feat:知识库检索已经命中了，但 vLLM 流式回答没有正常吐正文

apps/models_provider/impl/vllm_model_provider/model/llm.py

@@ -3,7 +3,7 @@
 from typing import Dict, List
 from urllib.parse import urlparse, ParseResult
 
-from langchain_core.messages import BaseMessage, get_buffer_string
+from langchain_core.messages import BaseMessage, get_buffer_string, AIMessageChunk
 
 from common.config.tokenizer_manage_config import TokenizerManage
 from models_provider.base_model_provider import MaxKBBaseModel
@@ -48,3 +48,18 @@ class VllmChatModel(MaxKBBaseModel, BaseChatOpenAI):
             tokenizer = TokenizerManage.get_tokenizer()
             return len(tokenizer.encode(text))
         return self.get_last_generation_info().get('output_tokens', 0)
+
+    def stream(self, input, config=None, *, stop=None, **kwargs):
+        has_content = False
+        for chunk in super().stream(input, config=config, stop=stop, **kwargs):
+            content = getattr(chunk, 'content', '') or ''
+            reasoning_content = (getattr(chunk, 'additional_kwargs', {}) or {}).get('reasoning_content', '')
+            if content or reasoning_content:
+                has_content = True
+                yield chunk
+        if not has_content:
+            result = self.invoke(input, config=config, stop=stop, **kwargs)
+            yield AIMessageChunk(
+                content=getattr(result, 'content', '') or '',
+                additional_kwargs=getattr(result, 'additional_kwargs', {}) or {}
+            )