add

base-config.yaml

@@ -31,12 +31,33 @@ platforms:
     max_tokens: 2000
     max_words: 1000
     max_context_messages: 20
+  qwen:
+    # 国内: https://dashscope.aliyuncs.com
+    # 海外: https://dashscope-intl.aliyuncs.com
+    url: https://dashscope.aliyuncs.com
+    api_key:
+    model: qwen-plus
+    temperature: 0.7
+    top_p: 0.8
+    max_tokens: 2000
+    max_words: 1000
+    max_context_messages: 20
+    # 是否开启深度思考模式（仅 qwq 系列支持）
+    enable_thinking: false
   deepseek:
     url: https://api.deepseek.com
     api_key:
     model:
     max_words: 1000
     max_context_messages: 20
+  gemini:
+    url: https://generativelanguage.googleapis.com
+    api_key:
+    model: gemini-2.0-flash
+    temperature: 1
+    max_tokens: 2000
+    max_words: 1000
+    max_context_messages: 20
   openai:
     url: https://api.openai.com
     api_key:
@@ -52,6 +73,8 @@ platforms:
     max_words: 1000
     max_tokens: 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
   xai:
     url: https://api.x.ai
     api_key:
@@ -60,14 +83,7 @@ platforms:
     max_tokens: 1000
     max_words: 2000
     max_context_messages: 20
-  gemini:
+
-    url: https://generativelanguage.googleapis.com
-    api_key:
-    model: gemini-2.0-flash
-    temperature: 1
-    max_tokens: 2000
-    max_words: 1000
-    max_context_messages: 20
 
 # additional prompt
 additional_prompt:

maubot_llmplus/aibot.py

@@ -10,7 +10,7 @@ from mautrix.util.config import BaseProxyConfig, ConfigUpdateHelper
 from maubot_llmplus.local_paltform import Ollama, LmStudio
 from maubot_llmplus.platforms import Platform
 from maubot_llmplus.plugin import AbsExtraConfigPlugin, Config
-from maubot_llmplus.thrid_platform import OpenAi, Anthropic, XAi, Deepseek, Gemini
+from maubot_llmplus.thrid_platform import OpenAi, Anthropic, XAi, Deepseek, Gemini, Qwen
 
 
 class AiBotPlugin(AbsExtraConfigPlugin):
@@ -124,15 +124,17 @@ class AiBotPlugin(AbsExtraConfigPlugin):
             await event.mark_read()
             await self.client.set_typing(event.room_id, timeout=99999)
             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
+
             chat_completion = await platform.create_chat_completion(self, event)
             self.log.debug(
                 f"发送结果 {chat_completion.message}, {chat_completion.model}, {chat_completion.finish_reason}")
-            # ai gpt调用
-            # 关闭typing提示
             await self.client.set_typing(event.room_id, timeout=0)
-            # 打开typing提示
             if chat_completion.result:
-                # if hasattr(chat_completion.message, 'content'):
                 resp_content = chat_completion.message['content']
                 response = TextMessageEventContent(msgtype=MessageType.TEXT, body=resp_content, format=Format.HTML,
                                                    formatted_body=markdown.render(resp_content))
@@ -150,6 +152,48 @@ class AiBotPlugin(AbsExtraConfigPlugin):
 
         return None
 
+    async def _handle_streaming(self, evt: MessageEvent, platform) -> None:
+        # 发送初始占位消息
+        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'])
+
+        accumulated = ""
+        last_edit_len = 0
+        EDIT_THRESHOLD = 50  # 每积累50个字符更新一次消息
+
+        try:
+            async for chunk in platform.create_chat_completion_stream(self, evt):
+                accumulated += chunk
+                if len(accumulated) - last_edit_len >= EDIT_THRESHOLD:
+                    display = accumulated + " ▌"
+                    new_content = TextMessageEventContent(
+                        msgtype=MessageType.TEXT,
+                        body=display,
+                        format=Format.HTML,
+                        formatted_body=markdown.render(display)
+                    )
+                    new_content.set_edit(response_event_id)
+                    await self.client.send_message(evt.room_id, 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}"
+
+        # 输出最终完整内容
+        if not accumulated:
+            accumulated = "（无响应）"
+        final_content = TextMessageEventContent(
+            msgtype=MessageType.TEXT,
+            body=accumulated,
+            format=Format.HTML,
+            formatted_body=markdown.render(accumulated)
+        )
+        final_content.set_edit(response_event_id)
+        await self.client.send_message(evt.room_id, final_content)
+
     def get_ai_platform(self) -> Platform:
         use_platform = self.config.cur_platform
         if use_platform == 'openai':
@@ -162,6 +206,8 @@ class AiBotPlugin(AbsExtraConfigPlugin):
             return Deepseek(self.config, self.http)
         if use_platform == 'gemini':
             return Gemini(self.config, self.http)
+        if use_platform == 'qwen':
+            return Qwen(self.config, self.http)
         if use_platform == 'local_ai#ollama':
             return Ollama(self.config, self.http)
         if use_platform == 'local_ai#lmstudio':
@@ -300,7 +346,7 @@ class AiBotPlugin(AbsExtraConfigPlugin):
                 self.config.cur_model = self.config['platforms'][argus.split("#")[0]]['model']
                 await event.react("✅")
         # 如果是openai或者是claude
-        elif argus == 'openai' or argus == 'anthropic' or argus == 'xai' or argus == 'deepseek' or argus == 'gemini':
+        elif argus == 'openai' or argus == 'anthropic' or argus == 'xai' or argus == 'deepseek' or argus == 'gemini' or argus == 'qwen':
             if argus == self.config.cur_platform:
                 await event.reply(f"current ai platform has be {argus}")
                 pass

maubot_llmplus/platforms.py

@@ -1,7 +1,7 @@
 import json
 from collections import deque
 from datetime import datetime
-from typing import Optional, List, Generator
+from typing import Optional, List, Generator, AsyncIterator
 
 from aiohttp import ClientSession
 from maubot import Plugin
@@ -55,6 +55,12 @@ class Platform:
     async def create_chat_completion(self, plugin: AbsExtraConfigPlugin, evt: MessageEvent) -> ChatCompletion:
         raise NotImplementedError()
 
+    async def create_chat_completion_stream(self, plugin: AbsExtraConfigPlugin, evt: MessageEvent) -> AsyncIterator[str]:
+        raise NotImplementedError()
+
+    def is_streaming_enabled(self) -> bool:
+        return False
+
     async def list_models(self) -> List[str]:
         raise NotImplementedError()
 

maubot_llmplus/thrid_platform.py

@@ -129,10 +129,22 @@ class OpenAi(Platform):
 
 class Anthropic(Platform):
     max_tokens: int
+    streaming: bool
 
     def __init__(self, config: BaseProxyConfig, http: ClientSession) -> None:
         super().__init__(config, http)
         self.max_tokens = self.config['max_tokens']
+        self.streaming = self.config.get('streaming', False)
+
+    def is_streaming_enabled(self) -> bool:
+        return self.streaming
+
+    def _build_request(self, full_chat_context: list) -> tuple:
+        endpoint = f"{self.url}/v1/messages"
+        headers = {"x-api-key": self.api_key, "anthropic-version": "2023-06-01", "content-type": "application/json"}
+        req_body = {"model": self.model, "max_tokens": self.max_tokens, "system": self.system_prompt,
+                    "messages": full_chat_context}
+        return endpoint, headers, req_body
 
     async def create_chat_completion(self, plugin: AbsExtraConfigPlugin, evt: MessageEvent) -> ChatCompletion:
         full_chat_context = []
@@ -140,10 +152,7 @@ class Anthropic(Platform):
         chat_context = await maubot_llmplus.platforms.get_chat_context(system_context, plugin, self, evt)
         full_chat_context.extend(list(chat_context))
 
-        endpoint = f"{self.url}/v1/messages"
+        endpoint, headers, req_body = self._build_request(full_chat_context)
-        headers = {"x-api-key": self.api_key, "anthropic-version": "2023-06-01", "content-type": "application/json"}
-        req_body = {"model": self.model, "max_tokens": self.max_tokens, "system": self.system_prompt,
-                    "messages": 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()}")
@@ -162,7 +171,34 @@ class Anthropic(Platform):
                 finish_reason=response_json['stop_reason'],
                 model=response_json['model']
             )
-        pass
+
+    async def create_chat_completion_stream(self, plugin: AbsExtraConfigPlugin, evt: MessageEvent):
+        full_chat_context = []
+        system_context = deque()
+        chat_context = await maubot_llmplus.platforms.get_chat_context(system_context, plugin, self, evt)
+        full_chat_context.extend(list(chat_context))
+
+        endpoint, headers, req_body = self._build_request(full_chat_context)
+        req_body["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 line_bytes in response.content:
+                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") == "content_block_delta":
+                        delta = data.get("delta", {})
+                        if delta.get("type") == "text_delta":
+                            yield delta.get("text", "")
+                except json.JSONDecodeError:
+                    pass
 
     async def list_models(self) -> List[str]:
         # 调用openai接口获取模型列表
@@ -324,3 +360,79 @@ class XAi(Platform):
 
     def get_type(self) -> str:
         return "xai"
+
+
+class Qwen(Platform):
+    max_tokens: int
+    temperature: float
+    top_p: float
+    enable_thinking: bool
+
+    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.enable_thinking = self.config['enable_thinking']
+
+    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))
+
+        parameters = {
+            "result_format": "message"
+        }
+        if self.max_tokens:
+            parameters["max_tokens"] = self.max_tokens
+        if self.temperature is not None:
+            parameters["temperature"] = self.temperature
+        if self.top_p is not None:
+            parameters["top_p"] = self.top_p
+        if self.enable_thinking:
+            parameters["enable_thinking"] = True
+
+        request_body = {
+            "model": self.model,
+            "input": {
+                "messages": full_context
+            },
+            "parameters": parameters
+        }
+
+        endpoint = f"{self.url}/api/v1/services/aigc/text-generation/generation"
+        headers = {
+            "Content-Type": "application/json",
+            "Authorization": f"Bearer {self.api_key}"
+        }
+
+        async with self.http.post(endpoint, headers=headers, data=json.dumps(request_body)) as response:
+            if response.status != 200:
+                return ChatCompletion(
+                    result=False,
+                    message={},
+                    finish_reason=f"Error: {await response.text()}",
+                    model=None
+                )
+            response_json = await response.json()
+            choice = response_json["output"]["choices"][0]
+            return ChatCompletion(
+                result=True,
+                message=choice["message"],
+                finish_reason=choice.get("finish_reason", "stop"),
+                model=response_json.get("model", self.model)
+            )
+
+    async def list_models(self) -> List[str]:
+        models = [
+            "qwen-max", "qwen-max-latest",
+            "qwen-plus", "qwen-plus-latest",
+            "qwen-turbo", "qwen-turbo-latest",
+            "qwen-long",
+            "qwen3-235b-a22b", "qwen3-30b-a3b",
+            "qwq-plus", "qwq-plus-latest",
+        ]
+        return [f"- {m}" for m in models]
+
+    def get_type(self) -> str:
+        return "qwen"