feat: enhance Live Mode with text input functionality and UI improvements

- Added a text input panel to allow users to send plain text messages while in Live Mode.
- Updated the LiveMode.vue component to handle text input and integrate it with WebSocket communication.
- Improved the layout and styling of the Live Mode interface for better user experience.
- Documented the new `text_input` message type in the Live API README.

.github/PULL_REQUEST_TEMPLATE.md

@@ -3,8 +3,8 @@
 
 ### Modifications / 改动点
 
-<!--Please summarize your changes: What core files were modified? What functionality was implemented?-->
 <!--请总结你的改动：哪些核心文件被修改了？实现了什么功能？-->
+<!--Please summarize your changes: What core files were modified? What functionality was implemented?-->
 
 - [x] This is NOT a breaking change. / 这不是一个破坏性变更。
 <!-- If your changes is a breaking change, please uncheck the checkbox above -->
@@ -21,14 +21,23 @@
 <!--If merged, your code will serve tens of thousands of users! Please double-check the following items before submitting.-->
 <!--如果分支被合并，您的代码将服务于数万名用户！在提交前，请核查一下几点内容。-->
 
-- [ ] 😊 If there are new features added in the PR, I have discussed it with the authors through issues/emails, etc. 
-  / 如果 PR 中有新加入的功能，已经通过 Issue / 邮件等方式和作者讨论过。
+- [ ] 😊 如果 PR 中有新加入的功能，已经通过 Issue / 邮件等方式和作者讨论过。
+  / If there are new features added in the PR, I have discussed it with the authors through issues/emails, etc.
 
-- [ ] 👀 My changes have been well-tested, **and "Verification Steps" and "Screenshots" have been provided above**.
-  / 我的更改经过了良好的测试，**并已在上方提供了“验证步骤”和“运行截图”**。
+- [ ] 👀 我的更改经过了良好的测试，**并已在上方提供了“验证步骤”和“运行截图”**。
+  / My changes have been well-tested, **and "Verification Steps" and "Screenshots" have been provided above**.
 
-- [ ] 🤓 I have ensured that no new dependencies are introduced, OR if new dependencies are introduced, they have been added to the appropriate locations in `requirements.txt` and `pyproject.toml`.
-  / 我确保没有引入新依赖库，或者引入了新依赖库的同时将其添加到 `requirements.txt` 和 `pyproject.toml` 文件相应位置。
+- [ ] 🤓 我确保没有引入新依赖库，或者引入了新依赖库的同时将其添加到 `requirements.txt` 和 `pyproject.toml` 文件相应位置。
+  / I have ensured that no new dependencies are introduced, OR if new dependencies are introduced, they have been added to the appropriate locations in `requirements.txt` and `pyproject.toml`.
 
-- [ ] 😮 My changes do not introduce malicious code.
-  / 我的更改没有引入恶意代码。
+- [ ] 😮 我的更改没有引入恶意代码。
+  / My changes do not introduce malicious code.
+
+- [ ] ⚠️ 我已认真阅读并理解以上所有内容，确保本次提交符合规范。
+  / I have read and understood all the above and confirm this PR follows the rules.
+
+- [ ] 🚀 我确保本次开发**基于 dev 分支**，并将代码合并至**开发分支**（除非极其紧急，才允许合并到主分支）。
+   / I confirm that this development is **based on the dev branch** and will be merged into the **development branch**, unless it is extremely urgent to merge into the main branch.
+
+- [ ] ⚠️ 我**没有**认真阅读以上内容，直接提交。
+  / I **did not** read the above carefully before submitting.

astrbot/cli/__init__.py

@@ -1 +1 @@
-__version__ = "4.20.1"
+__version__ = "4.20.0"

astrbot/core/astr_agent_run_util.py

@@ -326,6 +326,7 @@ async def run_live_agent(
 
     # 创建队列
     text_queue: asyncio.Queue[str | None] = asyncio.Queue()
+    delta_queue: asyncio.Queue[str | None] = asyncio.Queue()
     # audio_queue stored bytes or (text, bytes)
     audio_queue: asyncio.Queue[bytes | tuple[str, bytes] | None] = asyncio.Queue()
 
@@ -334,6 +335,7 @@ async def run_live_agent(
         _run_agent_feeder(
             agent_runner,
             text_queue,
+            delta_queue,
             max_step,
             show_tool_use,
             show_tool_call_result,
@@ -353,32 +355,63 @@ async def run_live_agent(
 
     # 3. 主循环：从 audio_queue 读取音频并 yield
     try:
-        while True:
-            queue_item = await audio_queue.get()
+        delta_done = False
+        audio_done = False
+        while not (delta_done and audio_done):
+            task_sources: dict[asyncio.Task, str] = {}
+            if not delta_done:
+                task = asyncio.create_task(delta_queue.get())
+                task_sources[task] = "delta"
+            if not audio_done:
+                task = asyncio.create_task(audio_queue.get())
+                task_sources[task] = "audio"
 
-            if queue_item is None:
-                break
+            done, pending = await asyncio.wait(
+                list(task_sources),
+                return_when=asyncio.FIRST_COMPLETED,
+            )
 
-            text = None
-            if isinstance(queue_item, tuple):
-                text, audio_data = queue_item
-            else:
-                audio_data = queue_item
+            for task in pending:
+                task.cancel()
+            if pending:
+                await asyncio.gather(*pending, return_exceptions=True)
 
-            if not first_chunk_received:
-                # 记录首帧延迟（从开始处理到收到第一个音频块）
-                tts_first_frame_time = time.time() - tts_start_time
-                first_chunk_received = True
+            for task in done:
+                source = task_sources[task]
+                queue_item = task.result()
+                if source == "delta":
+                    if queue_item is None:
+                        delta_done = True
+                        continue
+                    yield MessageChain(
+                        chain=[Plain(queue_item)], type="live_text_delta"
+                    )
+                    continue
 
-            # 将音频数据封装为 MessageChain
-            import base64
+                if queue_item is None:
+                    audio_done = True
+                    continue
 
-            audio_b64 = base64.b64encode(audio_data).decode("utf-8")
-            comps: list[BaseMessageComponent] = [Plain(audio_b64)]
-            if text:
-                comps.append(Json(data={"text": text}))
-            chain = MessageChain(chain=comps, type="audio_chunk")
-            yield chain
+                text = None
+                if isinstance(queue_item, tuple):
+                    text, audio_data = queue_item
+                else:
+                    audio_data = queue_item
+
+                if not first_chunk_received:
+                    # 记录首帧延迟（从开始处理到收到第一个音频块）
+                    tts_first_frame_time = time.time() - tts_start_time
+                    first_chunk_received = True
+
+                # 将音频数据封装为 MessageChain
+                import base64
+
+                audio_b64 = base64.b64encode(audio_data).decode("utf-8")
+                comps: list[BaseMessageComponent] = [Plain(audio_b64)]
+                if text:
+                    comps.append(Json(data={"text": text}))
+                chain = MessageChain(chain=comps, type="audio_chunk")
+                yield chain
 
     except Exception as e:
         logger.error(f"[Live Agent] 运行时发生错误: {e}", exc_info=True)
@@ -421,6 +454,7 @@ async def run_live_agent(
 async def _run_agent_feeder(
     agent_runner: AgentRunner,
     text_queue: asyncio.Queue,
+    delta_queue: asyncio.Queue,
     max_step: int,
     show_tool_use: bool,
     show_tool_call_result: bool,
@@ -440,9 +474,13 @@ async def _run_agent_feeder(
             if chain is None:
                 continue
 
+            if chain.type == "reasoning":
+                continue
+
             # 提取文本
             text = chain.get_plain_text()
             if text:
+                await delta_queue.put(text)
                 buffer += text
 
                 # 分句逻辑：匹配标点符号
@@ -477,6 +515,7 @@ async def _run_agent_feeder(
     finally:
         # 发送结束信号
         await text_queue.put(None)
+        await delta_queue.put(None)
 
 
 async def _safe_tts_stream_wrapper(

astrbot/core/config/default.py

@@ -5,7 +5,7 @@ from typing import Any, TypedDict
 
 from astrbot.core.utils.astrbot_path import get_astrbot_data_path
 
-VERSION = "4.20.1"
+VERSION = "4.20.0"
 DB_PATH = os.path.join(get_astrbot_data_path(), "data_v4.db")
 
 WEBHOOK_SUPPORTED_PLATFORMS = [

astrbot/dashboard/routes/live_chat.py

@@ -130,16 +130,6 @@ class LiveChatRoute(Route):
 
     async def live_chat_ws(self) -> None:
         """Legacy Live Chat WebSocket 处理器（默认 ct=live）"""
-        await self._unified_ws_loop(force_ct="live")
-
-    async def unified_chat_ws(self) -> None:
-        """Unified Chat WebSocket 处理器（支持 ct=live/chat）"""
-        await self._unified_ws_loop(force_ct=None)
-
-    async def _unified_ws_loop(self, force_ct: str | None = None) -> None:
-        """统一 WebSocket 循环"""
-        # WebSocket 不能通过 header 传递 token，需要从 query 参数获取
-        # 注意：WebSocket 上下文使用 websocket.args 而不是 request.args
         token = websocket.args.get("token")
         if not token:
             await websocket.close(1008, "Missing authentication token")
@@ -156,6 +146,49 @@ class LiveChatRoute(Route):
             await websocket.close(1008, "Invalid token")
             return
 
+        await self.run_ws_session(username=username, force_ct="live")
+
+    async def unified_chat_ws(self) -> None:
+        """Unified Chat WebSocket 处理器（支持 ct=live/chat）"""
+        token = websocket.args.get("token")
+        if not token:
+            await websocket.close(1008, "Missing authentication token")
+            return
+
+        try:
+            jwt_secret = self.config["dashboard"].get("jwt_secret")
+            payload = jwt.decode(token, jwt_secret, algorithms=["HS256"])
+            username = payload["username"]
+        except jwt.ExpiredSignatureError:
+            await websocket.close(1008, "Token expired")
+            return
+        except jwt.InvalidTokenError:
+            await websocket.close(1008, "Invalid token")
+            return
+
+        await self.run_ws_session(username=username, force_ct=None)
+
+    async def _unified_ws_loop(self, force_ct: str | None = None) -> None:
+        """统一 WebSocket 循环"""
+        # Keep the legacy entry point for internal call sites.
+        token = websocket.args.get("token")
+        if not token:
+            await websocket.close(1008, "Missing authentication token")
+            return
+        try:
+            jwt_secret = self.config["dashboard"].get("jwt_secret")
+            payload = jwt.decode(token, jwt_secret, algorithms=["HS256"])
+            username = payload["username"]
+        except jwt.ExpiredSignatureError:
+            await websocket.close(1008, "Token expired")
+            return
+        except jwt.InvalidTokenError:
+            await websocket.close(1008, "Invalid token")
+            return
+        await self.run_ws_session(username=username, force_ct=force_ct)
+
+    async def run_ws_session(self, username: str, force_ct: str | None = None) -> None:
+        """Run a live/unified websocket session for an authenticated username."""
         session_id = f"webchat_live!{username}!{uuid.uuid4()}"
         live_session = LiveChatSession(session_id, username)
         self.sessions[session_id] = live_session
@@ -690,6 +723,16 @@ class LiveChatRoute(Route):
 
         elif msg_type == "end_speaking":
             # 结束说话
+            if session.is_processing:
+                await websocket.send_json(
+                    {
+                        "t": "error",
+                        "data": "Session is busy",
+                        "code": "PROCESSING_ERROR",
+                    }
+                )
+                return
+
             stamp = message.get("stamp")
             if not stamp:
                 logger.warning("[Live Chat] end_speaking 缺少 stamp")
@@ -703,45 +746,59 @@ class LiveChatRoute(Route):
             # 处理音频：STT -> LLM -> TTS
             await self._process_audio(session, audio_path, assemble_duration)
 
+        elif msg_type == "text_input":
+            if session.is_processing:
+                await websocket.send_json(
+                    {
+                        "t": "error",
+                        "data": "Session is busy",
+                        "code": "PROCESSING_ERROR",
+                    }
+                )
+                return
+
+            user_text = message.get("text")
+            if not isinstance(user_text, str):
+                user_text = message.get("message")
+
+            if not isinstance(user_text, str) or not user_text.strip():
+                await websocket.send_json(
+                    {
+                        "t": "error",
+                        "data": "message must be non-empty text",
+                        "code": "INVALID_MESSAGE_FORMAT",
+                    }
+                )
+                return
+
+            await self._process_live_user_text(
+                session,
+                user_text=user_text.strip(),
+                initial_metrics={"input_type": "text"},
+                processing_start_time=time.time(),
+            )
+
         elif msg_type == "interrupt":
             # 用户打断
             session.should_interrupt = True
             logger.info(f"[Live Chat] 用户打断: {session.username}")
 
-    async def _process_audio(
-        self, session: LiveChatSession, audio_path: str, assemble_duration: float
+    async def _process_live_user_text(
+        self,
+        session: LiveChatSession,
+        user_text: str,
+        initial_metrics: dict[str, Any] | None = None,
+        processing_start_time: float | None = None,
     ) -> None:
-        """处理音频：STT -> LLM -> 流式 TTS"""
+        """处理 Live 用户文本：走 run_live_agent pipeline 并回传流式 TTS."""
         try:
-            # 发送 WAV 组装耗时
-            await websocket.send_json(
-                {"t": "metrics", "data": {"wav_assemble_time": assemble_duration}}
-            )
-            wav_assembly_finish_time = time.time()
+            if initial_metrics:
+                await websocket.send_json({"t": "metrics", "data": initial_metrics})
 
+            processing_start = processing_start_time or time.time()
             session.is_processing = True
             session.should_interrupt = False
 
-            # 1. STT - 语音转文字
-            ctx = self.plugin_manager.context
-            stt_provider = ctx.provider_manager.stt_provider_insts[0]
-
-            if not stt_provider:
-                logger.error("[Live Chat] STT Provider 未配置")
-                await websocket.send_json({"t": "error", "data": "语音识别服务未配置"})
-                return
-
-            await websocket.send_json(
-                {"t": "metrics", "data": {"stt": stt_provider.meta().type}}
-            )
-
-            user_text = await stt_provider.get_text(audio_path)
-            if not user_text:
-                logger.warning("[Live Chat] STT 识别结果为空")
-                return
-
-            logger.info(f"[Live Chat] STT 结果: {user_text}")
-
             await websocket.send_json(
                 {
                     "t": "user_msg",
@@ -761,7 +818,6 @@ class LiveChatRoute(Route):
                 "action_type": "live",  # 标记为 live mode
             }
 
-            # 将消息放入队列
             await queue.put((session.username, cid, payload))
 
             # 3. 等待响应并流式发送 TTS 音频
@@ -776,11 +832,9 @@ class LiveChatRoute(Route):
                         # 用户打断，停止处理
                         logger.info("[Live Chat] 检测到用户打断")
                         await websocket.send_json({"t": "stop_play"})
-                        # 保存消息并标记为被打断
                         await self._save_interrupted_message(
                             session, user_text, bot_text
                         )
-                        # 清空队列中未处理的消息
                         while not back_queue.empty():
                             try:
                                 back_queue.get_nowait()
@@ -805,6 +859,7 @@ class LiveChatRoute(Route):
 
                     result_type = result.get("type")
                     result_chain_type = result.get("chain_type")
+                    result_streaming = bool(result.get("streaming", False))
                     data = result.get("data", "")
 
                     if result_chain_type == "agent_stats":
@@ -827,29 +882,41 @@ class LiveChatRoute(Route):
                     if result_chain_type == "tts_stats":
                         try:
                             stats = json.loads(data)
-                            await websocket.send_json(
-                                {
-                                    "t": "metrics",
-                                    "data": stats,
-                                }
-                            )
+                            await websocket.send_json({"t": "metrics", "data": stats})
                         except Exception as e:
                             logger.error(f"[Live Chat] 解析 TTSStats 失败: {e}")
                         continue
 
+                    if result_chain_type == "live_text_delta":
+                        if data:
+                            await websocket.send_json(
+                                {
+                                    "t": "bot_delta_chunk",
+                                    "data": {"text": data},
+                                }
+                            )
+                        continue
+
                     if result_type == "plain":
-                        # 普通文本消息
+                        if (
+                            result_streaming
+                            and data
+                            and result_chain_type != "reasoning"
+                        ):
+                            await websocket.send_json(
+                                {
+                                    "t": "bot_delta_chunk",
+                                    "data": {"text": data},
+                                }
+                            )
                         bot_text += data
 
                     elif result_type == "audio_chunk":
-                        # 流式音频数据
                         if not audio_playing:
                             audio_playing = True
                             logger.debug("[Live Chat] 开始播放音频流")
-
-                            # Calculate latency from wav assembly finish to first audio chunk
                             speak_to_first_frame_latency = (
-                                time.time() - wav_assembly_finish_time
+                                time.time() - processing_start
                             )
                             await websocket.send_json(
                                 {
@@ -869,19 +936,15 @@ class LiveChatRoute(Route):
                                 }
                             )
 
-                        # 发送音频数据给前端
                         await websocket.send_json(
                             {
                                 "t": "response",
-                                "data": data,  # base64 编码的音频数据
+                                "data": data,
                             }
                         )
 
                     elif result_type in ["complete", "end"]:
-                        # 处理完成
                         logger.info(f"[Live Chat] Bot 回复完成: {bot_text}")
-
-                        # 如果没有音频流，发送 bot 消息文本
                         if not audio_playing:
                             await websocket.send_json(
                                 {
@@ -893,11 +956,8 @@ class LiveChatRoute(Route):
                                 }
                             )
 
-                        # 发送结束标记
                         await websocket.send_json({"t": "end"})
-
-                        # 发送总耗时
-                        wav_to_tts_duration = time.time() - wav_assembly_finish_time
+                        wav_to_tts_duration = time.time() - processing_start
                         await websocket.send_json(
                             {
                                 "t": "metrics",
@@ -909,13 +969,65 @@ class LiveChatRoute(Route):
                 webchat_queue_mgr.remove_back_queue(message_id)
 
         except Exception as e:
-            logger.error(f"[Live Chat] 处理音频失败: {e}", exc_info=True)
+            logger.error(f"[Live Chat] 处理文本失败: {e}", exc_info=True)
             await websocket.send_json({"t": "error", "data": f"处理失败: {str(e)}"})
 
         finally:
             session.is_processing = False
             session.should_interrupt = False
 
+    async def _process_audio(
+        self, session: LiveChatSession, audio_path: str, assemble_duration: float
+    ) -> None:
+        """处理音频：STT -> LLM -> 流式 TTS"""
+        try:
+            await websocket.send_json(
+                {
+                    "t": "metrics",
+                    "data": {
+                        "wav_assemble_time": assemble_duration,
+                        "input_type": "audio",
+                    },
+                }
+            )
+            wav_assembly_finish_time = time.time()
+
+            # 1. STT - 语音转文字
+            ctx = self.plugin_manager.context
+            stt_provider = ctx.provider_manager.stt_provider_insts[0]
+
+            if not stt_provider:
+                logger.error("[Live Chat] STT Provider 未配置")
+                await websocket.send_json({"t": "error", "data": "语音识别服务未配置"})
+                return
+
+            await websocket.send_json(
+                {
+                    "t": "metrics",
+                    "data": {
+                        "stt": stt_provider.meta().type,
+                    },
+                }
+            )
+
+            user_text = await stt_provider.get_text(audio_path)
+            if not user_text:
+                logger.warning("[Live Chat] STT 识别结果为空")
+                return
+
+            logger.info(f"[Live Chat] STT 结果: {user_text}")
+
+            await self._process_live_user_text(
+                session,
+                user_text=user_text,
+                initial_metrics=None,
+                processing_start_time=wav_assembly_finish_time,
+            )
+
+        except Exception as e:
+            logger.error(f"[Live Chat] 处理音频失败: {e}", exc_info=True)
+            await websocket.send_json({"t": "error", "data": f"处理失败: {str(e)}"})
+
     async def _save_interrupted_message(
         self, session: LiveChatSession, user_text: str, bot_text: str
     ) -> None:

astrbot/dashboard/routes/open_api.py

@@ -19,6 +19,7 @@ from astrbot.core.utils.datetime_utils import to_utc_isoformat
 
 from .api_key import ALL_OPEN_API_SCOPES
 from .chat import ChatRoute
+from .live_chat import LiveChatRoute
 from .route import Response, Route, RouteContext
 
 
@@ -29,12 +30,14 @@ class OpenApiRoute(Route):
         db: BaseDatabase,
         core_lifecycle: AstrBotCoreLifecycle,
         chat_route: ChatRoute,
+        live_chat_route: LiveChatRoute,
     ) -> None:
         super().__init__(context)
         self.db = db
         self.core_lifecycle = core_lifecycle
         self.platform_manager = core_lifecycle.platform_manager
         self.chat_route = chat_route
+        self.live_chat_route = live_chat_route
 
         self.routes = {
             "/v1/chat": ("POST", self.chat_send),
@@ -46,6 +49,7 @@ class OpenApiRoute(Route):
         }
         self.register_routes()
         self.app.websocket("/api/v1/chat/ws")(self.chat_ws)
+        self.app.websocket("/api/v1/live/ws")(self.live_ws)
 
     @staticmethod
     def _resolve_open_username(
@@ -534,6 +538,39 @@ class OpenApiRoute(Route):
         except Exception as e:
             logger.debug("Open API WS connection closed: %s", e)
 
+    async def live_ws(self) -> None:
+        authed, auth_err = await self._authenticate_chat_ws_api_key()
+        if not authed:
+            await self._send_chat_ws_error(auth_err or "Unauthorized", "UNAUTHORIZED")
+            await websocket.close(1008, auth_err or "Unauthorized")
+            return
+
+        username, username_err = self._resolve_open_username(
+            websocket.args.get("username")
+        )
+        if username_err or not username:
+            await self._send_chat_ws_error(
+                username_err or "Invalid username",
+                "BAD_USER",
+            )
+            await websocket.close(1008, username_err or "Invalid username")
+            return
+
+        ct = websocket.args.get("ct")
+        force_ct = ct.strip() if isinstance(ct, str) and ct.strip() else "live"
+        if force_ct not in {"live", "chat"}:
+            await self._send_chat_ws_error(
+                "ct must be 'live' or 'chat'",
+                "INVALID_MESSAGE",
+            )
+            await websocket.close(1008, "Invalid ct")
+            return
+
+        await self.live_chat_route.run_ws_session(
+            username=username,
+            force_ct=force_ct,
+        )
+
     async def upload_file(self):
         return await self.chat_route.post_file()
 

astrbot/dashboard/server.py

@@ -115,11 +115,13 @@ class AstrBotDashboard:
         self.ar = AuthRoute(self.context)
         self.api_key_route = ApiKeyRoute(self.context, db)
         self.chat_route = ChatRoute(self.context, db, core_lifecycle)
+        self.live_chat_route = LiveChatRoute(self.context, db, core_lifecycle)
         self.open_api_route = OpenApiRoute(
             self.context,
             db,
             core_lifecycle,
             self.chat_route,
+            self.live_chat_route,
         )
         self.chatui_project_route = ChatUIProjectRoute(self.context, db)
         self.tools_root = ToolsRoute(self.context, core_lifecycle)
@@ -138,7 +140,6 @@ class AstrBotDashboard:
         self.kb_route = KnowledgeBaseRoute(self.context, core_lifecycle)
         self.platform_route = PlatformRoute(self.context, core_lifecycle)
         self.backup_route = BackupRoute(self.context, db, core_lifecycle)
-        self.live_chat_route = LiveChatRoute(self.context, db, core_lifecycle)
 
         self.app.add_url_rule(
             "/api/plug/<path:subpath>",
@@ -244,6 +245,7 @@ class AstrBotDashboard:
         scope_map = {
             "/api/v1/chat": "chat",
             "/api/v1/chat/ws": "chat",
+            "/api/v1/live/ws": "chat",
             "/api/v1/chat/sessions": "chat",
             "/api/v1/configs": "config",
             "/api/v1/file": "file",

changelogs/v4.20.1.md

@@ -1,93 +0,0 @@
-## What's Changed
-
-### 新增
-
-- 补充 MiniMax Provider。（[#6318](https://github.com/AstrBotDevs/AstrBot/pull/6318)）
-- 新增 WebUI ChatUI 页面的会话批量删除功能。（[#6160](https://github.com/AstrBotDevs/AstrBot/pull/6160)）
-- 新增 WebUI ChatUI 配置发送快捷键。（[#6272](https://github.com/AstrBotDevs/AstrBot/pull/6272)）
-
-### 优化
-
-- 优化 UMO 处理兼容性。（[#5996](https://github.com/AstrBotDevs/AstrBot/pull/5996)）
-- 重构 `_extract_session_id`，改进聊天类型分支处理。（#5775）
-- 优化聊天组件行为，使用 `shiki` 进行代码块渲染。（[#6286](https://github.com/AstrBotDevs/AstrBot/pull/6286)）
-- 优化 WebUI 主题配色与视觉体验。（[#6263](https://github.com/AstrBotDevs/AstrBot/pull/6263)）
-- 优化 OneBot @ 组件后处理，避免消息文本解析空格问题。（[#6238](https://github.com/AstrBotDevs/AstrBot/pull/6238)）
-
-### 修复
-
-- 修复创建新 Provider 后未同步 `providers_config` 的问题。（[#6388](https://github.com/AstrBotDevs/AstrBot/pull/6388)）
-- 修复 API 返回 `null choices` 时的 `TypeError`。（[#6313](https://github.com/AstrBotDevs/AstrBot/pull/6313)）
-- 修复 QQ Webhook 重试回调重复触发的问题。（[#6320](https://github.com/AstrBotDevs/AstrBot/pull/6320)）
-- 修复流式模式下 `delta` 为 `None` 导致工具调用时报错的问题。（[#6365](https://github.com/AstrBotDevs/AstrBot/pull/6365)）
-- 修复模型服务链接说明文字错误。（[#6296](https://github.com/AstrBotDevs/AstrBot/pull/6296)）
-- 修复 AI 在 tool-calling 模式设为 `skills-like` 时发送媒体失败的问题。（[#6317](https://github.com/AstrBotDevs/AstrBot/pull/6317)）
-- 修复 Telegram 适配器中 GIF 被错误转成静态图的问题。（[#6329](https://github.com/AstrBotDevs/AstrBot/pull/6329)）
-- 将 Provider 图标来源替换为 jsDelivr CDN 地址，修复部分环境下图标加载问题。（[#6340](https://github.com/AstrBotDevs/AstrBot/pull/6340)）
-- 修复 QQ 官方表情消息未解析为可读文本的问题。（[#6355](https://github.com/AstrBotDevs/AstrBot/pull/6355)）
-- 修复 WebChat 队列异常时流式结果页面崩溃的问题。（[#6123](https://github.com/AstrBotDevs/AstrBot/pull/6123)）
-- 修复子代理 handoff 工具在插件过滤时丢失的问题。（[#6155](https://github.com/AstrBotDevs/AstrBot/pull/6155)）
-- 修复 Cron 提示文案缺少空格及 `utcnow()` 的弃用警告问题。（[#6192](https://github.com/AstrBotDevs/AstrBot/pull/6192)）
-- 修复 WebUI 启动时 Sidebar hash 导航抖动/定位问题。（[#6159](https://github.com/AstrBotDevs/AstrBot/pull/6159)）
-- 修复启动重试过程中移除已移除 API Key 的 `ValueError` 报错。（[#6193](https://github.com/AstrBotDevs/AstrBot/pull/6193)）
-- 修复 README 启动命令引用更新为 `astrbot run`。（[#6189](https://github.com/AstrBotDevs/AstrBot/pull/6189)）
-- 修复 `Plain.toDict()` 在 `@` 提及场景下空白字符丢失的问题。（[#6244](https://github.com/AstrBotDevs/AstrBot/pull/6244)）
-- 修复 provider 依赖重复定义问题。（[#6247](https://github.com/AstrBotDevs/AstrBot/pull/6247)）
-- 修复 Telegram 中普通回复被误判为线程的处理问题。（[#6174](https://github.com/AstrBotDevs/AstrBot/pull/6174)）
-
-### 其他
-
-- 调整 `astrbot.service` 及 CI 配置，升级 GitHub Actions 版本。
-
----
-
-## What's Changed (EN)
-
-### New Features
-
-- Added OpenRouter chat completion provider adapter with support for custom headers ([#6436](https://github.com/AstrBotDevs/AstrBot/pull/6436)).
-- Added MiniMax provider ([#6318](https://github.com/AstrBotDevs/AstrBot/pull/6318)).
-- Added batch conversation deletion in WebChat ([#6160](https://github.com/AstrBotDevs/AstrBot/pull/6160)).
-- Added send shortcut settings and localization support for WebChat input ([#6272](https://github.com/AstrBotDevs/AstrBot/pull/6272)).
-- Added local temporary directory binding in YAML config ([#6191](https://github.com/AstrBotDevs/AstrBot/pull/6191)).
-
-### Improvements
-
-- Improved UMO processing compatibility ([#5996](https://github.com/AstrBotDevs/AstrBot/pull/5996)).
-- Refactored `_extract_session_id` for chat type handling (#5775).
-- Improved chat component behavior and uses `shiki` for code-block rendering ([#6286](https://github.com/AstrBotDevs/AstrBot/pull/6286)).
-- Improved WebUI theme color and visual behavior ([#6263](https://github.com/AstrBotDevs/AstrBot/pull/6263)).
-- Improved OneBot `@` component spacing handling ([#6238](https://github.com/AstrBotDevs/AstrBot/pull/6238)).
-- Improved PR checklist validation and closure messaging.
-
-### Bug Fixes
-
-- Fixed missing `providers_config` sync after creating new providers ([#6388](https://github.com/AstrBotDevs/AstrBot/pull/6388)).
-- Fixed `TypeError` when API returns null choices ([#6313](https://github.com/AstrBotDevs/AstrBot/pull/6313)).
-- Fixed repeated QQ webhook retry callbacks ([#6320](https://github.com/AstrBotDevs/AstrBot/pull/6320)).
-- Fixed tool-calling streaming null `delta` handling to prevent `AttributeError` ([#6365](https://github.com/AstrBotDevs/AstrBot/pull/6365)).
-- Fixed model service link wording in docs/config ([#6296](https://github.com/AstrBotDevs/AstrBot/pull/6296)).
-- Fixed AI media sending failure when tool-calling mode is set to `skills-like` ([#6317](https://github.com/AstrBotDevs/AstrBot/pull/6317)).
-- Fixed GIF being sent as static image in Telegram adapter ([#6329](https://github.com/AstrBotDevs/AstrBot/pull/6329)).
-- Replaced npm registry URLs with jsDelivr CDN for provider icons ([#6340](https://github.com/AstrBotDevs/AstrBot/pull/6340)).
-- Fixed QQ official face message parsing to readable text ([#6355](https://github.com/AstrBotDevs/AstrBot/pull/6355)).
-- Fixed WebChat stream-result crash on queue errors ([#6123](https://github.com/AstrBotDevs/AstrBot/pull/6123)).
-- Preserved subagent handoff tools during plugin filtering ([#6155](https://github.com/AstrBotDevs/AstrBot/pull/6155)).
-- Fixed cron prompt spacing and deprecated `utcnow()` usage ([#6192](https://github.com/AstrBotDevs/AstrBot/pull/6192)).
-- Fixed unstable sidebar hash navigation on startup ([#6159](https://github.com/AstrBotDevs/AstrBot/pull/6159)).
-- Fixed `ValueError` in retry loop when removing an already removed API key ([#6193](https://github.com/AstrBotDevs/AstrBot/pull/6193)).
-- Updated startup command to `astrbot run` across READMEs ([#6189](https://github.com/AstrBotDevs/AstrBot/pull/6189)).
-- Preserved whitespace in `Plain.toDict()` for @ mentions ([#6244](https://github.com/AstrBotDevs/AstrBot/pull/6244)).
-- Removed duplicate dependencies entries ([#6247](https://github.com/AstrBotDevs/AstrBot/pull/6247)).
-- Fixed Telegram normal reply being treated as topic thread ([#6174](https://github.com/AstrBotDevs/AstrBot/pull/6174)).
-
-### Documentation
-
-- Updated `rainyun` backup/access documentation ([#6427](https://github.com/AstrBotDevs/AstrBot/pull/6427)).
-- Updated `package.md` and platform docs, including Matrix and Wecom AI bot documentation.
-- Fixed Discord invite link in community docs.
-
-### Chores
-
-- Updated PR templates/checklist workflow, repository service config, and automated checks.
-- Refreshed repository automation and formatting maintenance, and removed obsolete changelog scripts.

dashboard/src/main.ts

@@ -98,14 +98,28 @@ axios.interceptors.request.use((config) => {
 // Some parts of the UI use fetch directly; without this, those requests will 401.
 const _origFetch = window.fetch.bind(window);
 window.fetch = (input: RequestInfo | URL, init?: RequestInit) => {
+  const requestUrl = (() => {
+    if (typeof input === 'string') return input;
+    if (input instanceof URL) return input.toString();
+    return input.url;
+  })();
+
+  let shouldAttachAuth = false;
+  try {
+    const resolvedUrl = new URL(requestUrl, window.location.origin);
+    shouldAttachAuth = resolvedUrl.origin === window.location.origin;
+  } catch (_) {
+    shouldAttachAuth = requestUrl.startsWith('/');
+  }
+
   const token = localStorage.getItem('token');
-  if (!token) return _origFetch(input, init);
+  const locale = localStorage.getItem('astrbot-locale');
+  if (!token && !locale) return _origFetch(input, init);
 
   const headers = new Headers(init?.headers || (typeof input !== 'string' && 'headers' in input ? (input as Request).headers : undefined));
-  if (!headers.has('Authorization')) {
+  if (shouldAttachAuth && token && !headers.has('Authorization')) {
     headers.set('Authorization', `Bearer ${token}`);
   }
-  const locale = localStorage.getItem('astrbot-locale');
   if (locale && !headers.has('Accept-Language')) {
     headers.set('Accept-Language', locale);
   }

docs/en/dev/openapi.md

@@ -29,6 +29,7 @@ X-API-Key: abk_xxx
 ## Common Endpoints
 
 - `POST /api/v1/chat`: send chat message (SSE stream, server generates UUID when `session_id` is omitted)
+- `GET /api/v1/live/ws`: Live API WebSocket (API Key auth, requires `username` query parameter, optional `ct=live|chat`)
 - `GET /api/v1/chat/sessions`: list sessions for a specific `username` with pagination
 - `GET /api/v1/configs`: list available config files
 - `POST /api/v1/file`: upload attachment
@@ -49,3 +50,7 @@ curl -N 'http://localhost:6185/api/v1/chat' \
 Use the interactive docs:
 
 - https://docs.astrbot.app/scalar.html
+
+For the full Live API wire protocol, see:
+
+- `docs/live-api/README.md`

docs/live-api/README.md

@@ -0,0 +1,434 @@
+# AstrBot Live API Protocol
+
+This document describes the current WebSocket protocol for AstrBot Live API.
+
+## Endpoint
+
+- Legacy JWT endpoint: `/api/live_chat/ws`
+- Legacy unified JWT endpoint: `/api/unified_chat/ws`
+- Open API endpoint: `/api/v1/live/ws`
+
+## Authentication
+
+### Legacy dashboard endpoints
+
+Pass a dashboard JWT in the `token` query parameter.
+
+Example:
+
+```text
+ws://localhost:6185/api/live_chat/ws?token=<dashboard_jwt>
+```
+
+### Open API endpoint
+
+Use an API key and provide `username` in the query string.
+
+Examples:
+
+```text
+ws://localhost:6185/api/v1/live/ws?api_key=<api_key>&username=alice
+ws://localhost:6185/api/v1/live/ws?api_key=<api_key>&username=alice&ct=chat
+```
+
+`ct` values:
+
+- `live`: voice conversation mode
+- `chat`: unified chat mode over the same WebSocket transport
+
+The Open API endpoint reuses the `chat` API key scope.
+
+## Transport
+
+- Protocol: WebSocket
+- Payload format: UTF-8 JSON text frames
+- Audio upload format in `live` mode:
+  - client sends raw PCM frames encoded as Base64
+  - sample rate: `16000`
+  - channels: `1`
+  - sample width: `16-bit`
+
+## Top-Level Envelope
+
+### Client to server
+
+```json
+{
+  "t": "message_type",
+  "...": "message specific fields"
+}
+```
+
+When using the unified socket, the client can also include:
+
+```json
+{
+  "ct": "live|chat",
+  "t": "message_type"
+}
+```
+
+### Server to client
+
+Legacy `live` mode uses:
+
+```json
+{
+  "t": "message_type",
+  "data": {}
+}
+```
+
+Unified `chat` mode uses:
+
+```json
+{
+  "ct": "chat",
+  "type": "message_type",
+  "data": {}
+}
+```
+
+Some forwarded `chat` frames may also contain `t`, `streaming`, `chain_type`, `message_id`, or `session_id`.
+
+## Live Mode
+
+### Client messages
+
+#### `start_speaking`
+
+Start a voice capture segment.
+
+```json
+{
+  "t": "start_speaking",
+  "stamp": "seg_001"
+}
+```
+
+#### `speaking_part`
+
+Send one audio frame.
+
+```json
+{
+  "t": "speaking_part",
+  "data": "<base64_pcm_bytes>"
+}
+```
+
+#### `end_speaking`
+
+Finish the current voice capture segment.
+
+```json
+{
+  "t": "end_speaking",
+  "stamp": "seg_001"
+}
+```
+
+#### `text_input`
+
+Send a plain text input directly while using `ct=live`. The server will still route through Live mode with TTS and interrupt handling.
+
+```json
+{
+  "t": "text_input",
+  "text": "Hello, what is the weather today?"
+}
+```
+
+#### `interrupt`
+
+Interrupt the current model or TTS response.
+
+```json
+{
+  "t": "interrupt"
+}
+```
+
+### Server messages
+
+#### `metrics`
+
+Performance and provider metadata.
+
+Example:
+
+```json
+{
+  "t": "metrics",
+  "data": {
+    "wav_assemble_time": 0.12,
+    "stt": "whisper_api",
+    "llm_ttft": 0.84,
+    "tts_total_time": 1.72
+  }
+}
+```
+
+#### `user_msg`
+
+STT result from the uploaded audio.
+
+```json
+{
+  "t": "user_msg",
+  "data": {
+    "text": "Hello there",
+    "ts": 1710000000000
+  }
+}
+```
+
+#### `bot_delta_chunk`
+
+Raw model text delta. This is the token or chunk level stream and is not sentence segmented.
+
+```json
+{
+  "t": "bot_delta_chunk",
+  "data": {
+    "text": "Hel"
+  }
+}
+```
+
+Notes:
+
+- This event is generated directly from the model streaming path.
+- It is independent from TTS chunking.
+- Consumers should append `data.text` to a local buffer.
+
+#### `bot_text_chunk`
+
+Text associated with the current TTS chunk. This is usually sentence or phrase segmented.
+
+```json
+{
+  "t": "bot_text_chunk",
+  "data": {
+    "text": "Hello there."
+  }
+}
+```
+
+Notes:
+
+- This event is aligned to TTS output, not raw token streaming.
+- It may be coarser than `bot_delta_chunk`.
+
+#### `response`
+
+One TTS audio chunk, Base64 encoded.
+
+```json
+{
+  "t": "response",
+  "data": "<base64_audio_bytes>"
+}
+```
+
+#### `bot_msg`
+
+Final bot text when the response completed without audio streaming.
+
+```json
+{
+  "t": "bot_msg",
+  "data": {
+    "text": "Final reply text",
+    "ts": 1710000001234
+  }
+}
+```
+
+#### `stop_play`
+
+Stop client-side audio playback because the response was interrupted.
+
+```json
+{
+  "t": "stop_play"
+}
+```
+
+#### `end`
+
+Marks the end of the current response turn.
+
+```json
+{
+  "t": "end"
+}
+```
+
+#### `error`
+
+Recoverable or terminal processing error.
+
+```json
+{
+  "t": "error",
+  "data": "error message"
+}
+```
+
+## Unified Chat Mode
+
+Set `ct=chat` on the Open API endpoint or include `"ct": "chat"` in each client frame when using `/api/unified_chat/ws`.
+
+### Client messages
+
+#### `bind`
+
+Subscribe to an existing webchat session.
+
+```json
+{
+  "ct": "chat",
+  "t": "bind",
+  "session_id": "session_001"
+}
+```
+
+#### `send`
+
+Send a chat request.
+
+```json
+{
+  "ct": "chat",
+  "t": "send",
+  "username": "alice",
+  "session_id": "session_001",
+  "message_id": "msg_001",
+  "message": [
+    {
+      "type": "plain",
+      "text": "Please summarize this"
+    }
+  ],
+  "selected_provider": "openai_chat_completion",
+  "selected_model": "gpt-4.1-mini",
+  "enable_streaming": true
+}
+```
+
+`message` uses the same message-part schema as `POST /api/v1/chat`.
+
+#### `interrupt`
+
+Interrupt the current chat response.
+
+```json
+{
+  "ct": "chat",
+  "t": "interrupt"
+}
+```
+
+### Server messages
+
+#### `session_bound`
+
+Acknowledges a successful `bind`.
+
+```json
+{
+  "ct": "chat",
+  "type": "session_bound",
+  "session_id": "session_001",
+  "message_id": "ws_sub_xxx"
+}
+```
+
+#### Forwarded streaming events
+
+The server forwards the normal webchat queue payloads. Common examples:
+
+```json
+{
+  "ct": "chat",
+  "type": "plain",
+  "data": "Hello",
+  "streaming": true,
+  "chain_type": null,
+  "message_id": "msg_001"
+}
+```
+
+```json
+{
+  "ct": "chat",
+  "type": "image",
+  "data": "[IMAGE]file.jpg",
+  "streaming": false,
+  "message_id": "msg_001"
+}
+```
+
+```json
+{
+  "ct": "chat",
+  "type": "agent_stats",
+  "data": {
+    "time_to_first_token": 0.8
+  }
+}
+```
+
+```json
+{
+  "ct": "chat",
+  "type": "message_saved",
+  "data": {
+    "id": 123,
+    "created_at": "2026-03-16T10:00:00Z"
+  }
+}
+```
+
+```json
+{
+  "ct": "chat",
+  "type": "end",
+  "data": "",
+  "streaming": false,
+  "message_id": "msg_001"
+}
+```
+
+#### Chat errors
+
+```json
+{
+  "ct": "chat",
+  "t": "error",
+  "code": "INVALID_MESSAGE_FORMAT",
+  "data": "message must be list"
+}
+```
+
+## Recommended Client Strategy
+
+For `live` mode:
+
+1. Append every `bot_delta_chunk.data.text` into a raw transcript buffer.
+2. Use `bot_text_chunk` only when you need text aligned with audio playback.
+3. Decode and play each `response` audio chunk in arrival order.
+4. Reset per-turn buffers after `end`.
+
+For `chat` mode:
+
+1. Treat `plain + streaming=true` as incremental text.
+2. Treat `complete` or `end` as the end of a response turn.
+3. Persist `message_saved` metadata if you need server-side history IDs.
+
+## Compatibility Notes
+
+- `bot_text_chunk` remains sentence or phrase segmented for TTS compatibility.
+- `bot_delta_chunk` is the new delta-level text event for real-time rendering.
+- The legacy JWT endpoints and the new Open API endpoint share the same runtime behavior after authentication.

docs/public/openapi.json

@@ -257,6 +257,56 @@
         }
       }
     },
+    "/api/v1/live/ws": {
+      "get": {
+        "tags": [
+          "Open API"
+        ],
+        "summary": "Live API WebSocket",
+        "description": "WebSocket endpoint for Live API. Authenticate with API Key using query parameter `api_key` or header `Authorization: Bearer <api_key>`, and pass `username` as a query parameter. Use `ct=live` for voice mode or `ct=chat` for unified chat mode. See docs/live-api/README.md for the full frame-level protocol.",
+        "security": [
+          {
+            "ApiKeyHeader": []
+          }
+        ],
+        "parameters": [
+          {
+            "name": "username",
+            "in": "query",
+            "required": true,
+            "schema": {
+              "type": "string"
+            },
+            "description": "Target username for the live session."
+          },
+          {
+            "name": "ct",
+            "in": "query",
+            "schema": {
+              "type": "string",
+              "enum": [
+                "live",
+                "chat"
+              ],
+              "default": "live"
+            },
+            "description": "Session mode. `live` for voice conversation, `ct=chat` for the unified chat WebSocket."
+          }
+        ],
+        "responses": {
+          "101": {
+            "description": "WebSocket protocol switch"
+          },
+          "401": {
+            "$ref": "#/components/responses/Unauthorized"
+          },
+          "403": {
+            "$ref": "#/components/responses/Forbidden"
+          }
+        },
+        "x-websocket": true
+      }
+    },
     "/api/v1/im/message": {
       "post": {
         "tags": [

docs/zh/dev/openapi.md

@@ -46,6 +46,7 @@ X-API-Key: abk_xxx
 调用 AstrBot 内建的 Agent 进行对话交互。支持插件调用、工具调用等能力，与 IM 端对话能力一致。
 
 - `POST /api/v1/chat`：发送对话消息（SSE 流式返回，不传 `session_id` 会自动创建 UUID）
+- `GET /api/v1/live/ws`：Live API WebSocket（API Key 鉴权，查询参数必须包含 `username`，可选 `ct=live|chat`）
 - `GET /api/v1/chat/sessions`：分页获取指定 `username` 的会话
 - `GET /api/v1/configs`：获取可用配置文件列表
 
@@ -148,3 +149,7 @@ curl -N 'http://localhost:6185/api/v1/chat' \
 交互式 API 文档请查看：
 
 - https://docs.astrbot.app/scalar.html
+
+Live API 协议说明请查看：
+
+- `docs/live-api/README.md`

openapi.json

@@ -257,6 +257,56 @@
         }
       }
     },
+    "/api/v1/live/ws": {
+      "get": {
+        "tags": [
+          "Open API"
+        ],
+        "summary": "Live API WebSocket",
+        "description": "WebSocket endpoint for Live API. Authenticate with API Key using query parameter `api_key` or header `Authorization: Bearer <api_key>`, and pass `username` as a query parameter. Use `ct=live` for voice mode or `ct=chat` for unified chat mode. See docs/live-api/README.md for the full frame-level protocol.",
+        "security": [
+          {
+            "ApiKeyHeader": []
+          }
+        ],
+        "parameters": [
+          {
+            "name": "username",
+            "in": "query",
+            "required": true,
+            "schema": {
+              "type": "string"
+            },
+            "description": "Target username for the live session."
+          },
+          {
+            "name": "ct",
+            "in": "query",
+            "schema": {
+              "type": "string",
+              "enum": [
+                "live",
+                "chat"
+              ],
+              "default": "live"
+            },
+            "description": "Session mode. `live` for voice conversation, `chat` for the unified chat WebSocket."
+          }
+        ],
+        "responses": {
+          "101": {
+            "description": "WebSocket protocol switch"
+          },
+          "401": {
+            "$ref": "#/components/responses/Unauthorized"
+          },
+          "403": {
+            "$ref": "#/components/responses/Forbidden"
+          }
+        },
+        "x-websocket": true
+      }
+    },
     "/api/v1/im/message": {
       "post": {
         "tags": [

pyproject.toml

@@ -1,6 +1,6 @@
 [project]
 name = "AstrBot"
-version = "4.20.1"
+version = "4.20.0"
 description = "Easy-to-use multi-platform LLM chatbot and development framework"
 readme = "README.md"
 requires-python = ">=3.12"