refactor(server,web): remove Socket.IO entirely, unify on REST polling

- task event stream (_append_event) is now the sole realtime channel for web and CLI
- status snapshots: idle 5s polling of /api/status (was status_update push); operation initiator gets state from REST responses
- terminal panel: REST polling (list 5s, output 1.5s, prefix-matched incremental xterm writes)
- multi-tab passive sync deliberately degrades to polling-based perception
- delete socket_handlers/broadcast/useLegacySocket (2103-line dead file); extensions.py keeps only run_background
- remove socket token chain (/api/socket-token, prune/consume_socket_token, pending_socket_tokens)
- app.run(threaded=True) replaces socketio.run; reapers switched to plain threading
- drop flask-socketio/socket.io-client/websockets dependencies
- docs: AGENTS.md section 12.5 (replacement map + hard constraints), plus CLI rewrite companion doc updates

Co-authored-by: Astrion powered by Kimi-K3 <astrion-agent@users.noreply.github.com>

AGENTS.md

@@ -23,7 +23,7 @@
   - `server/app.py`: 推荐的 Web 服务入口（封装并转发到 `server/app_legacy.py`）
   - `web_server.py`: 兼容入口，已标记 deprecated，但仍可启动
 - **后端核心目录**
-  - `server/`: Flask 业务主线（chat/task/status/context 已拆分为子包：`server/chat/`、`server/status/`、`server/tasks/`、`server/context/`（用户资源/身份/广播/个性化，原 `server/context.py` 已拆包并保留兼容 re-export）；REST 任务轮询为主，Socket.IO 主要用于兼容与实时辅助通道）
+  - `server/`: Flask 业务主线（chat/task/status/context 已拆分为子包：`server/chat/`、`server/status/`、`server/tasks/`、`server/context/`（用户资源/身份/个性化，原 `server/context.py` 已拆包并保留兼容 re-export）；**REST 任务轮询为唯一实时通道**——Socket.IO 已于 2026-09-11 整体移除，详见 §12.5）
   - `server/runtime/`: 公共任务入口（2026-09 Gateway 化阶段二新增）：`context.py` 定义 RuntimeContext 三层模型（TrustedPrincipal/TaskParams/InternalDirectives），`service.py` 提供 RuntimeService（create_task/cancel/guidance/queue/get_task_events）；契约见 `docs/runtime_contract.md`
   - `core/`: 终端与工具编排（`main_terminal.py`、`web_terminal.py`、`main_terminal_parts/*`；其中 `main_terminal_parts/context/` 和 `main_terminal_parts/tools_definition` 已拆分为 base + mixin 子包）
   - `modules/`: 可复用能力模块（terminal/file/memory/sub_agent/upload_security/user 等；`file_manager`、`persistent_terminal`、`terminal_ops`、`mcp_client_manager` 已拆分为子包）
@@ -31,7 +31,7 @@
   - `utils/`: API client、日志、上下文与对话工具等公共函数（`api_client`、`tool_result_formatter`、`context_manager`、`conversation_manager` 已拆分为子包；原入口文件保留为兼容入口）
 - **前端目录**
   - `static/src/`: Vue 3 + TS 前端
-  - `cli/src/`: React 19 + Ink 6 + TypeScript CLI 前端（正在重写中）
+  - `cli/src/`: opentui（Zig 内核 + Yoga）+ React 19 CLI 前端（2026-09-11 起正式替代 Ink 版；旧 Ink 实现存档于 `cli/legacy-ink-src/` 仅作字段参考）
   - 监控动画相关核心文件：
     - `static/src/components/chat/monitor/MonitorDirector.ts`
     - `static/src/stores/monitor.ts`
@@ -117,12 +117,12 @@
 - 开发监听（当前脚本是 build watch）：`npm run dev`
 - Lint：`npm run lint`
 
-### CLI（React / Ink）
+### CLI（opentui / bun）
-- 安装依赖：`npm --prefix cli install`
+- 运行环境：bun ≥1.3（opentui FFI；node ≥26.4 亦可，本机 node v24 不可用）
-- 开发启动：`npm run cli` 或 `npm --prefix cli run dev`
+- 开发启动：`bun cli/src/main.tsx`（或 `npm --prefix cli run dev`）
-- 构建：`npm run cli:build`
+- 类型检查：`cd cli && ./node_modules/.bin/tsc --noEmit`
-- 类型检查：`npm run cli:typecheck`
+- 可执行命令名：`astrion`（已装到 `/opt/homebrew/bin/astrion` → `cli/bin/astrion` symlink，bun wrapper 支持 symlink 解析；`--version`/`--help` 快速退出不启 TUI；后续 `bun build --compile` 单文件分发）
-- 可执行命令名（构建后）：`agents` / `agents-cli`
+- 依赖安装：沙箱网络受限时不跑 npm/bun install，从 `cli-redesign-demo/node_modules` 铺平（版本以 `cli/package.json` 为准）
 
 ## 3) 测试现状（不要再写过时命令）
 
@@ -133,8 +133,8 @@
 - `test_system_message.py` 依赖外部 `MOONSHOT_API_KEY` 与网络，不属于离线稳定 CI 用例。
 - 当前仓库未发现 `pytest.ini`/`pyproject.toml`/`tox.ini`；不要默认要求 `pytest` 作为唯一入口（仅作为冒烟测试的便捷运行器）。
 - CLI 当前最小可复现验证：
-  - `npm --prefix cli run typecheck`
+  - `cd cli && ./node_modules/.bin/tsc --noEmit`
-  - `npm --prefix cli run build`
+  - 无头冒烟：`testRender` 抓帧脚本（写完即用即删，参考 cli-redesign-demo 的验证模式； gateway 指向 127.0.0.1:9 不碰真实服务）
 - 若改动 `server/chat/`、`server/status/`、`server/tasks/` 等后端接口适配，补充：
   - `python3 -m py_compile server/chat/*.py server/status/*.py server/tasks/*.py`
   - `python -m pytest test/test_server_refactor_smoke.py -q`
@@ -146,7 +146,7 @@
 - **文件编辑方式**：修改文件时优先使用 `apply_patch` 或其他原生文件编辑工具；尽量不要用 `bash`/`python` 脚本批量改文件，除非原生工具明显不适合。
 - **后端改动优先级**：先改 `modules/`、`server/` 内对应模块，最后才动入口。
 - **前端改动优先级**：按 `static/src` 现有分层改（`app/`、`stores/`、`components/`、`composables/`）。
-- **CLI 改动优先级**：优先在 `cli/src/App.tsx`、`cli/src/components.tsx`、`cli/src/eventMapper.ts`、`cli/src/api.ts` 内做最小闭环修改。
+- **CLI 改动优先级**：`cli/src/` 分层——入口编排 `main.tsx`/`boot.tsx`；主界面 `app.tsx`；Gateway 客户端 `gateway.ts`（Bearer）；会话运行时 `runtime.ts`（发消息/事件轮询/审批弹出）；菜单状态机 `menuState.ts`；渲染分发 `menu.tsx`；通用件 `components.tsx`/`selector.tsx`；数据层 `data.ts`；文案 `i18n/`。新交互先定归属层再动手。
 - 涉及 monitor 动画/事件联动时，至少同步检查：
   - `MonitorDirector.ts`（动画与场景执行）
   - `stores/monitor.ts`（事件队列与状态机）
@@ -173,12 +173,14 @@
 - 运行根目录前端构建时，默认使用 `npm run build --silent 2>&1 | tail -n 5`。
 - 构建/安装/Lint 类命令（如 `npm run build`、`npm install`、`npm run lint`）因沙箱权限失败时（如报 `EPERM` / `Operation not permitted` / 沙箱拒绝写入），直接向用户说明失败原因并申请权限或调整路径授权，禁止尝试任何绕过沙箱的做法（如换执行方式规避限制、向其他路径写入等）。
 
-### CLI 当前交互约束（2026-05-15）
+### CLI 当前交互约束（2026-09-11 重写）
 
-- CLI 连接的是现有本地 Web API（默认 `127.0.0.1:8091`），不是独立 agent runtime。
+- **CLI 全部走 Gateway host Bearer 通道**（`docs/runtime_protocol.md` §6）：token 读 `~/.astrion/astrion/host/data/host_api_token`，请求带 `Authorization: Bearer`；**禁止** Web 会话通道（Cookie/CSRF/host-login）。工作区经 `X-Astrion-Workspace-Id` 头绑定（server/gateway_auth.py 解析）。
-- 启动 CLI 时应清屏、连接本地服务、创建新会话，并将输入区固定在底部。
+- **启动指令 `astrion`**：cwd = 工作区，运行内不可切换；cwd 未注册为工作区时先提示并询问创建（确认面板 Esc=取消退出）。
-- 当前目录若不在工作区中，应先弹出“是否添加到工作区”的选择。
+- **服务自启动**：CLI 不与 Web 端启动绑定——探测无服务时自动 spawn `python3 -m server.app --port 8091 --thinking-mode`（detached+unref，显式端口）；服务在运行但无 host token = 旧版本服务，提示重启（token 由服务端启动时生成，`initialize_system` 内 host 分支）；请求 401 自动重读 token 重试一次（自愈）。**边界：CLI 只在探测无服务时启动新实例，绝不 kill/重启已在运行的进程。**
-- 思考内容当前默认隐藏，只显示“思考中 / 思考完成”标题；相关折叠代码保留，后续可继续修。
+- **审批面板自动弹出**：收到 `tool_approval_required` 事件即弹出（不依赖手动 /approvals）；←→ 选 运行/拒绝/切无限制，Enter 裁决（decision 端点）。
+- **多语言读 OS 语言设置**（`cli/src/i18n/`，LC_ALL/LANG 检测 zh/en，启动时定死）；CLI 内所有文案禁按字符长度硬编码布局（选中反色只落文字节点，项间固定间距）。
+- 输入语义：Enter=发送，Shift+Enter=换行（opentui textarea 默认相反，Composer 已用 keyBindings 覆盖）。
 - 不要在未获得用户要求的情况下运行交互式 TUI 压测或长时间模拟输入，以免刷屏占满上下文。
 
 ## 5.5) 前端设计风格统一规范（强制）
@@ -238,7 +240,7 @@
 - **key 奇偶强校验**：en-US 聚合器用 `DeepString<typeof zhCN>` 约束，en 缺/多 key 直接 tsc 报错；新增命名空间须在 `zh-CN.ts` / `en-US.ts` 同步注册。
 - **防回退栏杆**：`npm run lint` 先跑 `scripts/i18n_audit.mjs`（剥离注释后查裸中文，独立命令 `lint:text`）；存量文件列在 `scripts/i18n_baseline.txt`，**迁移完一个文件就删一行**，删除后该文件永久受栏杆保护。
 - **语言切换**：个人空间 → 外观 → 界面语言；默认 zh-CN（不跟随浏览器），持久化 key `agents_ui_locale`。
-- **边界**：后端下发文字（API error、通知、工具结果摘要）不做多语言，前端原样显示；CLI 暂不纳入。
+- **边界**：后端下发文字（API error、通知、工具结果摘要）不做多语言，前端原样显示；CLI 独立成体系（`cli/src/i18n/`，读 OS 语言设置，不跟随 web 个人空间）。
 
 ## 6) Git 工作流（开发 + Review）
 
@@ -460,7 +462,7 @@ AI 执行以下流程时，每一步都要向用户说明在做什么：
 
 ### 11.2 子智能体执行机制
 
-- 子智能体在主进程内 `asyncio.Task`，跑在独立后台事件循环线程里（避开 Flask-SocketIO threading 冲突）。工具调用复用主进程沙箱/容器链路，网络调用走 `utils.api_client.APIClient`。
+- 子智能体在主进程内 `asyncio.Task`，跑在独立后台事件循环线程里（历史原因是避开 Flask-SocketIO threading 冲突；Socket.IO 已移除，该线程模型仍保留）。工具调用复用主进程沙箱/容器链路，网络调用走 `utils.api_client.APIClient`。
 - **模型请求重试（2026-08-26 起）**：`_run_loop` 对 `_call_model` 包重试循环，与主智能体 `run_streaming_attempts` 同构——最多 5 次尝试（`_SUB_AGENT_MAX_API_RETRIES=4`）、间隔 10s（`_SUB_AGENT_RETRY_DELAY_SECONDS`，用 `asyncio.sleep` 分段等待并响应软停止/取消）。重试条件：**仅当零接收**（未收到任何文本/思考/工具调用）才重试；已开始收到内容后断流（`SubAgentModelCallError.received_any=True`）直接失败。失败终态分模式：多智能体模式下 5 次全失败 → 转为 idle + `_forward_output_to_master` 报错（等 Team Leader 重新下达指令），输出期间断开 → 直接 failed 并同步向主智能体报错；传统模式一律 `_write_failure`。
 - **工具「正在调用」进度事件（2026-08-26 起）**：`_call_model` 在工具名+id 首个流式 chunk 到达时即 emit `status="calling"` 进度事件（与后续 running/completed 共用同一 tool_call id），前端按 id 原地更新条目；`RunnerDetailPanel.vue` / `SubAgentActivityDialog.vue` 的 normalizeStatus 识别 `calling`（显示 spinner + 「调用中」），并支持同一 id 历史条目跨组原地更新（多工具 calling 事件交错场景）。
 - 子智能体在多智能体模式下：
@@ -570,7 +572,7 @@ AI 执行以下流程时，每一步都要向用户说明在做什么：
 
 ## 12) 对话级主任务门闸与单写者不变量（2026-08-12）
 
-> 事故背景：一个对话并发运行了两个主聊天任务（socketio 用户任务 + 完成通知轮询器派发的通知任务），交叉写入共享 `conversation_history`，产生 `assistant→assistant→tool→tool` 乱序段，最终 API 400 `tool_call_id is not found`、通知永久丢失。本节机制即为修复该事故引入。
+> 事故背景：一个对话并发运行了两个主聊天任务（socketio 入口的用户任务 + 完成通知轮询器派发的通知任务），交叉写入共享 `conversation_history`，产生 `assistant→assistant→tool→tool` 乱序段，最终 API 400 `tool_call_id is not found`、通知永久丢失。本节机制即为修复该事故引入。
 
 ### 12.1 单写者不变量（核心约束）
 
@@ -594,6 +596,25 @@ AI 执行以下流程时，每一步都要向用户说明在做什么：
 2. **不要在 return 分支手写 `_tool_loop_active` 恢复**：`execute_tool_calls`（`server/chat_flow_tool_loop.py`）已改为守护包装（try/finally 复位，内层 `_execute_tool_calls_impl`），新增提前返回路径无需也不应手动操作该标志——并发交错「存旧值→置True→恢复旧值」正是此前标志卡死的原因。
 3. **不要依赖 build_messages 防御层掩盖并发问题**：`core/main_terminal_parts/context/messages.py` 的孤儿 tool 消息剥离只是「坏数据不再 400」的止血层，乱序段本身意味着历史已被污染；发现剥离 warning 日志应按事故排查，而不是视为正常。
 
+### 12.5 Socket.IO 已整体移除（2026-09-11）
+
+Web 端实时通道曾长期双轨（REST 任务轮询为主 + Socket.IO 辅助推送），本次清理后**轮询成为唯一通道**，Socket.IO 及其依赖（flask-socketio / socket.io-client / websockets）已全量移除。
+
+**事件出口唯一权威 = 任务事件流**（`TaskRecord.events`，`_append_event`）：运行期一切事件（text/thinking/tool/token/todo/edited_files/edit_summary/context_warning/quota_exceeded/conversation_changed 标题更新等）都在其中，前端任务轮询与 CLI 共用。历史遗留的 socket 推送只是同一事件的镜像，删除不丢数据。
+
+**各原 socket 功能的替代机制**（改代码前必查）：
+- `status_update`（原 13 处推送）→ 前端 `socketMethods.fetchStatusSnapshot` 空闲 5s 轮询 `/api/status`（运行期跳过，由任务事件流驱动）；发起方操作后的即时状态走 REST 响应载荷。
+- `quota_update/notice/exceeded` → `quota_exceeded` 本就在事件流；配额快照走 resource store 既有的 UsageQuotaPolling。
+- `conversation_list_update / conversation_changed / conversation_loaded` → 发起方操作后主动刷新（既有主路径）；**多标签页被动同步退化为 status 轮询间接感知**（不再即时，属刻意取舍）。
+- `system_ready` → `loadInitialData` 的 `/api/status` 一次性初始化已覆盖。
+- 终端面板实时输出 → `GET /api/terminals/<name>/output` REST 轮询（TerminalPanel 面板打开时输出 1.5s、列表 5s），快照前缀匹配增量写入 xterm。
+- 连接状态指示 → 既有 REST 心跳（`/api/health`，连续失败阈值置灰），与 socket 断开语义等价。
+
+**硬性约束**：
+1. 新增任何「实时通知」一律写任务事件流（`_append_event`）或扩 `/api/status` 载荷，**禁止**重新引入 socket 推送。
+2. `server/extensions.py` 只剩 `run_background`（daemon 线程）；不要再往里加推送类函数。
+3. `WebTerminal.broadcast()` / `terminal_manager.broadcast` 调用点保留但回调恒为 None（判空安全），新代码不要依赖它们产生用户可见效果。
+
 ## 13) 工具动态加载（tool_loading，2026-09 新增）
 
 > 设计文档：`docs/dynamic_tool_loading_plan.md`。注册表与状态辅助唯一权威：`core/tool_loading.py`。

core/web_terminal.py

@@ -687,7 +687,7 @@ class WebTerminal(MainTerminal):
         return "思考模式" if self.thinking_mode else "快速模式"
     
     def broadcast(self, event_type: str, data: Dict):
-        """广播事件到WebSocket"""
+        """经 message_callback 推送事件（Socket.IO 已移除，回调恒为 None 时 no-op）"""
         if self.message_callback:
             payload = dict(data or {})
             payload.setdefault('conversation_id', self.context_manager.current_conversation_id)

package.json

@@ -46,7 +46,6 @@
     "remark-parse": "^11.0.0",
     "remark-rehype": "^11.1.2",
     "sherpa-onnx": "^1.13.2",
-    "socket.io-client": "^4.7.5",
     "unified": "^11.0.5",
     "virtua": "^0.49.3",
     "vue": "^3.4.15",

requirements.txt

@@ -1,12 +1,10 @@
 # 运行时直接依赖（人读、宽松，不锁版本，便于开发期升级）。
 # 可复现安装 / release 打包请用 requirements.lock.txt（精确版本）。
 flask
-flask-socketio
 flask-cors
 werkzeug
 httpx[http2]
 openai
 cryptography
 pillow
-websockets
 pyyaml>=6.0

server/__init__.py

@@ -5,7 +5,7 @@
 """
 import importlib
 
-__all__ = ["app", "socketio", "run_server", "parse_arguments", "initialize_system", "resource_busy_page"]
+__all__ = ["app", "run_server", "parse_arguments", "initialize_system", "resource_busy_page"]
 
 
 def __getattr__(name):

server/app.py

@@ -3,7 +3,6 @@ import argparse
 
 from .app_legacy import (
     app,
-    socketio,
     run_server as _run_server,
     parse_arguments as _parse_arguments,
     initialize_system,
@@ -34,7 +33,6 @@ def main():
 
 __all__ = [
     "app",
-    "socketio",
     "run_server",
     "parse_arguments",
     "initialize_system",

server/app_legacy.py

@@ -11,7 +11,6 @@ import re
 import threading
 from typing import Dict, List, Optional, Callable, Any, Tuple
 from flask import Flask, request, jsonify, send_from_directory, session, redirect, send_file, abort
-from flask_socketio import SocketIO, emit, join_room, leave_room, disconnect
 from flask_cors import CORS
 from werkzeug.exceptions import RequestEntityTooLarge
 from pathlib import Path
@@ -41,7 +40,6 @@ from server.workflow_page import workflow_page_bp
 from server.workflow_runtime_api import workflow_runtime_bp
 from server.conversation_bootstrap import conversation_bootstrap_bp
 from server.gateway_api import gateway_bp
-from server.socket_handlers import socketio
 from server.security import attach_security_hooks
 from werkzeug.utils import secure_filename
 from werkzeug.routing import BaseConverter
@@ -201,9 +199,6 @@ if not ENABLE_VERBOSE_CONSOLE:
 # 抑制 Flask/Werkzeug 访问日志，只保留 brief_log 输出
 logging.getLogger('werkzeug').setLevel(logging.ERROR)
 logging.getLogger('werkzeug').disabled = True
-for noisy_logger in ('engineio.server', 'socketio.server'):
-    logging.getLogger(noisy_logger).setLevel(logging.ERROR)
-    logging.getLogger(noisy_logger).disabled = True
 # 静音子智能体模块错误日志（交由 brief_log 或前端提示处理）
 sub_agent_logger = logging.getLogger('modules.sub_agent.manager')
 sub_agent_logger.setLevel(logging.CRITICAL)
@@ -271,8 +266,6 @@ app.config['SESSION_COOKIE_SECURE'] = _cookie_secure_env in {"1", "true", "yes"}
 app.config['SESSION_COOKIE_HTTPONLY'] = True
 CORS(app)
 
-socketio.init_app(app, cors_allowed_origins='*', async_mode='threading', logger=False, engineio_logger=False)
-
 
 class EndpointFilter(logging.Filter):
     """过滤掉噪声请求日志。"""
@@ -329,7 +322,6 @@ MONITOR_SNAPSHOT_CHAR_LIMIT = state.MONITOR_SNAPSHOT_CHAR_LIMIT
 MONITOR_MEMORY_ENTRY_LIMIT = state.MONITOR_MEMORY_ENTRY_LIMIT
 RATE_LIMIT_BUCKETS = state.RATE_LIMIT_BUCKETS
 FAILURE_TRACKERS = state.FAILURE_TRACKERS
-pending_socket_tokens = state.pending_socket_tokens
 usage_trackers = state.usage_trackers
 
 MONITOR_SNAPSHOT_CACHE = state.MONITOR_SNAPSHOT_CACHE
@@ -350,7 +342,6 @@ CSRF_PROTECTED_PREFIXES = state.CSRF_PROTECTED_PREFIXES
 CSRF_EXEMPT_PATHS = state.CSRF_EXEMPT_PATHS
 FAILED_LOGIN_LIMIT = state.FAILED_LOGIN_LIMIT
 FAILED_LOGIN_LOCK_SECONDS = state.FAILED_LOGIN_LOCK_SECONDS
-SOCKET_TOKEN_TTL_SECONDS = state.SOCKET_TOKEN_TTL_SECONDS
 PROJECT_STORAGE_CACHE = state.PROJECT_STORAGE_CACHE
 PROJECT_STORAGE_CACHE_TTL_SECONDS = state.PROJECT_STORAGE_CACHE_TTL_SECONDS
 USER_IDLE_TIMEOUT_SECONDS = state.USER_IDLE_TIMEOUT_SECONDS
@@ -457,7 +448,7 @@ def start_background_jobs():
         return
     _idle_reaper_started = True
     _load_last_active_cache()
-    socketio.start_background_task(idle_reaper_loop)
+    threading.Thread(target=idle_reaper_loop, daemon=True).start()
     try:
         from .context import start_conversation_terminal_reaper
         start_conversation_terminal_reaper()
@@ -564,18 +555,8 @@ def generate_conversation_title_background(web_terminal: WebTerminal, conversati
             _title_debug_log("title_save_failed", conversation_id=conversation_id, safe_title=safe_title)
             return
         _title_debug_log("title_save_success", conversation_id=conversation_id, safe_title=safe_title)
-        try:
+        # 标题更新推送已随 WebSocket 移除：标题变更已写入任务事件流
-            socketio.emit('conversation_changed', {
+        # （chat_flow_helpers 的 conversation_changed 事件），前端轮询可见。
-                'conversation_id': conversation_id,
-                'title': safe_title
-            }, room=f"user_{username}")
-            socketio.emit('conversation_list_update', {
-                'action': 'updated',
-                'conversation_id': conversation_id
-            }, room=f"user_{username}")
-        except Exception as exc:
-            debug_log(f"[TitleGen] 推送标题更新失败: {exc}")
-            _title_debug_log("title_emit_exception", error=str(exc), conversation_id=conversation_id, username=username)
 
     try:
         asyncio.run(_runner())
@@ -744,31 +725,6 @@ def validate_csrf_request() -> bool:
         return False
 
 
-def prune_socket_tokens(now: Optional[float] = None):
-    current = now or time.time()
-    for token, meta in list(pending_socket_tokens.items()):
-        if meta.get("expires_at", 0) <= current:
-            pending_socket_tokens.pop(token, None)
-
-
-def consume_socket_token(token_value: Optional[str], username: Optional[str]) -> bool:
-    if not token_value or not username:
-        return False
-    prune_socket_tokens()
-    token_meta = pending_socket_tokens.pop(token_value, None)
-    if not token_meta:
-        return False
-    if token_meta.get("username") != username:
-        return False
-    if token_meta.get("expires_at", 0) <= time.time():
-        return False
-    fingerprint = token_meta.get("fingerprint") or ""
-    request_fp = (request.headers.get("User-Agent") or "")[:128]
-    if fingerprint and request_fp and not hmac.compare_digest(fingerprint, request_fp):
-        return False
-    return True
-
-
 def format_tool_result_notice(tool_name: str, tool_call_id: Optional[str], content: str) -> str:
     """将工具执行结果转为系统消息文本，方便在对话中回传。"""
     header = tr("legacy.tool_result_header", tool=tool_name)
@@ -1047,27 +1003,6 @@ def detect_tool_failure(result_data: Any) -> bool:
         return True
     return False
 
-# 终端广播回调函数
-def terminal_broadcast(event_type, data):
-    """广播终端事件到所有订阅者"""
-    try:
-        # 对于全局事件，发送给所有连接的客户端
-        if event_type in ('token_update', 'todo_updated', 'edited_files_updated'):
-            socketio.emit(event_type, data)  # 全局广播，不限制房间
-            debug_log(f"全局广播{event_type}: {data}")
-        else:
-            # 其他终端事件发送到终端订阅者房间
-            socketio.emit(event_type, data, room='terminal_subscribers')
-            
-            # 如果是特定会话的事件，也发送到该会话的专属房间
-            if 'session' in data:
-                session_room = f"terminal_{data['session']}"
-                socketio.emit(event_type, data, room=session_room)
-        
-        debug_log(f"终端广播: {event_type} - {data}")
-    except Exception as e:
-        debug_log(f"终端广播错误: {e}")
-
 
 # Routes removed; now provided by Blueprints in server/auth.py and server/files.py
 
@@ -1103,6 +1038,15 @@ def initialize_system(path: str, thinking_mode: bool = False):
         print(f"{OUTPUT_FORMATS['success']} 预设子智能体角色同步完成")
     except Exception as _e:
         print(f"{OUTPUT_FORMATS['warning']} 预设子智能体角色同步失败: {_e}")
+    # host 模式启动即生成 CLI Bearer token（原惰性生成依赖首个 Bearer 请求触发，
+    # CLI 无 token 不发请求会形成鸡生蛋死锁；启动时生成根治）
+    if TERMINAL_SANDBOX_MODE == "host":
+        try:
+            from server.gateway_auth import get_or_create_host_api_token
+            if get_or_create_host_api_token():
+                print(f"{OUTPUT_FORMATS['success']} Host API token 已就绪（CLI Bearer 通道）")
+        except Exception as _e:
+            print(f"{OUTPUT_FORMATS['warning']} Host API token 生成失败: {_e}")
     _mode_label = "宿主机模式（单用户）" if TERMINAL_SANDBOX_MODE == "host" else "多用户模式（Web）"
     print(f"{OUTPUT_FORMATS['success']} Web 系统初始化完成（{_mode_label}）")
 
@@ -1116,13 +1060,13 @@ def run_server(path: str, thinking_mode: bool = False, port: int = DEFAULT_PORT,
         app.config['SESSION_COOKIE_NAME'] = f"agents_session_{port}"
     initialize_system(path, thinking_mode)
     start_background_jobs()
-    socketio.run(
+    # Socket.IO 已移除，直接使用 Werkzeug 内建服务器（threaded 以支持并发轮询）
-        app,
+    app.run(
         host=WEB_SERVER_HOST,
         port=port,
         debug=debug,
         use_reloader=debug,
-        allow_unsafe_werkzeug=True
+        threaded=True,
     )
 
 

server/auth.py

@@ -351,9 +351,6 @@ def logout():
                 state.container_manager.release_container(key, reason="logout")
             except Exception:
                 pass
-        for token_value, meta in list(state.pending_socket_tokens.items()):
-            if meta.get("username") == username:
-                state.pending_socket_tokens.pop(token_value, None)
     auth_debug_log(f"[auth_debug] {request.method} /logout after_clear session={_session_debug_snapshot()}")
     if request.method == 'GET':
         resp = make_response(redirect('/login?logged_out=1'))

server/chat/approval.py

@@ -36,11 +36,10 @@ from config.model_profiles import get_model_context_window
 from server.auth_helpers import resolve_admin_policy, get_current_user_record, get_current_username
 from server.gateway_auth import api_login_or_host_token_required
 from server.context import with_terminal, get_gui_manager, get_upload_guard, build_upload_error_response, ensure_conversation_loaded, get_or_create_usage_tracker
-from server.security import rate_limited, prune_socket_tokens
+from server.security import rate_limited
 from server.utils_common import debug_log
-from server.state import PROJECT_MAX_STORAGE_MB, pending_socket_tokens, SOCKET_TOKEN_TTL_SECONDS
+from server.state import PROJECT_MAX_STORAGE_MB
 from server.runtime import runtime_service
-from server.extensions import socketio
 from server.monitor import get_cached_monitor_snapshot
 
 from modules.i18n import tr

server/chat/files.py

@@ -34,11 +34,10 @@ from config.model_profiles import get_model_context_window
 
 from server.auth_helpers import api_login_required, resolve_admin_policy, get_current_user_record, get_current_username
 from server.context import with_terminal, get_gui_manager, get_upload_guard, build_upload_error_response, ensure_conversation_loaded, get_or_create_usage_tracker
-from server.security import rate_limited, prune_socket_tokens
+from server.security import rate_limited
 from server.utils_common import debug_log
-from server.state import PROJECT_MAX_STORAGE_MB, pending_socket_tokens, SOCKET_TOKEN_TTL_SECONDS
+from server.state import PROJECT_MAX_STORAGE_MB
 from server.state import tool_approval_manager, user_question_manager
-from server.extensions import socketio
 from server.monitor import get_cached_monitor_snapshot
 from server.utils_common import sanitize_filename_preserve_unicode
 

server/chat/misc.py

@@ -35,11 +35,10 @@ from config.model_profiles import get_model_context_window
 
 from server.auth_helpers import api_login_required, resolve_admin_policy, get_current_user_record, get_current_username
 from server.context import with_terminal, get_gui_manager, get_upload_guard, build_upload_error_response, ensure_conversation_loaded, get_or_create_usage_tracker
-from server.security import rate_limited, prune_socket_tokens
+from server.security import rate_limited
 from server.utils_common import debug_log
-from server.state import PROJECT_MAX_STORAGE_MB, pending_socket_tokens, SOCKET_TOKEN_TTL_SECONDS
+from server.state import PROJECT_MAX_STORAGE_MB
 from server.state import tool_approval_manager, user_question_manager
-from server.extensions import socketio
 from server.monitor import get_cached_monitor_snapshot
 
 from modules.i18n import tr
@@ -147,9 +146,8 @@ def tool_settings(terminal: WebTerminal, workspace: UserWorkspace, username: str
             }), 403
         terminal.set_tool_category_enabled(category, enabled)
         snapshot = terminal.get_tool_settings_snapshot()
-        socketio.emit('tool_settings_updated', {
+        # tool_settings 变更不再经 WebSocket 广播：发起方直接使用本响应中的
-            'categories': snapshot
+        # categories，其他标签页在下次操作或 status 轮询时同步。
-        }, room=f"user_{username}")
         return jsonify({
             "success": True,
             "categories": snapshot

server/chat/permission.py

@@ -46,11 +46,10 @@ from config.model_profiles import get_model_context_window
 
 from server.auth_helpers import api_login_required, resolve_admin_policy, get_current_user_record, get_current_username
 from server.context import with_terminal, get_gui_manager, get_upload_guard, build_upload_error_response, ensure_conversation_loaded, get_or_create_usage_tracker, get_user_resources
-from server.security import rate_limited, prune_socket_tokens
+from server.security import rate_limited
 from server.utils_common import debug_log
-from server.state import PROJECT_MAX_STORAGE_MB, pending_socket_tokens, SOCKET_TOKEN_TTL_SECONDS
+from server.state import PROJECT_MAX_STORAGE_MB
 from server.state import tool_approval_manager, user_question_manager
-from server.extensions import socketio
 from server.monitor import get_cached_monitor_snapshot
 import os
 import re
@@ -189,8 +188,6 @@ def update_permission_mode(terminal: WebTerminal, workspace: UserWorkspace, user
                 "error": str(exc),
                 "message": tr("chat_permission.update_failed")
             }), 500
-        status = terminal.get_status()
-        socketio.emit('status_update', status, room=f"user_{username}")
         return jsonify({
             "success": True,
             "mode": target_mode,
@@ -221,8 +218,6 @@ def update_permission_mode(terminal: WebTerminal, workspace: UserWorkspace, user
         }), 500
 
     session["permission_mode"] = applied_mode
-    status = terminal.get_status()
-    socketio.emit('status_update', status, room=f"user_{username}")
     return jsonify({
         "success": True,
         "mode": applied_mode,
@@ -284,8 +279,6 @@ def update_execution_mode(terminal: WebTerminal, workspace: UserWorkspace, usern
             _sync_workspace_terminal_mode(username, workspace, "execution_mode", target_mode)
         except Exception as exc:
             return jsonify({"success": False, "error": str(exc), "message": tr("chat_permission.execution_update_failed")}), 500
-        status = terminal.get_status()
-        socketio.emit('status_update', status, room=f"user_{username}")
         return jsonify({
             "success": True,
             "state": {
@@ -310,8 +303,6 @@ def update_execution_mode(terminal: WebTerminal, workspace: UserWorkspace, usern
         _sync_workspace_terminal_mode(username, workspace, "execution_mode", state.get("mode", target_mode))
     except Exception as exc:
         return jsonify({"success": False, "error": str(exc), "message": tr("chat_permission.execution_update_failed")}), 500
-    status = terminal.get_status()
-    socketio.emit('status_update', status, room=f"user_{username}")
     return jsonify({
         "success": True,
         "state": state,
@@ -368,8 +359,6 @@ def update_network_permission(terminal: WebTerminal, workspace: UserWorkspace, u
             _sync_workspace_terminal_mode(username, workspace, "network_permission", target_mode)
         except Exception as exc:
             return jsonify({"success": False, "error": str(exc), "message": tr("chat_permission.network_update_failed")}), 500
-        status = terminal.get_status()
-        socketio.emit('status_update', status, room=f"user_{username}")
         return jsonify({
             "success": True,
             "mode": target_mode,
@@ -391,8 +380,6 @@ def update_network_permission(terminal: WebTerminal, workspace: UserWorkspace, u
         _sync_workspace_terminal_mode(username, workspace, "network_permission", applied)
     except Exception as exc:
         return jsonify({"success": False, "error": str(exc), "message": tr("chat_permission.network_update_failed")}), 500
-    status = terminal.get_status()
-    socketio.emit('status_update', status, room=f"user_{username}")
     return jsonify({
         "success": True,
         "mode": applied,
@@ -490,8 +477,6 @@ def update_work_mode(terminal: WebTerminal, workspace: UserWorkspace, username:
     except Exception:
         pass
 
-    status = terminal.get_status()
-    socketio.emit('status_update', status, room=f"user_{username}")
     return jsonify({
         "success": True,
         "mode": result.get("mode") or target_mode,

server/chat/settings.py

@@ -39,11 +39,10 @@ from config.model_profiles import get_model_context_window
 
 from server.auth_helpers import api_login_required, resolve_admin_policy, get_current_user_record, get_current_username
 from server.context import with_terminal, get_gui_manager, get_upload_guard, build_upload_error_response, ensure_conversation_loaded, get_or_create_usage_tracker
-from server.security import rate_limited, prune_socket_tokens
+from server.security import rate_limited
 from server.utils_common import debug_log
-from server.state import PROJECT_MAX_STORAGE_MB, pending_socket_tokens, SOCKET_TOKEN_TTL_SECONDS
+from server.state import PROJECT_MAX_STORAGE_MB
 from server.state import tool_approval_manager, user_question_manager
-from server.extensions import socketio
 from server.monitor import get_cached_monitor_snapshot
 
 from modules.i18n import tr
@@ -104,9 +103,6 @@ def update_thinking_mode(terminal: WebTerminal, workspace: UserWorkspace, userna
             except Exception as exc:
                 logger.error(f"[API] 保存思考模式到对话失败: {exc}")
         
-        status = terminal.get_status()
-        socketio.emit('status_update', status, room=f"user_{username}")
-        
         return jsonify({
             "success": True,
             "data": {
@@ -170,9 +166,6 @@ def update_reasoning_effort(terminal: WebTerminal, workspace: UserWorkspace, use
             except Exception as exc:
                 logger.error(f"[API] 保存推理强度到对话失败: {exc}")
 
-        status = terminal.get_status()
-        socketio.emit('status_update', status, room=f"user_{username}")
-
         return jsonify({
             "success": True,
             "data": {
@@ -261,9 +254,6 @@ def update_model(terminal: WebTerminal, workspace: UserWorkspace, username: str)
                     f"与 terminal 当前对话 {current_cid} 不一致"
                 )
 
-        status = terminal.get_status()
-        socketio.emit('status_update', status, room=f"user_{username}")
-
         return jsonify({
             "success": True,
             "data": {
@@ -377,11 +367,6 @@ def update_personalization_settings(terminal: WebTerminal, workspace: UserWorksp
                     )
                 except Exception as meta_exc:
                     debug_log(f"应用个性化偏好失败: 同步对话元数据异常 {meta_exc}")
-            try:
-                status = terminal.get_status()
-                socketio.emit('status_update', status, room=f"user_{username}")
-            except Exception as status_exc:
-                debug_log(f"广播个性化状态失败: {status_exc}")
         except Exception as exc:
             debug_log(f"应用个性化偏好失败: {exc}")
         config_out = dict(config)

server/chat/terminal.py

@@ -35,11 +35,10 @@ from config.model_profiles import get_model_context_window
 
 from server.auth_helpers import api_login_required, resolve_admin_policy, get_current_user_record, get_current_username
 from server.context import with_terminal, get_gui_manager, get_upload_guard, build_upload_error_response, ensure_conversation_loaded, get_or_create_usage_tracker
-from server.security import rate_limited, prune_socket_tokens
+from server.security import rate_limited
 from server.utils_common import debug_log
-from server.state import PROJECT_MAX_STORAGE_MB, pending_socket_tokens, SOCKET_TOKEN_TTL_SECONDS
+from server.state import PROJECT_MAX_STORAGE_MB
 from server.state import tool_approval_manager, user_question_manager
-from server.extensions import socketio
 from server.monitor import get_cached_monitor_snapshot
 
 from modules.i18n import tr
@@ -59,24 +58,29 @@ def get_terminals(terminal: WebTerminal, workspace: UserWorkspace, username: str
     else:
         return jsonify({"sessions": [], "active": None, "total": 0})
 
-@chat_bp.route('/api/socket-token', methods=['GET'])
+@chat_bp.route('/api/terminals/<session_name>/output')
 @api_login_required
-def issue_socket_token():
+@with_terminal
-    """生成一次性 WebSocket token，供握手阶段使用。"""
+def get_terminal_output_rest(terminal: WebTerminal, workspace: UserWorkspace, username: str, session_name: str):
-    username = get_current_username()
+    """获取终端输出历史（TerminalPanel 轮询端点，替代原 WebSocket get_terminal_output）。"""
-    prune_socket_tokens()
+    policy = resolve_admin_policy(get_current_user_record())
-    now = time.time()
+    if policy.get("ui_blocks", {}).get("block_realtime_terminal"):
-    for token_value, meta in list(pending_socket_tokens.items()):
+        return jsonify({"success": False, "error": tr("chat_terminal.realtime_terminal_admin_disabled")}), 403
-        if meta.get("username") == username:
+    if not terminal.terminal_manager:
-            pending_socket_tokens.pop(token_value, None)
+        return jsonify({"success": False, "error": "Terminal system not initialized"}), 400
-    token_value = secrets.token_urlsafe(32)
+    try:
-    pending_socket_tokens[token_value] = {
+        lines = int(request.args.get("lines", 100))
-        "username": username,
+    except (TypeError, ValueError):
-        "expires_at": now + SOCKET_TOKEN_TTL_SECONDS,
+        lines = 100
-        "fingerprint": (request.headers.get('User-Agent') or '')[:128],
+    lines = max(1, min(lines, 2000))
-    }
+    result = terminal.terminal_manager.get_terminal_output(session_name, lines)
-    return jsonify({
+    if result.get("success"):
-        "success": True,
+        return jsonify({
-        "token": token_value,
+            "success": True,
-        "expires_in": SOCKET_TOKEN_TTL_SECONDS
+            "session": session_name,
-    })
+            "output": result.get("output", ""),
+            "is_interactive": result.get("is_interactive", False),
+            "last_command": result.get("last_command", ""),
+            "last_event_time": result.get("last_event_time"),
+        })
+    return jsonify({"success": False, "error": result.get("error", "unknown error")}), 404

server/chat_flow.py

@@ -66,10 +66,10 @@ from .utils_common import (
     CHUNK_FRONTEND_LOG_FILE,
     STREAMING_DEBUG_LOG_FILE,
 )
-from .security import rate_limited, format_tool_result_notice, compact_web_search_result, consume_socket_token, prune_socket_tokens, validate_csrf_request, requires_csrf_protection, get_csrf_token
+from .security import rate_limited, format_tool_result_notice, compact_web_search_result, validate_csrf_request, requires_csrf_protection, get_csrf_token
 from .main_task_gate import acquire_adopted_main_task_gate, release_main_task_gate
 from .monitor import cache_monitor_snapshot, get_cached_monitor_snapshot
-from .extensions import socketio, run_background
+from .extensions import run_background
 from .state import (
     MONITOR_FILE_TOOLS,
     MONITOR_MEMORY_TOOLS,
@@ -77,7 +77,6 @@ from .state import (
     MONITOR_MEMORY_ENTRY_LIMIT,
     RATE_LIMIT_BUCKETS,
     FAILURE_TRACKERS,
-    pending_socket_tokens,
     usage_trackers,
     MONITOR_SNAPSHOT_CACHE,
     MONITOR_SNAPSHOT_CACHE_LIMIT,
@@ -118,7 +117,6 @@ def generate_conversation_title_background(web_terminal: WebTerminal, conversati
         conversation_id=conversation_id,
         user_message=user_message,
         username=username,
-        socketio_instance=socketio,
         title_prompt_path=TITLE_PROMPT_PATH,
         debug_logger=debug_log,
         title_model=title_model,

server/chat_flow_helpers.py

@@ -119,7 +119,6 @@ def generate_conversation_title_background(
     conversation_id: str,
     user_message: str,
     username: str,
-    socketio_instance,
     title_prompt_path,
     debug_logger,
     title_model: str = "",
@@ -186,21 +185,6 @@ def generate_conversation_title_background(
             debug_logger(f"[TitleGen] 添加任务事件失败: {exc}")
             _title_debug_log("title_task_event_exception", error=str(exc), conversation_id=conversation_id, username=username)
 
-        try:
-            socketio_instance.emit(
-                'conversation_changed',
-                {'conversation_id': conversation_id, 'title': safe_title},
-                room=f"user_{username}",
-            )
-            socketio_instance.emit(
-                'conversation_list_update',
-                {'action': 'updated', 'conversation_id': conversation_id},
-                room=f"user_{username}",
-            )
-        except Exception as exc:
-            debug_logger(f"[TitleGen] 推送标题更新失败: {exc}")
-            _title_debug_log("title_emit_exception", error=str(exc), conversation_id=conversation_id, username=username)
-
     try:
         asyncio.run(_runner())
     except Exception as exc:

server/chat_flow_runtime.py

@@ -58,9 +58,8 @@ from .utils_common import (
     CHUNK_FRONTEND_LOG_FILE,
     STREAMING_DEBUG_LOG_FILE,
 )
-from .security import rate_limited, format_tool_result_notice, compact_web_search_result, consume_socket_token, prune_socket_tokens, validate_csrf_request, requires_csrf_protection, get_csrf_token
+from .security import rate_limited, format_tool_result_notice, compact_web_search_result, validate_csrf_request, requires_csrf_protection, get_csrf_token
 from .monitor import cache_monitor_snapshot, get_cached_monitor_snapshot
-from .extensions import socketio
 from .state import (
     MONITOR_FILE_TOOLS,
     MONITOR_MEMORY_TOOLS,
@@ -68,7 +67,6 @@ from .state import (
     MONITOR_MEMORY_ENTRY_LIMIT,
     RATE_LIMIT_BUCKETS,
     FAILURE_TRACKERS,
-    pending_socket_tokens,
     usage_trackers,
     MONITOR_SNAPSHOT_CACHE,
     MONITOR_SNAPSHOT_CACHE_LIMIT,
@@ -110,7 +108,6 @@ def generate_conversation_title_background(web_terminal: WebTerminal, conversati
         conversation_id=conversation_id,
         user_message=user_message,
         username=username,
-        socketio_instance=socketio,
         title_prompt_path=TITLE_PROMPT_PATH,
         debug_logger=debug_log,
         title_model=title_model,

server/chat_flow_task_main.py

@@ -67,10 +67,10 @@ from .utils_common import (
     CHUNK_FRONTEND_LOG_FILE,
     STREAMING_DEBUG_LOG_FILE,
 )
-from .security import rate_limited, compact_web_search_result, consume_socket_token, prune_socket_tokens, validate_csrf_request, requires_csrf_protection, get_csrf_token
+from .security import rate_limited, compact_web_search_result, validate_csrf_request, requires_csrf_protection, get_csrf_token
 from .main_task_gate import try_acquire_main_task_gate, release_main_task_gate
 from .monitor import cache_monitor_snapshot, get_cached_monitor_snapshot
-from .extensions import emit_event, run_background
+from .extensions import run_background
 from .state import (
     MONITOR_FILE_TOOLS,
     MONITOR_MEMORY_TOOLS,
@@ -78,7 +78,6 @@ from .state import (
     MONITOR_MEMORY_ENTRY_LIMIT,
     RATE_LIMIT_BUCKETS,
     FAILURE_TRACKERS,
-    pending_socket_tokens,
     usage_trackers,
     MONITOR_SNAPSHOT_CACHE,
     MONITOR_SNAPSHOT_CACHE_LIMIT,
@@ -477,7 +476,7 @@ def _persist_and_echo_preceding_notice(
     message: str,
     payload: Dict[str, Any],
 ) -> None:
-    """预写一条「前置完成通知」：写入对话历史 + socketio 回显（不触发新一轮工作）。
+    """预写一条「前置完成通知」：写入对话历史 + 任务事件流回显（不触发新一轮工作）。
 
     用于「通知池」一次取出多条时，前 N-1 条随同一后续任务一起呈现。
     这些通知必须落历史，否则模型看不到、刷新后也会丢失。
@@ -503,8 +502,8 @@ def _persist_and_echo_preceding_notice(
             cm.add_conversation("user", message, metadata=metadata)
     except Exception as exc:
         debug_log(f"[CompletionNotice] 前置通知写入历史失败: {exc}")
-    # socketio 回显（在线客户端即时可见）；轮询客户端由后续任务事件流回放，
+    # 回显经 sender 通道（任务内=任务事件流；poll 轮询器=no-op）；
-    # 前端按消息内容 dedup，两条通道不会双显。
+    # 轮询客户端统一由后续任务事件流回放，前端按消息内容 dedup 不会双显。
     try:
         echo_payload = {
             "message": message,
@@ -553,8 +552,8 @@ async def _dispatch_completion_user_notice(
     if message_source not in _VALID_USER_MESSAGE_SOURCES:
         message_source = "user"
 
-    # 先把前置通知写入历史并 socketio 回显（在线客户端即时可见）。
+    # 先把前置通知写入历史并经 sender 回显（Socket.IO 移除后无即时推送，
-    # 轮询客户端则通过后续任务事件流回放（见 _run_chat_task 的 preceding_user_notices 注入）。
+    # 统一通过后续任务事件流回放，见 _run_chat_task 的 preceding_user_notices 注入）。
     for item in preceding_notices:
         _persist_and_echo_preceding_notice(
             web_terminal=web_terminal,
@@ -576,7 +575,7 @@ async def _dispatch_completion_user_notice(
             **ui_defaults,
             "timestamp": datetime.now().isoformat(),
         }
-        # 轮询客户端通过后续任务事件流回放前置通知（在线客户端已由上面的 socketio 回显覆盖）。
+        # 轮询客户端通过后续任务事件流回放前置通知。
         preceding_list = None
         if preceding_notices:
             preceding_list = [
@@ -965,14 +964,13 @@ async def poll_completion_notifications(*, web_terminal, workspace, conversation
 
     把子智能体完成通知与后台 run_command 完成通知合并到同一条链路：
     - 每轮把两路所有待通知项一次性取出（池化），按时间排序；
-    - 这一批里前 N-1 条作为「预写通知」随同一个后续任务一起注入（写历史 + 事件流 + socketio），
+    - 这一批里前 N-1 条作为「预写通知」随同一个后续任务一起注入（写历史 + 事件流），
       不各自触发新一轮工作；
     - 仅最后 1 条作为后续 chat task 的触发消息，启动一次「工作 → 停止」循环；
     - 该后续任务结束后回到 handle_task_with_sender 结尾重新 spawn 本轮询器，继续消费剩余通知。
 
     这样多个后台任务同时完成时，会被合并成一次（而非每条都触发一轮停止-再工作）。
     """
-    from .extensions import emit_event
 
     sub_manager = getattr(web_terminal, "sub_agent_manager", None)
     bg_manager = getattr(web_terminal, "background_command_manager", None)
@@ -985,10 +983,9 @@ async def poll_completion_notifications(*, web_terminal, workspace, conversation
     start_wait = time.time()
 
     def sender(event_type, data):
-        try:
+        # WebSocket 推送已移除：事件均经任务事件流（_append_event）由轮询消费；
-            emit_event(event_type, data, room=f"user_{username}")
+        # 此 sender 仅保留签名以兼容既有调用点，不再做任何实时推送。
-        except Exception:
+        return
-            pass
 
     loop_count = 0
     ma_debug("poll_completion_notifications_start", conversation_id=conversation_id)
@@ -1115,16 +1112,13 @@ async def poll_multi_agent_notifications(*, web_terminal, workspace, conversatio
     与 poll_completion_notifications 完全分离，避免多智能体消息和传统后台
     通知竞争 task_manager 的单工作区互斥。
     """
-    from .extensions import emit_event
 
     max_wait_time = 3600
     start_wait = time.time()
 
     def sender(event_type, data):
-        try:
+        # WebSocket 推送已移除：事件均经任务事件流（_append_event）由轮询消费。
-            emit_event(event_type, data, room=f"user_{username}")
+        return
-        except Exception:
-            pass
 
     import threading as _threading
     ma_debug(
@@ -1258,7 +1252,7 @@ async def _dispatch_multi_agent_idle_messages(
 ):
     """主智能体空闲时，把多智能体 pending 消息作为新一轮用户消息分发。
 
-    所有消息先持久化到对话历史；前 N-1 条作为前置通知通过 socketio 推送；
+    所有消息先持久化到对话历史；前 N-1 条作为前置通知写入任务事件流；
     最后一条创建 task_type="notice" 任务触发 Team Leader 新一轮工作。
     """
     ma_debug(
@@ -1341,8 +1335,8 @@ async def _dispatch_multi_agent_idle_messages(
             error=str(exc),
         )
 
-    # 2) 前置通知的 socketio emit 已在上述注入 inject_multi_agent_master_message 内
+    # 2) 前置通知的回显已在上述注入 inject_multi_agent_master_message 内
-    #    同时完成，避免在多智能体消息场景重复发送相同事件。
+    #    同时完成（经 sender/事件流），避免在多智能体消息场景重复发送相同事件。
 
     # 3) 最后一条触发新一轮工作
     ui_defaults = dict(_user_message_ui_defaults(message_source, auto_user_message_event=True))
@@ -2170,12 +2164,6 @@ async def handle_task_with_sender(
             quota_allowed, quota_info = web_terminal.record_model_call(bool(thinking_expected))
         if not quota_allowed:
             quota_type = 'thinking' if thinking_expected else 'fast'
-            emit_event('quota_notice', {
-                'type': quota_type,
-                'reset_at': quota_info.get('reset_at'),
-                'limit': quota_info.get('limit'),
-                'count': quota_info.get('count')
-            }, room=f"user_{getattr(web_terminal, 'username', '')}")
             sender('quota_exceeded', {
                 'type': quota_type,
                 'reset_at': quota_info.get('reset_at')

server/chat_flow_tool_loop.py

@@ -533,11 +533,6 @@ async def _handle_submit_plan(*, web_terminal, arguments: Dict[str, Any], sender
                 })
             except Exception:
                 pass
-            try:
-                from .extensions import emit_event
-                emit_event('status_update', web_terminal.get_status(), room=f"user_{username}")
-            except Exception:
-                pass
             switch_note = tr("tool_loop.plan_switch_note_ok")
         except Exception as exc:
             switch_note = tr("tool_loop.plan_switch_note_failed", error=exc)

server/context/__init__.py

@@ -12,14 +12,10 @@ _SYMBOL_MODULE = {
     "RuntimeIdentity": "server.context.identity",
     "_resolve_user_role": "server.context.identity",
     # broadcast
-    "make_terminal_callback": "server.context.broadcast",
-    "attach_user_broadcast": "server.context.broadcast",
-    "_wrap_callback_with_conversation_id": "server.context.broadcast",
     # personalization
     "_apply_workspace_personalization_preferences": "server.context.personalization",
     # usage
     "get_or_create_usage_tracker": "server.context.usage",
-    "emit_user_quota_update": "server.context.usage",
     # upload
     "get_gui_manager": "server.context.upload",
     "get_upload_guard": "server.context.upload",
@@ -55,9 +51,6 @@ __all__ = [
     "apply_conversation_overrides",
     "reset_system_state",
     "get_or_create_usage_tracker",
-    "emit_user_quota_update",
-    "make_terminal_callback",
-    "attach_user_broadcast",
     "reap_idle_conversation_terminals",
     "start_conversation_terminal_reaper",
 ]

server/context/broadcast.py

@@ -1,47 +0,0 @@
-"""终端事件广播与回调包装（用户房间广播 + conversation_id 注入）。"""
-from __future__ import annotations
-
-from typing import Optional
-
-from core.web_terminal import WebTerminal
-from server.utils_common import debug_log
-
-
-def make_terminal_callback(username: str):
-    """生成面向指定用户的广播函数（独立 Gateway 进程下 socket 未初始化时静默跳过）"""
-    from server.extensions import emit_event
-    def _callback(event_type, data):
-        emit_event(event_type, data, room=f"user_{username}")
-    return _callback
-
-
-def attach_user_broadcast(terminal: WebTerminal, username: str):
-    """确保终端的广播函数指向当前用户的房间。
-
-    对话级 terminal 的回调会额外包装注入 conversation_id（见 _wrap_callback_with_conversation_id）。
-    """
-    callback = make_terminal_callback(username)
-    callback = _wrap_callback_with_conversation_id(
-        callback, getattr(terminal, "_bound_conversation_id", None)
-    )
-    terminal.message_callback = callback
-    if terminal.terminal_manager:
-        terminal.terminal_manager.broadcast = callback
-
-
-def _wrap_callback_with_conversation_id(callback, conversation_id: Optional[str]):
-    """包装广播回调，为 dict 类型的事件数据注入 conversation_id（setdefault，不覆盖已有值）。
-
-    对话级 terminal 的广播（shell 输出、terminal 列表等）仍发到用户房间，
-    前端按 conversation_id 过滤，避免同工作区多个对话的终端事件互相串扰。
-    """
-    if not callback or not conversation_id:
-        return callback
-
-    def _wrapped(event_type, data):
-        if isinstance(data, dict):
-            data = dict(data)
-            data.setdefault("conversation_id", conversation_id)
-        return callback(event_type, data)
-
-    return _wrapped

server/context/reaper.py

@@ -157,5 +157,4 @@ def start_conversation_terminal_reaper():
     if _conversation_terminal_reaper_started:
         return
     _conversation_terminal_reaper_started = True
-    from server.extensions import socketio
+    threading.Thread(target=_conversation_terminal_reaper_loop, daemon=True).start()
-    socketio.start_background_task(_conversation_terminal_reaper_loop)

server/context/resources.py

@@ -55,9 +55,8 @@ def _get_current_user_role(record=None) -> str:
     return get_current_user_role(record)
 
 from server.context.identity import NoWorkspaceError, RuntimeIdentity, _resolve_user_role
-from server.context.broadcast import make_terminal_callback, attach_user_broadcast
 from server.context.personalization import _apply_workspace_personalization_preferences
-from server.context.usage import get_or_create_usage_tracker, emit_user_quota_update
+from server.context.usage import get_or_create_usage_tracker
 
 
 
@@ -316,14 +315,12 @@ def get_user_resources(
                 project_path=str(project_path),
                 thinking_mode=thinking_mode,
                 run_mode=run_mode,
-                message_callback=make_terminal_callback("host"),
+                message_callback=None,
                 data_dir=str(data_dir),
                 container_session=container_handle,
                 usage_tracker=usage_tracker,
                 conversation_id=conversation_id,
             )
-            if terminal.terminal_manager:
-                terminal.terminal_manager.broadcast = terminal.message_callback
             state.user_terminals[term_key] = terminal
             terminal.username = "host"
             terminal.user_role = "admin"
@@ -335,7 +332,6 @@ def get_user_resources(
                 session_set('host_workspace_id', getattr(workspace, "workspace_id", None))
         else:
             terminal.update_container_session(container_handle)
-            attach_user_broadcast(terminal, "host")
             terminal.username = "host"
             terminal.user_role = "admin"
             if can_write_session:
@@ -476,12 +472,10 @@ def get_user_resources(
             usage_tracker=usage_tracker,
             conversation_id=conversation_id,
         )
-        if terminal.terminal_manager:
-            terminal.terminal_manager.broadcast = terminal.message_callback
         state.user_terminals[term_key] = terminal
         terminal.username = username
         terminal.user_role = "api" if is_api_user else _resolve_user_role(identity, record)
-        terminal.quota_update_callback = (lambda metric=None: emit_user_quota_update(username)) if not is_api_user else None
+        terminal.quota_update_callback = None
         if can_write_session:
             session_set('run_mode', terminal.run_mode)
             session_set('thinking_mode', terminal.thinking_mode)
@@ -489,10 +483,9 @@ def get_user_resources(
             session_set('workspace_id', getattr(workspace, "workspace_id", None))
     else:
         terminal.update_container_session(container_handle)
-        attach_user_broadcast(terminal, username)
         terminal.username = username
         terminal.user_role = "api" if is_api_user else _resolve_user_role(identity, record)
-        terminal.quota_update_callback = (lambda metric=None: emit_user_quota_update(username)) if not is_api_user else None
+        terminal.quota_update_callback = None
         if can_write_session:
             session_set('workspace_id', getattr(workspace, "workspace_id", None))
 

server/context/usage.py

@@ -1,4 +1,4 @@
-"""配额追踪器（UsageTracker）的获取与广播。"""
+"""配额追踪器（UsageTracker）的获取。"""
 from __future__ import annotations
 
 from typing import Optional, TYPE_CHECKING
@@ -24,17 +24,3 @@ def get_or_create_usage_tracker(username: Optional[str], workspace: Optional['mo
     tracker = UsageTracker(str(workspace.data_dir), role=role or "user")
     state.usage_trackers[username] = tracker
     return tracker
-
-
-def emit_user_quota_update(username: Optional[str]):
-    from server.extensions import emit_event
-    if not username:
-        return
-    tracker = get_or_create_usage_tracker(username)
-    if not tracker:
-        return
-    try:
-        snapshot = tracker.get_quota_snapshot()
-        emit_event('quota_update', {'quotas': snapshot}, room=f"user_{username}")
-    except Exception:
-        pass

server/conversation.py

@@ -26,7 +26,6 @@ from io import BytesIO
 from typing import Dict, Any, Optional, List, Tuple
 
 from flask import Blueprint, request, jsonify, session, send_file
-from flask_socketio import emit
 from werkzeug.utils import secure_filename
 import zipfile
 
@@ -72,7 +71,7 @@ from utils.conversation_manager import ConversationManager
 from utils.api_client import APIClient
 
 from .auth_helpers import api_login_required, resolve_admin_policy, get_current_user_record, get_current_username
-from .context import with_terminal, get_terminal_for_sid, get_gui_manager, get_upload_guard, build_upload_error_response, ensure_conversation_loaded, reset_system_state, get_user_resources, get_or_create_usage_tracker
+from .context import with_terminal, get_gui_manager, get_upload_guard, build_upload_error_response, ensure_conversation_loaded, reset_system_state, get_user_resources, get_or_create_usage_tracker
 from .utils_common import (
     build_review_lines,
     debug_log,
@@ -86,7 +85,6 @@ from .utils_common import (
     CHUNK_FRONTEND_LOG_FILE,
     STREAMING_DEBUG_LOG_FILE,
 )
-from .extensions import socketio
 from .state import (
     RECENT_UPLOAD_EVENT_LIMIT,
     RECENT_UPLOAD_FEED_LIMIT,
@@ -807,18 +805,7 @@ def create_conversation(terminal: WebTerminal, workspace: UserWorkspace, usernam
             except Exception as exc:
                 debug_log(f"[Versioning] create_conversation apply default failed: {exc}")
             perf_log("create_conversation after default versioning", elapsed_ms=(time.perf_counter() - t0) * 1000, extra={"conv_id": result.get("conversation_id")})
-            # 广播对话列表更新事件
+            # 对话列表/切换推送已随 WebSocket 移除：发起方依据响应刷新，其他标签页由轮询同步。
-            socketio.emit('conversation_list_update', {
-                'action': 'created',
-                'conversation_id': result["conversation_id"]
-            }, room=f"user_{username}")
-
-            if not result.get("safe_navigation"):
-                # 安全导航创建不代表后端 terminal 当前对话已切换，因此不广播 conversation_changed。
-                socketio.emit('conversation_changed', {
-                    'conversation_id': result["conversation_id"],
-                    'title': tr("conversation.default_title")
-                }, room=f"user_{username}")
 
             perf_log("create_conversation route done", elapsed_ms=(time.perf_counter() - t0) * 1000, extra={"conv_id": result.get("conversation_id")})
             return jsonify(result), 201
@@ -941,22 +928,9 @@ def load_conversation(conversation_id, terminal: WebTerminal, workspace: UserWor
                 }
 
             if not result.get("safe_navigation"):
-                # 安全导航只改变前端查看对象，不代表后端 terminal 当前上下文已切换。
+                # 安全导航只改变前端查看对象；对话切换/加载推送已随 WebSocket 移除，
-                socketio.emit('conversation_changed', {
+                # 发起方依据响应刷新，其他标签页由轮询同步。
-                    'conversation_id': conversation_id,
+                pass
-                    'title': result.get("title", tr("conversation.unknown_title")),
-                    'messages_count': result.get("messages_count", 0)
-                }, room=f"user_{username}")
-
-                # 广播系统状态更新（因为当前对话改变了）
-                status = terminal.get_status()
-                socketio.emit('status_update', status, room=f"user_{username}")
-
-                # 清理和重置相关UI状态
-                socketio.emit('conversation_loaded', {
-                    'conversation_id': conversation_id,
-                    'clear_ui': True  # 提示前端清理当前UI状态
-                }, room=f"user_{username}")
 
             return jsonify(result)
         else:
@@ -1007,23 +981,9 @@ def delete_conversation(conversation_id, terminal: WebTerminal, workspace: UserW
         result = terminal.delete_conversation(conversation_id)
         
         if result["success"]:
-            # 广播对话列表更新事件
+            # 对话删除/清空推送已随 WebSocket 移除：发起方依据响应刷新，
-            socketio.emit('conversation_list_update', {
+            # 其他标签页由对话列表轮询同步。is_current 变量保留供下方响应使用。
-                'action': 'deleted',
+            pass
-                'conversation_id': conversation_id
-            }, room=f"user_{username}")
-            
-            # 如果删除的是当前对话，广播对话清空事件
-            if is_current:
-                socketio.emit('conversation_changed', {
-                    'conversation_id': None,
-                    'title': None,
-                    'cleared': True
-                }, room=f"user_{username}")
-                
-                # 更新系统状态
-                status = terminal.get_status()
-                socketio.emit('status_update', status, room=f"user_{username}")
             
             return jsonify(result)
         else:
@@ -1705,14 +1665,7 @@ def restore_conversation_versioning_checkpoint(conversation_id, terminal: WebTer
         session['run_mode'] = terminal.run_mode
         session['thinking_mode'] = terminal.thinking_mode
 
-        socketio.emit('conversation_list_update', {
+        # 版本恢复后的列表/切换推送已随 WebSocket 移除：发起方依据响应刷新。
-            'action': 'version_restored',
-            'conversation_id': target_conversation_id
-        }, room=f"user_{username}")
-        socketio.emit('conversation_changed', {
-            'conversation_id': target_conversation_id,
-            'title': (cm.load_conversation(target_conversation_id) or {}).get("title", tr("conversation.version_restored_title")),
-        }, room=f"user_{username}")
 
         return jsonify({
             "success": True,
@@ -1766,17 +1719,9 @@ def compress_conversation(conversation_id, terminal: WebTerminal, workspace: Use
             status_code = 404 if _is_not_found_message(result.get("error", "")) else (409 if result.get("in_progress") else 400)
             return jsonify(result), status_code
 
-        # in-place 压缩：对话 id 不变。通知前端当前对话内容已变化（历史前缀被标记），刷新展示。
+        # in-place 压缩：对话 id 不变。对话内容变化推送已随 WebSocket 移除，
+        # 发起方依据响应（guide_inserted 等字段）刷新历史。
         load_result = terminal.load_conversation(normalized_id)
-        if load_result.get("success"):
-            socketio.emit('conversation_list_update', {
-                'action': 'compressed',
-                'conversation_id': normalized_id
-            }, room=f"user_{username}")
-            socketio.emit('conversation_loaded', {
-                'conversation_id': normalized_id,
-                'clear_ui': True
-            }, room=f"user_{username}")
 
         response_payload = {
             "success": True,
@@ -1802,29 +1747,7 @@ def compress_conversation(conversation_id, terminal: WebTerminal, workspace: Use
                 )
                 response_payload["auto_task_started"] = False
                 response_payload["guide_inserted"] = True
-                # 通知前端实时显示这条 compact 消息（覆盖 socket 与 in-place 未刷新场景）
+                # 引导语已写入历史，前端依据响应 guide_inserted 刷新历史即可看到。
-                try:
-                    emit(
-                        "user_message",
-                        {
-                            "message": guide_message,
-                            "images": [],
-                            "videos": [],
-                            "media_refs": [],
-                            "message_source": "compression_handoff",
-                            "visibility": "compact",
-                            "starts_work": False,
-                            "metadata": {
-                                "message_source": "compression_handoff",
-                                "visibility": "compact",
-                                "starts_work": False,
-                            },
-                            "conversation_id": normalized_id,
-                        },
-                        room=f"user_{username}",
-                    )
-                except Exception as emit_exc:
-                    debug_log(f"[Compression] 发送 user_message 事件失败: {emit_exc}")
             except Exception as exc:
                 debug_log(f"[Compression] 追加引导语消息失败: {exc}")
                 response_payload["auto_task_started"] = False
@@ -2267,21 +2190,6 @@ def duplicate_conversation(conversation_id, terminal: WebTerminal, workspace: Us
         new_conversation_id = result["duplicate_conversation_id"]
         load_result = terminal.load_conversation(new_conversation_id)
 
-        if load_result.get("success"):
-            socketio.emit('conversation_list_update', {
-                'action': 'duplicated',
-                'conversation_id': new_conversation_id
-            }, room=f"user_{username}")
-            socketio.emit('conversation_changed', {
-                'conversation_id': new_conversation_id,
-                'title': load_result.get('title', tr('conversation.duplicated_title')),
-                'messages_count': load_result.get('messages_count', 0)
-            }, room=f"user_{username}")
-            socketio.emit('conversation_loaded', {
-                'conversation_id': new_conversation_id,
-                'clear_ui': True
-            }, room=f"user_{username}")
-
         response_payload = {
             "success": True,
             "duplicate_conversation_id": new_conversation_id,
@@ -2466,23 +2374,8 @@ def get_current_conversation(terminal: WebTerminal, workspace: UserWorkspace, us
             "error": str(e)
         }), 500
     
-@socketio.on('send_command')
-def handle_command(data):
-    """处理系统命令"""
-    command = data.get('command', '')
-    
-    username, terminal, _ = get_terminal_for_sid(request.sid)
-    if not terminal:
-        emit('error', {'message': 'System not initialized'})
-        return
-    record_user_activity(username)
-    
-    result = _execute_system_command(terminal, command)
-    emit('command_result', result)
-
-
 def _execute_system_command(terminal: WebTerminal, command: str) -> Dict[str, Any]:
-    """执行系统命令，供 WebSocket 与 REST API 复用。"""
+    """执行系统命令（REST API 专用；原 WebSocket send_command 通道已移除）。"""
     command = (command or '').strip()
     if command.startswith('/'):
         command = command[1:]

server/extensions.py

@@ -1,35 +1,16 @@
-"""Flask/SocketIO 扩展实例。"""
+"""后台任务辅助（WebSocket 移除后仅剩后台线程启动器）。"""
 import threading
 
-from flask_socketio import SocketIO
-
-# 统一的 SocketIO 实例，使用线程模式以兼容现有逻辑
-socketio = SocketIO(cors_allowed_origins="*", async_mode='threading', logger=False, engineio_logger=False)
-
-
-def emit_event(event, data, room=None, **kwargs):
-    """Web 模式下经 socketio 实时推送；独立 Gateway 进程（socketio 未绑定 app）静默跳过。
-
-    事件的权威记录是任务事件流（TaskRecord.events，_append_event 落盘在前）；
-    socket 推送只是在线客户端的实时增量通道——独立进程没有 socket 客户端，
-    跳过不丢数据。Web 模式下推送失败也静默（对齐原各调用点的 try/except 语义）。
-    """
-    if socketio.server is None:
-        return
-    try:
-        socketio.emit(event, data, room=room, **kwargs)
-    except Exception:
-        pass
-
 
 def run_background(fn, *args, **kwargs):
-    """启动后台任务：Web 模式走 socketio.start_background_task（兼容其线程模型）；
+    """以 daemon 线程启动后台任务。
-    独立 Gateway 进程（socketio 未初始化）降级为 daemon 线程。"""
+
-    if socketio.server is not None:
+    历史上该函数会在 Web 模式优先走 socketio.start_background_task；
-        return socketio.start_background_task(fn, *args, **kwargs)
+    Socket.IO 移除后统一为 daemon 线程（与原降级路径行为一致）。
+    """
     t = threading.Thread(target=fn, args=args, kwargs=kwargs, daemon=True)
     t.start()
     return t
 
 
-__all__ = ["socketio", "emit_event", "run_background"]
+__all__ = ["run_background"]

server/main_task_gate.py

@@ -1,6 +1,6 @@
 """对话级主任务门闸（单写者防护）。
 
-背景（2026-08-12「平行时空」事故）：socketio 入口的主聊天任务不在
+背景（2026-08-12「平行时空」事故）：socketio 入口（已移除）的主聊天任务不在
 task_manager 注册，`create_chat_task` 的单对话互斥对它们不可见；完成通知
 轮询器又只凭 `_tool_loop_active`（仅覆盖工具执行窗口）判断对话是否空闲，
 于是在主任务两次工具循环的间隙里派发了通知任务——两个主任务并发交叉写入

server/multi_agent.py

@@ -428,15 +428,8 @@ def create_multi_agent_conversation():
     except Exception:
         pass
 
-    # 触发对话列表更新事件
+    # 对话列表更新不再经 WebSocket 广播：发起方依据本响应刷新列表，
-    try:
+    # 其他标签页由对话列表轮询同步。
-        from server.app_legacy import socketio
-        socketio.emit('conversation_list_update', {
-            'action': 'created',
-            'conversation_id': conversation_id,
-        }, room=f"user_{username}")
-    except Exception:
-        pass
 
     return jsonify({
         "success": True,

server/security.py

@@ -188,31 +188,6 @@ def validate_csrf_request() -> bool:
         return False
 
 
-def prune_socket_tokens(now: Optional[float] = None):
-    current = now or time.time()
-    for token, meta in list(state.pending_socket_tokens.items()):
-        if meta.get("expires_at", 0) <= current:
-            state.pending_socket_tokens.pop(token, None)
-
-
-def consume_socket_token(token_value: Optional[str], username: Optional[str]) -> bool:
-    if not token_value or not username:
-        return False
-    prune_socket_tokens()
-    token_meta = state.pending_socket_tokens.pop(token_value, None)
-    if not token_meta:
-        return False
-    if token_meta.get("username") != username:
-        return False
-    if token_meta.get("expires_at", 0) <= time.time():
-        return False
-    fingerprint = token_meta.get("fingerprint") or ""
-    request_fp = (request.headers.get("User-Agent") or "")[:128]
-    if fingerprint and request_fp and not hmac.compare_digest(fingerprint, request_fp):
-        return False
-    return True
-
-
 def format_tool_result_notice(tool_name: str, tool_call_id: Optional[str], content: str) -> str:
     """将工具执行结果转为系统消息文本，方便在对话中回传。"""
     header = f"[工具结果] {tool_name}"
@@ -301,8 +276,6 @@ __all__ = [
     "get_csrf_token",
     "requires_csrf_protection",
     "validate_csrf_request",
-    "prune_socket_tokens",
-    "consume_socket_token",
     "format_tool_result_notice",
     "compact_web_search_result",
     "attach_security_hooks",
@@ -318,8 +291,6 @@ __all__ = [
     "get_csrf_token",
     "requires_csrf_protection",
     "validate_csrf_request",
-    "prune_socket_tokens",
-    "consume_socket_token",
     "format_tool_result_notice",
     "compact_web_search_result",
 ]

server/socket_handlers.py

@@ -1,406 +0,0 @@
-from __future__ import annotations
-import asyncio, time, json, re
-from typing import Dict, Any
-from flask import request
-from flask_socketio import emit, join_room, leave_room, disconnect
-
-from .extensions import socketio
-from modules.i18n import tr
-from .auth_helpers import get_current_username, resolve_admin_policy
-from .context import (
-    get_terminal_for_sid,
-    ensure_conversation_loaded,
-    reset_system_state,
-    get_user_resources,
-)
-from .utils_common import debug_log, log_frontend_chunk, log_streaming_debug_entry
-from .state import connection_users, stop_flags, terminal_rooms, pending_socket_tokens, user_manager, get_stop_flag, set_stop_flag, clear_stop_flag
-from .usage import record_user_activity
-from .chat_flow import start_chat_task
-from .security import consume_socket_token, prune_socket_tokens
-from config import OUTPUT_FORMATS, AGENT_VERSION
-from config.model_profiles import model_supports_image, model_supports_video
-
-@socketio.on('connect')
-def handle_connect(auth):
-    """客户端连接"""
-    debug_log(f"[WebSocket] 客户端连接: {request.sid}")
-    username = get_current_username()
-    token_value = (auth or {}).get('socket_token') if isinstance(auth, dict) else None
-    if not username or not consume_socket_token(token_value, username):
-        emit('error', {'message': tr("socket.not_logged_in")})
-        disconnect()
-        return
-    
-    emit('connected', {'status': 'Connected to server'})
-    connection_users[request.sid] = username
-    
-    # 清理可能存在的停止标志和状态
-    stop_flags.pop(request.sid, None)
-    # 将旧的 username 级别任务映射到新的 sid，便于重新停止
-    user_entry = get_stop_flag(None, username)
-    if user_entry:
-        set_stop_flag(request.sid, username, user_entry)
-    
-    join_room(f"user_{username}")
-    join_room(f"user_{username}_terminal")
-    if request.sid not in terminal_rooms:
-        terminal_rooms[request.sid] = set()
-    terminal_rooms[request.sid].update({f"user_{username}", f"user_{username}_terminal"})
-    
-    terminal, workspace = get_user_resources(username)
-    if terminal:
-        reset_system_state(terminal)
-        emit('system_ready', {
-            'project_path': str(workspace.project_path),
-            'thinking_mode': bool(getattr(terminal, "thinking_mode", False)),
-            'version': AGENT_VERSION
-        }, room=request.sid)
-        
-        if terminal.terminal_manager:
-            terminals = terminal.terminal_manager.get_terminal_list()
-            emit('terminal_list_update', {
-                'terminals': terminals,
-                'active': terminal.terminal_manager.active_terminal
-            }, room=request.sid)
-            
-            if terminal.terminal_manager.active_terminal:
-                for name, term in terminal.terminal_manager.terminals.items():
-                    emit('terminal_started', {
-                        'session': name,
-                        'working_dir': str(term.working_dir),
-                        'shell': term.shell_command,
-                        'time': term.start_time.isoformat() if term.start_time else None
-                    }, room=request.sid)
-
-@socketio.on('disconnect')
-def handle_disconnect():
-    """客户端断开"""
-    debug_log(f"[WebSocket] 客户端断开: {request.sid}")
-    username = connection_users.pop(request.sid, None)
-    # 若同一用户仍有其他活跃连接，不因断开而停止任务
-    has_other_connection = False
-    if username:
-        for sid, user in connection_users.items():
-            if user == username:
-                has_other_connection = True
-                break
-
-    # 检查是否有通过 REST API 创建的运行中任务
-    # 如果有，说明使用轮询模式，不应该停止任务
-    has_rest_api_task = False
-    if username and not has_other_connection:
-        try:
-            from .tasks import task_manager
-            running_tasks = [t for t in task_manager.list_tasks(username) if t.status == "running"]
-            if running_tasks:
-                has_rest_api_task = True
-                debug_log(f"[WebSocket] 用户 {username} 有运行中的 REST API 任务，不停止")
-        except Exception as e:
-            debug_log(f"[WebSocket] 检查 REST API 任务失败: {e}")
-
-    task_info = get_stop_flag(request.sid, username)
-    # 只有在没有其他连接且没有 REST API 任务时才停止
-    if isinstance(task_info, dict) and not has_other_connection and not has_rest_api_task:
-        task_info['stop'] = True
-        pending_task = task_info.get('task')
-        if pending_task and not pending_task.done():
-            debug_log(f"disconnect: cancel task for {request.sid}")
-            pending_task.cancel()
-        terminal = task_info.get('terminal')
-        if terminal:
-            reset_system_state(terminal)
-
-    # 清理停止标志（只清理 sid 级别的，不清理 user 级别的）
-    if request.sid in stop_flags:
-        stop_flags.pop(request.sid, None)
-
-    # 从所有房间移除
-    for room in list(terminal_rooms.get(request.sid, [])):
-        leave_room(room)
-    if request.sid in terminal_rooms:
-        del terminal_rooms[request.sid]
-
-    if username:
-        leave_room(f"user_{username}")
-        leave_room(f"user_{username}_terminal")
-
-@socketio.on('stop_task')
-def handle_stop_task():
-    """处理停止任务请求"""
-    debug_log(f"[停止] 收到停止请求: {request.sid}")
-    username = connection_users.get(request.sid)
-    task_info = get_stop_flag(request.sid, username)
-    if not isinstance(task_info, dict):
-        task_info = {'stop': False, 'task': None, 'terminal': None}
-    # 标记停止标志，让任务内部检测并优雅停止
-    task_info['stop'] = True
-    # 注释掉直接取消任务，改为通过停止标志让任务内部处理
-    # pending_task = task_info.get('task')
-    # if pending_task and not pending_task.done():
-    #     debug_log(f"正在取消任务: {request.sid}")
-    #     pending_task.cancel()
-    debug_log(f"设置停止标志: {request.sid}")
-    if task_info.get('terminal'):
-        reset_system_state(task_info['terminal'])
-    set_stop_flag(request.sid, username, task_info)
-
-    emit('stop_requested', {
-        'message': tr("socket.stop_received")
-    })
-
-@socketio.on('terminal_subscribe')
-def handle_terminal_subscribe(data):
-    """订阅终端事件"""
-    session_name = data.get('session')
-    subscribe_all = data.get('all', False)
-
-    conv_id = (data.get('conversation_id') or '').strip() or None
-    username, terminal, _ = get_terminal_for_sid(request.sid, conversation_id=conv_id)
-    if not username or not terminal or not terminal.terminal_manager:
-        emit('error', {'message': 'Terminal system not initialized'})
-        return
-    policy = resolve_admin_policy(user_manager.get_user(username))
-    if policy.get("ui_blocks", {}).get("block_realtime_terminal"):
-        emit('error', {'message': tr("socket.terminal_disabled")})
-        return
-    
-    if request.sid not in terminal_rooms:
-        terminal_rooms[request.sid] = set()
-    
-    if subscribe_all:
-        # 订阅所有终端事件
-        room_name = f"user_{username}_terminal"
-        join_room(room_name)
-        terminal_rooms[request.sid].add(room_name)
-        debug_log(f"[Terminal] {request.sid} 订阅所有终端事件")
-        
-        # 发送当前终端状态
-        emit('terminal_subscribed', {
-            'type': 'all',
-            'terminals': terminal.terminal_manager.get_terminal_list()
-        })
-    elif session_name:
-        # 订阅特定终端会话
-        room_name = f'user_{username}_terminal_{session_name}'
-        join_room(room_name)
-        terminal_rooms[request.sid].add(room_name)
-        debug_log(f"[Terminal] {request.sid} 订阅终端: {session_name}")
-        
-        # 发送该终端的当前输出
-        output_result = terminal.terminal_manager.get_terminal_output(session_name, 100)
-        if output_result['success']:
-            emit('terminal_history', {
-                'session': session_name,
-                'output': output_result['output']
-            })
-
-@socketio.on('terminal_unsubscribe')
-def handle_terminal_unsubscribe(data):
-    """取消订阅终端事件"""
-    session_name = data.get('session')
-    username = connection_users.get(request.sid)
-    
-    if session_name:
-        room_name = f'user_{username}_terminal_{session_name}' if username else f'terminal_{session_name}'
-        leave_room(room_name)
-        if request.sid in terminal_rooms:
-            terminal_rooms[request.sid].discard(room_name)
-        debug_log(f"[Terminal] {request.sid} 取消订阅终端: {session_name}")
-
-@socketio.on('get_terminal_output')
-def handle_get_terminal_output(data):
-    """获取终端输出历史"""
-    session_name = data.get('session')
-    lines = data.get('lines', 50)
-
-    conv_id = (data.get('conversation_id') or '').strip() or None
-    username, terminal, _ = get_terminal_for_sid(request.sid, conversation_id=conv_id)
-    if not terminal or not terminal.terminal_manager:
-        emit('error', {'message': 'Terminal system not initialized'})
-        return
-    policy = resolve_admin_policy(user_manager.get_user(username))
-    if policy.get("ui_blocks", {}).get("block_realtime_terminal"):
-        emit('error', {'message': tr("socket.terminal_disabled")})
-        return
-    
-    result = terminal.terminal_manager.get_terminal_output(session_name, lines)
-    
-    if result['success']:
-        emit('terminal_output_history', {
-            'session': session_name,
-            'output': result['output'],
-            'is_interactive': result.get('is_interactive', False),
-            'last_command': result.get('last_command', ''),
-            'last_event_time': result.get('last_event_time')
-        })
-    else:
-        emit('error', {'message': result['error']})
-
-@socketio.on('send_message')
-def handle_message(data):
-    """
-    【已废弃】WebSocket 聊天入口，已由 REST API (/api/tasks) 替代
-
-    保留此处理器用于向后兼容，但建议前端使用 REST API 轮询模式。
-    新的架构优势：
-    - 支持分布式部署（任务存储可扩展为 Redis）
-    - 更好的错误恢复（页面刷新后可恢复任务状态）
-    - 减少 WebSocket 连接压力
-    """
-    username, terminal, workspace = get_terminal_for_sid(request.sid)
-    if not terminal:
-        emit('error', {'message': 'System not initialized'})
-        return
-
-    # 返回废弃提示
-    emit('error', {
-        'message': tr("socket.ws_chat_deprecated"),
-        'code': 'DEPRECATED',
-        'migration_guide': tr("socket.ws_migration_guide")
-    })
-    return
-
-    # 以下代码保留用于紧急回退
-    # ========================================
-    message = (data.get('message') or '').strip()
-    images = data.get('images') or []
-    videos = data.get('videos') or []
-    if not message and not images and not videos:
-        emit('error', {'message': tr("socket.empty_message")})
-        return
-    current_model = getattr(terminal, "model_key", None) or ""
-    if images and not model_supports_image(current_model):
-        emit('error', {'message': tr("socket.image_not_supported")})
-        return
-    if videos and not model_supports_video(current_model):
-        emit('error', {'message': tr("socket.video_not_supported")})
-        return
-    if images and videos:
-        emit('error', {'message': tr("socket.image_video_separate")})
-        return
-
-    debug_log(f"[WebSocket] 收到消息: {message}")
-    debug_log(f"\n{'='*80}\n新任务开始: {message}\n{'='*80}")
-    record_user_activity(username)
-
-    requested_conversation_id = data.get('conversation_id')
-    try:
-        conversation_id, created_new = ensure_conversation_loaded(terminal, requested_conversation_id)
-    except RuntimeError as exc:
-        emit('error', {'message': str(exc)})
-        return
-    try:
-        conv_data = terminal.context_manager._get_conversation_manager_for_id(conversation_id).load_conversation(conversation_id) or {}
-    except Exception:
-        conv_data = {}
-    title = conv_data.get('title', tr("conversation.default_title"))
-
-    socketio.emit('conversation_resolved', {
-        'conversation_id': conversation_id,
-        'title': title,
-        'created': created_new
-    }, room=request.sid)
-
-    if created_new:
-        socketio.emit('conversation_list_update', {
-            'action': 'created',
-            'conversation_id': conversation_id
-        }, room=f"user_{username}")
-        socketio.emit('conversation_changed', {
-            'conversation_id': conversation_id,
-            'title': title
-        }, room=request.sid)
-
-    client_sid = request.sid
-
-    def send_to_client(event_type, data):
-        """发送消息到客户端"""
-        socketio.emit(event_type, data, room=client_sid)
-
-    # 模型活动事件：用于刷新"在线"心跳（回复/工具调用都算活动）
-    activity_events = {
-        'ai_message_start', 'thinking_start', 'thinking_chunk', 'thinking_end',
-        'text_start', 'text_chunk', 'text_end',
-        'tool_preparing', 'tool_start', 'update_action',
-        'append_payload', 'modify_payload', 'system_message',
-        'task_complete'
-    }
-    last_model_activity = 0.0
-
-    def send_with_activity(event_type, data):
-        """模型产生输出或调用工具时刷新活跃时间，防止长回复被误判下线。"""
-        nonlocal last_model_activity
-        if event_type in activity_events:
-            now = time.time()
-            # 轻量节流：1 秒内多次事件只记一次
-            if now - last_model_activity >= 1.0:
-                record_user_activity(username)
-                last_model_activity = now
-        send_to_client(event_type, data)
-
-    # 传递客户端ID
-    images = data.get('images') or []
-    videos = data.get('videos') or []
-    start_chat_task(terminal, message, images, send_with_activity, client_sid, workspace, username, videos)
-
-
-# WS 客户端日志事件的轻量限流（防已认证用户洪泛写盘；username -> 时间戳滑窗）
-_WS_CLIENT_LOG_LIMIT = 30  # 每 60 秒最多 30 条
-_WS_CLIENT_LOG_WINDOW = 60.0
-_ws_client_log_buckets: Dict[str, list] = {}
-_WS_CLIENT_LOG_MAX_KEYS = 10000
-
-
-def _ws_client_log_allowed(username: str) -> bool:
-    now = time.time()
-    if len(_ws_client_log_buckets) > _WS_CLIENT_LOG_MAX_KEYS:
-        # 全局回收：清空过老桶（简单策略，防止桶表无限膨胀）
-        for key in [k for k, v in _ws_client_log_buckets.items() if not v or now - v[-1] > _WS_CLIENT_LOG_WINDOW]:
-            _ws_client_log_buckets.pop(key, None)
-        if len(_ws_client_log_buckets) > _WS_CLIENT_LOG_MAX_KEYS:
-            _ws_client_log_buckets.clear()
-    bucket = _ws_client_log_buckets.setdefault(username, [])
-    while bucket and now - bucket[0] > _WS_CLIENT_LOG_WINDOW:
-        bucket.pop(0)
-    if len(bucket) >= _WS_CLIENT_LOG_LIMIT:
-        return False
-    bucket.append(now)
-    return True
-
-
-@socketio.on('client_chunk_log')
-def handle_client_chunk_log(data):
-    """前端chunk日志上报"""
-    username = connection_users.get(request.sid)
-    if not username or not _ws_client_log_allowed(username):
-        return
-    conversation_id = data.get('conversation_id')
-    chunk_index = int(data.get('index') or data.get('chunk_index') or 0)
-    elapsed = float(data.get('elapsed') or 0.0)
-    length = int(data.get('length') or len(data.get('content') or ""))
-    client_ts = float(data.get('ts') or 0.0)
-    log_frontend_chunk(conversation_id, chunk_index, elapsed, length, client_ts)
-
-
-@socketio.on('client_stream_debug_log')
-def handle_client_stream_debug_log(data):
-    """前端流式调试日志"""
-    username = connection_users.get(request.sid)
-    if not username or not _ws_client_log_allowed(username):
-        return
-    if not isinstance(data, dict):
-        return
-    entry = dict(data)
-    entry.setdefault('server_ts', time.time())
-    log_streaming_debug_entry(entry)
-
-# 在 web_server.py 中添加以下对话管理API接口
-# 添加在现有路由之后，@socketio 事件处理之前
-
-# ==========================================
-# 对话管理API接口
-# ==========================================
-
-
-# conversation routes moved to server/conversation.py

server/state.py

@@ -48,7 +48,6 @@ MONITOR_SNAPSHOT_CHAR_LIMIT = 60000
 MONITOR_MEMORY_ENTRY_LIMIT = 256
 RATE_LIMIT_BUCKETS: Dict[str, deque] = defaultdict(deque)
 FAILURE_TRACKERS: Dict[str, Dict[str, float]] = {}
-pending_socket_tokens: Dict[str, Dict[str, Any]] = {}
 usage_trackers: Dict[str, UsageTracker] = {}
 active_login_nonces: Dict[str, set] = defaultdict(set)
 
@@ -71,7 +70,6 @@ CSRF_PROTECTED_PREFIXES = ("/api/",)
 CSRF_EXEMPT_PATHS = {"/api/csrf-token"}
 FAILED_LOGIN_LIMIT = 5
 FAILED_LOGIN_LOCK_SECONDS = 300
-SOCKET_TOKEN_TTL_SECONDS = 45
 USER_IDLE_TIMEOUT_SECONDS = int(os.environ.get("USER_IDLE_TIMEOUT_SECONDS", "900"))
 LAST_ACTIVE_FILE = Path(LOGS_DIR).expanduser().resolve() / "last_active.json"
 _last_active_lock = threading.Lock()
@@ -104,7 +102,6 @@ __all__ = [
     "MONITOR_MEMORY_ENTRY_LIMIT",
     "RATE_LIMIT_BUCKETS",
     "FAILURE_TRACKERS",
-    "pending_socket_tokens",
     "usage_trackers",
     "active_login_nonces",
     "tool_approval_manager",
@@ -123,7 +120,6 @@ __all__ = [
     "CSRF_EXEMPT_PATHS",
     "FAILED_LOGIN_LIMIT",
     "FAILED_LOGIN_LOCK_SECONDS",
-    "SOCKET_TOKEN_TTL_SECONDS",
     "USER_IDLE_TIMEOUT_SECONDS",
     "LAST_ACTIVE_FILE",
     "_last_active_lock",

server/status/app.py

@@ -18,7 +18,7 @@ from pathlib import Path
 from flask import Blueprint, jsonify, request, send_file, session
 
 from server.auth_helpers import api_login_required, resolve_admin_policy
-from server.context import with_terminal, attach_user_broadcast
+from server.context import with_terminal
 from server.state import (
     PROJECT_STORAGE_CACHE,
     PROJECT_STORAGE_CACHE_TTL_SECONDS,

server/status/base.py

@@ -13,7 +13,7 @@ from pathlib import Path
 from flask import Blueprint, jsonify, request, send_file, session
 
 from server.auth_helpers import api_login_required, resolve_admin_policy
-from server.context import with_terminal, attach_user_broadcast
+from server.context import with_terminal
 from server.state import (
     PROJECT_STORAGE_CACHE,
     PROJECT_STORAGE_CACHE_TTL_SECONDS,

server/status/docker.py

@@ -18,7 +18,7 @@ from pathlib import Path
 from flask import Blueprint, jsonify, request, send_file, session
 
 from server.auth_helpers import api_login_required, resolve_admin_policy
-from server.context import with_terminal, attach_user_broadcast
+from server.context import with_terminal
 from server.state import (
     PROJECT_STORAGE_CACHE,
     PROJECT_STORAGE_CACHE_TTL_SECONDS,
@@ -141,12 +141,6 @@ def select_docker_project():
     except RuntimeError as exc:
         session["workspace_id"] = previous_workspace_id
         return jsonify({"success": False, "error": str(exc)}), 503
-    terminal = state.user_terminals.get(f"{username}::{session['workspace_id']}")
-    if terminal:
-        try:
-            attach_user_broadcast(terminal, username)
-        except Exception:
-            pass
     default_workspace_id = user_manager._get_user_default_workspace_id(username)
     return jsonify({
         "success": True,

server/status/file_open.py

@@ -15,7 +15,7 @@ from pathlib import Path
 from flask import Blueprint, jsonify, request, send_file, session
 
 from server.auth_helpers import api_login_required, resolve_admin_policy
-from server.context import with_terminal, attach_user_broadcast
+from server.context import with_terminal
 from server.state import (
     PROJECT_STORAGE_CACHE,
     PROJECT_STORAGE_CACHE_TTL_SECONDS,

server/status/git.py

@@ -13,7 +13,7 @@ from pathlib import Path
 from flask import Blueprint, jsonify, request, send_file, session
 
 from server.auth_helpers import api_login_required, resolve_admin_policy
-from server.context import with_terminal, attach_user_broadcast
+from server.context import with_terminal
 from server.state import (
     PROJECT_STORAGE_CACHE,
     PROJECT_STORAGE_CACHE_TTL_SECONDS,

server/tasks/models.py

@@ -912,15 +912,8 @@ class TaskManager:
                         data.setdefault("conversation_id", rec.conversation_id)
                     if rec.workspace_id:
                         data.setdefault("workspace_id", rec.workspace_id)
-                # 记录事件
+                # 记录事件到任务事件流（轮询通道，唯一出口）
                 self._append_event(rec, event_type, data)
-                # 在线用户仍然收到实时推送（房间 user_{username}）；
-                # 安全包装：socketio 未绑定（独立 Gateway 进程）时静默跳过
-                try:
-                    from server.extensions import emit_event
-                    emit_event(event_type, data, room=f"user_{username}")
-                except Exception:
-                    pass
 
             # 轮询模式需要把 context_manager 的回调切到当前任务 sender，
             # 否则 token_update 等事件只走 websocket，前端任务轮询拿不到实时更新。
@@ -1043,7 +1036,6 @@ class TaskManager:
                 )
                 # 统一发送 task_stopped，携带后台任务状态
                 try:
-                    from server.extensions import emit_event
                     stopped_payload = {
                         'message': tr("task_main.task_stopped"),
                         'reason': 'user_requested',
@@ -1052,10 +1044,8 @@ class TaskManager:
                         'has_running_sub_agents': bg_state["has_running_sub_agents"],
                         'has_running_background_commands': bg_state["has_running_background_commands"],
                     }
-                    # 先写权威事件流（轮询客户端可见），再做实时推送——
+                    # 事件流（轮询通道）为唯一权威出口，不再做 WebSocket 实时推送。
-                    # 推送失败（含 socketio 未绑定）不得影响事件流记录
                     self._append_event(rec, "task_stopped", stopped_payload)
-                    emit_event('task_stopped', stopped_payload, room=f"user_{rec.username}")
                     debug_log(
                         f"[TaskRun] 已发送 task_stopped: task_id={rec.task_id}, "
                         f"has_bg={has_bg}, room=user_{rec.username}"

static/src/app/computed.ts

@@ -58,7 +58,6 @@ const AVATAR_TERMINAL_STATUS = new Set([
 export const computed = {
   ...mapWritableState(useConnectionStore, [
     'isConnected',
-    'socket',
     'stopRequested',
     'projectPath',
     'agentVersion',

static/src/app/lifecycle.ts

@@ -47,6 +47,7 @@ export async function mounted() {
   const initialDataPromise = this.loadInitialData();
   this.startProjectGitSummaryIdleRefresh?.();
   this.startConnectionHeartbeat();
+  this.startStatusIdleRefresh?.();
   this.fetchTerminalCount();
   this.startTerminalCountIdleRefresh();
   this.checkTutorialPrompt();
@@ -123,6 +124,7 @@ export function beforeUnmount() {
   this.teardownMobileViewportWatcher();
   this.stopProjectGitSummaryIdleRefresh?.();
   this.stopTerminalCountIdleRefresh?.();
+  this.stopStatusIdleRefresh?.();
   this.resourceStopContainerStatsPolling();
   this.resourceStopProjectStoragePolling();
   this.resourceStopUsageQuotaPolling();

static/src/app/methods/ui/panel.ts

@@ -49,9 +49,6 @@ export const panelMethods = {
   },
   toggleTerminalPanel() {
     this.terminalPanelOpen = !this.terminalPanelOpen;
-    if (this.terminalPanelOpen) {
-      this.subscribeTerminalEvents();
-    }
   },
   closeTerminalPanel() {
     this.terminalPanelOpen = false;

static/src/app/methods/ui/socket.ts

@@ -30,6 +30,51 @@ export const socketMethods = {
     // 主网页端已切换为 REST + 轮询模式，这里保留空实现用于兼容旧调用
     this.socket = null;
   },
+  // 状态快照轮询：替代原 WebSocket status_update 推送。
+  // 任务运行期跳过（运行期状态由任务事件流驱动），空闲期每 5s 一次。
+  // 配额轮询已由 resource store 的 UsageQuotaPolling 独立承担，这里不重复。
+  async fetchStatusSnapshot() {
+    try {
+      const res = await fetch('/api/status', { cache: 'no-store' });
+      if (!res.ok) return;
+      const status = await res.json();
+      this.applyStatusSnapshot(status);
+      // 显式新建路由（/new 等）与独立全屏路由不接管当前对话与运行模式
+      const onExplicitNewRoute =
+        typeof this.isExplicitNewConversationRoute === 'function' &&
+        this.isExplicitNewConversationRoute();
+      const onIndependentRoute =
+        typeof this.isConversationIndependentRoute === 'function' &&
+        this.isConversationIndependentRoute();
+      if (status.conversation && status.conversation.current_id) {
+        if (this.initialRouteResolved && !this.currentConversationId && !onExplicitNewRoute && !onIndependentRoute) {
+          this.currentConversationId = status.conversation.current_id;
+        }
+      }
+      if (!onExplicitNewRoute) {
+        if (typeof status.run_mode === 'string') {
+          this.runMode = status.run_mode;
+        } else if (typeof status.thinking_mode !== 'undefined') {
+          this.runMode = status.thinking_mode ? 'thinking' : 'fast';
+        }
+      }
+    } catch {
+      // 断线判定由连接心跳负责，这里静默失败
+    }
+  },
+  startStatusIdleRefresh() {
+    if (this.statusRefreshTimer) return;
+    this.statusRefreshTimer = window.setInterval(() => {
+      if (!this.isConnected) return;
+      if (typeof this.isOutputActive === 'function' && this.isOutputActive()) return;
+      this.fetchStatusSnapshot();
+    }, 5000);
+  },
+  stopStatusIdleRefresh() {
+    if (!this.statusRefreshTimer) return;
+    window.clearInterval(this.statusRefreshTimer);
+    this.statusRefreshTimer = null;
+  },
   async checkConnectionHealth() {
     if (this.connectionHeartbeatInFlight) {
       connectionDiag('log', 'health-skip-inflight', {

static/src/app/methods/ui/terminal.ts

@@ -34,11 +34,6 @@ export const terminalMethods = {
     }
     this.toggleTerminalPanel();
   },
-  subscribeTerminalEvents() {
-    const socket = this.socket;
-    if (!socket) return;
-    socket.emit('terminal_subscribe', { all: true, conversation_id: this.currentConversationId || undefined });
-  },
   setTerminalSessions(sessions: Record<string, { working_dir?: string; shell?: string }>) {
     this.terminalSessions = sessions;
   },
@@ -47,9 +42,7 @@ export const terminalMethods = {
   },
   switchTerminalSession(name: string) {
     this.terminalActiveSession = name;
-    if (this.socket) {
+    // 终端输出由 TerminalPanel 的 REST 轮询获取，无需主动拉取
-      this.socket.emit('get_terminal_output', { session: name, lines: 0, conversation_id: this.currentConversationId || undefined });
-    }
   },
   async fetchTerminalCount() {
     try {

static/src/components/panels/TerminalPanel.vue

@@ -37,7 +37,6 @@
 <script setup lang="ts">
 import { ref, computed, watch, onMounted, onBeforeUnmount, nextTick } from 'vue';
 import { Terminal } from 'xterm';
-import { io as createSocketClient } from 'socket.io-client';
 import 'xterm/css/xterm.css';
 import CloseButton from '@/components/common/CloseButton.vue';
 
@@ -53,33 +52,22 @@ const emit = defineEmits<{
   (event: 'close'): void;
 }>();
 
-// ---- 本地状态（自建 socket，不依赖父组件） ----
+// ---- 本地状态（REST 轮询驱动，不依赖父组件） ----
 const sessions = ref<Record<string, { working_dir?: string; shell?: string }>>({});
 const activeSession = ref('');
 const sessionLogs = ref<Record<string, string>>({});
 const sessionHydrated = ref<Record<string, boolean>>({});
 
-let socket: ReturnType<typeof createSocketClient> | null = null;
-
 const terminalContainer = ref<HTMLElement | null>(null);
 let term: Terminal | null = null;
 let themeObserver: MutationObserver | null = null;
 let resizeObserver: ResizeObserver | null = null;
 const _historyReady = ref<Record<string, boolean>>({});
-// 历史加载完成后记录最后一条事件的时间戳，用于去重刷新后回放的旧 terminal_output 事件
+let listPollTimer: number | null = null;
-const _historyLastEventTime = ref<Record<string, number>>({});
+let outputPollTimer: number | null = null;
 
 const sessionKeys = computed(() => Object.keys(sessions.value));
 
-// 对话级隔离：广播类终端事件（started/list_update/output/input/closed/reset/switched）
-// 由后端注入 conversation_id，仅处理当前对话的事件；
-// 响应类事件（terminal_subscribed / *_history）是请求的直接回应，无此字段，不过滤。
-function acceptTerminalBroadcast(data: any): boolean {
-  const cid = data?.conversation_id;
-  if (!cid || !props.conversationId) return false;
-  return cid === props.conversationId;
-}
-
 // ---- 主题适配 ----
 function getCurrentTheme(): 'light' | 'dark' {
   const theme = document.documentElement.getAttribute('data-theme');
@@ -217,13 +205,9 @@ function switchToSession(name: string) {
   activeSession.value = name;
   if (!sessionHydrated.value[name]) {
     _historyReady.value = { ..._historyReady.value, [name]: false };
-    const updatedTimes = { ..._historyLastEventTime.value };
-    delete updatedTimes[name];
-    _historyLastEventTime.value = updatedTimes;
-    if (socket?.connected) {
-      socket.emit('get_terminal_output', { session: name, lines: 0, conversation_id: props.conversationId });
-    }
   }
+  // 切到新会话立即拉一次输出（不等下一轮轮询）
+  void fetchActiveOutput();
   // renderSessionLog 由 watch(activeSession) 统一触发，此处不重复调用
 }
 
@@ -238,221 +222,121 @@ watch(sessionKeys, (keys) => {
   }
 });
 
-// 工作区切换时重连
+// 工作区/对话切换时重置状态并重新拉取（对话级 terminal：各对话 shell 互相独立）
+function resetPanelState() {
+  sessions.value = {};
+  activeSession.value = '';
+  sessionLogs.value = {};
+  sessionHydrated.value = {};
+  _historyReady.value = {};
+  if (term) term.clear();
+}
+
 watch(() => props.workspaceId, (newId, oldId) => {
   if (newId && oldId && newId !== oldId) {
-    // 断开旧连接，清空状态，重新连接
+    resetPanelState();
-    if (socket) {
+    void fetchTerminalList();
-      socket.disconnect();
-      socket = null;
-    }
-    sessions.value = {};
-    activeSession.value = '';
-    sessionLogs.value = {};
-    sessionHydrated.value = {};
-    _historyReady.value = {};
-    _historyLastEventTime.value = {};
-    if (term) term.clear();
-    initSocket();
   }
 });
 
-// 对话切换时重置并重新订阅（对话级 terminal：各对话 shell 互相独立）
 watch(() => props.conversationId, (newId, oldId) => {
   if (newId !== oldId) {
-    sessions.value = {};
+    resetPanelState();
-    activeSession.value = '';
+    void fetchTerminalList();
-    sessionLogs.value = {};
-    sessionHydrated.value = {};
-    _historyReady.value = {};
-    _historyLastEventTime.value = {};
-    if (term) term.clear();
-    if (socket?.connected) {
-      socket.emit('terminal_subscribe', { all: true, conversation_id: newId });
-    }
   }
 });
 
-// ---- socket 初始化 ----
+// ---- REST 轮询（替代原 WebSocket 订阅与事件推送） ----
-async function initSocket() {
+async function fetchTerminalList() {
-
+  try {
-  socket = createSocketClient('/', {
+    const cid = encodeURIComponent(props.conversationId || '');
-    transports: ['websocket', 'polling'],
+    const res = await fetch(`/api/terminals?conversation_id=${cid}`, { cache: 'no-store' });
-    autoConnect: false
+    if (!res.ok) return;
-  });
+    const data = await res.json();
-
+    const list = Array.isArray(data?.sessions) ? data.sessions : [];
-  const assignToken = async (): Promise<boolean> => {
+    const map: Record<string, { working_dir?: string; shell?: string }> = {};
-    const w = window as any;
+    for (const t of list) {
-    if (typeof w.requestSocketToken !== 'function') {
+      const name = t.session_name || t.name || t.session || t.id;
-      console.warn('[TerminalPanel] requestSocketToken 不可用');
+      if (name) map[name] = { working_dir: t.working_dir, shell: t.shell || 'bash' };
-      return false;
     }
-    try {
+    sessions.value = map;
-      const token = await w.requestSocketToken();
+    const keys = Object.keys(map);
-      (socket as any).auth = { socket_token: token };
+    if (keys.length === 0) {
-      return true;
+      activeSession.value = '';
-    } catch (e) {
+    } else if (!activeSession.value || !map[activeSession.value]) {
-      console.error('[TerminalPanel] 获取 token 失败:', e);
+      // 新终端出现 / 当前终端被关闭：自动切换
-      return false;
+      switchToSession(keys[0]);
     }
-  };
+  } catch {
-
+    // 断线判定由全局连接心跳负责，轮询静默失败
-  socket.io.on('reconnect_attempt', () => assignToken());
-
-  const ready = await assignToken();
-  if (!ready) {
-    console.error('[TerminalPanel] token 获取失败，放弃 socket');
-    socket = null;
-    return;
   }
+}
 
-  socket.on('connect', () => {
+async function fetchActiveOutput() {
-    socket!.emit('terminal_subscribe', { all: true, conversation_id: props.conversationId });
+  const s = activeSession.value;
-  });
+  if (!s) return;
+  try {
+    const cid = encodeURIComponent(props.conversationId || '');
+    const res = await fetch(
+      `/api/terminals/${encodeURIComponent(s)}/output?lines=1000&conversation_id=${cid}`,
+      { cache: 'no-store' }
+    );
+    if (!res.ok) return;
+    const data = await res.json();
+    if (!data?.success) return;
+    const output = typeof data.output === 'string' ? data.output : '';
+    applyOutputSnapshot(s, output);
+  } catch {
+    // 静默失败，下一轮重试
+  }
+}
 
-  socket.on('disconnect', () => {
+// 快照驱动渲染：输出是纯追加时只写增量（避免闪烁）；
-  });
+// reset/截断/窗口滚动导致前缀不匹配时全量重绘。
+function applyOutputSnapshot(session: string, output: string) {
+  const prev = sessionLogs.value[session] || '';
+  sessionHydrated.value = { ...sessionHydrated.value, [session]: true };
+  _historyReady.value = { ..._historyReady.value, [session]: true };
+  if (output === prev) return;
+  if (prev && output.startsWith(prev)) {
+    const delta = output.slice(prev.length);
+    sessionLogs.value[session] = output;
+    if (session === activeSession.value) appendToTerm(delta);
+  } else {
+    sessionLogs.value[session] = output;
+    if (session === activeSession.value) renderSessionLog();
+  }
+}
 
-  socket.on('terminal_subscribed', (data: any) => {
+function startPolling() {
-    if (data?.terminals) {
+  stopPolling();
-      const map: Record<string, { working_dir?: string; shell?: string }> = {};
+  void fetchTerminalList();
-      for (const t of data.terminals) {
+  listPollTimer = window.setInterval(() => void fetchTerminalList(), 5000);
-        map[t.name] = { working_dir: t.working_dir, shell: 'bash' };
+  outputPollTimer = window.setInterval(() => void fetchActiveOutput(), 1500);
-      }
+}
-      sessions.value = map;
-      if (data.terminals.length > 0 && !activeSession.value) {
-        switchToSession(data.terminals[0].name);
-      }
-    }
-  });
 
-  socket.on('terminal_started', (data: any) => {
+function stopPolling() {
-    if (!acceptTerminalBroadcast(data)) return;
+  if (listPollTimer !== null) {
-    sessions.value = {
+    window.clearInterval(listPollTimer);
-      ...sessions.value,
+    listPollTimer = null;
-      [data.session]: { working_dir: data.working_dir, shell: 'bash' }
+  }
-    };
+  if (outputPollTimer !== null) {
-    if (!activeSession.value) switchToSession(data.session);
+    window.clearInterval(outputPollTimer);
-  });
+    outputPollTimer = null;
-
+  }
-  socket.on('terminal_list_update', (data: any) => {
-    if (!acceptTerminalBroadcast(data)) return;
-    if (data?.terminals) {
-      const map: Record<string, { working_dir?: string; shell?: string }> = {};
-      for (const t of data.terminals) {
-        map[t.name] = { working_dir: t.working_dir, shell: 'bash' };
-      }
-      sessions.value = map;
-      if (data.active && !activeSession.value) switchToSession(data.active);
-    }
-  });
-
-  socket.on('terminal_output', (data: any) => {
-    if (!acceptTerminalBroadcast(data)) return;
-    const s = data.session;
-    if (!s) return;
-    const delta = (data.data || '') as string;
-
-    // 去重：刷新页面后后端会回放历史快照（terminal_output_history）
-    // 同时可能回放最近的 terminal_output 事件，导致已有内容被重复写入。
-    // 用历史快照的 last_event_time 作为截止线，跳过更早的事件。
-    const historyTime = _historyLastEventTime.value[s] || 0;
-    const eventTime = typeof data.timestamp === 'number' ? data.timestamp : 0;
-    if (historyTime > 0 && eventTime > 0 && eventTime <= historyTime) {
-      return;
-    }
-
-    // 先把输出暂存到 sessionLogs；历史加载完成后再写入终端，
-    // 避免新终端历史为空时 _historyReady 始终为 false 导致永远写不进去。
-    sessionLogs.value[s] = (sessionLogs.value[s] || '') + delta;
-    if (s === activeSession.value && term && _historyReady.value[s]) {
-      term.write(delta);
-    } else {
-    }
-    if (!activeSession.value) switchToSession(s);
-  });
-
-  socket.on('terminal_input', (data: any) => {
-    if (!acceptTerminalBroadcast(data)) return;
-    // 原样显示后端输出，此处不重复写入
-    // 仅用于新终端自动切换
-    if (!activeSession.value && data.session) switchToSession(data.session);
-  });
-
-  socket.on('terminal_closed', (data: any) => {
-    if (!acceptTerminalBroadcast(data)) return;
-    const updated = { ...sessions.value };
-    delete updated[data.session];
-    sessions.value = updated;
-    const remaining = Object.keys(updated);
-    if (activeSession.value === data.session) {
-      activeSession.value = remaining.length > 0 ? remaining[0] : '';
-    }
-  });
-
-  socket.on('terminal_reset', (data: any) => {
-    if (!acceptTerminalBroadcast(data)) return;
-    const target = data.session || activeSession.value;
-    if (target) {
-      sessionLogs.value = { ...sessionLogs.value, [target]: '' };
-      sessionHydrated.value = { ...sessionHydrated.value, [target]: true };
-      const updatedTimes = { ..._historyLastEventTime.value };
-      delete updatedTimes[target];
-      _historyLastEventTime.value = updatedTimes;
-      if (target === activeSession.value) renderSessionLog();
-    }
-  });
-
-  socket.on('terminal_switched', (data: any) => {
-    if (!acceptTerminalBroadcast(data)) return;
-    if (data.current) switchToSession(data.current);
-  });
-
-  const handleHistory = (data: any) => {
-    const s = data.session || data.active;
-    if (!s) return;
-    const raw = data.output || data.data || '';
-    let payload = '';
-    if (typeof raw === 'string') {
-      payload = raw;
-    } else if (Array.isArray(raw)) {
-      payload = raw.join('\n');
-    }
-    if (payload) {
-      sessionLogs.value[s] = payload;
-    }
-    // 无论历史是否为空，都要标记加载完成；否则新建终端在收到第一条输出前
-    // _historyReady 一直为 false，所有实时输出都会被跳过，终端 seemingly 卡住。
-    sessionHydrated.value = { ...sessionHydrated.value, [s]: true };
-    _historyReady.value = { ..._historyReady.value, [s]: true };
-    // 记录历史快照最后一条事件的时间戳，用于后续 terminal_output 去重
-    const lastEventTime = typeof data.last_event_time === 'number' ? data.last_event_time : 0;
-    _historyLastEventTime.value = { ..._historyLastEventTime.value, [s]: lastEventTime };
-    if (s === activeSession.value) {
-      renderSessionLog();
-    }
-  };
-
-  socket.on('terminal_output_history', handleHistory);
-  socket.on('terminal_history', handleHistory);
-
-  socket.connect();
 }
 
 // ---- 生命周期 ----
 onMounted(() => {
   nextTick(() => {
     initTerminal();
-    initSocket();
+    startPolling();
   });
 });
 
 onBeforeUnmount(() => {
+  stopPolling();
   disposeTerminal();
-  if (socket) {
-    socket.disconnect();
-    socket = null;
-  }
 });
 </script>
 

static/src/stores/connection.ts

@@ -1,10 +1,8 @@
 import { defineStore } from 'pinia';
-import type { Socket } from 'socket.io-client';
 import type { ReasoningEffort } from './personalization';
 
 interface ConnectionState {
   isConnected: boolean;
-  socket: Socket | null;
   stopRequested: boolean;
   projectPath: string;
   agentVersion: string;
@@ -16,7 +14,6 @@ interface ConnectionState {
 export const useConnectionStore = defineStore('connection', {
   state: (): ConnectionState => ({
     isConnected: false,
-    socket: null,
     stopRequested: false,
     projectPath: '',
     agentVersion: '',
@@ -25,9 +22,6 @@ export const useConnectionStore = defineStore('connection', {
     reasoningEffort: null
   }),
   actions: {
-    setSocket(socket: Socket | null) {
-      this.socket = socket;
-    },
     setConnected(value: boolean) {
       this.isConnected = value;
     },