Merge pull request #1094 from CNMrSunshine/dev

修复配置文件中禁用LPMM无效的问题一个工作记忆问题和麦麦重复回复的问题
2025-07-06 11:42:24 +08:00
parent 61e708fe86 f624547034
commit d84fb0614c
6 changed files with 91 additions and 51 deletions
--- a/src/chat/focus_chat/info_processors/working_memory_processor.py
+++ b/src/chat/focus_chat/info_processors/working_memory_processor.py
@@ -71,6 +71,7 @@ class WorkingMemoryProcessor(BaseProcessor):
        """
        working_memory = None
        chat_info = ""
+        chat_obs = None
        try:
            for observation in observations:
                if isinstance(observation, WorkingMemoryObservation):
@@ -79,10 +80,15 @@ class WorkingMemoryProcessor(BaseProcessor):
                    chat_info = observation.get_observe_info()
                    chat_obs = observation
                    # 检查是否有待压缩内容
-            if chat_obs.compressor_prompt:
+            if chat_obs and chat_obs.compressor_prompt:
                logger.debug(f"{self.log_prefix} 压缩聊天记忆")
                await self.compress_chat_memory(working_memory, chat_obs)

+            # 检查working_memory是否为None
+            if working_memory is None:
+                logger.debug(f"{self.log_prefix} 没有找到工作记忆观察，跳过处理")
+                return []
+
            all_memory = working_memory.get_all_memories()
            if not all_memory:
                logger.debug(f"{self.log_prefix} 目前没有工作记忆，跳过提取")
@@ -183,6 +189,11 @@ class WorkingMemoryProcessor(BaseProcessor):
            working_memory: 工作记忆对象
            obs: 聊天观察对象
        """
+        # 检查working_memory是否为None
+        if working_memory is None:
+            logger.warning(f"{self.log_prefix} 工作记忆对象为None，无法压缩聊天记忆")
+            return
+            
        try:
            summary_result, _ = await self.llm_model.generate_response_async(obs.compressor_prompt)
            if not summary_result:
@@ -235,6 +246,11 @@ class WorkingMemoryProcessor(BaseProcessor):
            memory_id1: 第一个记忆ID
            memory_id2: 第二个记忆ID
        """
+        # 检查working_memory是否为None
+        if working_memory is None:
+            logger.warning(f"{self.log_prefix} 工作记忆对象为None，无法合并记忆")
+            return
+            
        try:
            merged_memory = await working_memory.merge_memory(memory_id1, memory_id2)
            logger.debug(f"{self.log_prefix} 合并后的记忆梗概: {merged_memory.brief}")
--- a/src/chat/focus_chat/planners/planner_simple.py
+++ b/src/chat/focus_chat/planners/planner_simple.py
@@ -29,6 +29,7 @@ def init_prompt():
 {chat_context_description}，以下是具体的聊天内容：
 {chat_content_block}
 {moderation_prompt}
+
 现在请你根据聊天内容选择合适的action:

 {action_options_text}
--- a/src/chat/knowledge/knowledge_lib.py
+++ b/src/chat/knowledge/knowledge_lib.py
@@ -5,60 +5,68 @@ from src.chat.knowledge.mem_active_manager import MemoryActiveManager
 from src.chat.knowledge.qa_manager import QAManager
 from src.chat.knowledge.kg_manager import KGManager
 from src.chat.knowledge.global_logger import logger
+from src.config.config import global_config as bot_global_config
 # try:
 #     import quick_algo
 # except ImportError:
 #     print("quick_algo not found, please install it first")

-logger.info("正在初始化Mai-LPMM\n")
-logger.info("创建LLM客户端")
-llm_client_list = dict()
-for key in global_config["llm_providers"]:
-    llm_client_list[key] = LLMClient(
-        global_config["llm_providers"][key]["base_url"],
-        global_config["llm_providers"][key]["api_key"],
+# 检查LPMM知识库是否启用
+if bot_global_config.lpmm_knowledge.enable:
+    logger.info("正在初始化Mai-LPMM\n")
+    logger.info("创建LLM客户端")
+    llm_client_list = dict()
+    for key in global_config["llm_providers"]:
+        llm_client_list[key] = LLMClient(
+            global_config["llm_providers"][key]["base_url"],
+            global_config["llm_providers"][key]["api_key"],
+        )
+
+    # 初始化Embedding库
+    embed_manager = EmbeddingManager(llm_client_list[global_config["embedding"]["provider"]])
+    logger.info("正在从文件加载Embedding库")
+    try:
+        embed_manager.load_from_file()
+    except Exception as e:
+        logger.warning("此消息不会影响正常使用：从文件加载Embedding库时，{}".format(e))
+        # logger.warning("如果你是第一次导入知识，或者还未导入知识，请忽略此错误")
+    logger.info("Embedding库加载完成")
+    # 初始化KG
+    kg_manager = KGManager()
+    logger.info("正在从文件加载KG")
+    try:
+        kg_manager.load_from_file()
+    except Exception as e:
+        logger.warning("此消息不会影响正常使用：从文件加载KG时，{}".format(e))
+        # logger.warning("如果你是第一次导入知识，或者还未导入知识，请忽略此错误")
+    logger.info("KG加载完成")
+
+    logger.info(f"KG节点数量：{len(kg_manager.graph.get_node_list())}")
+    logger.info(f"KG边数量：{len(kg_manager.graph.get_edge_list())}")
+
+
+    # 数据比对：Embedding库与KG的段落hash集合
+    for pg_hash in kg_manager.stored_paragraph_hashes:
+        key = PG_NAMESPACE + "-" + pg_hash
+        if key not in embed_manager.stored_pg_hashes:
+            logger.warning(f"KG中存在Embedding库中不存在的段落：{key}")
+
+    # 问答系统（用于知识库）
+    qa_manager = QAManager(
+        embed_manager,
+        kg_manager,
+        llm_client_list[global_config["embedding"]["provider"]],
+        llm_client_list[global_config["qa"]["llm"]["provider"]],
+        llm_client_list[global_config["qa"]["llm"]["provider"]],
    )

-# 初始化Embedding库
-embed_manager = EmbeddingManager(llm_client_list[global_config["embedding"]["provider"]])
-logger.info("正在从文件加载Embedding库")
-try:
-    embed_manager.load_from_file()
-except Exception as e:
-    logger.warning("此消息不会影响正常使用：从文件加载Embedding库时，{}".format(e))
-    # logger.warning("如果你是第一次导入知识，或者还未导入知识，请忽略此错误")
-logger.info("Embedding库加载完成")
-# 初始化KG
-kg_manager = KGManager()
-logger.info("正在从文件加载KG")
-try:
-    kg_manager.load_from_file()
-except Exception as e:
-    logger.warning("此消息不会影响正常使用：从文件加载KG时，{}".format(e))
-    # logger.warning("如果你是第一次导入知识，或者还未导入知识，请忽略此错误")
-logger.info("KG加载完成")
-
-logger.info(f"KG节点数量：{len(kg_manager.graph.get_node_list())}")
-logger.info(f"KG边数量：{len(kg_manager.graph.get_edge_list())}")
-
-
-# 数据比对：Embedding库与KG的段落hash集合
-for pg_hash in kg_manager.stored_paragraph_hashes:
-    key = PG_NAMESPACE + "-" + pg_hash
-    if key not in embed_manager.stored_pg_hashes:
-        logger.warning(f"KG中存在Embedding库中不存在的段落：{key}")
-
-# 问答系统（用于知识库）
-qa_manager = QAManager(
-    embed_manager,
-    kg_manager,
-    llm_client_list[global_config["embedding"]["provider"]],
-    llm_client_list[global_config["qa"]["llm"]["provider"]],
-    llm_client_list[global_config["qa"]["llm"]["provider"]],
-)
-
-# 记忆激活（用于记忆库）
-inspire_manager = MemoryActiveManager(
-    embed_manager,
-    llm_client_list[global_config["embedding"]["provider"]],
-)
+    # 记忆激活（用于记忆库）
+    inspire_manager = MemoryActiveManager(
+        embed_manager,
+        llm_client_list[global_config["embedding"]["provider"]],
+    )
+else:
+    logger.info("LPMM知识库已禁用，跳过初始化")
+    # 创建空的占位符对象，避免导入错误
+    qa_manager = None
+    inspire_manager = None
--- a/src/chat/replyer/default_generator.py
+++ b/src/chat/replyer/default_generator.py
@@ -956,6 +956,11 @@ async def get_prompt_info(message: str, threshold: float):
    logger.debug(f"获取知识库内容，元消息：{message[:30]}...，消息长度: {len(message)}")
    # 从LPMM知识库获取知识
    try:
+        # 检查LPMM知识库是否启用
+        if qa_manager is None:
+            logger.debug("LPMM知识库已禁用，跳过知识获取")
+            return ""
+            
        found_knowledge_from_lpmm = qa_manager.get_knowledge(message)

        end_time = time.time()
--- a/src/experimental/PFC/pfc_KnowledgeFetcher.py
+++ b/src/experimental/PFC/pfc_KnowledgeFetcher.py
@@ -35,6 +35,11 @@ class KnowledgeFetcher:

        logger.debug(f"[私聊][{self.private_name}]正在从LPMM知识库中获取知识")
        try:
+            # 检查LPMM知识库是否启用
+            if qa_manager is None:
+                logger.debug(f"[私聊][{self.private_name}]LPMM知识库已禁用，跳过知识获取")
+                return "未找到匹配的知识"
+                
            knowledge_info = qa_manager.get_knowledge(query)
            logger.debug(f"[私聊][{self.private_name}]LPMM知识库查询结果: {knowledge_info:150}")
            return knowledge_info
--- a/src/tools/not_using/lpmm_get_knowledge.py
+++ b/src/tools/not_using/lpmm_get_knowledge.py
@@ -36,6 +36,11 @@ class SearchKnowledgeFromLPMMTool(BaseTool):
            query = function_args.get("query")
            # threshold = function_args.get("threshold", 0.4)

+            # 检查LPMM知识库是否启用
+            if qa_manager is None:
+                logger.debug("LPMM知识库已禁用，跳过知识获取")
+                return {"type": "info", "id": query, "content": "LPMM知识库已禁用"}
+
            # 调用知识库搜索

            knowledge_info = qa_manager.get_knowledge(query)