源文本无法拼出定界标签。
 
 #### Token 影响