修复错误代码

base-config.yaml

@@ -31,6 +31,8 @@ platforms:
     max_tokens: 2000
     max_words: 1000
     max_context_messages: 20
+    # 是否开启流式输出（开启后 Element 中消息会逐步更新）
+    streaming: false
   qwen:
     # 国内: https://dashscope.aliyuncs.com
     # 海外: https://dashscope-intl.aliyuncs.com
@@ -44,12 +46,16 @@ platforms:
     max_context_messages: 20
     # 是否开启深度思考模式（仅 qwq 系列支持）
     enable_thinking: false
+    # 是否开启流式输出（开启后 Element 中消息会逐步更新）
+    streaming: false
   deepseek:
     url: https://api.deepseek.com
     api_key:
     model:
     max_words: 1000
     max_context_messages: 20
+    # 是否开启流式输出（开启后 Element 中消息会逐步更新）
+    streaming: false
   gemini:
     url: https://generativelanguage.googleapis.com
     api_key:
@@ -58,6 +64,8 @@ platforms:
     max_tokens: 2000
     max_words: 1000
     max_context_messages: 20
+    # 是否开启流式输出（开启后 Element 中消息会逐步更新）
+    streaming: false
   openai:
     url: https://api.openai.com
     api_key:
@@ -66,6 +74,8 @@ platforms:
     max_words: 1000
     max_context_messages: 20
     temperature: 1
+    # 是否开启流式输出（开启后 Element 中消息会逐步更新）
+    streaming: false
   anthropic:
     url: https://api.anthropic.com
     api_key:
@@ -83,6 +93,8 @@ platforms:
     max_tokens: 1000
     max_words: 2000
     max_context_messages: 20
+    # 是否开启流式输出（开启后 Element 中消息会逐步更新）
+    streaming: false
 
 
 # additional prompt

maubot_llmplus/aibot.py

@@ -1,3 +1,4 @@
+import asyncio
 import re
 
 from typing import Type
@@ -126,7 +127,6 @@ class AiBotPlugin(AbsExtraConfigPlugin):
             platform = self.get_ai_platform()
 
             if platform.is_streaming_enabled():
-                await self.client.set_typing(event.room_id, timeout=0)
                 await self._handle_streaming(event, platform)
                 return
 
@@ -153,18 +153,37 @@ class AiBotPlugin(AbsExtraConfigPlugin):
         return None
 
     async def _handle_streaming(self, evt: MessageEvent, platform) -> None:
-        # 发送初始占位消息
+        # 发送初始占位消息；on_message 已设 typing=on，等收到第一个 chunk 再关掉
         placeholder = TextMessageEventContent(
             msgtype=MessageType.TEXT, body="▌", format=Format.HTML, formatted_body="▌"
         )
         response_event_id = await evt.respond(placeholder, in_thread=self.config['reply_in_thread'])
+        self.log.debug("Streaming: placeholder sent")
 
         accumulated = ""
         last_edit_len = 0
-        EDIT_THRESHOLD = 50  # 每积累50个字符更新一次消息
+        first_chunk = True
+        EDIT_THRESHOLD = 100
+
+        async def send_edit(content: TextMessageEventContent) -> None:
+            # shield 防止 wait_for 超时时 cancel send_task，保护 mautrix 内部锁不残留
+            send_task = asyncio.ensure_future(self.client.send_message(evt.room_id, content))
+            try:
+                await asyncio.wait_for(asyncio.shield(send_task), timeout=8.0)
+            except asyncio.TimeoutError:
+                self.log.debug("Streaming: edit timed out, waiting naturally")
+                await send_task
+            except Exception as e:
+                self.log.warning(f"Streaming: edit error: {e}")
+                if not send_task.done():
+                    await send_task
 
         try:
             async for chunk in platform.create_chat_completion_stream(self, evt):
+                if first_chunk:
+                    # 收到第一个 chunk 才关掉 typing，等待期间用户可见 typing 指示器
+                    await self.client.set_typing(evt.room_id, timeout=0)
+                    first_chunk = False
                 accumulated += chunk
                 if len(accumulated) - last_edit_len >= EDIT_THRESHOLD:
                     display = accumulated + " ▌"
@@ -175,14 +194,19 @@ class AiBotPlugin(AbsExtraConfigPlugin):
                         formatted_body=markdown.render(display)
                     )
                     new_content.set_edit(response_event_id)
-                    await self.client.send_message(evt.room_id, new_content)
+                    self.log.debug(f"Streaming: mid-edit, accumulated={len(accumulated)}")
+                    await send_edit(new_content)
                     last_edit_len = len(accumulated)
         except Exception as e:
             self.log.exception(f"Streaming error: {e}")
             if not accumulated:
                 accumulated = f"Streaming error: {e}"
+        finally:
+            if first_chunk:
+                await self.client.set_typing(evt.room_id, timeout=0)
+
+        self.log.debug(f"Streaming: loop done, total={len(accumulated)}")
 
-        # 输出最终完整内容
         if not accumulated:
             accumulated = "（无响应）"
         final_content = TextMessageEventContent(
@@ -192,7 +216,9 @@ class AiBotPlugin(AbsExtraConfigPlugin):
             formatted_body=markdown.render(accumulated)
         )
         final_content.set_edit(response_event_id)
-        await self.client.send_message(evt.room_id, final_content)
+        self.log.debug("Streaming: sending final edit")
+        await send_edit(final_content)
+        self.log.debug("Streaming: final edit done")
 
     def get_ai_platform(self) -> Platform:
         use_platform = self.config.cur_platform

maubot_llmplus/local_paltform.py

@@ -1,3 +1,4 @@
+import asyncio
 import json
 
 from typing import List
@@ -11,12 +12,17 @@ import maubot_llmplus
 import maubot_llmplus.platforms
 from maubot_llmplus.platforms import Platform, ChatCompletion
 from maubot_llmplus.plugin import AbsExtraConfigPlugin
+from maubot_llmplus.thrid_platform import _read_openai_sse
 
 
 class Ollama(Platform):
 
     def __init__(self, config: BaseProxyConfig, http: ClientSession) -> None:
         super().__init__(config, http)
+        self.streaming = self.config.get('streaming', False)
+
+    def is_streaming_enabled(self) -> bool:
+        return self.streaming
 
     async def create_chat_completion(self, plugin: AbsExtraConfigPlugin, evt: MessageEvent) -> ChatCompletion:
         full_context = []
@@ -27,20 +33,52 @@ class Ollama(Platform):
         req_body = {'model': self.model, 'messages': full_context, 'stream': False}
         headers = {'Content-Type': 'application/json'}
         async with self.http.post(endpoint, headers=headers, json=req_body) as response:
-            # plugin.log.debug(f"响应内容：{response.status}, {await response.json()}")
             if response.status != 200:
                 return ChatCompletion(
+                    result=False,
                     message={},
                     finish_reason=f"http status {response.status}",
                     model=None
                 )
             response_json = await response.json()
             return ChatCompletion(
+                result=True,
                 message=response_json['message'],
                 finish_reason='success',
                 model=response_json['model']
             )
 
+    async def create_chat_completion_stream(self, plugin: AbsExtraConfigPlugin, evt: MessageEvent):
+        full_context = []
+        context = await maubot_llmplus.platforms.get_context(plugin, self, evt)
+        full_context.extend(list(context))
+
+        endpoint = f"{self.url}/api/chat"
+        req_body = {'model': self.model, 'messages': full_context, 'stream': True}
+        headers = {'Content-Type': 'application/json'}
+        async with self.http.post(endpoint, headers=headers, json=req_body) as response:
+            if response.status != 200:
+                raise ValueError(f"Error: http status {response.status}")
+            while True:
+                try:
+                    line_bytes = await asyncio.wait_for(response.content.readline(), timeout=60.0)
+                except asyncio.TimeoutError:
+                    break
+                if not line_bytes:
+                    break
+                line = line_bytes.decode("utf-8").strip()
+                if not line:
+                    continue
+                try:
+                    data = json.loads(line)
+                    if data.get("done"):
+                        break
+                    content = data.get("message", {}).get("content", "")
+                    if content:
+                        yield content
+                except json.JSONDecodeError:
+                    pass
+
     async def list_models(self) -> List[str]:
         full_url = f"{self.url}/api/tags"
         async with self.http.get(full_url) as response:
@@ -53,13 +91,16 @@ class Ollama(Platform):
         return "local_ai"
 
 
-class LmStudio(Platform) :
+class LmStudio(Platform):
     temperature: int
 
     def __init__(self, config: BaseProxyConfig, http: ClientSession) -> None:
         super().__init__(config, http)
-        self.temperature = self.config['temperature']
-        pass
+        self.temperature = float(self.config['temperature']) if self.config.get('temperature') is not None else None
+        self.streaming = self.config.get('streaming', False)
+
+    def is_streaming_enabled(self) -> bool:
+        return self.streaming
 
     async def create_chat_completion(self, plugin: AbsExtraConfigPlugin, evt: MessageEvent) -> ChatCompletion:
         full_context = []
@@ -72,9 +113,9 @@ class LmStudio(Platform) :
         async with self.http.post(
                 endpoint, headers=headers, data=json.dumps(req_body)
         ) as response:
-            # plugin.log.debug(f"响应内容：{response.status}, {await response.json()}")
             if response.status != 200:
                 return ChatCompletion(
+                    result=False,
                     message={},
                     finish_reason=f"Error: {await response.text()}",
                     model=None
@@ -82,11 +123,26 @@ class LmStudio(Platform) :
             response_json = await response.json()
             choice = response_json["choices"][0]
             return ChatCompletion(
+                result=True,
                 message=choice["message"],
                 finish_reason=choice["finish_reason"],
                 model=choice.get("model", None)
             )
 
+    async def create_chat_completion_stream(self, plugin: AbsExtraConfigPlugin, evt: MessageEvent):
+        full_context = []
+        context = await maubot_llmplus.platforms.get_context(plugin, self, evt)
+        full_context.extend(list(context))
+
+        endpoint = f"{self.url}/v1/chat/completions"
+        headers = {"content-type": "application/json"}
+        req_body = {"model": self.model, "messages": full_context, "temperature": self.temperature, "stream": True}
+        async with self.http.post(endpoint, headers=headers, data=json.dumps(req_body)) as response:
+            if response.status != 200:
+                raise ValueError(f"Error: {await response.text()}")
+            async for chunk in _read_openai_sse(response):
+                yield chunk
+
     async def list_models(self) -> List[str]:
         full_url = f"{self.url}/v1/models"
         async with self.http.get(full_url) as response:

maubot_llmplus/thrid_platform.py

@@ -1,4 +1,6 @@
+import asyncio
 import json
+import logging
 from collections import deque
 
 from typing import List
@@ -12,10 +14,41 @@ from maubot_llmplus.platforms import Platform, ChatCompletion
 from maubot_llmplus.plugin import AbsExtraConfigPlugin
 
 
+async def _read_openai_sse(response):
+    """读取 OpenAI 兼容格式的 SSE 流，yield 每个 delta content"""
+    while True:
+        try:
+            line_bytes = await asyncio.wait_for(response.content.readline(), timeout=60.0)
+        except asyncio.TimeoutError:
+            break
+        if not line_bytes:
+            break
+        line = line_bytes.decode("utf-8").strip()
+        if not line.startswith("data: "):
+            continue
+        data_str = line[6:]
+        if data_str == "[DONE]":
+            break
+        try:
+            data = json.loads(data_str)
+            choices = data.get("choices", [])
+            if choices:
+                content = choices[0].get("delta", {}).get("content", "")
+                if content:
+                    yield content
+        except json.JSONDecodeError:
+            pass
+
+
+
 class Deepseek(Platform):
 
     def __init__(self, config: BaseProxyConfig, http: ClientSession):
         super().__init__(config, http)
+        self.streaming = self.config.get('streaming', False)
+
+    def is_streaming_enabled(self) -> bool:
+        return self.streaming
 
     async def create_chat_completion(self, plugin: AbsExtraConfigPlugin, evt: MessageEvent) -> ChatCompletion:
         full_context = []
@@ -51,6 +84,28 @@ class Deepseek(Platform):
                 model=response_json.get("model", None)
             )
 
+    async def create_chat_completion_stream(self, plugin: AbsExtraConfigPlugin, evt: MessageEvent):
+        full_context = []
+        context = await maubot_llmplus.platforms.get_context(plugin, self, evt)
+        full_context.extend(list(context))
+
+        headers = {
+            "Content-Type": "application/json",
+            "Authorization": f"Bearer {self.api_key}"
+        }
+        data = {
+            "model": self.model,
+            "messages": full_context,
+            "stream": True,
+        }
+
+        endpoint = f"{self.url}/chat/completions"
+        async with self.http.post(endpoint, headers=headers, data=json.dumps(data)) as response:
+            if response.status != 200:
+                raise ValueError(f"Error: {await response.text()}")
+            async for chunk in _read_openai_sse(response):
+                yield chunk
+
     async def list_models(self) -> List[str]:
         models = ["deepseek-chat", "deepseek-reasoner"]
         return [f"- {m}" for m in models]
@@ -60,12 +115,16 @@ class Deepseek(Platform):
 
 class OpenAi(Platform):
     max_tokens: int
-    temperature: int
+    temperature: float
 
     def __init__(self, config: BaseProxyConfig, http: ClientSession) -> None:
         super().__init__(config, http)
-        self.max_tokens = self.config['max_tokens']
-        self.temperature = self.config['temperature']
+        self.max_tokens = int(self.config['max_tokens']) if self.config.get('max_tokens') else None
+        self.temperature = float(self.config['temperature']) if self.config.get('temperature') is not None else None
+        self.streaming = self.config.get('streaming', False)
+
+    def is_streaming_enabled(self) -> bool:
+        return self.streaming
 
     async def create_chat_completion(self, plugin: AbsExtraConfigPlugin, evt: MessageEvent) -> ChatCompletion:
         full_context = []
@@ -96,7 +155,6 @@ class OpenAi(Platform):
         async with self.http.post(
                 endpoint, headers=headers, data=json.dumps(data)
         ) as response:
-            # plugin.log.debug(f"响应内容：{response.status}, {await response.json()}")
             if response.status != 200:
                 return ChatCompletion(
                     result=False,
@@ -113,6 +171,37 @@ class OpenAi(Platform):
                 model=choice.get("model", None)
             )
 
+    async def create_chat_completion_stream(self, plugin: AbsExtraConfigPlugin, evt: MessageEvent):
+        full_context = []
+        context = await maubot_llmplus.platforms.get_context(plugin, self, evt)
+        full_context.extend(list(context))
+
+        headers = {
+            "Content-Type": "application/json",
+            "Authorization": f"Bearer {self.api_key}"
+        }
+        data = {
+            "model": self.model,
+            "messages": full_context,
+            "stream": True,
+        }
+
+        if 'max_tokens' in self.config and self.max_tokens:
+            if 'gpt-5' in self.model:
+                data["max_completion_tokens"] = self.max_tokens
+            else:
+                data["max_tokens"] = self.max_tokens
+
+        if 'temperature' in self.config and self.temperature:
+            data["temperature"] = self.temperature
+
+        endpoint = f"{self.url}/v1/chat/completions"
+        async with self.http.post(endpoint, headers=headers, data=json.dumps(data)) as response:
+            if response.status != 200:
+                raise ValueError(f"Error: {await response.text()}")
+            async for chunk in _read_openai_sse(response):
+                yield chunk
+
     async def list_models(self) -> List[str]:
         # 调用openai接口获取模型列表
         full_url = f"{self.url}/v1/models"
@@ -133,7 +222,7 @@ class Anthropic(Platform):
 
     def __init__(self, config: BaseProxyConfig, http: ClientSession) -> None:
         super().__init__(config, http)
-        self.max_tokens = self.config['max_tokens']
+        self.max_tokens = int(self.config['max_tokens']) if self.config.get('max_tokens') else None
         self.streaming = self.config.get('streaming', False)
 
     def is_streaming_enabled(self) -> bool:
@@ -155,7 +244,6 @@ class Anthropic(Platform):
         endpoint, headers, req_body = self._build_request(full_chat_context)
 
         async with self.http.post(endpoint, headers=headers, data=json.dumps(req_body)) as response:
-            # plugin.log.debug(f"响应内容：{response.status}, {await response.json()}")
             if response.status != 200:
                 return ChatCompletion(
                     result=False,
@@ -184,15 +272,21 @@ class Anthropic(Platform):
         async with self.http.post(endpoint, headers=headers, data=json.dumps(req_body)) as response:
             if response.status != 200:
                 raise ValueError(f"Error: {await response.text()}")
-            async for line_bytes in response.content:
+            while True:
+                try:
+                    line_bytes = await asyncio.wait_for(response.content.readline(), timeout=60.0)
+                except asyncio.TimeoutError:
+                    break
+                if not line_bytes:
+                    break
                 line = line_bytes.decode("utf-8").strip()
                 if not line.startswith("data: "):
                     continue
                 data_str = line[6:]
-                if data_str == "[DONE]":
-                    break
                 try:
                     data = json.loads(data_str)
+                    if data.get("type") == "message_stop":
+                        break
                     if data.get("type") == "content_block_delta":
                         delta = data.get("delta", {})
                         if delta.get("type") == "text_delta":
@@ -201,7 +295,6 @@ class Anthropic(Platform):
                     pass
 
     async def list_models(self) -> List[str]:
-        # 调用openai接口获取模型列表
         full_url = f"{self.url}/v1/models"
         headers = {
             'anthropic-version': "2023-06-01",
@@ -223,12 +316,14 @@ class Gemini(Platform):
 
     def __init__(self, config: BaseProxyConfig, http: ClientSession) -> None:
         super().__init__(config, http)
-        self.max_tokens = self.config['max_tokens']
-        self.temperature = self.config['temperature']
+        self.max_tokens = int(self.config['max_tokens']) if self.config.get('max_tokens') else None
+        self.temperature = float(self.config['temperature']) if self.config.get('temperature') is not None else None
+        self.streaming = self.config.get('streaming', False)
 
-    async def create_chat_completion(self, plugin: AbsExtraConfigPlugin, evt: MessageEvent) -> ChatCompletion:
-        context = await maubot_llmplus.platforms.get_context(plugin, self, evt)
+    def is_streaming_enabled(self) -> bool:
+        return self.streaming
 
+    def _build_gemini_request(self, context) -> tuple:
         system_parts = []
         contents = []
         for msg in context:
@@ -255,12 +350,17 @@ class Gemini(Platform):
         if self.temperature:
             request_body["generationConfig"]["temperature"] = self.temperature
 
-        endpoint = f"{self.url}/v1beta/models/{self.model}:generateContent"
         headers = {
             "Content-Type": "application/json",
             "x-goog-api-key": self.api_key
         }
+        return request_body, headers
 
+    async def create_chat_completion(self, plugin: AbsExtraConfigPlugin, evt: MessageEvent) -> ChatCompletion:
+        context = await maubot_llmplus.platforms.get_context(plugin, self, evt)
+        request_body, headers = self._build_gemini_request(context)
+
+        endpoint = f"{self.url}/v1beta/models/{self.model}:generateContent"
         async with self.http.post(endpoint, headers=headers, data=json.dumps(request_body)) as response:
             if response.status != 200:
                 return ChatCompletion(
@@ -279,6 +379,47 @@ class Gemini(Platform):
                 model=response_json.get("modelVersion", self.model)
             )
 
+    async def create_chat_completion_stream(self, plugin: AbsExtraConfigPlugin, evt: MessageEvent):
+        context = await maubot_llmplus.platforms.get_context(plugin, self, evt)
+        request_body, headers = self._build_gemini_request(context)
+
+        endpoint = f"{self.url}/v1beta/models/{self.model}:streamGenerateContent?alt=sse"
+        async with self.http.post(endpoint, headers=headers, data=json.dumps(request_body)) as response:
+            if response.status != 200:
+                raise ValueError(f"Error: {await response.text()}")
+            # 与 Anthropic 保持一致：内联 while 循环，避免双层异步生成器代理
+            while True:
+                try:
+                    line_bytes = await asyncio.wait_for(response.content.readline(), timeout=60.0)
+                except asyncio.TimeoutError:
+                    break
+                if not line_bytes:
+                    break
+                line = line_bytes.decode("utf-8").strip()
+                if not line.startswith("data: "):
+                    continue
+                data_str = line[6:]
+                try:
+                    data = json.loads(data_str)
+                    candidates = data.get("candidates", [])
+                    if not candidates:
+                        continue
+                    candidate = candidates[0]
+                    # 先 yield 文本，再判断是否结束（对齐 OpenAI [DONE] 逻辑）
+                    parts = candidate.get("content", {}).get("parts", [])
+                    for part in parts:
+                        text = part.get("text", "")
+                        if text:
+                            yield text
+                    finish_reason = candidate.get("finishReason")
+                    if finish_reason:
+                        logging.getLogger("instance/aibot").debug(
+                            f"Gemini stream finished: finishReason={finish_reason}"
+                        )
+                        break
+                except json.JSONDecodeError:
+                    pass
+
     async def list_models(self) -> List[str]:
         full_url = f"{self.url}/v1beta/models"
         headers = {"x-goog-api-key": self.api_key}
@@ -302,8 +443,12 @@ class XAi(Platform):
 
     def __init__(self, config: BaseProxyConfig, http: ClientSession) -> None:
         super().__init__(config, http)
-        self.temperature = self.config['temperature']
-        self.max_tokens = self.config['max_tokens']
+        self.temperature = float(self.config['temperature']) if self.config.get('temperature') is not None else None
+        self.max_tokens = int(self.config['max_tokens']) if self.config.get('max_tokens') else None
+        self.streaming = self.config.get('streaming', False)
+
+    def is_streaming_enabled(self) -> bool:
+        return self.streaming
 
     async def create_chat_completion(self, plugin: AbsExtraConfigPlugin, evt: MessageEvent) -> ChatCompletion:
         full_context = []
@@ -328,7 +473,6 @@ class XAi(Platform):
 
         endpoint = f"{self.url}/v1/chat/completions"
         async with self.http.post(url=endpoint, data=json.dumps(request_body), headers=headers) as response:
-            # plugin.log.debug(f"响应内容：{response.status}, {await response.json()}")
             if response.status != 200:
                 return ChatCompletion(
                     result=False,
@@ -345,10 +489,35 @@ class XAi(Platform):
                 model=response_json["model"]
             )
 
-        pass
+    async def create_chat_completion_stream(self, plugin: AbsExtraConfigPlugin, evt: MessageEvent):
+        full_context = []
+        context = await maubot_llmplus.platforms.get_context(plugin, self, evt)
+        full_context.extend(list(context))
+
+        headers = {
+            "Content-Type": "application/json",
+            "Authorization": f"Bearer {self.api_key}"
+        }
+        request_body = {
+            "messages": full_context,
+            "model": self.model,
+            "stream": True,
+        }
+
+        if 'max_tokens' in self.config and self.max_tokens:
+            request_body["max_tokens"] = self.max_tokens
+
+        if 'temperature' in self.config and self.temperature:
+            request_body["temperature"] = self.temperature
+
+        endpoint = f"{self.url}/v1/chat/completions"
+        async with self.http.post(url=endpoint, data=json.dumps(request_body), headers=headers) as response:
+            if response.status != 200:
+                raise ValueError(f"Error: {await response.text()}")
+            async for chunk in _read_openai_sse(response):
+                yield chunk
 
     async def list_models(self) -> List[str]:
-        # 调用openai接口获取模型列表
         full_url = f"{self.url}/v1/models"
         headers = {'Content-Type': 'application/json', 'Authorization': f"Bearer {self.api_key}"}
         async with self.http.get(full_url, headers=headers) as response:
@@ -356,7 +525,6 @@ class XAi(Platform):
                 return []
             response_data = await response.json()
             return [f"- {m['id']}" for m in response_data["data"]]
-        pass
 
     def get_type(self) -> str:
         return "xai"
@@ -370,16 +538,16 @@ class Qwen(Platform):
 
     def __init__(self, config: BaseProxyConfig, http: ClientSession) -> None:
         super().__init__(config, http)
-        self.max_tokens = self.config['max_tokens']
-        self.temperature = self.config['temperature']
-        self.top_p = self.config['top_p']
+        self.max_tokens = int(self.config['max_tokens']) if self.config.get('max_tokens') else None
+        self.temperature = float(self.config['temperature']) if self.config.get('temperature') is not None else None
+        self.top_p = float(self.config['top_p']) if self.config.get('top_p') is not None else None
         self.enable_thinking = self.config['enable_thinking']
+        self.streaming = self.config.get('streaming', False)
 
-    async def create_chat_completion(self, plugin: AbsExtraConfigPlugin, evt: MessageEvent) -> ChatCompletion:
-        full_context = []
-        context = await maubot_llmplus.platforms.get_context(plugin, self, evt)
-        full_context.extend(list(context))
+    def is_streaming_enabled(self) -> bool:
+        return self.streaming
 
+    def _build_qwen_request(self, full_context: list) -> tuple:
         parameters = {
             "result_format": "message"
         }
@@ -405,6 +573,14 @@ class Qwen(Platform):
             "Content-Type": "application/json",
             "Authorization": f"Bearer {self.api_key}"
         }
+        return endpoint, headers, request_body
+
+    async def create_chat_completion(self, plugin: AbsExtraConfigPlugin, evt: MessageEvent) -> ChatCompletion:
+        full_context = []
+        context = await maubot_llmplus.platforms.get_context(plugin, self, evt)
+        full_context.extend(list(context))
+
+        endpoint, headers, request_body = self._build_qwen_request(full_context)
 
         async with self.http.post(endpoint, headers=headers, data=json.dumps(request_body)) as response:
             if response.status != 200:
@@ -423,6 +599,40 @@ class Qwen(Platform):
                 model=response_json.get("model", self.model)
             )
 
+    async def create_chat_completion_stream(self, plugin: AbsExtraConfigPlugin, evt: MessageEvent):
+        full_context = []
+        context = await maubot_llmplus.platforms.get_context(plugin, self, evt)
+        full_context.extend(list(context))
+
+        endpoint, headers, request_body = self._build_qwen_request(full_context)
+        # DashScope SSE 流式：增加 header 和 incremental_output 参数（每次只返回增量）
+        headers["X-DashScope-SSE"] = "enable"
+        request_body["parameters"]["incremental_output"] = True
+
+        async with self.http.post(endpoint, headers=headers, data=json.dumps(request_body)) as response:
+            if response.status != 200:
+                raise ValueError(f"Error: {await response.text()}")
+            while True:
+                try:
+                    line_bytes = await asyncio.wait_for(response.content.readline(), timeout=60.0)
+                except asyncio.TimeoutError:
+                    break
+                if not line_bytes:
+                    break
+                line = line_bytes.decode("utf-8").strip()
+                if not line.startswith("data:"):
+                    continue
+                data_str = line[5:].strip()
+                try:
+                    data = json.loads(data_str)
+                    choices = data.get("output", {}).get("choices", [])
+                    if choices:
+                        content = choices[0].get("message", {}).get("content", "")
+                        if content:
+                            yield content
+                except json.JSONDecodeError:
+                    pass
+
     async def list_models(self) -> List[str]:
         models = [
             "qwen-max", "qwen-max-latest",