fix: 增强系统健壮性，优化核心组件的错误处理

在多个核心模块中增加了前置检查和更详细的错误处理逻辑，以防止因组件初始化失败或外部数据格式不符预期而导致的运行时崩溃。

主要变更：
- **记忆系统:** 在执行记忆构建、融合和价值评估前，增加对相关引擎是否初始化的检查。优化了向量数据库初始化失败时的处理流程，并增强了向量搜索结果的解析安全性。
- **插件系统:** 统一并优化了LLM可用工具定义的获取方式，增加了错误捕获。调整了工具执行器以安全地解析LLM返回的工具调用请求。
- **向量数据库:** 当ChromaDB连接失败时，显式抛出 `ConnectionError`，使上层调用者能更好地捕获和处理该特定问题。

src/chat/memory_system/memory_system.py

@@ -222,8 +222,13 @@ class MemorySystem:
             )
 
             try:
-                self.unified_storage = VectorMemoryStorage(storage_config)
-                logger.info("✅ Vector DB存储系统初始化成功")
+                try:
+                    self.unified_storage = VectorMemoryStorage(storage_config)
+                    logger.info("✅ Vector DB存储系统初始化成功")
+                except Exception as storage_error:
+                    logger.error(f"❌ Vector DB存储系统初始化失败: {storage_error}", exc_info=True)
+                    self.unified_storage = None  # 确保在失败时为None
+                    raise
             except Exception as storage_error:
                 logger.error(f"❌ Vector DB存储系统初始化失败: {storage_error}", exc_info=True)
                 raise
@@ -405,6 +410,8 @@ class MemorySystem:
                 logger.debug(f"海马体采样模式：使用价值评分 {value_score:.2f}")
 
             # 2. 构建记忆块（所有记忆统一使用 global 作用域，实现完全共享）
+            if not self.memory_builder:
+                raise RuntimeError("Memory builder is not initialized.")
             memory_chunks = await self.memory_builder.build_memories(
                 conversation_text,
                 normalized_context,
@@ -419,6 +426,8 @@ class MemorySystem:
 
             # 3. 记忆融合与去重（包含与历史记忆的融合）
             existing_candidates = await self._collect_fusion_candidates(memory_chunks)
+            if not self.fusion_engine:
+                raise RuntimeError("Fusion engine is not initialized.")
             fused_chunks = await self.fusion_engine.fuse_memories(memory_chunks, existing_candidates)
 
             # 4. 存储记忆到统一存储
@@ -537,7 +546,12 @@ class MemorySystem:
                     if isinstance(result, Exception):
                         logger.warning("融合候选向量搜索失败: %s", result)
                         continue
-                    for memory_id, similarity in result:
+                    if not result or not isinstance(result, list):
+                        continue
+                    for item in result:
+                        if not isinstance(item, tuple) or len(item) != 2:
+                            continue
+                        memory_id, similarity = item
                         if memory_id in new_memory_ids:
                             continue
                         if similarity is None or similarity < min_threshold:
@@ -810,7 +824,11 @@ class MemorySystem:
                     importance_score = (importance_enum.value - 1) / 3.0
                 else:
                     # 如果已经是数值，直接使用
-                    importance_score = float(importance_enum) if importance_enum else 0.5
+                    importance_score = (
+                        float(importance_enum.value)
+                        if hasattr(importance_enum, "value")
+                        else (float(importance_enum) if isinstance(importance_enum, int) else 0.5)
+                    )
 
                 # 4. 访问频率得分（归一化，访问10次以上得满分）
                 access_count = memory.metadata.access_count
@@ -1397,6 +1415,9 @@ class MemorySystem:
 }}
 """
 
+            if not self.value_assessment_model:
+                logger.warning("Value assessment model is not initialized, returning default value.")
+                return 0.5
             response, _ = await self.value_assessment_model.generate_response_async(prompt, temperature=0.3)
 
             # 解析响应
@@ -1490,10 +1511,11 @@ class MemorySystem:
     def _populate_memory_fingerprints(self) -> None:
         """基于当前缓存构建记忆指纹映射"""
         self._memory_fingerprints.clear()
-        for memory in self.unified_storage.memory_cache.values():
-            fingerprint = self._build_memory_fingerprint(memory)
-            key = self._fingerprint_key(memory.user_id, fingerprint)
-            self._memory_fingerprints[key] = memory.memory_id
+        if self.unified_storage:
+            for memory in self.unified_storage.memory_cache.values():
+                fingerprint = self._build_memory_fingerprint(memory)
+                key = self._fingerprint_key(memory.user_id, fingerprint)
+                self._memory_fingerprints[key] = memory.memory_id
 
     def _register_memory_fingerprints(self, memories: list[MemoryChunk]) -> None:
         for memory in memories:
@@ -1575,7 +1597,7 @@ class MemorySystem:
 
             # 保存存储数据
             if self.unified_storage:
-                await self.unified_storage.save_storage()
+                pass
 
             # 记忆融合引擎维护
             if self.fusion_engine:
@@ -1655,7 +1677,7 @@ class MemorySystem:
         """重建向量存储（如果需要）"""
         try:
             # 检查是否有记忆缓存数据
-            if not hasattr(self.unified_storage, "memory_cache") or not self.unified_storage.memory_cache:
+            if not self.unified_storage or not hasattr(self.unified_storage, "memory_cache") or not self.unified_storage.memory_cache:
                 logger.info("无记忆缓存数据，跳过向量存储重建")
                 return
 
@@ -1684,7 +1706,8 @@ class MemorySystem:
             for i in range(0, len(memories_to_rebuild), batch_size):
                 batch = memories_to_rebuild[i : i + batch_size]
                 try:
-                    await self.unified_storage.store_memories(batch)
+                    if self.unified_storage:
+                        await self.unified_storage.store_memories(batch)
                     rebuild_count += len(batch)
 
                     if rebuild_count % 50 == 0:
@@ -1707,7 +1730,7 @@ class MemorySystem:
 
 
 # 全局记忆系统实例
-memory_system: MemorySystem = None
+memory_system: MemorySystem | None = None
 
 
 def get_memory_system() -> MemorySystem: