fix:function call bug (#268)

* 优化function call消息处理

* fix:用户说话重复记录bug

* 2025-3-10-优化llm intent (#254)

Co-authored-by: 欣南科技 <huangrongzhuang@xin-nan.com>

* update:回复上版提示词，无需再二次识别歌曲名

---------

Co-authored-by: 玄凤科技 <eric230308@gmail.com>
Co-authored-by: hrz <1710360675@qq.com>
Co-authored-by: Jiao Haoyang <108573524+XuSenfeng@users.noreply.github.com>

main/xiaozhi-server/core/providers/intent/base.py

@@ -5,6 +5,7 @@ from config.logger import setup_logging
 TAG = __name__
 logger = setup_logging()
 
+
 class IntentProviderBase(ABC):
     def __init__(self, config):
         self.config = config
@@ -17,9 +18,9 @@ class IntentProviderBase(ABC):
     def set_llm(self, llm):
         self.llm = llm
         logger.bind(tag=TAG).debug("Set LLM for intent provider")
-        
+
     @abstractmethod
-    async def detect_intent(self, dialogue_history: List[Dict]) -> str:
+    async def detect_intent(self, dialogue_history: List[Dict], text: str) -> str:
         """
         检测用户最后一句话的意图
         Args:

main/xiaozhi-server/core/providers/intent/intent_llm/intent_llm.py

@@ -37,7 +37,7 @@ class IntentProvider(IntentProviderBase):
         )
         return prompt
     
-    async def detect_intent(self, dialogue_history: List[Dict]) -> str:
+    async def detect_intent(self, dialogue_history: List[Dict], text:str) -> str:
         if not self.llm:
             raise ValueError("LLM provider not set")
 
@@ -48,9 +48,8 @@ class IntentProvider(IntentProviderBase):
                 msgStr += f"User: {msg.content}\n"
             elif msg.role== "assistant":
                 msgStr += f"Assistant: {msg.content}\n"
-
+        msgStr += f"User: {text}\n"
         user_prompt = f"请分析用户的意图：\n{msgStr}"
-        
         # 使用LLM进行意图识别
         intent = self.llm.response_no_stream(
             system_prompt=self.promot,

main/xiaozhi-server/core/providers/intent/nointent/nointent.py

@@ -5,12 +5,14 @@ from config.logger import setup_logging
 TAG = __name__
 logger = setup_logging()
 
+
 class IntentProvider(IntentProviderBase):
-    async def detect_intent(self, dialogue_history: List[Dict]) -> str:
+    async def detect_intent(self, dialogue_history: List[Dict], text: str) -> str:
         """
         默认的意图识别实现，始终返回继续聊天
         Args:
             dialogue_history: 对话历史记录列表
+            text: 本次对话记录
         Returns:
             固定返回"继续聊天"
         """

main/xiaozhi-server/core/providers/llm/ollama/ollama.py

@@ -51,30 +51,8 @@ class LLMProvider(LLMProviderBase):
                 tools=functions,
             )
             
-            current_function_call = None
-            current_content = ""
-            
             for chunk in stream:
-                delta = chunk.choices[0].delta
-                
-                if delta.content:
-                    current_content += delta.content
-                    yield {"type": "content", "content": delta.content}
-                
-                if delta.tool_calls:
-                    tool_call = delta.tool_calls[0]
-                    # Handle the function call data using proper attribute access
-                    if not current_function_call:
-                        current_function_call = {
-                            "function": {
-                                "name": tool_call.function.name,
-                                "arguments": tool_call.function.arguments
-                            }
-                        }
-            
-            if current_function_call:
-                logger.bind(tag=TAG).debug(f"ollama Function call detected: {current_function_call}")
-                yield {"type": "function_call", "function_call": current_function_call}
+                yield chunk.choices[0].delta.content, chunk.choices[0].delta.tool_calls
 
         except Exception as e:
             logger.bind(tag=TAG).error(f"Error in Ollama function call: {e}")

main/xiaozhi-server/core/providers/llm/openai/openai.py

@@ -53,30 +53,8 @@ class LLMProvider(LLMProviderBase):
                 tools=functions,
             )
             
-            current_function_call = None
-            current_content = ""
-            
             for chunk in stream:
-                delta = chunk.choices[0].delta
-                
-                if delta.content:
-                    current_content += delta.content
-                    yield {"type": "content", "content": delta.content}
-                
-                if delta.tool_calls:
-                    tool_call = delta.tool_calls[0]
-                    # Handle the function call data using proper attribute access
-                    if not current_function_call:
-                        current_function_call = {
-                            "function": {
-                                "name": tool_call.function.name,
-                                "arguments": tool_call.function.arguments
-                            }
-                        }
-            
-            if current_function_call:
-                logger.bind(tag=TAG).debug(f"openai Function call detected: {current_function_call}")
-                yield {"type": "function_call", "function_call": current_function_call}
+                yield chunk.choices[0].delta.content, chunk.choices[0].delta.tool_calls
                 
         except Exception as e:
             self.logger.bind(tag=TAG).error(f"Error in function call streaming: {e}")