From 1b9e6f6b3aa40f93c2a98c070ef16eff4052db42 Mon Sep 17 00:00:00 2001 From: qing <1776177662@qq.com> Date: Sat, 1 Aug 2026 02:56:29 +0800 Subject: [PATCH] fix(botops): resolve duplicate AI response rendering in ChatView MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit Root cause: aibox.py think_and_respond recursively calls itself after tool execution, and each recursive call's complete callback carries only the final-round full_response. The old finish_ai logic treated complete as a full-text alignment signal — when full_text != self._ai_text (which always happened after recursive tool calls because _ai_text accumulated all chunks while complete only held the last round), it triggered remove+remount, causing visible duplicate/flash of AI responses. A secondary issue was dialogue_ended arriving before complete, which made a second finish_ai call remount the message a second time. Changes: - Remove the full-text replacement branch in finish_ai; chunk streaming is now the authoritative content source, complete is a finalization signal only - Add _ai_finalized flag: start_ai resets it, finish_ai sets it; repeated finish_ai calls (dialogue_ended + complete) return early to prevent remount - append_ai drops late chunks after finalization to avoid dirty appends - Empty pre-built widget without chunks still gets filled via w.update(full_text) - Full lost-frame fallback (no widget + full_text) is preserved - Update BotOps/README.md protocol section to reflect new complete semantics --- BotOps/README.md | 7 +++++- BotOps/widgets/chat_view.py | 50 +++++++++++++++++++++++++------------ 2 files changed, 40 insertions(+), 17 deletions(-) diff --git a/BotOps/README.md b/BotOps/README.md index 3de9e43..9ca5ec1 100644 --- a/BotOps/README.md +++ b/BotOps/README.md @@ -58,4 +58,9 @@ BotOps/ 与 aibox 服务端协议完全兼容（详见 aibox/client.py 注释）： - 发送：`{"type":"message","content":...}` / `{"type":"command","command":"/..."}` / `{"type":"interrupt",...}` -- 接收：`chunk`（**增量片段**，服务端 `chunk_buffer` 发送后即清零，前端直接拼接）/ `complete`（**完整全文**，用于对齐已显示内容）/ `line`（过程行：🔧 工具调用提示、工具输出、💬 助手前缀，路由到任务侧栏）/ `reasoning`（思考流，仅服务端 debug_level≥3 发送）/ `dialogue_ended` / `dialogue_reset` / `task_added` / `task_process` / `task_status` / `task_error` / `interrupt_ack` / `error` / `info` / `command_result` / `connection_ack` / `notification` / `pong` +- 接收： + - `chunk`（**增量片段**）：服务端 `chunk_buffer` 发送后即清零，前端直接 `append_ai` 拼接，为内容权威来源 + - `complete`（**收尾信号**）：chunk 流式已完整渲染，complete 仅用于收尾；仅在完全丢帧（未收到任何 chunk）时用 full_text 补建消息，**不再做全文替换对齐**（避免服务端递归调用工具后 complete 只含最后一轮内容导致重复渲染） + - `line`（过程行）：🔧 工具调用提示、工具输出、💬 助手前缀，路由到任务侧栏；助手前缀/思考提示在聊天区已由 AI 消息框承载，不重复展示 + - `reasoning`（思考流）：仅服务端 debug_level≥3 发送，记入任务侧栏 + - `dialogue_ended` / `dialogue_reset` / `task_added` / `task_process` / `task_status` / `task_error` / `interrupt_ack` / `error` / `info` / `command_result` / `connection_ack` / `notification` / `pong` diff --git a/BotOps/widgets/chat_view.py b/BotOps/widgets/chat_view.py index 718ef2c..445b295 100644 --- a/BotOps/widgets/chat_view.py +++ b/BotOps/widgets/chat_view.py @@ -1,8 +1,9 @@ """聊天视图：用户/AI/系统消息 + AI 回复流式 Markdown 渲染。 -服务端协议（aibox/main.py output_callback 原样透传 msg_type）： - - chunk 增量片段（服务端 chunk_buffer 发送后即清零），直接 append - - complete 完整全文，用于对齐：与已显示文本不一致时整体替换（防丢帧/乱序） +服务端协议（aibox/aibox.py output_callback 原样透传 msg_type）： + - chunk 增量片段（服务端 chunk_buffer 发送后即清零），直接 append + - complete 收尾信号：chunk 流式是内容权威来源，complete 仅用于收尾； + 仅当流式完全丢帧（未建 widget 且无内容）时，用 full_text 补建 """ from textual.containers import VerticalScroll from textual.widgets import Markdown, Static @@ -20,6 +21,7 @@ class ChatView(VerticalScroll): super().__init__(**kwargs) self._ai_widget: Markdown | None = None self._ai_text = "" + self._ai_finalized = True # 初始无活跃 AI 消息 @property def ai_text(self) -> str: @@ -42,6 +44,7 @@ class ChatView(VerticalScroll): if self._ai_widget is not None: return self._ai_text = "" + self._ai_finalized = False self._ai_widget = Markdown("", classes="msg-ai") await self.mount(self._ai_widget) self._anchor_bottom() @@ -50,6 +53,9 @@ class ChatView(VerticalScroll): """追加增量片段（服务端 chunk 为增量语义，直接拼接）。""" if not delta: return + if self._ai_finalized: + # 收尾后收到的延迟 chunk，丢弃以避免在上一条回复后追加脏数据 + return if self._ai_widget is None: await self.start_ai() self._ai_widget.append(delta) @@ -60,31 +66,43 @@ class ChatView(VerticalScroll): async def finish_ai(self, full_text: str = ""): """收尾一条 AI 消息。 - - full_text 为 complete 携带的完整全文：与已显示不一致时整体替换对齐； - - full_text 为空且未收到任何内容（如发送失败）：移除预建的空消息框。 + chunk 流式追加是内容权威来源，complete 只做收尾： + - 已收尾则直接返回（防止 dialogue_ended + complete 双发导致重复挂载） + - widget 存在且有流式内容：保持已渲染内容不变，仅清空引用 + - widget 存在但无内容（预建空框）且有 full_text：用 full_text 填充 + - widget 存在但无内容也无 full_text：移除空框 + - widget 不存在（流式丢帧/start_ai 未调用）且有 full_text：补建 """ + if self._ai_finalized: + return w = self._ai_widget self._ai_widget = None - if w is None: - # 流式期间丢帧，只在 complete 收到全文：补建完整消息 + + if w is not None: + if self._ai_text: + # 有流式内容，chunk 已完整渲染，直接保留，不做替换 + pass + elif full_text: + # 预建了空 widget 但没收到任何 chunk：用 full_text 填充 + w.update(full_text) + self._ai_text = full_text + else: + # 预建后没有任何输出：移除空框 + await w.remove() + else: + # widget 不存在（start_ai 未调用），有 full_text 则补建 if full_text: self._ai_text = full_text await self.mount(Markdown(full_text, classes="msg-ai")) self._anchor_bottom() - return - if full_text and full_text != self._ai_text: - # 已显示内容与全文不一致（丢字/乱序/丢帧），整体替换对齐 - await w.remove() - self._ai_text = full_text - await self.mount(Markdown(full_text, classes="msg-ai")) - elif not full_text and not self._ai_text: - # 预建后没有任何输出：移除空框 - await w.remove() + self._ai_text = "" + self._ai_finalized = True self.call_after_refresh(self.scroll_end, animate=False) async def reset(self): """清空聊天区（/clear）。""" self._ai_widget = None self._ai_text = "" + self._ai_finalized = True await self.remove_children()