From 19cfa37097787184be1bbe8edbf9fec799ca783f Mon Sep 17 00:00:00 2001 From: Windpicker-owo <3431391539@qq.com> Date: Tue, 11 Nov 2025 19:47:31 +0800 Subject: [PATCH] =?UTF-8?q?feat(memory=5Fmanager):=20=E5=A2=9E=E5=BC=BA?= =?UTF-8?q?=E9=85=8D=E7=BD=AE=E6=97=A5=E5=BF=97=E8=AE=B0=E5=BD=95=EF=BC=8C?= =?UTF-8?q?=E6=B7=BB=E5=8A=A0=E6=90=9C=E7=B4=A2=E6=9D=83=E9=87=8D=E5=92=8C?= =?UTF-8?q?=E9=98=88=E5=80=BC=E8=BF=87=E6=BB=A4=E6=94=AF=E6=8C=81?= MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit --- src/memory_graph/manager.py | 36 ++++++++- src/memory_graph/tools/memory_tools.py | 103 ++++++++++++++++++------- 2 files changed, 111 insertions(+), 28 deletions(-) diff --git a/src/memory_graph/manager.py b/src/memory_graph/manager.py index f9bf5ed6a..5100d7b23 100644 --- a/src/memory_graph/manager.py +++ b/src/memory_graph/manager.py @@ -141,7 +141,29 @@ class MemoryManager: # 检查配置值 expand_depth = self.config.search_max_expand_depth expand_semantic_threshold = self.config.search_expand_semantic_threshold - logger.info(f"📊 配置检查: search_max_expand_depth={expand_depth}, search_expand_semantic_threshold={expand_semantic_threshold}") + search_top_k = self.config.search_top_k + # 读取权重配置 + search_vector_weight = self.config.search_vector_weight + search_importance_weight = self.config.search_importance_weight + search_recency_weight = self.config.search_recency_weight + # 读取阈值过滤配置 + search_min_importance = self.config.search_min_importance + search_similarity_threshold = self.config.search_similarity_threshold + + logger.info( + f"📊 配置检查: search_max_expand_depth={expand_depth}, " + f"search_expand_semantic_threshold={expand_semantic_threshold}, " + f"search_top_k={search_top_k}" + ) + logger.info( + f"📊 权重配置: vector={search_vector_weight}, " + f"importance={search_importance_weight}, " + f"recency={search_recency_weight}" + ) + logger.info( + f"📊 阈值过滤: min_importance={search_min_importance}, " + f"similarity_threshold={search_similarity_threshold}" + ) self.tools = MemoryTools( vector_store=self.vector_store, @@ -150,6 +172,12 @@ class MemoryManager: embedding_generator=self.embedding_generator, max_expand_depth=expand_depth, # 从配置读取图扩展深度 expand_semantic_threshold=expand_semantic_threshold, # 从配置读取图扩展语义阈值 + search_top_k=search_top_k, # 从配置读取默认 top_k + search_vector_weight=search_vector_weight, # 从配置读取向量权重 + search_importance_weight=search_importance_weight, # 从配置读取重要性权重 + search_recency_weight=search_recency_weight, # 从配置读取时效性权重 + search_min_importance=search_min_importance, # 从配置读取最小重要性阈值 + search_similarity_threshold=search_similarity_threshold, # 从配置读取相似度阈值 ) self._initialized = True @@ -348,7 +376,7 @@ class MemoryManager: async def search_memories( self, query: str, - top_k: int = 10, + top_k: int | None = None, memory_types: list[str] | None = None, time_range: tuple[datetime, datetime] | None = None, min_importance: float = 0.0, @@ -384,6 +412,10 @@ class MemoryManager: await self.initialize() try: + # 使用配置的默认值 + if top_k is None: + top_k = getattr(self.config, "search_top_k", 10) + # 准备搜索参数 params = { "query": query, diff --git a/src/memory_graph/tools/memory_tools.py b/src/memory_graph/tools/memory_tools.py index 195d2d01b..be03db4fb 100644 --- a/src/memory_graph/tools/memory_tools.py +++ b/src/memory_graph/tools/memory_tools.py @@ -37,6 +37,14 @@ class MemoryTools: embedding_generator: EmbeddingGenerator | None = None, max_expand_depth: int = 1, expand_semantic_threshold: float = 0.3, + search_top_k: int = 10, + # 新增：搜索权重配置 + search_vector_weight: float = 0.65, + search_importance_weight: float = 0.25, + search_recency_weight: float = 0.10, + # 新增：阈值过滤配置 + search_min_importance: float = 0.3, + search_similarity_threshold: float = 0.5, ): """ 初始化工具集 @@ -48,15 +56,37 @@ class MemoryTools: embedding_generator: 嵌入生成器（可选） max_expand_depth: 图扩展深度的默认值（从配置读取） expand_semantic_threshold: 图扩展时语义相似度阈值（从配置读取） + search_top_k: 默认检索返回数量（从配置读取） + search_vector_weight: 向量相似度权重（从配置读取） + search_importance_weight: 重要性权重（从配置读取） + search_recency_weight: 时效性权重（从配置读取） + search_min_importance: 最小重要性阈值（从配置读取） + search_similarity_threshold: 向量相似度阈值（从配置读取） """ self.vector_store = vector_store self.graph_store = graph_store self.persistence_manager = persistence_manager self._initialized = False - self.max_expand_depth = max_expand_depth # 保存配置的默认值 - self.expand_semantic_threshold = expand_semantic_threshold # 保存配置的语义阈值 + self.max_expand_depth = max_expand_depth + self.expand_semantic_threshold = expand_semantic_threshold + self.search_top_k = search_top_k + + # 保存权重配置 + self.base_vector_weight = search_vector_weight + self.base_importance_weight = search_importance_weight + self.base_recency_weight = search_recency_weight + + # 保存阈值过滤配置 + self.search_min_importance = search_min_importance + self.search_similarity_threshold = search_similarity_threshold - logger.info(f"MemoryTools 初始化: max_expand_depth={max_expand_depth}, expand_semantic_threshold={expand_semantic_threshold}") + logger.info( + f"MemoryTools 初始化: max_expand_depth={max_expand_depth}, " + f"expand_semantic_threshold={expand_semantic_threshold}, " + f"search_top_k={search_top_k}, " + f"权重配置: vector={search_vector_weight}, importance={search_importance_weight}, recency={search_recency_weight}, " + f"阈值过滤: min_importance={search_min_importance}, similarity_threshold={search_similarity_threshold}" + ) # 初始化组件 self.extractor = MemoryExtractor() @@ -465,7 +495,7 @@ class MemoryTools: """ try: query = params.get("query", "") - top_k = params.get("top_k", 10) + top_k = params.get("top_k", self.search_top_k) # 使用配置的默认值 expand_depth = params.get("expand_depth", self.max_expand_depth) use_multi_query = params.get("use_multi_query", True) prefer_node_types = params.get("prefer_node_types", []) # 🆕 优先节点类型 @@ -610,7 +640,7 @@ class MemoryTools: if activation_score == 0.0 and memory.activation > 0.0: activation_score = memory.activation - # 🆕 动态权重计算：根据记忆类型和节点类型自适应调整 + # 🆕 动态权重计算：使用配置的基础权重 + 根据记忆类型微调 memory_type = memory.memory_type.value if hasattr(memory.memory_type, 'value') else str(memory.memory_type) # 检测记忆的主要节点类型 @@ -621,36 +651,48 @@ class MemoryTools: dominant_node_type = max(node_types_count.items(), key=lambda x: x[1])[0] if node_types_count else "unknown" - # 根据节点类型动态调整权重 + # 根据记忆类型和节点类型计算调整系数（在配置权重基础上微调） if dominant_node_type in ["ATTRIBUTE", "REFERENCE"] or memory_type == "FACT": - # 事实性记忆（如文档地址、配置信息）：语义相似度最重要 - weights = { - "similarity": 0.70, # 语义相似度 70% ⬆️ - "importance": 0.25, # 重要性 25% ⬆️ - "recency": 0.05, # 时效性 5%（事实不随时间失效） + # 事实性记忆：提升相似度权重，降低时效性权重 + type_adjustments = { + "similarity": 1.08, # 相似度提升 8% + "importance": 1.0, # 重要性保持 + "recency": 0.5, # 时效性降低 50%（事实不随时间失效） } elif memory_type in ["CONVERSATION", "EPISODIC"] or dominant_node_type == "EVENT": - # 对话/事件记忆：时效性更重要 - weights = { - "similarity": 0.55, # 语义相似度 55% ⬆️ - "importance": 0.20, # 重要性 20% ⬆️ - "recency": 0.25, # 时效性 25% ⬆️ + # 对话/事件记忆：提升时效性权重 + type_adjustments = { + "similarity": 0.85, # 相似度降低 15% + "importance": 0.8, # 重要性降低 20% + "recency": 2.5, # 时效性提升 150% } elif dominant_node_type == "ENTITY" or memory_type == "SEMANTIC": - # 实体/语义记忆：平衡各项 - weights = { - "similarity": 0.60, # 语义相似度 60% ⬆️ - "importance": 0.30, # 重要性 30% ⬆️ - "recency": 0.10, # 时效性 10% + # 实体/语义记忆：平衡调整 + type_adjustments = { + "similarity": 0.92, # 相似度微降 8% + "importance": 1.2, # 重要性提升 20% + "recency": 1.0, # 时效性保持 } else: - # 默认权重（保守策略，偏向语义） - weights = { - "similarity": 0.65, # 语义相似度 65% ⬆️ - "importance": 0.25, # 重要性 25% ⬆️ - "recency": 0.10, # 时效性 10% + # 默认不调整 + type_adjustments = { + "similarity": 1.0, + "importance": 1.0, + "recency": 1.0, } + # 应用调整后的权重（基于配置的基础权重） + weights = { + "similarity": self.base_vector_weight * type_adjustments["similarity"], + "importance": self.base_importance_weight * type_adjustments["importance"], + "recency": self.base_recency_weight * type_adjustments["recency"], + } + + # 归一化权重（确保总和为1.0） + total_weight = sum(weights.values()) + if total_weight > 0: + weights = {k: v / total_weight for k, v in weights.items()} + # 综合分数计算（🔥 移除激活度影响） final_score = ( similarity_score * weights["similarity"] + @@ -658,6 +700,15 @@ class MemoryTools: recency_score * weights["recency"] ) + # 🆕 阈值过滤：基于配置的最小重要性和相似度阈值 + if memory.importance < self.search_min_importance: + logger.debug(f"记忆 {memory.id[:8]} 重要性 {memory.importance:.2f} 低于阈值 {self.search_min_importance}，过滤") + continue + + if similarity_score < self.search_similarity_threshold: + logger.debug(f"记忆 {memory.id[:8]} 相似度 {similarity_score:.2f} 低于阈值 {self.search_similarity_threshold}，过滤") + continue + # 🆕 节点类型加权：对REFERENCE/ATTRIBUTE节点额外加分（促进事实性信息召回） if "REFERENCE" in node_types_count or "ATTRIBUTE" in node_types_count: final_score *= 1.1 # 10% 加成