perf:del dify stream in ai.yaml config.and enbale stream in lark.yaml.

fix:localagent remove_think bug

pkg/platform/sources/lark.py

@@ -407,7 +407,7 @@ class LarkAdapter(adapter.MessagePlatformAdapter):
         
         async def is_stream_output_supported() -> bool:
             is_stream = False
-            if self.config.get("enable-card-reply",None):
+            if self.config.get("enable-stream-reply",None):
                 is_stream = True
             self.is_stream = is_stream
 
@@ -603,8 +603,8 @@ class LarkAdapter(adapter.MessagePlatformAdapter):
         """
         lark_message = await self.message_converter.yiri2target(message, self.api_client)
 
-        if not is_final:
-            self.seq += 1
+
+        self.seq += 1
 
 
 
@@ -620,7 +620,7 @@ class LarkAdapter(adapter.MessagePlatformAdapter):
             'type': 'card_json',
             'data': {'card_id': self.card_id_dict[message_id], 'elements': {'content': text_message}},
         }
-        print(self.seq)
+
         request: ContentCardElementRequest = ContentCardElementRequest.builder() \
             .card_id(self.card_id_dict[message_id]) \
             .element_id("markdown_1") \

pkg/platform/sources/lark.yaml

@@ -65,23 +65,16 @@ spec:
       type: string
       required: true
       default: ""
-    - name: enable-card-reply
+    - name: enable-stream-reply
       label:
-        en_US: Enable Card Reply Mode
-        zh_Hans: 启用飞书卡片回复模式
+        en_US: Enable Stream Reply Mode
+        zh_Hans: 启用飞书流式回复模式
       description:
-        en_US: If enabled, the bot will use the card of lark reply mode
-        zh_Hans: 如果启用，将使用飞书卡片方式来回复内容
+        en_US: If enabled, the bot will use the stream of lark reply mode
+        zh_Hans: 如果启用，将使用飞书流式方式来回复内容
       type: boolean
       required: true
       default: false
-    - name: card_template_id
-      label:
-        en_US: card template id
-        zh_Hans: 卡片模板ID
-      type: string
-      required: true
-      default: "填写你的卡片template_id"
 execution:
   python:
     path: ./lark.py

pkg/provider/modelmgr/requesters/chatcmpl.py

@@ -64,7 +64,6 @@ class OpenAIChatCompletions(requester.LLMAPIRequester):
         reasoning_content = chatcmpl_message['reasoning_content'] if 'reasoning_content' in chatcmpl_message else None
 
         # deepseek的reasoner模型
-        print(pipeline_config['trigger'].get('misc', '').get('remove_think'))
         if pipeline_config['trigger'].get('misc', '').get('remove_think'):
             pass
         else:
@@ -79,6 +78,7 @@ class OpenAIChatCompletions(requester.LLMAPIRequester):
         self,
         pipeline_config: dict[str, typing.Any],
         chat_completion: chat_completion.ChatCompletion,
+        idx: int,
     ) -> llm_entities.MessageChunk:
 
         # 处理流式chunk和完整响应的差异
@@ -106,7 +106,7 @@ class OpenAIChatCompletions(requester.LLMAPIRequester):
             else:
                 delta['content'] = delta['content']
         else:
-            if reasoning_content is not None:
+            if reasoning_content is not None and idx == 0:
                 delta['content']  += f'<think>\n{reasoning_content}'
             elif reasoning_content is None:
                 if self.is_content:
@@ -165,11 +165,10 @@ class OpenAIChatCompletions(requester.LLMAPIRequester):
             pipeline_config = query.pipeline_config
             async for chunk in self._req_stream(args, extra_body=extra_args):
                 # 处理流式消息
-                delta_message = await self._make_msg_chunk(pipeline_config,chunk)
+                delta_message = await self._make_msg_chunk(pipeline_config,chunk,chunk_idx)
                 if delta_message.content:
                     current_content += delta_message.content
                     delta_message.content = current_content
-                    print(current_content)
                     # delta_message.all_content = current_content
                 if delta_message.tool_calls:
                     for tool_call in delta_message.tool_calls:
@@ -324,7 +323,6 @@ class OpenAIChatCompletions(requester.LLMAPIRequester):
                 extra_args=extra_args,
             ):
                 yield item
-                print(item)
 
         except asyncio.TimeoutError:
             raise errors.RequesterError('请求超时')

pkg/provider/runners/difysvapi.py

@@ -115,9 +115,7 @@ class DifyServiceAPIRunner(runner.RequestRunner):
 
         stream_output_pending_chunk = ''
 
-        batch_pending_max_size = self.pipeline_config['ai']['dify-service-api'].get(
-            'output-batch-size', 0
-        )  # 积累一定量的消息更新消息一次
+        batch_pending_max_size = 64  # 积累一定量的消息更新消息一次
 
         batch_pending_index = 0
 
@@ -255,14 +253,13 @@ class DifyServiceAPIRunner(runner.RequestRunner):
 
             if chunk['event'] == 'agent_message' or chunk['event'] == 'message_end':
                 if chunk['event'] == 'message_end':
-                    print(chunk['event'])
                     # break
                     is_final = True
                 else:
                     is_final = False
                     pending_agent_message += chunk['answer']
                 if is_stream:
-                    if batch_pending_index % 64 == 0 or is_final:
+                    if batch_pending_index % 32 == 0 or is_final:
                         yield llm_entities.MessageChunk(
                             role='assistant',
                             content=self._try_convert_thinking(pending_agent_message),
@@ -276,7 +273,7 @@ class DifyServiceAPIRunner(runner.RequestRunner):
                         role='assistant',
                         content=self._try_convert_thinking(pending_agent_message),
                     )
-                pending_agent_message = ''
+
 
                 if chunk['event'] == 'agent_thought':
                     if chunk['tool'] != '' and chunk['observation'] != '':  # 工具调用结果，跳过
@@ -312,7 +309,7 @@ class DifyServiceAPIRunner(runner.RequestRunner):
                                 ],
                             )
                         yield msg
-                if chunk['event'] == 'message_file':
+                elif chunk['event'] == 'message_file':
                     if chunk['type'] == 'image' and chunk['belongs_to'] == 'assistant':
                         base_url = self.dify_client.base_url
 
@@ -330,8 +327,10 @@ class DifyServiceAPIRunner(runner.RequestRunner):
                                 role='assistant',
                                 content=[llm_entities.ContentElement.from_image_url(image_url)],
                             )
-                if chunk['event'] == 'error':
+                elif chunk['event'] == 'error':
                     raise errors.DifyAPIError('dify 服务错误: ' + chunk['message'])
+                else:
+                    pending_agent_message = ''
 
         if chunk is None:
             raise errors.DifyAPIError('Dify API 没有返回任何响应，请检查网络连接和API配置')

templates/metadata/pipeline/ai.yaml

@@ -128,20 +128,7 @@ stages:
             label:
               en_US: Remove
               zh_Hans: 移除
-      - name: enable-streaming
-        label:
-          en_US: enable streaming mode
-          zh_Hans: 开启流式输出
-        type: boolean
-        required: true
-        default: false
-      - name: output-batch-size
-        label:
-          en_US: output batch size
-          zh_Hans: 输出批次大小(积累多少条消息后一起输出)
-        type: integer
-        required: true
-        default: 10
+
 
   - name: dashscope-app-api
     label: