@yh-ui/ai-sdk 0.1.21

package/dist/vector-store.d.ts

@@ -0,0 +1,74 @@
+/**
+ * YH-UI AI SDK - 向量存储抽象层
+ *
+ * 支持多种后端，便于对接 Pinecone / Weaviate / Qdrant 等生产级向量库。
+ * 内置内存实现，可选接入外部向量服务。
+ */
+export interface VectorDocument {
+    id: string;
+    content: string;
+    embedding: number[];
+    metadata?: Record<string, unknown>;
+}
+export interface VectorSearchResult {
+    id: string;
+    content: string;
+    metadata?: Record<string, unknown>;
+    score: number;
+    embedding?: number[];
+}
+export interface VectorStoreConfig {
+    /** 命名空间/索引名 */
+    namespace?: string;
+    /** 向量维度 (部分后端需要) */
+    dimension?: number;
+}
+/**
+ * 向量存储统一接口
+ * 实现此接口即可接入 Pinecone / Weaviate / Qdrant 等
+ */
+export interface IVectorStore {
+    /** 添加文档（含向量） */
+    add(documents: VectorDocument[], namespace?: string): Promise<void>;
+    /** 相似度检索 */
+    search(queryEmbedding: number[], options: {
+        topK?: number;
+        threshold?: number;
+        namespace?: string;
+        filter?: Record<string, unknown>;
+    }): Promise<VectorSearchResult[]>;
+    /** 按 ID 删除 */
+    delete(ids: string[], namespace?: string): Promise<void>;
+    /** 清空命名空间 */
+    clear(namespace?: string): Promise<void>;
+}
+/**
+ * 内存向量存储（内置实现，适用于开发与小规模）
+ * 使用余弦相似度检索
+ */
+export declare function createInMemoryVectorStore(_config?: VectorStoreConfig): IVectorStore;
+/**
+ * Pinecone 适配器接口（由用户或可选包实现）
+ * 安装 @pinecone-database/pinecone 后可实现
+ */
+export type PineconeVectorStoreConfig = VectorStoreConfig & {
+    apiKey: string;
+    indexName: string;
+    environment?: string;
+};
+/**
+ * Weaviate 适配器接口
+ */
+export type WeaviateVectorStoreConfig = VectorStoreConfig & {
+    url: string;
+    apiKey?: string;
+    className?: string;
+};
+/**
+ * Qdrant 适配器接口
+ */
+export type QdrantVectorStoreConfig = VectorStoreConfig & {
+    url: string;
+    apiKey?: string;
+    collectionName?: string;
+};

package/dist/vector-store.mjs

@@ -0,0 +1,55 @@
+export function createInMemoryVectorStore(_config) {
+  const store = /* @__PURE__ */ new Map();
+  function cosineSimilarity(a, b) {
+    if (a.length !== b.length) return 0;
+    let dot = 0;
+    let normA = 0;
+    let normB = 0;
+    for (let i = 0; i < a.length; i++) {
+      dot += a[i] * b[i];
+      normA += a[i] * a[i];
+      normB += b[i] * b[i];
+    }
+    const denom = Math.sqrt(normA) * Math.sqrt(normB);
+    return denom === 0 ? 0 : dot / denom;
+  }
+  return {
+    async add(documents, namespace = "default") {
+      const list = store.get(namespace) || [];
+      for (const doc of documents) {
+        const idx = list.findIndex((d) => d.id === doc.id);
+        if (idx >= 0) list[idx] = doc;
+        else list.push(doc);
+      }
+      store.set(namespace, list);
+    },
+    async search(queryEmbedding, options = {}) {
+      const topK = options.topK ?? 10;
+      const threshold = options.threshold ?? 0;
+      const namespaceOpt = options.namespace ?? "default";
+      const list = store.get(namespaceOpt) || [];
+      const scored = list.map((doc) => ({
+        ...doc,
+        score: cosineSimilarity(doc.embedding, queryEmbedding)
+      })).filter((d) => d.score >= threshold).sort((a, b) => b.score - a.score).slice(0, topK);
+      return scored.map(({ id, content, metadata, score, embedding }) => ({
+        id,
+        content,
+        metadata,
+        score,
+        embedding
+      }));
+    },
+    async delete(ids, namespace = "default") {
+      const list = store.get(namespace) || [];
+      const set = new Set(ids);
+      store.set(
+        namespace,
+        list.filter((d) => !set.has(d.id))
+      );
+    },
+    async clear(namespace = "default") {
+      store.set(namespace, []);
+    }
+  };
+}