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

@@ -21,7 +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。-->
 
-- [ ] 😊 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.
-- [ ] 👀 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**.
-- [ ] 🤓 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了 `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`.
-- [ ] 😮 我的更改没有引入恶意代码。/ My changes do not introduce malicious code.
+- [ ] 😊 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.
+
+- [ ] 👀 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**.
+
+- [ ] 🤓 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`.
+
+- [ ] 😮 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.

.github/workflows/pr-checklist-check.yml

@@ -0,0 +1,45 @@
+name: PR Checklist Check
+
+on:
+  pull_request_target:
+    types: [opened, edited, reopened, synchronize]
+
+jobs:
+  check:
+    runs-on: ubuntu-latest
+
+    permissions:
+      pull-requests: write
+      issues: write
+
+    steps:
+      - name: Check checklist
+        id: check
+        uses: actions/github-script@v7
+        with:
+          script: |
+            const body = context.payload.pull_request.body || "";
+            const regex = /-\s*\[\s*x\s*\].*没有.*认真阅读/i;
+            const bad = regex.test(body);
+            core.setOutput("bad", bad);
+
+      - name: Close PR
+        if: steps.check.outputs.bad == 'true'
+        uses: actions/github-script@v7
+        with:
+          script: |
+            const pr = context.payload.pull_request;
+
+            await github.rest.issues.createComment({
+              owner: context.repo.owner,
+              repo: context.repo.repo,
+              issue_number: pr.number,
+              body: `检测到你勾选了“我没有认真阅读”，PR 已关闭。`
+            });
+
+            await github.rest.pulls.update({
+              owner: context.repo.owner,
+              repo: context.repo.repo,
+              pull_number: pr.number,
+              state: "closed"
+            });

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/astr_main_agent_resources.py

@@ -188,7 +188,12 @@ class KnowledgeBaseQueryTool(FunctionTool[AstrAgentContext]):
 @dataclass
 class SendMessageToUserTool(FunctionTool[AstrAgentContext]):
     name: str = "send_message_to_user"
-    description: str = "Directly send message to the user. Only use this tool when you need to proactively message the user. Otherwise you can directly output the reply in the conversation."
+    description: str = (
+        "Send message to the user. "
+        "Supports various message types including `plain`, `image`, `record`, `video`, `file`, and `mention_user`. "
+        "Use this tool to send media files (`image`, `record`, `video`, `file`), "
+        "or when you need to proactively message the user(such as cron job). For normal text replies, you can output directly."
+    )
 
     parameters: dict = Field(
         default_factory=lambda: {

astrbot/core/computer/tools/neo_skills.py

@@ -164,7 +164,10 @@ class CreateSkillPayloadTool(NeoSkillToolBase):
             "type": "object",
             "properties": {
                 "payload": {
-                    "anyOf": [{"type": "object"}, {"type": "array"}],
+                    "anyOf": [
+                        {"type": "object"},
+                        {"type": "array", "items": {"type": "object"}},
+                    ],
                     "description": (
                         "Skill payload JSON. Typical schema: {skill_markdown, inputs, outputs, meta}. "
                         "This only stores content and returns payload_ref; it does not create a candidate or release."

astrbot/core/config/default.py

@@ -1132,6 +1132,18 @@ CONFIG_METADATA_2 = {
                         "proxy": "",
                         "custom_headers": {},
                     },
+                    "MiniMax": {
+                        "id": "minimax",
+                        "provider": "minimax",
+                        "type": "openai_chat_completion",
+                        "provider_type": "chat_completion",
+                        "enable": True,
+                        "key": [],
+                        "api_base": "https://api.minimaxi.com/v1",
+                        "timeout": 120,
+                        "proxy": "",
+                        "custom_headers": {},
+                    },
                     "xAI": {
                         "id": "xai",
                         "provider": "xai",

astrbot/core/platform/sources/qqofficial/qqofficial_platform_adapter.py

@@ -391,6 +391,47 @@ class QQOfficialPlatformAdapter(Platform):
                 else:
                     msg.append(File(name=filename, file=url, url=url))
 
+    @staticmethod
+    def _parse_face_message(content: str) -> str:
+        """Parse QQ official face message format and convert to readable text.
+
+        QQ official face message format:
+        <faceType=4,faceId="",ext="eyJ0ZXh0IjoiW+a7oeWktOmXruWPt10ifQ==">
+
+        The ext field contains base64-encoded JSON with a 'text' field
+        describing the emoji (e.g., '[满头问号]').
+
+        Args:
+            content: The message content that may contain face tags.
+
+        Returns:
+            Content with face tags replaced by readable emoji descriptions.
+        """
+        import base64
+        import json
+        import re
+
+        def replace_face(match):
+            face_tag = match.group(0)
+            # Extract ext field from the face tag
+            ext_match = re.search(r'ext="([^"]*)"', face_tag)
+            if ext_match:
+                try:
+                    ext_encoded = ext_match.group(1)
+                    # Decode base64 and parse JSON
+                    ext_decoded = base64.b64decode(ext_encoded).decode("utf-8")
+                    ext_data = json.loads(ext_decoded)
+                    emoji_text = ext_data.get("text", "")
+                    if emoji_text:
+                        return f"[表情:{emoji_text}]"
+                except Exception:
+                    pass
+            # Fallback if parsing fails
+            return "[表情]"
+
+        # Match face tags: <faceType=...>
+        return re.sub(r"<faceType=\d+[^>]*>", replace_face, content)
+
     @staticmethod
     def _parse_from_qqofficial(
         message: botpy.message.Message
@@ -416,7 +457,10 @@ class QQOfficialPlatformAdapter(Platform):
                 abm.group_id = message.group_openid
             else:
                 abm.sender = MessageMember(message.author.user_openid, "")
-            abm.message_str = message.content.strip()
+            # Parse face messages to readable text
+            abm.message_str = QQOfficialPlatformAdapter._parse_face_message(
+                message.content.strip()
+            )
             abm.self_id = "unknown_selfid"
             msg.append(At(qq="qq_official"))
             msg.append(Plain(abm.message_str))
@@ -432,10 +476,12 @@ class QQOfficialPlatformAdapter(Platform):
             else:
                 abm.self_id = ""
 
-            plain_content = message.content.replace(
-                "<@!" + str(abm.self_id) + ">",
-                "",
-            ).strip()
+            plain_content = QQOfficialPlatformAdapter._parse_face_message(
+                message.content.replace(
+                    "<@!" + str(abm.self_id) + ">",
+                    "",
+                ).strip()
+            )
 
             QQOfficialPlatformAdapter._append_attachments(msg, message.attachments)
             abm.message = msg

astrbot/core/platform/sources/qqofficial_webhook/qo_webhook_server.py

@@ -1,5 +1,6 @@
 import asyncio
 import logging
+import time
 from typing import cast
 
 import quart
@@ -39,6 +40,9 @@ class QQOfficialWebhook:
         self.client = botpy_client
         self.event_queue = event_queue
         self.shutdown_event = asyncio.Event()
+        # Deduplication cache for webhook retry callbacks.
+        self._seen_event_ids: dict[str, float] = {}
+        self._dedup_ttl: int = 60  # seconds
 
     async def initialize(self) -> None:
         logger.info("正在登录到 QQ 官方机器人...")
@@ -106,6 +110,22 @@ class QQOfficialWebhook:
             print(signed)
             return signed
 
+        event_id = msg.get("id")
+        if event_id:
+            now = time.monotonic()
+            # Lazily evict expired entries to prevent unbounded growth.
+            expired = [
+                k
+                for k, ts in self._seen_event_ids.items()
+                if now - ts > self._dedup_ttl
+            ]
+            for k in expired:
+                del self._seen_event_ids[k]
+            if event_id in self._seen_event_ids:
+                logger.debug(f"Duplicate webhook event {event_id!r}, skipping.")
+                return {"opcode": 12}
+            self._seen_event_ids[event_id] = now
+
         if event and opcode == BotWebSocket.WS_DISPATCH_EVENT:
             event = msg["t"].lower()
             try:

astrbot/core/platform/sources/telegram/tg_event.py

@@ -25,6 +25,16 @@ from astrbot.api.platform import AstrBotMessage, MessageType, PlatformMetadata
 from astrbot.core.utils.metrics import Metric
 
 
+def _is_gif(path: str) -> bool:
+    if path.lower().endswith(".gif"):
+        return True
+    try:
+        with open(path, "rb") as f:
+            return f.read(6) in (b"GIF87a", b"GIF89a")
+    except OSError:
+        return False
+
+
 class TelegramPlatformEvent(AstrMessageEvent):
     # Telegram 的最大消息长度限制
     MAX_MESSAGE_LENGTH = 4096
@@ -291,7 +301,13 @@ class TelegramPlatformEvent(AstrMessageEvent):
                         await client.send_message(text=chunk, **cast(Any, payload))
             elif isinstance(i, Image):
                 image_path = await i.convert_to_file_path()
-                await client.send_photo(photo=image_path, **cast(Any, payload))
+                if _is_gif(image_path):
+                    send_coro = client.send_animation
+                    media_kwarg = {"animation": image_path}
+                else:
+                    send_coro = client.send_photo
+                    media_kwarg = {"photo": image_path}
+                await send_coro(**media_kwarg, **cast(Any, payload))
             elif isinstance(i, File):
                 path = await i.get_file()
                 name = i.name or os.path.basename(path)
@@ -406,12 +422,20 @@ class TelegramPlatformEvent(AstrMessageEvent):
                 on_text(i.text)
             elif isinstance(i, Image):
                 image_path = await i.convert_to_file_path()
+                if _is_gif(image_path):
+                    action = ChatAction.UPLOAD_VIDEO
+                    send_coro = self.client.send_animation
+                    media_kwarg = {"animation": image_path}
+                else:
+                    action = ChatAction.UPLOAD_PHOTO
+                    send_coro = self.client.send_photo
+                    media_kwarg = {"photo": image_path}
                 await self._send_media_with_action(
                     self.client,
-                    ChatAction.UPLOAD_PHOTO,
-                    self.client.send_photo,
+                    action,
+                    send_coro,
                     user_name=user_name,
-                    photo=image_path,
+                    **media_kwarg,
                     **cast(Any, payload),
                 )
             elif isinstance(i, File):

astrbot/core/platform/sources/wecom_ai_bot/wecomai_adapter.py

@@ -440,9 +440,16 @@ class WecomAIBotAdapter(Platform):
         )
 
     def _extract_session_id(self, message_data: dict[str, Any]) -> str:
-        """从消息数据中提取会话ID"""
-        user_id = message_data.get("from", {}).get("userid", "default_user")
-        return format_session_id("wecomai", user_id)
+        """从消息数据中提取会话ID
+        群聊使用 chatid，单聊使用 userid
+        """
+        chattype = message_data.get("chattype", "single")
+        if chattype == "group":
+            chat_id = message_data.get("chatid", "default_group")
+            return format_session_id("wecomai", chat_id)
+        else:
+            user_id = message_data.get("from", {}).get("userid", "default_user")
+            return format_session_id("wecomai", user_id)
 
     async def _enqueue_message(
         self,

astrbot/core/provider/manager.py

@@ -808,6 +808,8 @@ class ProviderManager:
             config.save_config()
             # load instance
             await self.load_provider(new_config)
+            # sync in-memory config for API queries (e.g., embedding provider list)
+            self.providers_config = astrbot_config["provider"]
 
     async def terminate(self) -> None:
         if self._mcp_init_task and not self._mcp_init_task.done():

astrbot/core/provider/sources/groq_source.py

@@ -13,3 +13,11 @@ class ProviderGroq(ProviderOpenAIOfficial):
     ) -> None:
         super().__init__(provider_config, provider_settings)
         self.reasoning_key = "reasoning"
+
+    def _finally_convert_payload(self, payloads: dict) -> None:
+        """Groq rejects assistant history items that include reasoning_content."""
+        super()._finally_convert_payload(payloads)
+        for message in payloads.get("messages", []):
+            if message.get("role") == "assistant":
+                message.pop("reasoning_content", None)
+                message.pop("reasoning", None)

astrbot/core/provider/sources/openai_source.py

@@ -311,7 +311,7 @@ class ProviderOpenAIOfficial(Provider):
                 state.handle_chunk(chunk)
             except Exception as e:
                 logger.warning("Saving chunk state error: " + str(e))
-            if len(chunk.choices) == 0:
+            if not chunk.choices:
                 continue
             delta = chunk.choices[0].delta
             # logger.debug(f"chunk delta: {delta}")
@@ -322,7 +322,7 @@ class ProviderOpenAIOfficial(Provider):
             if reasoning:
                 llm_response.reasoning_content = reasoning
                 _y = True
-            if delta.content:
+            if delta and delta.content:
                 # Don't strip streaming chunks to preserve spaces between words
                 completion_text = self._normalize_content(delta.content, strip=False)
                 llm_response.result_chain = MessageChain(
@@ -345,7 +345,7 @@ class ProviderOpenAIOfficial(Provider):
     ) -> str:
         """Extract reasoning content from OpenAI ChatCompletion if available."""
         reasoning_text = ""
-        if len(completion.choices) == 0:
+        if not completion.choices:
             return reasoning_text
         if isinstance(completion, ChatCompletion):
             choice = completion.choices[0]
@@ -468,7 +468,7 @@ class ProviderOpenAIOfficial(Provider):
         """Parse OpenAI ChatCompletion into LLMResponse"""
         llm_response = LLMResponse("assistant")
 
-        if len(completion.choices) == 0:
+        if not completion.choices:
             raise Exception("API 返回的 completion 为空。")
         choice = completion.choices[0]
 

astrbot/core/provider/sources/openrouter_source.py

@@ -16,4 +16,7 @@ class ProviderOpenRouter(ProviderOpenAIOfficial):
         self.client._custom_headers["HTTP-Referer"] = (  # type: ignore
             "https://github.com/AstrBotDevs/AstrBot"
         )
-        self.client._custom_headers["X-TITLE"] = "AstrBot"  # type: ignore
+        self.client._custom_headers["X-OpenRouter-Title"] = "AstrBot"  # type: ignore
+        self.client._custom_headers["X-OpenRouter-Categories"] = (
+            "general-chat,personal-agent"  # type: ignore
+        )

astrbot/core/umop_config_router.py

@@ -25,12 +25,22 @@ class UmopConfigRouter:
         )
         self.umop_to_conf_id = sp_data
 
+    @staticmethod
+    def _split_umo(umo: str) -> tuple[str, str, str] | None:
+        """将 UMO 拆分为 3 个部分，同时保留 session_id 中的 ':'"""
+        if not isinstance(umo, str):
+            return None
+        parts = umo.split(":", 2)
+        if len(parts) != 3:
+            return None
+        return parts[0], parts[1], parts[2]
+
     def _is_umo_match(self, p1: str, p2: str) -> bool:
         """判断 p2 umo 是否逻辑包含于 p1 umo"""
-        p1_ls = p1.split(":")
-        p2_ls = p2.split(":")
+        p1_ls = self._split_umo(p1)
+        p2_ls = self._split_umo(p2)
 
-        if len(p1_ls) != 3 or len(p2_ls) != 3:
+        if p1_ls is None or p2_ls is None:
             return False  # 非法格式
 
         return all(p == "" or fnmatch.fnmatchcase(t, p) for p, t in zip(p1_ls, p2_ls))
@@ -62,7 +72,7 @@ class UmopConfigRouter:
 
         """
         for part in new_routing:
-            if not isinstance(part, str) or len(part.split(":")) != 3:
+            if self._split_umo(part) is None:
                 raise ValueError(
                     "umop keys must be strings in the format [platform_id]:[message_type]:[session_id], with optional wildcards * or empty for all",
                 )
@@ -81,7 +91,7 @@ class UmopConfigRouter:
             ValueError: 如果 umo 格式不正确
 
         """
-        if not isinstance(umo, str) or len(umo.split(":")) != 3:
+        if self._split_umo(umo) is None:
             raise ValueError(
                 "umop must be a string in the format [platform_id]:[message_type]:[session_id], with optional wildcards * or empty for all",
             )
@@ -99,7 +109,7 @@ class UmopConfigRouter:
             ValueError: 当 umo 格式不正确时抛出
         """
 
-        if not isinstance(umo, str) or len(umo.split(":")) != 3:
+        if self._split_umo(umo) is None:
             raise ValueError(
                 "umop must be a string in the format [platform_id]:[message_type]:[session_id], with optional wildcards * or empty for all",
             )

astrbot/dashboard/routes/chat.py

@@ -36,6 +36,20 @@ async def track_conversation(convs: dict, conv_id: str):
         convs.pop(conv_id, None)
 
 
+async def _poll_webchat_stream_result(back_queue, username: str):
+    try:
+        result = await asyncio.wait_for(back_queue.get(), timeout=1)
+    except asyncio.TimeoutError:
+        return None, False
+    except asyncio.CancelledError:
+        logger.debug(f"[WebChat] 用户 {username} 断开聊天长连接。")
+        return None, True
+    except Exception as e:
+        logger.error(f"WebChat stream error: {e}")
+        return None, False
+    return result, False
+
+
 class ChatRoute(Route):
     def __init__(
         self,
@@ -343,16 +357,12 @@ class ChatRoute(Route):
 
                 async with track_conversation(self.running_convs, webchat_conv_id):
                     while True:
-                        try:
-                            result = await asyncio.wait_for(back_queue.get(), timeout=1)
-                        except asyncio.TimeoutError:
-                            continue
-                        except asyncio.CancelledError:
-                            logger.debug(f"[WebChat] 用户 {username} 断开聊天长连接。")
+                        result, should_break = await _poll_webchat_stream_result(
+                            back_queue, username
+                        )
+                        if should_break:
                             client_disconnected = True
-                        except Exception as e:
-                            logger.error(f"WebChat stream error: {e}")
-
+                            break
                         if not result:
                             continue
 

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",

dashboard/package.json

@@ -36,7 +36,6 @@
     "remixicon": "3.5.0",
     "shiki": "^3.20.0",
     "stream-markdown": "^0.0.13",
-    "stream-monaco": "^0.0.17",
     "vee-validate": "4.11.3",
     "vite-plugin-vuetify": "2.1.3",
     "vue": "3.3.4",

dashboard/pnpm-lock.yaml

@@ -81,9 +81,6 @@ importers:
       stream-markdown:
         specifier: ^0.0.13
         version: 0.0.13(shiki@3.22.0)
-      stream-monaco:
-        specifier: ^0.0.17
-        version: 0.0.17(monaco-editor@0.52.2)
       vee-validate:
         specifier: 4.11.3
         version: 4.11.3(vue@3.3.4)
@@ -3300,6 +3297,7 @@ snapshots:
       '@shikijs/core': 3.22.0
       '@shikijs/types': 3.22.0
       '@shikijs/vscode-textmate': 10.0.2
+    optional: true
 
   '@shikijs/themes@3.22.0':
     dependencies:
@@ -3992,7 +3990,8 @@ snapshots:
       json-schema-traverse: 1.0.0
       require-from-string: 2.0.2
 
-  alien-signals@2.0.8: {}
+  alien-signals@2.0.8:
+    optional: true
 
   ansi-regex@5.0.1: {}
 
@@ -5443,6 +5442,7 @@ snapshots:
       alien-signals: 2.0.8
       monaco-editor: 0.52.2
       shiki: 3.22.0
+    optional: true
 
   stringify-entities@4.0.4:
     dependencies:

dashboard/src/components/chat/Chat.vue

@@ -74,7 +74,7 @@
                                 :stagedImagesUrl="stagedImagesUrl"
                                 :stagedAudioUrl="stagedAudioUrl"
                                 :stagedFiles="stagedNonImageFiles"
-                                :disabled="isStreaming"
+                                :disabled="false"
                                 :is-running="isStreaming || isConvRunning"
                                 :enableStreaming="enableStreaming"
                                 :isRecording="isRecording"
@@ -106,7 +106,7 @@
                                 :stagedImagesUrl="stagedImagesUrl"
                                 :stagedAudioUrl="stagedAudioUrl"
                                 :stagedFiles="stagedNonImageFiles"
-                                :disabled="isStreaming"
+                                :disabled="false"
                                 :is-running="isStreaming || isConvRunning"
                                 :enableStreaming="enableStreaming"
                                 :isRecording="isRecording"
@@ -137,7 +137,7 @@
                             :stagedImagesUrl="stagedImagesUrl"
                             :stagedAudioUrl="stagedAudioUrl"
                             :stagedFiles="stagedNonImageFiles"
-                            :disabled="isStreaming"
+                            :disabled="false"
                             :is-running="isStreaming || isConvRunning"
                             :enableStreaming="enableStreaming"
                             :isRecording="isRecording"
@@ -348,6 +348,12 @@ function setSendShortcut(mode: SendShortcut) {
     localStorage.setItem(SEND_SHORTCUT_STORAGE_KEY, mode);
 }
 
+function focusChatInput() {
+    nextTick(() => {
+        chatInputRef.value?.focusInput?.();
+    });
+}
+
 // 检测是否为手机端
 function checkMobile() {
     isMobile.value = window.innerWidth <= 768;
@@ -505,6 +511,7 @@ async function handleSelectConversation(sessionIds: string[]) {
     nextTick(() => {
         messageList.value?.scrollToBottom();
     });
+    focusChatInput();
 }
 
 function handleNewChat() {
@@ -514,6 +521,7 @@ function handleNewChat() {
     // 退出项目视图
     selectedProjectId.value = null;
     projectSessions.value = [];
+    focusChatInput();
 }
 
 async function handleDeleteConversation(sessionId: string) {
@@ -671,6 +679,11 @@ async function handleSendMessage() {
     const selectedProviderId = selection?.providerId || '';
     const selectedModelName = selection?.modelName || '';
 
+    // 点击发送后立即将消息区滚到底部，确保用户看到最新消息
+    nextTick(() => {
+        messageList.value?.scrollToBottom();
+    });
+
     await sendMsg(
         promptToSend,
         filesToSend,
@@ -680,6 +693,11 @@ async function handleSendMessage() {
         replyToSend
     );
 
+    // 发送流程结束后再兜底一次，处理异步渲染场景
+    nextTick(() => {
+        messageList.value?.scrollToBottom();
+    });
+
     // 如果在项目中创建了新会话，将其添加到项目
     if (isCreatingNewSession && currentProjectId && currSessionId.value) {
         await addSessionToProject(currSessionId.value, currentProjectId);

dashboard/src/components/chat/ChatInput.vue

@@ -95,7 +95,7 @@
                             {{ isRecording ? tm('voice.speaking') : tm('voice.startRecording') }}
                         </v-tooltip>
                     </v-btn>
-                    <v-btn icon v-if="isRunning" @click="$emit('stop')" variant="tonal" color="primary" class="send-btn">
+                    <v-btn icon v-if="isRunning && !canSend" @click="$emit('stop')" variant="tonal" color="primary" class="send-btn">
                         <v-icon icon="mdi-stop" variant="text" plain></v-icon>
                         <v-tooltip activator="parent" location="top">
                             {{ tm('input.stopGenerating') }}
@@ -373,6 +373,11 @@ function getCurrentSelection() {
     return providerModelMenuRef.value?.getCurrentSelection();
 }
 
+function focusInput() {
+    if (!inputField.value) return;
+    inputField.value.focus();
+}
+
 onMounted(() => {
     if (inputField.value) {
         inputField.value.addEventListener('paste', handlePaste);
@@ -388,7 +393,8 @@ onBeforeUnmount(() => {
 });
 
 defineExpose({
-    getCurrentSelection
+    getCurrentSelection,
+    focusInput
 });
 </script>
 

dashboard/src/components/chat/MessageList.vue

@@ -180,7 +180,7 @@
 
 <script>
 import { useI18n, useModuleI18n } from '@/i18n/composables';
-import { enableKatex, enableMermaid, setCustomComponents } from 'markstream-vue'
+import { enableKatex, enableMermaid, MarkdownCodeBlockNode, setCustomComponents } from 'markstream-vue'
 import 'markstream-vue/index.css'
 import 'katex/dist/katex.min.css'
 import 'highlight.js/styles/github.css';
@@ -194,8 +194,11 @@ import ActionRef from './message_list_comps/ActionRef.vue';
 enableKatex();
 enableMermaid();
 
-// 注册自定义 ref 组件
-setCustomComponents('message-list', { ref: RefNode });
+// 注册 message-list 专用组件：引用节点 + Shiki 代码块渲染
+setCustomComponents('message-list', {
+    ref: RefNode,
+    code_block: MarkdownCodeBlockNode
+});
 
 export default {
     name: 'MessageList',

dashboard/src/components/chat/StandaloneChat.vue

@@ -22,7 +22,7 @@
                         v-model:prompt="prompt"
                         :stagedImagesUrl="stagedImagesUrl"
                         :stagedAudioUrl="stagedAudioUrl"
-                        :disabled="isStreaming"
+                        :disabled="false"
                         :is-running="isStreaming || isConvRunning"
                         :enableStreaming="enableStreaming"
                         :isRecording="isRecording"

dashboard/src/components/chat/message_list_comps/MessagePartsRenderer.vue

@@ -63,8 +63,9 @@
         <!-- Text (Markdown) -->
         <MarkdownRender
             v-else-if="renderPart.part.type === 'plain' && renderPart.part.text && renderPart.part.text.trim()"
+            :key="`${renderPart.key}-${isDark ? 'dark' : 'light'}`"
             custom-id="message-list" :custom-html-tags="['ref']" :content="renderPart.part.text" :typewriter="false"
-            class="markdown-content" :is-dark="isDark" :monacoOptions="{ theme: isDark ? 'vs-dark' : 'vs-light' }" />
+            class="markdown-content" :is-dark="isDark" />
 
         <!-- Image -->
         <div v-else-if="renderPart.part.type === 'image' && renderPart.part.embedded_url" class="embedded-images">

dashboard/src/components/chat/message_list_comps/ReasoningBlock.vue

@@ -9,7 +9,7 @@
             </span>
         </div>
         <div v-if="isExpanded" class="reasoning-content animate-fade-in">
-            <MarkdownRender :content="reasoning" class="reasoning-text markdown-content"
+            <MarkdownRender :key="`reasoning-${isDark ? 'dark' : 'light'}`" :content="reasoning" class="reasoning-text markdown-content"
                 :typewriter="false" :is-dark="isDark" :style="isDark ? { opacity: '0.85' } : {}" />
         </div>
     </div>

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);
   }

dashboard/src/utils/providerUtils.js

@@ -9,33 +9,33 @@
  */
 export function getProviderIcon(type) {
   const icons = {
-    'openai': 'https://registry.npmmirror.com/@lobehub/icons-static-svg/latest/files/icons/openai.svg',
-    'azure': 'https://registry.npmmirror.com/@lobehub/icons-static-svg/latest/files/icons/azure.svg',
-    'xai': 'https://registry.npmmirror.com/@lobehub/icons-static-svg/latest/files/icons/xai.svg',
-    'anthropic': 'https://registry.npmmirror.com/@lobehub/icons-static-svg/latest/files/icons/anthropic.svg',
-    'ollama': 'https://registry.npmmirror.com/@lobehub/icons-static-svg/latest/files/icons/ollama.svg',
-    'google': 'https://registry.npmmirror.com/@lobehub/icons-static-svg/latest/files/icons/gemini-color.svg',
-    'deepseek': 'https://registry.npmmirror.com/@lobehub/icons-static-svg/latest/files/icons/deepseek.svg',
-    'modelscope': 'https://registry.npmmirror.com/@lobehub/icons-static-svg/latest/files/icons/modelscope.svg',
-    'zhipu': 'https://registry.npmmirror.com/@lobehub/icons-static-svg/latest/files/icons/zhipu.svg',
-    'nvidia': 'https://registry.npmmirror.com/@lobehub/icons-static-svg/latest/files/icons/nvidia-color.svg',
-    'siliconflow': 'https://registry.npmmirror.com/@lobehub/icons-static-svg/latest/files/icons/siliconcloud.svg',
-    'moonshot': 'https://registry.npmmirror.com/@lobehub/icons-static-svg/latest/files/icons/kimi.svg',
-    'ppio': 'https://registry.npmmirror.com/@lobehub/icons-static-svg/latest/files/icons/ppio.svg',
-    'dify': 'https://registry.npmmirror.com/@lobehub/icons-static-svg/latest/files/icons/dify-color.svg',
-    "coze": "https://registry.npmmirror.com/@lobehub/icons-static-svg/1.66.0/files/icons/coze.svg",
-    'dashscope': 'https://registry.npmmirror.com/@lobehub/icons-static-svg/latest/files/icons/alibabacloud-color.svg',
+    'openai': 'https://cdn.jsdelivr.net/npm/@lobehub/icons-static-svg@latest/icons/openai.svg',
+    'azure': 'https://cdn.jsdelivr.net/npm/@lobehub/icons-static-svg@latest/icons/azure.svg',
+    'xai': 'https://cdn.jsdelivr.net/npm/@lobehub/icons-static-svg@latest/icons/xai.svg',
+    'anthropic': 'https://cdn.jsdelivr.net/npm/@lobehub/icons-static-svg@latest/icons/anthropic.svg',
+    'ollama': 'https://cdn.jsdelivr.net/npm/@lobehub/icons-static-svg@latest/icons/ollama.svg',
+    'google': 'https://cdn.jsdelivr.net/npm/@lobehub/icons-static-svg@latest/icons/gemini-color.svg',
+    'deepseek': 'https://cdn.jsdelivr.net/npm/@lobehub/icons-static-svg@latest/icons/deepseek.svg',
+    'modelscope': 'https://cdn.jsdelivr.net/npm/@lobehub/icons-static-svg@latest/icons/modelscope.svg',
+    'zhipu': 'https://cdn.jsdelivr.net/npm/@lobehub/icons-static-svg@latest/icons/zhipu.svg',
+    'nvidia': 'https://cdn.jsdelivr.net/npm/@lobehub/icons-static-svg@latest/icons/nvidia-color.svg',
+    'siliconflow': 'https://cdn.jsdelivr.net/npm/@lobehub/icons-static-svg@latest/icons/siliconcloud.svg',
+    'moonshot': 'https://cdn.jsdelivr.net/npm/@lobehub/icons-static-svg@latest/icons/kimi.svg',
+    'ppio': 'https://cdn.jsdelivr.net/npm/@lobehub/icons-static-svg@latest/icons/ppio.svg',
+    'dify': 'https://cdn.jsdelivr.net/npm/@lobehub/icons-static-svg@latest/icons/dify-color.svg',
+    "coze": "https://cdn.jsdelivr.net/npm/@lobehub/icons-static-svg@1.66.0/icons/coze.svg",
+    'dashscope': 'https://cdn.jsdelivr.net/npm/@lobehub/icons-static-svg@latest/icons/alibabacloud-color.svg',
     'deerflow': 'https://cdn.jsdelivr.net/gh/bytedance/deer-flow@main/frontend/public/images/deer.svg',
-    'fastgpt': 'https://registry.npmmirror.com/@lobehub/icons-static-svg/latest/files/icons/fastgpt-color.svg',
-    'lm_studio': 'https://registry.npmmirror.com/@lobehub/icons-static-svg/latest/files/icons/lmstudio.svg',
-    'fishaudio': 'https://registry.npmmirror.com/@lobehub/icons-static-svg/latest/files/icons/fishaudio.svg',
-    'minimax': 'https://registry.npmmirror.com/@lobehub/icons-static-svg/latest/files/icons/minimax.svg',
-    '302ai': 'https://registry.npmmirror.com/@lobehub/icons-static-svg/1.53.0/files/icons/ai302-color.svg',
-    'microsoft': 'https://registry.npmmirror.com/@lobehub/icons-static-svg/latest/files/icons/microsoft.svg',
-    'vllm': 'https://registry.npmmirror.com/@lobehub/icons-static-svg/latest/files/icons/vllm.svg',
-    'groq': 'https://registry.npmmirror.com/@lobehub/icons-static-svg/latest/files/icons/groq.svg',
-    'aihubmix': 'https://registry.npmmirror.com/@lobehub/icons-static-svg/latest/files/icons/aihubmix-color.svg',
-    'openrouter': 'https://registry.npmmirror.com/@lobehub/icons-static-svg/latest/files/icons/openrouter.svg',
+    'fastgpt': 'https://cdn.jsdelivr.net/npm/@lobehub/icons-static-svg@latest/icons/fastgpt-color.svg',
+    'lm_studio': 'https://cdn.jsdelivr.net/npm/@lobehub/icons-static-svg@latest/icons/lmstudio.svg',
+    'fishaudio': 'https://cdn.jsdelivr.net/npm/@lobehub/icons-static-svg@latest/icons/fishaudio.svg',
+    'minimax': 'https://cdn.jsdelivr.net/npm/@lobehub/icons-static-svg@latest/icons/minimax.svg',
+    '302ai': 'https://cdn.jsdelivr.net/npm/@lobehub/icons-static-svg@1.53.0/icons/ai302-color.svg',
+    'microsoft': 'https://cdn.jsdelivr.net/npm/@lobehub/icons-static-svg@latest/icons/microsoft.svg',
+    'vllm': 'https://cdn.jsdelivr.net/npm/@lobehub/icons-static-svg@latest/icons/vllm.svg',
+    'groq': 'https://cdn.jsdelivr.net/npm/@lobehub/icons-static-svg@latest/icons/groq.svg',
+    'aihubmix': 'https://cdn.jsdelivr.net/npm/@lobehub/icons-static-svg@latest/icons/aihubmix-color.svg',
+    'openrouter': 'https://cdn.jsdelivr.net/npm/@lobehub/icons-static-svg@latest/icons/openrouter.svg',
     "tokenpony": "https://tokenpony.cn/tokenpony-web/logo.png",
     "compshare": "https://compshare.cn/favicon.ico"
   };

docs/en/community.md

@@ -6,7 +6,7 @@ This documentation may not cover all features comprehensively. If you have any q
 
 ### Discord
 
-<https://discord.gg/PxgzhmxJ>
+<https://discord.gg/hAVk6tgV36>
 
 ### GitHub
 

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/community.md

@@ -21,7 +21,7 @@
 
 ### Discord
 
-https://discord.gg/PxgzhmxJ
+https://discord.gg/hAVk6tgV36
 
 ### Astrbook
 

docs/zh/deploy/astrbot/package.md

@@ -13,5 +13,5 @@
 ```bash
 uv tool install astrbot
 astrbot init # 只需要在第一次部署时执行，后续启动不需要执行
-astrbot
+astrbot run
 ```

docs/zh/deploy/astrbot/rainyun.md

@@ -41,4 +41,4 @@ AstrBot 已经上架至雨云的预装软件列表，支持**一键安装** Astr
 
 ![创建NAT端口映射规则](https://files.astrbot.app/docs/source/images/rainyun/image-2.png)
 
-然后，内网端口填写 `6185`，点击 `创建映射规则`，这样就可以通过 `http://IP:上面设置好的外网端口` 访问 AstrBot 的管理面板了。
+然后，内网端口填写 `6185`，点击 `创建映射规则`，这样就可以通过 `http://IP:上面设置好的外网端口` 访问 AstrBot 的管理面板了。如果无法打开，请点击`备用地址`，通过备用地址访问管理面板。

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`

docs/zh/what-is-astrbot.md

@@ -23,7 +23,7 @@ AstrBot 是一个开源的一站式 Agentic 个人和群聊助手，可在 QQ、
 
 - 部署 AstrBot：阅读部署指南，快速在本地机器或云服务器上部署 AstrBot。
 - 连接 IM 平台：按照说明将 AstrBot 连接到您喜欢的 IM 平台，如 Discord、Telegram、Slack 等。
-- 配置 AI 模型：AstrBot 支持各种 AI 模型。请参阅 [连接模型服务](/config/providers/start)
+- 配置 AI 模型：AstrBot 支持各种 AI 模型。请参阅 [连接模型服务](/providers/start)
 
 ## 它是如何实现的？
 

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": [

scripts/astrbot.service

@@ -1,5 +1,7 @@
+# user service
 [Unit]
 Description=AstrBot Service
+Documentation=https://github.com/AstrBotDevs/AstrBot
 After=network-online.target
 Wants=network-online.target
 
@@ -9,6 +11,9 @@ WorkingDirectory=%h/.local/share/astrbot
 ExecStart=/usr/bin/sh -c '/usr/bin/astrbot run || { /usr/bin/astrbot init && /usr/bin/astrbot run; }'
 Restart=on-failure
 RestartSec=5
+StandardOutput=journal
+StandardError=journal
+SyslogIdentifier=astrbot-%u
 Environment=PYTHONUNBUFFERED=1
 
 [Install]

scripts/generate_changelog.py

@@ -1,253 +0,0 @@
-#!/usr/bin/env python3
-"""
-Auto-generate changelog from git commits using LLM.
-Usage: python scripts/generate_changelog.py [--version VERSION]
-"""
-
-import argparse
-import os
-import re
-import subprocess
-import sys
-from pathlib import Path
-
-
-def get_latest_tag():
-    """Get the latest git tag."""
-    result = subprocess.run(
-        ["git", "describe", "--tags", "--abbrev=0"],
-        capture_output=True,
-        text=True,
-        check=True,
-    )
-    return result.stdout.strip()
-
-
-def get_commits_since_tag(tag):
-    """Get all commit messages since the specified tag."""
-    result = subprocess.run(
-        ["git", "log", f"{tag}..HEAD", "--pretty=format:%H|%s|%b"],
-        capture_output=True,
-        text=True,
-        check=True,
-    )
-    commits = []
-    for line in result.stdout.strip().split("\n"):
-        if not line:
-            continue
-        parts = line.split("|", 2)
-        if len(parts) >= 2:
-            commit_hash = parts[0]
-            subject = parts[1]
-            body = parts[2] if len(parts) > 2 else ""
-            commits.append({"hash": commit_hash[:7], "subject": subject, "body": body})
-    return commits
-
-
-def extract_issue_number(text):
-    """Extract issue number from commit message."""
-    # Match #1234 or (#1234)
-    match = re.search(r"#(\d+)", text)
-    return match.group(1) if match else None
-
-
-def call_llm_for_changelog(commits, version):
-    """Call LLM to generate changelog from commits."""
-    try:
-        # Try to use OpenAI API or other LLM providers
-        import openai
-
-        # Build prompt
-        commits_text = "\n".join([f"- {c['subject']}" for c in commits])
-
-        prompt = f"""Based on the following git commit messages, generate a changelog document in BOTH Chinese and English.
-
-Commit messages:
-{commits_text}
-
-Please organize the changes into these categories:
-- 新增 (New Features)
-- 修复 (Bug Fixes)
-- 优化 (Improvements)
-- 其他 (Others)
-
-Format requirements:
-1. Start with Chinese version under "## What's Changed"
-2. Follow with English version under "## What's Changed (EN)"
-3. Use markdown format with proper bullet points
-4. Keep descriptions concise and user-friendly
-5. If a commit mentions an issue number (#1234), include it in the format ([#1234](https://github.com/AstrBotDevs/AstrBot/issues/1234))
-
-Example format:
-## What's Changed
-
-### 新增
-- 支持某某功能 ([#1234](https://github.com/AstrBotDevs/AstrBot/issues/1234))
-
-### 修复
-- 修复某某问题
-
-## What's Changed (EN)
-
-### New Features
-- Add support for something ([#1234](https://github.com/AstrBotDevs/AstrBot/issues/1234))
-
-### Bug Fixes
-- Fix something
-"""
-
-        client = openai.OpenAI(
-            api_key=os.getenv("OPENAI_API_KEY"),
-            base_url=os.getenv("OPENAI_BASE_URL", "https://api.openai.com/v1"),
-        )
-
-        response = client.chat.completions.create(
-            model=os.getenv("OPENAI_MODEL", "gpt-4"),
-            messages=[
-                {
-                    "role": "system",
-                    "content": "You are a helpful assistant that generates well-structured changelogs.",
-                },
-                {"role": "user", "content": prompt},
-            ],
-            temperature=0.3,
-        )
-
-        return response.choices[0].message.content
-
-    except ImportError:
-        print(
-            "Warning: openai package not installed. Install it with: pip install openai"
-        )
-        return generate_simple_changelog(commits)
-    except Exception as e:
-        print(f"Warning: Failed to call LLM API: {e}")
-        print("Falling back to simple changelog generation...")
-        return generate_simple_changelog(commits)
-
-
-def generate_simple_changelog(commits):
-    """Generate a simple changelog without LLM."""
-    sections = {
-        "feat": ("新增", "New Features", []),
-        "fix": ("修复", "Bug Fixes", []),
-        "perf": ("优化", "Improvements", []),
-        "docs": ("文档", "Documentation", []),
-        "refactor": ("重构", "Refactoring", []),
-        "test": ("测试", "Tests", []),
-        "chore": ("其他", "Chore", []),
-        "other": ("其他", "Others", []),
-    }
-
-    # Categorize commits by conventional commit type
-    for commit in commits:
-        subject = commit["subject"]
-        issue_num = extract_issue_number(subject)
-        issue_link = (
-            f" ([#{issue_num}](https://github.com/AstrBotDevs/AstrBot/issues/{issue_num}))"
-            if issue_num
-            else ""
-        )
-
-        # Detect conventional commit type
-        matched = False
-        for prefix in ["feat", "fix", "perf", "docs", "refactor", "test", "chore"]:
-            if subject.lower().startswith(f"{prefix}:") or subject.lower().startswith(
-                f"{prefix}("
-            ):
-                # Remove prefix for display
-                clean_subject = re.sub(
-                    r"^[a-z]+(\([^)]+\))?:\s*", "", subject, flags=re.IGNORECASE
-                )
-                sections[prefix][2].append(f"- {clean_subject}{issue_link}")
-                matched = True
-                break
-
-        if not matched:
-            sections["other"][2].append(f"- {subject}{issue_link}")
-
-    # Build Chinese version
-    changelog_zh = "## What's Changed\n\n"
-    for section_key in ["feat", "fix", "perf", "docs", "refactor", "test", "other"]:
-        zh_title, _, items = sections[section_key]
-        if items:
-            changelog_zh += f"### {zh_title}\n\n"
-            changelog_zh += "\n".join(items) + "\n\n"
-
-    # Build English version
-    changelog_en = "## What's Changed (EN)\n\n"
-    for section_key in ["feat", "fix", "perf", "docs", "refactor", "test", "other"]:
-        _, en_title, items = sections[section_key]
-        if items:
-            changelog_en += f"### {en_title}\n\n"
-            changelog_en += "\n".join(items) + "\n\n"
-
-    return changelog_zh + changelog_en
-
-
-def main() -> None:
-    parser = argparse.ArgumentParser(description="Generate changelog from git commits")
-    parser.add_argument(
-        "--version", help="Version number for the changelog (e.g., v4.13.3)"
-    )
-    parser.add_argument(
-        "--use-llm",
-        action="store_true",
-        help="Use LLM to generate changelog (requires OpenAI API key)",
-    )
-    args = parser.parse_args()
-
-    # Get latest tag
-    try:
-        latest_tag = get_latest_tag()
-        print(f"Latest tag: {latest_tag}")
-    except subprocess.CalledProcessError:
-        print("Error: No tags found in repository")
-        sys.exit(1)
-
-    # Get commits since tag
-    commits = get_commits_since_tag(latest_tag)
-    if not commits:
-        print(f"No commits found since {latest_tag}")
-        sys.exit(0)
-
-    print(f"Found {len(commits)} commits since {latest_tag}")
-
-    # Determine version
-    if args.version:
-        version = args.version
-    else:
-        # Auto-increment patch version
-        match = re.match(r"v(\d+)\.(\d+)\.(\d+)", latest_tag)
-        if match:
-            major, minor, patch = map(int, match.groups())
-            version = f"v{major}.{minor}.{patch + 1}"
-        else:
-            print(f"Warning: Could not parse version from tag {latest_tag}")
-            version = "vX.X.X"
-
-    print(f"Generating changelog for {version}...")
-
-    # Generate changelog
-    if args.use_llm:
-        changelog_content = call_llm_for_changelog(commits, version)
-    else:
-        changelog_content = generate_simple_changelog(commits)
-
-    # Save to file
-    changelog_dir = Path(__file__).parent.parent / "changelogs"
-    changelog_dir.mkdir(exist_ok=True)
-    changelog_file = changelog_dir / f"{version}.md"
-
-    with open(changelog_file, "w", encoding="utf-8") as f:
-        f.write(changelog_content)
-
-    print(f"\n✓ Changelog generated: {changelog_file}")
-    print("\nPreview:")
-    print("=" * 80)
-    print(changelog_content)
-    print("=" * 80)
-
-
-if __name__ == "__main__":
-    main()

tests/test_chat_route.py

@@ -0,0 +1,56 @@
+import asyncio
+
+import pytest
+
+from astrbot.dashboard.routes.chat import _poll_webchat_stream_result
+
+
+class _QueueThatRaises:
+    def __init__(self, exc: BaseException):
+        self._exc = exc
+
+    async def get(self):
+        raise self._exc
+
+
+class _QueueWithResult:
+    def __init__(self, result):
+        self._result = result
+
+    async def get(self):
+        return self._result
+
+
+@pytest.mark.asyncio
+async def test_poll_webchat_stream_result_breaks_on_cancelled_error():
+    result, should_break = await _poll_webchat_stream_result(
+        _QueueThatRaises(asyncio.CancelledError()),
+        "alice",
+    )
+
+    assert result is None
+    assert should_break is True
+
+
+@pytest.mark.asyncio
+async def test_poll_webchat_stream_result_continues_on_generic_exception():
+    result, should_break = await _poll_webchat_stream_result(
+        _QueueThatRaises(RuntimeError("boom")),
+        "alice",
+    )
+
+    assert result is None
+    assert should_break is False
+
+
+@pytest.mark.asyncio
+async def test_poll_webchat_stream_result_returns_queue_payload():
+    payload = {"type": "end", "data": ""}
+
+    result, should_break = await _poll_webchat_stream_result(
+        _QueueWithResult(payload),
+        "alice",
+    )
+
+    assert result == payload
+    assert should_break is False

tests/test_openai_source.py

@@ -2,6 +2,7 @@ from types import SimpleNamespace
 
 import pytest
 
+from astrbot.core.provider.sources.groq_source import ProviderGroq
 from astrbot.core.provider.sources.openai_source import ProviderOpenAIOfficial
 
 
@@ -32,6 +33,21 @@ def _make_provider(overrides: dict | None = None) -> ProviderOpenAIOfficial:
     )
 
 
+def _make_groq_provider(overrides: dict | None = None) -> ProviderGroq:
+    provider_config = {
+        "id": "test-groq",
+        "type": "groq_chat_completion",
+        "model": "qwen/qwen3-32b",
+        "key": ["test-key"],
+    }
+    if overrides:
+        provider_config.update(overrides)
+    return ProviderGroq(
+        provider_config=provider_config,
+        provider_settings={},
+    )
+
+
 @pytest.mark.asyncio
 async def test_handle_api_error_content_moderated_removes_images():
     provider = _make_provider(
@@ -198,6 +214,57 @@ def test_extract_error_text_candidates_truncates_long_response_text():
     )
 
 
+@pytest.mark.asyncio
+async def test_openai_payload_keeps_reasoning_content_in_assistant_history():
+    provider = _make_provider()
+    try:
+        payloads = {
+            "messages": [
+                {
+                    "role": "assistant",
+                    "content": [
+                        {"type": "think", "think": "step 1"},
+                        {"type": "text", "text": "final answer"},
+                    ],
+                }
+            ]
+        }
+
+        provider._finally_convert_payload(payloads)
+
+        assistant_message = payloads["messages"][0]
+        assert assistant_message["content"] == [{"type": "text", "text": "final answer"}]
+        assert assistant_message["reasoning_content"] == "step 1"
+    finally:
+        await provider.terminate()
+
+
+@pytest.mark.asyncio
+async def test_groq_payload_drops_reasoning_content_from_assistant_history():
+    provider = _make_groq_provider()
+    try:
+        payloads = {
+            "messages": [
+                {
+                    "role": "assistant",
+                    "content": [
+                        {"type": "think", "think": "step 1"},
+                        {"type": "text", "text": "final answer"},
+                    ],
+                }
+            ]
+        }
+
+        provider._finally_convert_payload(payloads)
+
+        assistant_message = payloads["messages"][0]
+        assert assistant_message["content"] == [{"type": "text", "text": "final answer"}]
+        assert "reasoning_content" not in assistant_message
+        assert "reasoning" not in assistant_message
+    finally:
+        await provider.terminate()
+
+
 @pytest.mark.asyncio
 async def test_handle_api_error_content_moderated_without_images_raises():
     provider = _make_provider(