feat(frontend): enhance chat stream handling with assistant message management

- Introduce mechanisms to mark assistant messages as completed and handle in-flight messages during user queries.
- Implement functions to prepare for new outgoing messages and manage the stopping of assistant messages.
- Update chat stream handling to ensure proper message state management and prevent reuse of incomplete messages.
- Integrate new functionalities into existing composables for improved chat experience.
This commit is contained in:
wizardchen
2026-06-18 18:38:55 +08:00
committed by lyingbug
parent c6b5d2e77b
commit b5acb89b4d
4 changed files with 75 additions and 11 deletions
+4
View File
@@ -29,6 +29,7 @@ export function useStream() {
const error = ref<string | null>(null)// 错误信息
const lastStreamRequest = ref<StreamRequestMeta | null>(null)
let controller = new AbortController()
let streamGeneration = 0
// 流式渲染缓冲
let buffer: string[] = []
@@ -36,6 +37,7 @@ export function useStream() {
// 启动流式请求
const startStream = async (params: { session_id: any; query: any; knowledge_base_ids?: string[]; knowledge_ids?: string[]; agent_enabled?: boolean; agent_id?: string; web_search_enabled?: boolean; enable_memory?: boolean; summary_model_id?: string; mcp_service_ids?: string[]; mentioned_items?: Array<{id: string; name: string; type: string; kb_type?: string}>; images?: Array<{data: string}>; attachment_uploads?: Array<{data: string; file_name: string; file_size: number}>; method: string; url: string; embed_token?: string; embed_session_sig?: string }) => {
const myGeneration = ++streamGeneration
// 重置状态
output.value = '';
error.value = null;
@@ -159,6 +161,7 @@ export function useStream() {
},
onmessage: (ev) => {
if (myGeneration !== streamGeneration) return
const parsed = JSON.parse(ev.data);
// Log first answer chunk for end-to-end TTFB measurement.
// Filter by event type so non-answer events (references, tool
@@ -197,6 +200,7 @@ export function useStream() {
// 停止流
const stopStream = () => {
streamGeneration++
controller.abort();
controller = new AbortController(); // 重置控制器(如需重新发起)
isStreaming.value = false;
@@ -67,6 +67,53 @@ export function useChatStreamHandler(options: UseChatStreamHandlerOptions) {
return undefined
}
/** Incomplete assistant row for the current turn (must be the list tail). */
const getTrailingIncompleteAssistant = () => {
const last = messagesList[messagesList.length - 1]
if (last?.role === 'assistant' && !last.is_completed) return last
return undefined
}
const markAssistantStopped = (message: ChatMessage) => {
if (!message || message.is_completed) return
message.is_completed = true
if (message.isAgentMode) {
if (!message.agentEventStream) message.agentEventStream = []
const stream = message.agentEventStream as ChatMessage[]
if (!stream.some((e) => e.type === 'stop')) {
stream.push({
type: 'stop',
timestamp: Date.now(),
reason: 'user_requested',
})
}
}
}
/** Finalize any in-flight assistant rows before a new user query is sent. */
const prepareForNewOutgoingMessage = () => {
for (const msg of messagesList) {
if (msg.role === 'assistant' && !msg.is_completed) {
markAssistantStopped(msg)
}
}
fullContent.value = ''
currentAssistantMessageId.value = ''
}
/** Mark the assistant row being stopped without clearing its id (stop API still needs it). */
const markInFlightAssistantStopped = (messageId?: string) => {
let target: ChatMessage | undefined
if (messageId) {
target = messagesList.find(
(m) => m.id === messageId || m.request_id === messageId,
)
}
if (!target) target = getTrailingIncompleteAssistant()
if (target) markAssistantStopped(target)
fullContent.value = ''
}
const extractKnowledgeReferences = (data: ChatMessage) => {
const dataPayload = data.data as ChatMessage | undefined
const refs =
@@ -93,7 +140,7 @@ export function useChatStreamHandler(options: UseChatStreamHandlerOptions) {
})
if (matched) return matched
return findLastMessage((item) => item.role === 'assistant' && !item.is_completed)
return getTrailingIncompleteAssistant()
}
const applyKnowledgeReferences = (data: ChatMessage) => {
@@ -744,8 +791,11 @@ export function useChatStreamHandler(options: UseChatStreamHandlerOptions) {
timestamp: Date.now(),
reason: dataPayload?.reason || 'user_requested',
})
message.is_completed = true
loading.value = false
isReplying.value = false
fullContent.value = ''
currentAssistantMessageId.value = ''
break
}
}
@@ -767,9 +817,7 @@ export function useChatStreamHandler(options: UseChatStreamHandlerOptions) {
if (data.response_type === 'agent_query') {
if (data.id) {
const earlyMsg = findLastMessage(
(item) => item.role === 'assistant' && !item.is_completed,
)
const earlyMsg = getTrailingIncompleteAssistant()
if (earlyMsg) earlyMsg.request_id = data.id
}
if (data.assistant_message_id) {
@@ -788,8 +836,9 @@ export function useChatStreamHandler(options: UseChatStreamHandlerOptions) {
)
const created = !existingMessage
if (!existingMessage) {
const assistantId = data.assistant_message_id as string | undefined
existingMessage = {
id: data.id,
id: assistantId || data.id,
request_id: data.id,
role: 'assistant',
content: '',
@@ -849,6 +898,8 @@ export function useChatStreamHandler(options: UseChatStreamHandlerOptions) {
handleAgentChunk(data)
if (data.response_type === 'stop') {
log('[Stop Event] Generation stopped')
const stoppedMessage = resolveActiveAssistantMessage(data)
if (stoppedMessage) markAssistantStopped(stoppedMessage)
loading.value = false
isReplying.value = false
currentAssistantMessageId.value = ''
@@ -924,5 +975,7 @@ export function useChatStreamHandler(options: UseChatStreamHandlerOptions) {
shouldShowGlobalTypingIndicator,
handleMsgList,
processStreamChunk,
prepareForNewOutgoingMessage,
markInFlightAssistantStopped,
}
}
@@ -118,6 +118,8 @@ export function useEmbedChatSession(options: {
shouldShowGlobalTypingIndicator,
handleMsgList,
processStreamChunk,
prepareForNewOutgoingMessage,
markInFlightAssistantStopped,
} = useChatStreamHandler({
messagesList,
loading,
@@ -220,6 +222,7 @@ export function useEmbedChatSession(options: {
const handleStopGeneration = () => {
stopStream()
markInFlightAssistantStopped(currentAssistantMessageId.value)
const messageId = currentAssistantMessageId.value
if (messageId) {
stopEmbedSession(
@@ -238,6 +241,8 @@ export function useEmbedChatSession(options: {
value: string,
opts: { webSearchEnabled?: boolean; imageFiles?: File[]; attachmentFiles?: File[] } = {},
) => {
stopStream()
prepareForNewOutgoingMessage()
const outboundQuery = buildQueryWithHostContext(value, options.hostContext?.value)
const visitorWebSearchEnabled = opts.webSearchEnabled ?? false
const imageFiles = (options.allowFileUpload ? opts.imageFiles : undefined) || []
+8 -6
View File
@@ -448,6 +448,8 @@ const {
shouldShowGlobalTypingIndicator,
handleMsgList,
processStreamChunk,
prepareForNewOutgoingMessage,
markInFlightAssistantStopped,
} = useChatStreamHandler({
messagesList,
loading,
@@ -488,10 +490,6 @@ const {
},
onAgentQuery: (data, existingMessage) => {
pendingStreamDebug.value = buildStreamDebugPayload();
if (data.id) {
const earlyMsg = findLastMessage((item) => item.role === 'assistant' && !item.is_completed);
if (earlyMsg) attachStreamDebugToMessage(earlyMsg);
}
if (existingMessage) attachStreamDebugToMessage(existingMessage);
},
onMessageCreated: (message) => attachStreamDebugToMessage(message),
@@ -557,13 +555,17 @@ const getmsgList = (data, isScrollType = false, scrollHeight) => {
// 处理停止生成事件 - 立即清除 loading 状态
const handleStopGeneration = () => {
console.log('[Stop Generation] Immediately clearing loading state');
stopStream();
loading.value = false;
isReplying.value = false;
// 注意:不在这里清空 currentAssistantMessageId,因为需要它来调用 API
// API 调用成功后,后端的 stop 事件会清空它
// 标记当前 assistant 为已结束,避免下一条 query 复用该消息行
markInFlightAssistantStopped(currentAssistantMessageId.value);
// 保留 currentAssistantMessageId,Input-field 仍需用它调用 stop API
};
const sendMsg = async (value, modelId = '', mentionedItems = [], imageFiles = [], attachmentFiles = []) => {
stopStream();
prepareForNewOutgoingMessage();
isReplying.value = true;
loading.value = true;