fix(chat): preserve retry attachments and per-answer context snapshots

backend/app/contracts.py

@@ -312,6 +312,7 @@ class ConversationListResponse(Contract):
 
 
 class ChatMessage(Contract):
+    context_captured: bool = False
     attachments: list[str] = Field(default_factory=list)
     workspace_context: WorkspaceContext | None = None
     activity: list[dict[str, Any]] = Field(default_factory=list)

backend/app/database/migrations.py

@@ -171,6 +171,7 @@ MIGRATIONS: list[str] = [
     """ALTER TABLE chat_conversations ADD COLUMN active_response_id TEXT;""",
     """ALTER TABLE chat_messages ADD COLUMN workspace_context_json TEXT;""",
     """ALTER TABLE chat_messages ADD COLUMN attachments_json TEXT NOT NULL DEFAULT '[]';""",
+    """ALTER TABLE chat_messages ADD COLUMN context_captured INTEGER NOT NULL DEFAULT 0;""",
 ]
 
 

backend/app/routes.py

@@ -506,6 +506,9 @@ async def chat(request: ChatRequest) -> StreamingResponse:
                     usage=usage,
                     activity=activity,
                     parent_message_id=user_message_id,
+                    workspace_context=request.workspace_context.model_dump() if request.workspace_context else None,
+                    attachments=request.attachments,
+                    context_captured=True,
                 )
 
     return StreamingResponse(stream(), media_type="text/event-stream")

backend/app/services/chat_history.py

@@ -39,6 +39,7 @@ def _message(row) -> ChatMessage:
         thinking=row["thinking"],
         activity=json.loads(row['activity_json']),
         attachments=json.loads(row['attachments_json']),
+        context_captured=bool(row['context_captured']),
         workspace_context=json.loads(row['workspace_context_json']) if row['workspace_context_json'] else None,
         citations=citations,
         tool_calls=json.loads(row["tool_calls_json"]),
@@ -130,6 +131,7 @@ def append_message(
     parent_message_id: str | None = None,
     workspace_context: dict | None = None,
     attachments: list[str] | None = None,
+    context_captured: bool = False,
 ) -> None:
     now = _now().isoformat()
     clean_title = (title or "").strip() or content[:30].strip() or "New conversation"
@@ -139,7 +141,7 @@ def append_message(
             _append_message_in_transaction(
                 conn, conversation_id, message_id=message_id, role=role, content=content,
                 title=clean_title, thinking=thinking, citations=citations, tool_calls=tool_calls,
-                usage=usage, now=now, activity=activity, parent_message_id=parent_message_id, workspace_context=workspace_context, attachments=attachments,
+                usage=usage, now=now, activity=activity, parent_message_id=parent_message_id, workspace_context=workspace_context, attachments=attachments, context_captured=context_captured,
             )
             conn.execute("COMMIT")
         except BaseException:
@@ -165,6 +167,7 @@ def _append_message_in_transaction(
     parent_message_id: str | None = None,
     workspace_context: dict | None = None,
     attachments: list[str] | None = None,
+    context_captured: bool = False,
 ) -> None:
     conversation = conn.execute(
         "SELECT 1 FROM chat_conversations WHERE conversation_id=?", (conversation_id,)
@@ -215,6 +218,7 @@ def _append_message_in_transaction(
     conn.execute('UPDATE chat_messages SET parent_message_id=?, activity_json=? WHERE message_id=?', (parent, json.dumps(activity or [], ensure_ascii=False), message_id))
     conn.execute('UPDATE chat_messages SET workspace_context_json=? WHERE message_id=?', (json.dumps(workspace_context, ensure_ascii=False) if workspace_context is not None else None, message_id))
     conn.execute('UPDATE chat_messages SET attachments_json=? WHERE message_id=?', (json.dumps(attachments or []),message_id))
+    conn.execute('UPDATE chat_messages SET context_captured=? WHERE message_id=?', (int(context_captured), message_id))
     # A late stream may be persisted, but must not steal the selected branch.
     response_id = conn.execute('SELECT active_response_id FROM chat_conversations WHERE conversation_id=?', (conversation_id,)).fetchone()[0]
     if active_leaf == parent and (role != 'assistant' or response_id is None or response_id == message_id):

backend/tests/test_chat_versions.py

@@ -50,3 +50,40 @@ def test_workspace_snapshots_and_agent_links_survive_history_reload():
     assert total == 2
     assert messages[0].workspace_context.model_dump() == snapshot
     assert messages[1].tool_calls == calls
+
+
+def test_regeneration_persists_context_per_answer_without_rewriting_original(monkeypatch):
+    import asyncio
+    from types import SimpleNamespace
+    from app.contracts import ChatRequest, Message, ModelEvent, ModelEventType
+    from app.routes import chat, utc_now
+    received=[]
+    class Adapter:
+        async def stream(self, request):
+            received.append(request)
+            yield ModelEvent(event=ModelEventType.text_delta, sequence=0, data={'text':'answer'}, timestamp=utc_now())
+            yield ModelEvent(event=ModelEventType.done, sequence=1, data={}, timestamp=utc_now())
+    monkeypatch.setattr('app.routes.provider_or_404',lambda _:SimpleNamespace(adapter=Adapter()))
+    # Keep attachment parsing out of this persistence test; the route must save raw IDs.
+    async def prepare(request, provider):
+        return request.model_copy(update={'attachments':[]})
+    monkeypatch.setattr('app.services.chat_attachments.prepare',prepare)
+    async def scenario():
+        history.create('Snapshots','snapshots')
+        for index,context in enumerate([{'file_path':'a.md','content':'A'},{'file_path':'b.md','content':'B'},None]):
+            req=ChatRequest(provider_id='test',model='test',use_rag=False,conversation_id='snapshots',
+                user_message_id='su',assistant_message_id=f'sa{index}',retry_message_id=f'sa{index-1}' if index else None,
+                messages=[Message(role='user',content='explain')],workspace_context=context,attachments=[f'file{index}.md'])
+            response=await chat(req)
+            _=[chunk async for chunk in response.body_iterator]
+        for index,path in enumerate(['a.md','b.md',None]):
+            history.select_version('snapshots',f'sa{index}')
+            messages,_=history.list_messages('snapshots',100,0)
+            assert messages[0].workspace_context.file_path=='a.md'
+            answer=messages[-1]
+            assert answer.context_captured
+            assert (answer.workspace_context.file_path if answer.workspace_context else None)==path
+            assert answer.attachments==[f'file{index}.md']
+        assert 'b.md' in received[1].system
+        assert received[2].system is None
+    asyncio.run(scenario())

docs/contracts/第二阶段接口契约-开发版.md

@@ -1609,3 +1609,7 @@ CUDA 组件：`GET /api/local-models/runtime-components/cuda` 返回 status、st
 `/api/media/attachments` 新增允许 DOCX、PPTX、PPT、PNG、JPG/JPEG、WebP 后缀。聊天通过 `ChatRequest.attachments` 提交最多 8 个持久化附件 ID，并通过 `ChatMessage.attachments` 恢复记录。`image_fallback_tools` 最多两个注册工具名，服务端固定 MCP 优先、Plugin 次之，不接受任意命令或远程下载 URL。
 
 内部模型 `Message.images` 使用有大小限制的 PNG/JPEG/WebP base64 data URI，Provider 适配器转换为各自原生协议。文档和音频提取为参考文本后才交给普通聊天，清除已解析的二进制附件标记，使文本上下文检测仍可工作。附件失败返回 `CHAT_ATTACHMENT_FAILED`，进度与截断提示使用 `ContextStatus`，不将失败附件当作已读取内容。
+
+#### 回答版本的上下文快照（2026-09-07）
+
+`ChatMessage.context_captured` 为布尔值，旧记录默认 false。新 assistant 消息保存本次请求的 `workspace_context` 和 `attachments`，并设置 context_captured 为 true；此时 null 文件上下文和空附件列表都是明确快照。重新生成不覆盖原 user 消息的快照。客户端恢复旧记录时仅在 context_captured 为 false 时回退到对应父用户消息。

docs/development/聊天按需检索与Markdown工具.md

@@ -93,3 +93,11 @@ AI 消息提供“重新生成”，用户消息提供“编辑”及“保存
 内置 `chat-operator` Skill 与 `chat-policy` Plugin 随 Host 注册。Plugin 的 `chat-policy.plan` 使用宿主白名单 handler 校验任务与预算，生成读取、执行和核验步骤，不执行任意插件代码。聊天委托创建运行前调用检查；启用的 Skill 加入聊天系统提示词并作为委托运行的 Skill，权限继续由 Agent 管理。用户禁用扩展后不会自动重新启用。
 
 新增验证：`test_chat_attachments.py` 覆盖 Office/Markdown 文本抽取、旧 PPT 文本记录、截断、音频任务路由、原生视觉优先及 MCP 超时后 Plugin 降级；浮窗测试覆盖尺寸记忆和重置。浏览器实测折叠设置、上传入口和拖动缩放。未使用真实外部模型或 MCP 服务进行付费调用验收。
+
+### 重试上下文与历史版本一致性（2026-09-07）
+
+编辑旧用户消息使用该消息的附件；重新生成回答使用该回答版本实际使用的附件和文件快照。旧版未记录回答快照时才回退到它的父用户消息，不读取会话末尾其他轮次的附件。重试不会消耗输入区尚未发送的新附件。
+
+新回答将 `workspace_context`、`attachments` 和 `context_captured=true` 一起持久化。原用户消息保持不变，因此同一问题的不同回答可以各自恢复生成时的文件内容；工作区浮窗显式传入当前文件时，以本次文件为准。`context_captured=true` 且 `workspace_context=null` 表示该版本明确未附带文件，继续对话或重试时不能回退到原用户消息的旧文件。数据库迁移为既有消息设置 false，保持旧记录可恢复。
+
+回归覆盖：两轮使用不同附件后编辑/重试第一轮、保留待发送附件、切换工作区文件后生成新版本、历史回读后继续重试、明确清空文件上下文、原版本快照保持不变。

frontend/src/contracts/index.ts

@@ -71,6 +71,7 @@ export interface Conversation {
 export interface WorkspaceContext { file_path: string; content: string }
 
 export interface ChatMessage {
+  context_captured?: boolean
   attachments?: string[]
   workspace_context?: WorkspaceContext
   activity?: Array<{ type: 'thinking'; text: string } | { type: 'tool'; tool_call_id: string }>

frontend/src/stores/chat.spec.ts

@@ -356,3 +356,38 @@ it('uploads attachments and includes their durable IDs in an attachment-only mes
   expect(store.pendingAttachments).toEqual([])
   upload.mockRestore()
 })
+
+it.each(['user', 'assistant'] as const)('retries older %s messages with their attachments, preserving pending uploads', async role => {
+  const s=useChatStore(); s.selectedProviderId='real'; s.selectedModel='model'
+  s.pendingAttachments=[{attachment_id:'first.md',name:'first.md'}]
+  await s.sendMessage('first'); vi.mocked(streamChat).mock.calls.at(-1)![1].onDone?.()
+  const old=s.messages[role === 'user' ? 0 : 1]!.message_id
+  s.pendingAttachments=[{attachment_id:'later.md',name:'later.md'}]
+  await s.sendMessage('later'); vi.mocked(streamChat).mock.calls.at(-1)![1].onDone?.()
+  s.pendingAttachments=[{attachment_id:'draft.md',name:'draft.md'}]
+  await s.retryMessage(old, role === 'user' ? 'edited first' : undefined)
+  expect(vi.mocked(streamChat).mock.calls.at(-1)![0].attachments).toEqual(['first.md'])
+  expect(s.pendingAttachments.map(a=>a.attachment_id)).toEqual(['draft.md'])
+})
+
+it('restores each answer context after history reload, including explicitly absent workspace context', async () => {
+  const s=useChatStore(); s.selectedProviderId='real'; s.selectedModel='model'
+  const first={file_path:'a.md',content:'A'}; const second={file_path:'b.md',content:'B'}
+  await s.sendMessage('explain',undefined,first); vi.mocked(streamChat).mock.calls.at(-1)![1].onDone?.()
+  const original=s.messages[1]!.message_id
+  await s.retryMessage(original,undefined,second); vi.mocked(streamChat).mock.calls.at(-1)![1].onDone?.()
+  expect(s.messages[0]!.workspace_context).toEqual(first)
+  expect(s.messages[1]!.workspace_context).toEqual(second)
+  vi.mocked(listConversationMessages).mockResolvedValue({items:JSON.parse(JSON.stringify(s.messages)),page:{total:2,limit:500,offset:0}})
+  await s.setActiveConversation(s.activeConversationId!)
+  await s.retryMessage(s.messages[1]!.message_id)
+  expect(vi.mocked(streamChat).mock.calls.at(-1)![0].workspace_context).toEqual(second)
+  vi.mocked(streamChat).mock.calls.at(-1)![1].onDone?.()
+  await s.retryMessage(s.messages[1]!.message_id,undefined,null)
+  vi.mocked(streamChat).mock.calls.at(-1)![1].onDone?.()
+  // API serializes absent captured context as null; do not fall back to the original user snapshot.
+  vi.mocked(listConversationMessages).mockResolvedValue({items:JSON.parse(JSON.stringify(s.messages)),page:{total:2,limit:500,offset:0}})
+  await s.setActiveConversation(s.activeConversationId!)
+  await s.sendMessage('continue')
+  expect(vi.mocked(streamChat).mock.calls.at(-1)![0].workspace_context).toBeUndefined()
+})

frontend/src/stores/chat.ts

@@ -181,9 +181,15 @@ export const useChatStore = defineStore('chat', () => {
   async function sendMessage(text: string, retryMessageId?: string, workspaceContext?: WorkspaceContext | null) {
     const content = text.trim() || (pendingAttachments.value.length ? '请分析附件内容' : '')
     if (!content || !canSend.value || !selectedProviderId.value || !selectedModel.value) return
-    const context = workspaceContext === undefined ? [...messages.value].reverse().find(m => m.role === 'user')?.workspace_context : workspaceContext
+    const targetIndex = retryMessageId ? messages.value.findIndex(m => m.message_id === retryMessageId) : messages.value.length - 1
+    if (retryMessageId && targetIndex < 0) return
+    const target = messages.value[targetIndex]
+    const source = target?.role === 'assistant' && !target.context_captured
+      ? messages.value[targetIndex - 1] : target
+    const context = workspaceContext === undefined ? source?.workspace_context : workspaceContext
     const snapshot = context ? { ...context } : undefined
-    const attachments = pendingAttachments.value.length ? pendingAttachments.value.map(a=>a.attachment_id) : ([...messages.value].reverse().find(m=>m.role==='user')?.attachments ?? [])
+    const attachments = !retryMessageId && pendingAttachments.value.length
+      ? pendingAttachments.value.map(a => a.attachment_id) : [...(source?.attachments ?? [])]
     const version = ++streamVersion
     isPreparing.value = true
     historyError.value = ''
@@ -217,6 +223,7 @@ export const useChatStore = defineStore('chat', () => {
     const aiMsg = reactive<ChatMessage>({
       message_id: crypto.randomUUID(), conversation_id: conversationId, role: 'assistant', content: '',
       created_at: new Date().toISOString(), citations: [], tool_calls: [], activity: [],
+      context_captured: true, workspace_context: snapshot, attachments: [...attachments],
     })
     if (retryTarget) {
       messages.value = messages.value.slice(0, retryIndex)
@@ -226,7 +233,7 @@ export const useChatStore = defineStore('chat', () => {
     if (!regenerate) messages.value.push(userMsg)
     messages.value.push(aiMsg)
     inputText.value = ''
-    pendingAttachments.value = []
+    if (!retryMessageId) pendingAttachments.value = []
     isStreaming.value = true
     conversation.updated_at = new Date().toISOString()
     conversation.message_count = messages.value.length
@@ -322,7 +329,7 @@ export const useChatStore = defineStore('chat', () => {
     const message = messages.value[index]
     if (!message) return
     const text = message.role === 'user' ? editedText : messages.value[index - 1]?.content
-    if (text?.trim()) await sendMessage(text, messageId, workspaceContext !== undefined ? workspaceContext : (message.role === 'user' ? message.workspace_context : messages.value[index - 1]?.workspace_context))
+    if (text?.trim()) await sendMessage(text, messageId, workspaceContext)
   }
 
   async function switchVersion(messageId: string) {