chore(align): sync Git with running container as authoritative source

Phase 0 SSE 流式聊天 landed in the running container and captured in Git so
the three sources (container := image := Git) agree.

Phase 0 — SSE 流式聊天（Web 端，IM bot 路径保持不变、共用引擎）:
- app/api/chat.py: 新增独立 SSE 端点
  POST /app/api/chat/conversations/{conv_id}/messages/stream
  基于 asyncio.Queue + create_task 将 _send_message_core 桥接为
  text/event-stream 生成器，吐出 delta / reasoning / done / error 事件
- app/api/chat_service.py: _send_message_core 增加 stream + on_event 参数；
  普通（无工具）聊天 round-0 主调用走 achat_complete(stream=True) 流式；
  done 事件携带 message_id / content / process_steps / used_skills /
  used_mcp / generating
- app/core/llm_client.py: 修复 _create_kwargs 硬编码 stream=False 的根因 bug
  （流式调用传 stream=True 导致 TypeError: multiple values for keyword），
  现 _create_kwargs(request_body, stream=True) 字段随参数变化
- frontend/src/pages/front/ChatPage.tsx + ChatPage.css + types/index.ts:
  发消息改为 fetch SSE 消费，实时增量渲染 AI 回复文本与思考链(reasoning)，
  done 事件用 message_id 定稿并渲染过程步骤；网络/流式失败自动回退非流式

Alignment performed 2026-07-24 18:25:11. No deletions; secrets/venv/node_modules/
caches were not touched (tracked-source-only copy).

backend/python/app/api/chat_service.py

@@ -1999,6 +1999,15 @@ async def _send_message_core(
 
     process_steps = []  # 过程步骤(思考/技能/MCP执行)，用于前端收拢式展示
 
+    async def _push_step(step: dict):
+        """追加过程步骤，并在流式场景实时推送给前端(step 事件)。"""
+        await _push_step(step)
+        if on_event is not None:
+            try:
+                await on_event({"type": "step", "step": step, "index": len(process_steps) - 1})
+            except Exception:
+                pass
+
 
 
 
@@ -4489,7 +4498,7 @@ async def _send_message_core(
 
                             })
 
-                            process_steps.append({
+                            await _push_step({
 
                                 "type": "skill",
 
@@ -4791,10 +4800,11 @@ async def _send_message_core(
 
                                 try:
 
-                                    if stream and on_event is not None and round_num == 0 and not (effective_mcp_ids or effective_skill_ids or effective_agent_ids):
+                                    if stream and on_event is not None:
 
                                         _sc = ""
                                         _sr = ""
+                                        _final_msg = None
                                         async for _kind, _text in await achat_complete(
                                             api_url, api_key, request_body,
                                             proxy=_proxy_for_url(api_url),
@@ -4807,6 +4817,14 @@ async def _send_message_core(
                                             elif _kind == "reasoning":
                                                 _sr += _text
                                                 await on_event({"type": "reasoning", "content": _text})
+                                            elif _kind == "message":
+                                                _final_msg = _text
+                                        if _final_msg is not None:
+                                            _final_msg = dict(_final_msg)
+                                            if "reasoning_content" in _final_msg:
+                                                _final_msg["reasoning"] = _final_msg.pop("reasoning_content")
+                                            message = _final_msg
+                                        else:
                                             message = {"content": _sc, "tool_calls": [], "reasoning": _sr}
                                         choice = {"message": message}
                                         data = {"choices": [choice], "usage": {}}
@@ -4983,7 +5001,7 @@ async def _send_message_core(
 
                                 if _think and tool_calls:
 
-                                    process_steps.append({
+                                    await _push_step({
 
                                         "type": "thinking",
 
@@ -5469,7 +5487,7 @@ async def _send_message_core(
 
                                                     called_mcp_names.append(_srv.get("name", ""))
 
-                                                process_steps.append({
+                                                await _push_step({
 
                                                     "type": "mcp",
 
@@ -5545,7 +5563,7 @@ async def _send_message_core(
 
                                                 called_skill_names.append(_sk.get("name", ""))
 
-                                            process_steps.append({
+                                            await _push_step({
 
                                                 "type": "skill",
 

backend/python/app/core/llm_client.py

@@ -90,6 +90,9 @@ async def achat_complete(api_url, api_key, request_body, *, proxy=None, timeout=
 
         async def _gen():
             client = build_openai_client(api_url, api_key, proxy=proxy, timeout=timeout)
+            _content_parts = []
+            _reasoning_parts = []
+            _tool_acc = {}  # index -> {id, type, name, arguments}
             try:
                 gen = await client.chat.completions.create(
                     **_create_kwargs(request_body, stream=True)
@@ -100,12 +103,49 @@ async def achat_complete(api_url, api_key, request_body, *, proxy=None, timeout=
                     delta = chunk.choices[0].delta
                     rc = getattr(delta, "reasoning_content", None)
                     if rc:
+                        _reasoning_parts.append(rc)
                         yield ("reasoning", rc)
                     ct = getattr(delta, "content", None)
                     if ct:
+                        _content_parts.append(ct)
                         yield ("content", ct)
+                    # 累积 OpenAI 流式 tool_calls（供后续工具循环执行）
+                    tcs = getattr(delta, "tool_calls", None)
+                    if tcs:
+                        for tc in tcs:
+                            _idx = getattr(tc, "index", 0)
+                            _slot = _tool_acc.setdefault(
+                                _idx, {"id": None, "type": "function", "name": "", "arguments": ""}
+                            )
+                            _tcid = getattr(tc, "id", None)
+                            if _tcid:
+                                _slot["id"] = _tcid
+                            _fn = getattr(tc, "function", None)
+                            if _fn is not None:
+                                _fn_name = getattr(_fn, "name", None)
+                                if _fn_name:
+                                    _slot["name"] += _fn_name
+                                _fn_args = getattr(_fn, "arguments", None)
+                                if _fn_args:
+                                    _slot["arguments"] += _fn_args
             finally:
                 await client.close()
+            # 流式结束时吐出已组装的完整 message（含 tool_calls），供调用方无缝衔接工具循环
+            _tool_calls = None
+            if _tool_acc:
+                _tool_calls = []
+                for _i in sorted(_tool_acc.keys()):
+                    _s = _tool_acc[_i]
+                    _tool_calls.append({
+                        "id": _s["id"],
+                        "type": "function",
+                        "function": {"name": _s["name"], "arguments": _s["arguments"]},
+                    })
+            yield ("message", {
+                "content": "".join(_content_parts),
+                "tool_calls": _tool_calls,
+                "reasoning_content": "".join(_reasoning_parts) or None,
+            })
 
         return _gen()
 

frontend/src/pages/front/ChatPage.tsx

@@ -548,6 +548,54 @@ const AiProcessBlock: React.FC<{ steps: any[]; autoExpand: boolean }> = ({ steps
   );
 };
 
+// 思考过程收拢式展示（缺陷5：默认折叠；fresh 流式消息自动展开，数秒后自动收拢）
+const ReasoningBlock: React.FC<{ text: string; fresh: boolean }> = ({ text, fresh }) => {
+  const [open, setOpen] = useState(fresh);
+  const [isFresh, setFreshState] = useState(fresh);
+  const touchedRef = useRef(false);
+  const timerRef = useRef<number | null>(null);
+  useEffect(() => {
+    if (!fresh) {
+      setOpen(false);
+      setFreshState(false);
+      return;
+    }
+    setOpen(true);
+    setFreshState(true);
+    touchedRef.current = false;
+    timerRef.current = window.setTimeout(() => {
+      if (!touchedRef.current) {
+        setOpen(false);
+        setTimeout(() => setFreshState(false), 500);
+      }
+    }, 5000);
+    return () => { if (timerRef.current) window.clearTimeout(timerRef.current); };
+  }, [fresh]);
+  if (!text) return null;
+  const toggle = () => {
+    touchedRef.current = true;
+    setFreshState(false);
+    if (timerRef.current) window.clearTimeout(timerRef.current);
+    setOpen(o => !o);
+  };
+  return (
+    <details className={`ai-process ai-process-reasoning${isFresh ? ' ai-process-fresh' : ''}`} open={open}>
+      <summary className="ai-process-summary" onClick={(e) => { e.preventDefault(); toggle(); }}>
+        <span className="ai-process-badge">🤔</span>
+        AI 思考过程{open ? '（收拢中... 点击收起）' : '（点击展开）'}
+      </summary>
+      <div className="ai-process-list">
+        <div className="ai-process-item ai-process-thinking">
+          <span className="ai-process-ico">💡</span>
+          <div className="ai-process-body">
+            <p className="ai-process-text">{text}</p>
+          </div>
+        </div>
+      </div>
+    </details>
+  );
+};
+
 // ===== Main Component =====
 const ChatPage: React.FC = () => {
   const { id } = useParams<{ id?: string }>();
@@ -561,6 +609,9 @@ const ChatPage: React.FC = () => {
   const sendingTimerRef = useRef<ReturnType<typeof setTimeout> | null>(null);
   // 正在 SSE 流式渲染的助手消息 id；非空时隐藏"思考中"占位气泡
   const [streamingId, setStreamingId] = useState<string | null>(null);
+  // 流式中断控制器（缺陷4：支持「停止生成」）
+  const abortRef = useRef<AbortController | null>(null);
+  const stopStreaming = () => { try { abortRef.current?.abort(); } catch { /* noop */ } };
   const [conversations, setConversations] = useState<Conversation[]>([]);
   const [messages, setMessages] = useState<Message[]>([]);
   const [loadingConversations, setLoadingConversations] = useState(true);
@@ -1009,6 +1060,8 @@ const ChatPage: React.FC = () => {
     assistantId: string,
   ): Promise<void> => {
     const token = localStorage.getItem('aiflow_token');
+    const controller = new AbortController();
+    abortRef.current = controller;
     const resp = await fetch(`/app/api/chat/conversations/${convId}/messages/stream`, {
       method: 'POST',
       headers: {
@@ -1016,6 +1069,7 @@ const ChatPage: React.FC = () => {
         ...(token ? { Authorization: `Bearer ${token}` } : {}),
       },
       body: JSON.stringify(payload),
+      signal: controller.signal,
     });
     if (!resp.ok || !resp.body) {
       throw new Error(`SSE 连接失败 (${resp.status})`);
@@ -1026,9 +1080,32 @@ const ChatPage: React.FC = () => {
     let buffer = '';
     let accReasoning = '';
     let gotDone = false;
+    let pendingDelta = '';
+    let reasoningDirty = false;
+    let rafId: number | null = null;
 
-    const patchAssistant = (patch: Partial<Message>) => {
+    // 缺陷4：用 requestAnimationFrame 合并同一帧内的多个 delta/reasoning，避免逐 token 重渲染抖动
-      setMessages(prev => prev.map(m => (m.id === assistantId ? { ...m, ...patch } : m)));
+    const flushBuffered = () => {
+      rafId = null;
+      if (!pendingDelta && !reasoningDirty) return;
+      const piece = pendingDelta;
+      pendingDelta = '';
+      const r = reasoningDirty ? accReasoning : null;
+      reasoningDirty = false;
+      setMessages(prev => prev.map(m => {
+        if (m.id !== assistantId) return m;
+        const patch: any = {};
+        if (piece) patch.content = (m.content || '') + piece;
+        if (r !== null) patch.reasoning = r;
+        return { ...m, ...patch };
+      }));
+    };
+    const scheduleFlush = () => {
+      if (rafId == null) rafId = requestAnimationFrame(flushBuffered);
+    };
+    const finalizeFlush = () => {
+      if (rafId != null) { cancelAnimationFrame(rafId); rafId = null; }
+      flushBuffered();
     };
 
     while (true) {
@@ -1044,13 +1121,27 @@ const ChatPage: React.FC = () => {
         let evt: any;
         try { evt = JSON.parse(dataLine.slice(6)); } catch { continue; }
         if (evt.type === 'delta') {
-          const piece = evt.content || '';
+          pendingDelta += evt.content || '';
-          setMessages(prev => prev.map(m => (m.id === assistantId ? { ...m, content: (m.content || '') + piece } : m)));
+          scheduleFlush();
         } else if (evt.type === 'reasoning') {
           accReasoning += evt.content || '';
-          patchAssistant({ reasoning: accReasoning } as any);
+          reasoningDirty = true;
+          scheduleFlush();
+        } else if (evt.type === 'step') {
+          // 缺陷2：实时增量过程步骤，前端随事件收拢式渲染
+          const step = evt.step;
+          const idx = typeof evt.index === 'number' ? evt.index : -1;
+          setMessages(prev => prev.map(m => {
+            if (m.id !== assistantId) return m;
+            const steps: any[] = (m as any).process_steps || [];
+            const next = steps.slice();
+            if (idx >= 0 && idx < next.length) next[idx] = step;
+            else next.push(step);
+            return { ...m, process_steps: next };
+          }));
         } else if (evt.type === 'done') {
           gotDone = true;
+          finalizeFlush();
           const finalId = evt.message_id || assistantId;
           setMessages(prev => prev.map(m => {
             if (m.id !== assistantId) return m;
@@ -1065,6 +1156,7 @@ const ChatPage: React.FC = () => {
             };
           }));
           setStreamingId(null);
+          abortRef.current = null;
           if (evt.generating) startMediaPolling(convId, finalId);
           setFreshMsgIds(prev => new Set(prev).add(finalId));
           loadConversations();
@@ -1079,12 +1171,26 @@ const ChatPage: React.FC = () => {
             }).catch(() => {});
           }
         } else if (evt.type === 'error') {
-          throw new Error(evt.message || '生成失败');
+          // 缺陷4：容错——保留已生成内容，不强删，标记中断后优雅收尾
+          finalizeFlush();
+          setMessages(prev => prev.map(m => {
+            if (m.id !== assistantId) return m;
+            const note = '\n\n_(生成中断：' + (evt.message || '未知错误') + ')_';
+            return { ...m, content: (m.content || '') + note };
+          }));
+          setStreamingId(null);
+          abortRef.current = null;
+          setFreshMsgIds(prev => new Set(prev).add(assistantId));
+          loadConversations();
+          gotDone = true;
+          return;
         }
       }
     }
     if (!gotDone) {
       // 流结束但未收到 done：已累积的 delta 仍可展示，但视为异常兜底
+      finalizeFlush();
+      abortRef.current = null;
       setStreamingId(null);
       throw new Error('流式响应未正常结束');
     }
@@ -1405,20 +1511,7 @@ const ChatPage: React.FC = () => {
                     <AiProcessBlock steps={steps} autoExpand={freshMsgIds.has(msg.id)} />
                   )}
                   {msg.role === 'assistant' && (msg as any).reasoning && (
-                    <details className="ai-process ai-process-reasoning" open>
+                    <ReasoningBlock text={(msg as any).reasoning} fresh={freshMsgIds.has(msg.id)} />
-                      <summary className="ai-process-summary">
-                        <span className="ai-process-badge">🤔</span>
-                        AI 思考过程（实时）
-                      </summary>
-                      <div className="ai-process-list">
-                        <div className="ai-process-item ai-process-thinking">
-                          <span className="ai-process-ico">💡</span>
-                          <div className="ai-process-body">
-                            <p className="ai-process-text">{(msg as any).reasoning}</p>
-                          </div>
-                        </div>
-                      </div>
-                    </details>
                   )}
                   <div className={`message-bubble ${msg.role}`}>
                     {clarify ? (
@@ -1708,11 +1801,11 @@ const ChatPage: React.FC = () => {
                   </button>
                   <button
                     className="send-btn"
-                    onClick={handleSend}
+                    onClick={streamingId ? stopStreaming : handleSend}
-                    disabled={!inputValue.trim() || sending}
+                    disabled={!streamingId && (!inputValue.trim() || sending)}
-                    title="发送 (Enter)"
+                    title={streamingId ? '停止生成' : '发送 (Enter)'}
                   >
-                    {sending ? <IconSpinner /> : <IconSend />}
+                    {streamingId ? <IconStop /> : (sending ? <IconSpinner /> : <IconSend />)}
                   </button>
                 </div>
               </div>