fix: In the runner, every 8 tokens yield

pkg/provider/modelmgr/requesters/chatcmpl.py

@@ -189,7 +189,6 @@ class OpenAIChatCompletions(requester.ProviderAPIRequester):
                 delta_message.is_final = True
                 delta_message.content = current_content
 
-            if chunk_idx % 64 == 0 or delta_message.is_final:
             yield delta_message
             # return
 

pkg/provider/modelmgr/requesters/giteeaichatcmpl.py

@@ -195,5 +195,4 @@ class GiteeAIChatCompletions(chatcmpl.OpenAIChatCompletions):
                 delta_message.is_final = True
                 delta_message.content = current_content
 
-            if chunk_idx % 64 == 0 or delta_message.is_final:
             yield delta_message

pkg/provider/modelmgr/requesters/modelscopechatcmpl.py

@@ -286,7 +286,6 @@ class ModelScopeChatCompletions(requester.ProviderAPIRequester):
                 delta_message.is_final = True
                 delta_message.content = current_content
 
-            if chunk_idx % 64 == 0 or delta_message.is_final:
             yield delta_message
             # return
 

pkg/provider/runners/localagent.py

@@ -111,6 +111,7 @@ class LocalAgentRunner(runner.RequestRunner):
         else:
             # 流式输出，需要处理工具调用
             tool_calls_map: dict[str, llm_entities.ToolCall] = {}
+            msg_idx = 0
             async for msg in query.use_llm_model.requester.invoke_llm_stream(
                 query,
                 query.use_llm_model,
@@ -118,7 +119,8 @@ class LocalAgentRunner(runner.RequestRunner):
                 query.use_funcs,
                 extra_args=query.use_llm_model.model_entity.extra_args,
             ):
-                assert isinstance(msg, llm_entities.MessageChunk)
+                msg_idx = msg_idx + 1
+                if msg_idx % 8 == 0 or msg.is_final:
                     yield msg
                 if msg.tool_calls:
                     for tool_call in msg.tool_calls: