@langchain/classic 1.0.0-alpha.1

package/dist/retrievers/multi_vector.cjs

@@ -0,0 +1,67 @@
+const require_rolldown_runtime = require('../_virtual/rolldown_runtime.cjs');
+const require_storage_encoder_backed = require('../storage/encoder_backed.cjs');
+const __langchain_core_retrievers = require_rolldown_runtime.__toESM(require("@langchain/core/retrievers"));
+
+//#region src/retrievers/multi_vector.ts
+var multi_vector_exports = {};
+require_rolldown_runtime.__export(multi_vector_exports, { MultiVectorRetriever: () => MultiVectorRetriever });
+/**
+* A retriever that retrieves documents from a vector store and a document
+* store. It uses the vector store to find relevant documents based on a
+* query, and then retrieves the full documents from the document store.
+* @example
+* ```typescript
+* const retriever = new MultiVectorRetriever({
+*   vectorstore: new FaissStore(),
+*   byteStore: new InMemoryStore<Unit8Array>(),
+*   idKey: "doc_id",
+*   childK: 20,
+*   parentK: 5,
+* });
+*
+* const retrieverResult = await retriever.invoke("justice breyer");
+* console.log(retrieverResult[0].pageContent.length);
+* ```
+*/
+var MultiVectorRetriever = class extends __langchain_core_retrievers.BaseRetriever {
+	static lc_name() {
+		return "MultiVectorRetriever";
+	}
+	lc_namespace = [
+		"langchain",
+		"retrievers",
+		"multi_vector"
+	];
+	vectorstore;
+	docstore;
+	idKey;
+	childK;
+	parentK;
+	constructor(args) {
+		super(args);
+		this.vectorstore = args.vectorstore;
+		if (args.byteStore) this.docstore = require_storage_encoder_backed.createDocumentStoreFromByteStore(args.byteStore);
+		else if (args.docstore) this.docstore = args.docstore;
+		else throw new Error("byteStore and docstore are undefined. Please provide at least one.");
+		this.idKey = args.idKey ?? "doc_id";
+		this.childK = args.childK;
+		this.parentK = args.parentK;
+	}
+	async _getRelevantDocuments(query) {
+		const subDocs = await this.vectorstore.similaritySearch(query, this.childK);
+		const ids = [];
+		for (const doc of subDocs) if (doc.metadata[this.idKey] && !ids.includes(doc.metadata[this.idKey])) ids.push(doc.metadata[this.idKey]);
+		const docs = await this.docstore.mget(ids);
+		return docs.filter((doc) => doc !== void 0).slice(0, this.parentK);
+	}
+};
+
+//#endregion
+exports.MultiVectorRetriever = MultiVectorRetriever;
+Object.defineProperty(exports, 'multi_vector_exports', {
+  enumerable: true,
+  get: function () {
+    return multi_vector_exports;
+  }
+});
+//# sourceMappingURL=multi_vector.cjs.map

package/dist/retrievers/multi_vector.cjs.map

@@ -0,0 +1 @@
+{"version":3,"file":"multi_vector.cjs","names":["BaseRetriever","args: MultiVectorRetrieverInput","createDocumentStoreFromByteStore","query: string","ids: string[]"],"sources":["../../src/retrievers/multi_vector.ts"],"sourcesContent":["import {\n  BaseRetriever,\n  type BaseRetrieverInput,\n} from \"@langchain/core/retrievers\";\nimport type { VectorStoreInterface } from \"@langchain/core/vectorstores\";\nimport { Document } from \"@langchain/core/documents\";\nimport { BaseStore, type BaseStoreInterface } from \"@langchain/core/stores\";\nimport { createDocumentStoreFromByteStore } from \"../storage/encoder_backed.js\";\n\n/**\n * Arguments for the MultiVectorRetriever class.\n */\nexport interface MultiVectorRetrieverInput extends BaseRetrieverInput {\n  vectorstore: VectorStoreInterface;\n  /** @deprecated Prefer `byteStore`. */\n  docstore?: BaseStoreInterface<string, Document>;\n  byteStore?: BaseStore<string, Uint8Array>;\n  idKey?: string;\n  childK?: number;\n  parentK?: number;\n}\n\n/**\n * A retriever that retrieves documents from a vector store and a document\n * store. It uses the vector store to find relevant documents based on a\n * query, and then retrieves the full documents from the document store.\n * @example\n * ```typescript\n * const retriever = new MultiVectorRetriever({\n *   vectorstore: new FaissStore(),\n *   byteStore: new InMemoryStore<Unit8Array>(),\n *   idKey: \"doc_id\",\n *   childK: 20,\n *   parentK: 5,\n * });\n *\n * const retrieverResult = await retriever.invoke(\"justice breyer\");\n * console.log(retrieverResult[0].pageContent.length);\n * ```\n */\nexport class MultiVectorRetriever extends BaseRetriever {\n  static lc_name() {\n    return \"MultiVectorRetriever\";\n  }\n\n  lc_namespace = [\"langchain\", \"retrievers\", \"multi_vector\"];\n\n  public vectorstore: VectorStoreInterface;\n\n  public docstore: BaseStoreInterface<string, Document>;\n\n  protected idKey: string;\n\n  protected childK?: number;\n\n  protected parentK?: number;\n\n  constructor(args: MultiVectorRetrieverInput) {\n    super(args);\n    this.vectorstore = args.vectorstore;\n    if (args.byteStore) {\n      this.docstore = createDocumentStoreFromByteStore(args.byteStore);\n    } else if (args.docstore) {\n      this.docstore = args.docstore;\n    } else {\n      throw new Error(\n        \"byteStore and docstore are undefined. Please provide at least one.\"\n      );\n    }\n    this.idKey = args.idKey ?? \"doc_id\";\n    this.childK = args.childK;\n    this.parentK = args.parentK;\n  }\n\n  async _getRelevantDocuments(query: string): Promise<Document[]> {\n    const subDocs = await this.vectorstore.similaritySearch(query, this.childK);\n    const ids: string[] = [];\n    for (const doc of subDocs) {\n      if (doc.metadata[this.idKey] && !ids.includes(doc.metadata[this.idKey])) {\n        ids.push(doc.metadata[this.idKey]);\n      }\n    }\n    const docs = await this.docstore.mget(ids);\n    return docs\n      .filter((doc) => doc !== undefined)\n      .slice(0, this.parentK) as Document[];\n  }\n}\n"],"mappings":";;;;;;;;;;;;;;;;;;;;;;;;;AAwCA,IAAa,uBAAb,cAA0CA,0CAAc;CACtD,OAAO,UAAU;AACf,SAAO;CACR;CAED,eAAe;EAAC;EAAa;EAAc;CAAe;CAE1D,AAAO;CAEP,AAAO;CAEP,AAAU;CAEV,AAAU;CAEV,AAAU;CAEV,YAAYC,MAAiC;EAC3C,MAAM,KAAK;EACX,KAAK,cAAc,KAAK;AACxB,MAAI,KAAK,WACP,KAAK,WAAWC,gEAAiC,KAAK,UAAU;WACvD,KAAK,UACd,KAAK,WAAW,KAAK;MAErB,OAAM,IAAI,MACR;EAGJ,KAAK,QAAQ,KAAK,SAAS;EAC3B,KAAK,SAAS,KAAK;EACnB,KAAK,UAAU,KAAK;CACrB;CAED,MAAM,sBAAsBC,OAAoC;EAC9D,MAAM,UAAU,MAAM,KAAK,YAAY,iBAAiB,OAAO,KAAK,OAAO;EAC3E,MAAMC,MAAgB,CAAE;AACxB,OAAK,MAAM,OAAO,QAChB,KAAI,IAAI,SAAS,KAAK,UAAU,CAAC,IAAI,SAAS,IAAI,SAAS,KAAK,OAAO,EACrE,IAAI,KAAK,IAAI,SAAS,KAAK,OAAO;EAGtC,MAAM,OAAO,MAAM,KAAK,SAAS,KAAK,IAAI;AAC1C,SAAO,KACJ,OAAO,CAAC,QAAQ,QAAQ,OAAU,CAClC,MAAM,GAAG,KAAK,QAAQ;CAC1B;AACF"}

package/dist/retrievers/multi_vector.d.cts

@@ -0,0 +1,51 @@
+import { VectorStoreInterface } from "@langchain/core/vectorstores";
+import { BaseRetriever, BaseRetrieverInput } from "@langchain/core/retrievers";
+import { Document } from "@langchain/core/documents";
+import { BaseStore, BaseStoreInterface } from "@langchain/core/stores";
+
+//#region src/retrievers/multi_vector.d.ts
+
+/**
+ * Arguments for the MultiVectorRetriever class.
+ */
+interface MultiVectorRetrieverInput extends BaseRetrieverInput {
+  vectorstore: VectorStoreInterface;
+  /** @deprecated Prefer `byteStore`. */
+  docstore?: BaseStoreInterface<string, Document>;
+  byteStore?: BaseStore<string, Uint8Array>;
+  idKey?: string;
+  childK?: number;
+  parentK?: number;
+}
+/**
+ * A retriever that retrieves documents from a vector store and a document
+ * store. It uses the vector store to find relevant documents based on a
+ * query, and then retrieves the full documents from the document store.
+ * @example
+ * ```typescript
+ * const retriever = new MultiVectorRetriever({
+ *   vectorstore: new FaissStore(),
+ *   byteStore: new InMemoryStore<Unit8Array>(),
+ *   idKey: "doc_id",
+ *   childK: 20,
+ *   parentK: 5,
+ * });
+ *
+ * const retrieverResult = await retriever.invoke("justice breyer");
+ * console.log(retrieverResult[0].pageContent.length);
+ * ```
+ */
+declare class MultiVectorRetriever extends BaseRetriever {
+  static lc_name(): string;
+  lc_namespace: string[];
+  vectorstore: VectorStoreInterface;
+  docstore: BaseStoreInterface<string, Document>;
+  protected idKey: string;
+  protected childK?: number;
+  protected parentK?: number;
+  constructor(args: MultiVectorRetrieverInput);
+  _getRelevantDocuments(query: string): Promise<Document[]>;
+}
+//#endregion
+export { MultiVectorRetriever, MultiVectorRetrieverInput };
+//# sourceMappingURL=multi_vector.d.cts.map

package/dist/retrievers/multi_vector.d.cts.map

@@ -0,0 +1 @@
+{"version":3,"file":"multi_vector.d.cts","names":["BaseRetriever","BaseRetrieverInput","VectorStoreInterface","Document","BaseStore","BaseStoreInterface","MultiVectorRetrieverInput","Uint8Array","MultiVectorRetriever","Promise"],"sources":["../../src/retrievers/multi_vector.d.ts"],"sourcesContent":["import { BaseRetriever, type BaseRetrieverInput } from \"@langchain/core/retrievers\";\nimport type { VectorStoreInterface } from \"@langchain/core/vectorstores\";\nimport { Document } from \"@langchain/core/documents\";\nimport { BaseStore, type BaseStoreInterface } from \"@langchain/core/stores\";\n/**\n * Arguments for the MultiVectorRetriever class.\n */\nexport interface MultiVectorRetrieverInput extends BaseRetrieverInput {\n    vectorstore: VectorStoreInterface;\n    /** @deprecated Prefer `byteStore`. */\n    docstore?: BaseStoreInterface<string, Document>;\n    byteStore?: BaseStore<string, Uint8Array>;\n    idKey?: string;\n    childK?: number;\n    parentK?: number;\n}\n/**\n * A retriever that retrieves documents from a vector store and a document\n * store. It uses the vector store to find relevant documents based on a\n * query, and then retrieves the full documents from the document store.\n * @example\n * ```typescript\n * const retriever = new MultiVectorRetriever({\n *   vectorstore: new FaissStore(),\n *   byteStore: new InMemoryStore<Unit8Array>(),\n *   idKey: \"doc_id\",\n *   childK: 20,\n *   parentK: 5,\n * });\n *\n * const retrieverResult = await retriever.invoke(\"justice breyer\");\n * console.log(retrieverResult[0].pageContent.length);\n * ```\n */\nexport declare class MultiVectorRetriever extends BaseRetriever {\n    static lc_name(): string;\n    lc_namespace: string[];\n    vectorstore: VectorStoreInterface;\n    docstore: BaseStoreInterface<string, Document>;\n    protected idKey: string;\n    protected childK?: number;\n    protected parentK?: number;\n    constructor(args: MultiVectorRetrieverInput);\n    _getRelevantDocuments(query: string): Promise<Document[]>;\n}\n"],"mappings":";;;;;;;;;AAOA;AAA0C,UAAzBM,yBAAAA,SAAkCL,kBAAT,CAAA;EAAA,WACzBC,EAAAA,oBAAAA;EAAoB;EAEa,QAAnCG,CAAAA,EAAAA,kBAAAA,CAAAA,MAAAA,EAA2BF,QAA3BE,CAAAA;EAAkB,SACCE,CAAAA,EAAlBH,SAAkBG,CAAAA,MAAAA,EAAAA,UAAAA,CAAAA;EAAU,KAA5BH,CAAAA,EAAAA,MAAAA;EAAS,MAJ0BH,CAAAA,EAAAA,MAAAA;EAAkB,OAAA,CAAA,EAAA,MAAA;AA2BrE;;;;;;;;;AAA+D;;;;;;;;;;cAA1CO,oBAAAA,SAA6BR,aAAAA;;;eAGjCE;YACHG,2BAA2BF;;;;oBAInBG;wCACoBG,QAAQN"}

package/dist/retrievers/multi_vector.d.ts

@@ -0,0 +1,51 @@
+import { Document } from "@langchain/core/documents";
+import { BaseStore, BaseStoreInterface } from "@langchain/core/stores";
+import { VectorStoreInterface } from "@langchain/core/vectorstores";
+import { BaseRetriever, BaseRetrieverInput } from "@langchain/core/retrievers";
+
+//#region src/retrievers/multi_vector.d.ts
+
+/**
+ * Arguments for the MultiVectorRetriever class.
+ */
+interface MultiVectorRetrieverInput extends BaseRetrieverInput {
+  vectorstore: VectorStoreInterface;
+  /** @deprecated Prefer `byteStore`. */
+  docstore?: BaseStoreInterface<string, Document>;
+  byteStore?: BaseStore<string, Uint8Array>;
+  idKey?: string;
+  childK?: number;
+  parentK?: number;
+}
+/**
+ * A retriever that retrieves documents from a vector store and a document
+ * store. It uses the vector store to find relevant documents based on a
+ * query, and then retrieves the full documents from the document store.
+ * @example
+ * ```typescript
+ * const retriever = new MultiVectorRetriever({
+ *   vectorstore: new FaissStore(),
+ *   byteStore: new InMemoryStore<Unit8Array>(),
+ *   idKey: "doc_id",
+ *   childK: 20,
+ *   parentK: 5,
+ * });
+ *
+ * const retrieverResult = await retriever.invoke("justice breyer");
+ * console.log(retrieverResult[0].pageContent.length);
+ * ```
+ */
+declare class MultiVectorRetriever extends BaseRetriever {
+  static lc_name(): string;
+  lc_namespace: string[];
+  vectorstore: VectorStoreInterface;
+  docstore: BaseStoreInterface<string, Document>;
+  protected idKey: string;
+  protected childK?: number;
+  protected parentK?: number;
+  constructor(args: MultiVectorRetrieverInput);
+  _getRelevantDocuments(query: string): Promise<Document[]>;
+}
+//#endregion
+export { MultiVectorRetriever, MultiVectorRetrieverInput };
+//# sourceMappingURL=multi_vector.d.ts.map

package/dist/retrievers/multi_vector.d.ts.map

@@ -0,0 +1 @@
+{"version":3,"file":"multi_vector.d.ts","names":["BaseRetriever","BaseRetrieverInput","VectorStoreInterface","Document","BaseStore","BaseStoreInterface","MultiVectorRetrieverInput","Uint8Array","MultiVectorRetriever","Promise"],"sources":["../../src/retrievers/multi_vector.d.ts"],"sourcesContent":["import { BaseRetriever, type BaseRetrieverInput } from \"@langchain/core/retrievers\";\nimport type { VectorStoreInterface } from \"@langchain/core/vectorstores\";\nimport { Document } from \"@langchain/core/documents\";\nimport { BaseStore, type BaseStoreInterface } from \"@langchain/core/stores\";\n/**\n * Arguments for the MultiVectorRetriever class.\n */\nexport interface MultiVectorRetrieverInput extends BaseRetrieverInput {\n    vectorstore: VectorStoreInterface;\n    /** @deprecated Prefer `byteStore`. */\n    docstore?: BaseStoreInterface<string, Document>;\n    byteStore?: BaseStore<string, Uint8Array>;\n    idKey?: string;\n    childK?: number;\n    parentK?: number;\n}\n/**\n * A retriever that retrieves documents from a vector store and a document\n * store. It uses the vector store to find relevant documents based on a\n * query, and then retrieves the full documents from the document store.\n * @example\n * ```typescript\n * const retriever = new MultiVectorRetriever({\n *   vectorstore: new FaissStore(),\n *   byteStore: new InMemoryStore<Unit8Array>(),\n *   idKey: \"doc_id\",\n *   childK: 20,\n *   parentK: 5,\n * });\n *\n * const retrieverResult = await retriever.invoke(\"justice breyer\");\n * console.log(retrieverResult[0].pageContent.length);\n * ```\n */\nexport declare class MultiVectorRetriever extends BaseRetriever {\n    static lc_name(): string;\n    lc_namespace: string[];\n    vectorstore: VectorStoreInterface;\n    docstore: BaseStoreInterface<string, Document>;\n    protected idKey: string;\n    protected childK?: number;\n    protected parentK?: number;\n    constructor(args: MultiVectorRetrieverInput);\n    _getRelevantDocuments(query: string): Promise<Document[]>;\n}\n"],"mappings":";;;;;;;;;AAOA;AAA0C,UAAzBM,yBAAAA,SAAkCL,kBAAT,CAAA;EAAA,WACzBC,EAAAA,oBAAAA;EAAoB;EAEa,QAAnCG,CAAAA,EAAAA,kBAAAA,CAAAA,MAAAA,EAA2BF,QAA3BE,CAAAA;EAAkB,SACCE,CAAAA,EAAlBH,SAAkBG,CAAAA,MAAAA,EAAAA,UAAAA,CAAAA;EAAU,KAA5BH,CAAAA,EAAAA,MAAAA;EAAS,MAJ0BH,CAAAA,EAAAA,MAAAA;EAAkB,OAAA,CAAA,EAAA,MAAA;AA2BrE;;;;;;;;;AAA+D;;;;;;;;;;cAA1CO,oBAAAA,SAA6BR,aAAAA;;;eAGjCE;YACHG,2BAA2BF;;;;oBAInBG;wCACoBG,QAAQN"}

package/dist/retrievers/multi_vector.js

@@ -0,0 +1,61 @@
+import { __export } from "../_virtual/rolldown_runtime.js";
+import { createDocumentStoreFromByteStore } from "../storage/encoder_backed.js";
+import { BaseRetriever } from "@langchain/core/retrievers";
+
+//#region src/retrievers/multi_vector.ts
+var multi_vector_exports = {};
+__export(multi_vector_exports, { MultiVectorRetriever: () => MultiVectorRetriever });
+/**
+* A retriever that retrieves documents from a vector store and a document
+* store. It uses the vector store to find relevant documents based on a
+* query, and then retrieves the full documents from the document store.
+* @example
+* ```typescript
+* const retriever = new MultiVectorRetriever({
+*   vectorstore: new FaissStore(),
+*   byteStore: new InMemoryStore<Unit8Array>(),
+*   idKey: "doc_id",
+*   childK: 20,
+*   parentK: 5,
+* });
+*
+* const retrieverResult = await retriever.invoke("justice breyer");
+* console.log(retrieverResult[0].pageContent.length);
+* ```
+*/
+var MultiVectorRetriever = class extends BaseRetriever {
+	static lc_name() {
+		return "MultiVectorRetriever";
+	}
+	lc_namespace = [
+		"langchain",
+		"retrievers",
+		"multi_vector"
+	];
+	vectorstore;
+	docstore;
+	idKey;
+	childK;
+	parentK;
+	constructor(args) {
+		super(args);
+		this.vectorstore = args.vectorstore;
+		if (args.byteStore) this.docstore = createDocumentStoreFromByteStore(args.byteStore);
+		else if (args.docstore) this.docstore = args.docstore;
+		else throw new Error("byteStore and docstore are undefined. Please provide at least one.");
+		this.idKey = args.idKey ?? "doc_id";
+		this.childK = args.childK;
+		this.parentK = args.parentK;
+	}
+	async _getRelevantDocuments(query) {
+		const subDocs = await this.vectorstore.similaritySearch(query, this.childK);
+		const ids = [];
+		for (const doc of subDocs) if (doc.metadata[this.idKey] && !ids.includes(doc.metadata[this.idKey])) ids.push(doc.metadata[this.idKey]);
+		const docs = await this.docstore.mget(ids);
+		return docs.filter((doc) => doc !== void 0).slice(0, this.parentK);
+	}
+};
+
+//#endregion
+export { MultiVectorRetriever, multi_vector_exports };
+//# sourceMappingURL=multi_vector.js.map

package/dist/retrievers/multi_vector.js.map

@@ -0,0 +1 @@
+{"version":3,"file":"multi_vector.js","names":["args: MultiVectorRetrieverInput","query: string","ids: string[]"],"sources":["../../src/retrievers/multi_vector.ts"],"sourcesContent":["import {\n  BaseRetriever,\n  type BaseRetrieverInput,\n} from \"@langchain/core/retrievers\";\nimport type { VectorStoreInterface } from \"@langchain/core/vectorstores\";\nimport { Document } from \"@langchain/core/documents\";\nimport { BaseStore, type BaseStoreInterface } from \"@langchain/core/stores\";\nimport { createDocumentStoreFromByteStore } from \"../storage/encoder_backed.js\";\n\n/**\n * Arguments for the MultiVectorRetriever class.\n */\nexport interface MultiVectorRetrieverInput extends BaseRetrieverInput {\n  vectorstore: VectorStoreInterface;\n  /** @deprecated Prefer `byteStore`. */\n  docstore?: BaseStoreInterface<string, Document>;\n  byteStore?: BaseStore<string, Uint8Array>;\n  idKey?: string;\n  childK?: number;\n  parentK?: number;\n}\n\n/**\n * A retriever that retrieves documents from a vector store and a document\n * store. It uses the vector store to find relevant documents based on a\n * query, and then retrieves the full documents from the document store.\n * @example\n * ```typescript\n * const retriever = new MultiVectorRetriever({\n *   vectorstore: new FaissStore(),\n *   byteStore: new InMemoryStore<Unit8Array>(),\n *   idKey: \"doc_id\",\n *   childK: 20,\n *   parentK: 5,\n * });\n *\n * const retrieverResult = await retriever.invoke(\"justice breyer\");\n * console.log(retrieverResult[0].pageContent.length);\n * ```\n */\nexport class MultiVectorRetriever extends BaseRetriever {\n  static lc_name() {\n    return \"MultiVectorRetriever\";\n  }\n\n  lc_namespace = [\"langchain\", \"retrievers\", \"multi_vector\"];\n\n  public vectorstore: VectorStoreInterface;\n\n  public docstore: BaseStoreInterface<string, Document>;\n\n  protected idKey: string;\n\n  protected childK?: number;\n\n  protected parentK?: number;\n\n  constructor(args: MultiVectorRetrieverInput) {\n    super(args);\n    this.vectorstore = args.vectorstore;\n    if (args.byteStore) {\n      this.docstore = createDocumentStoreFromByteStore(args.byteStore);\n    } else if (args.docstore) {\n      this.docstore = args.docstore;\n    } else {\n      throw new Error(\n        \"byteStore and docstore are undefined. Please provide at least one.\"\n      );\n    }\n    this.idKey = args.idKey ?? \"doc_id\";\n    this.childK = args.childK;\n    this.parentK = args.parentK;\n  }\n\n  async _getRelevantDocuments(query: string): Promise<Document[]> {\n    const subDocs = await this.vectorstore.similaritySearch(query, this.childK);\n    const ids: string[] = [];\n    for (const doc of subDocs) {\n      if (doc.metadata[this.idKey] && !ids.includes(doc.metadata[this.idKey])) {\n        ids.push(doc.metadata[this.idKey]);\n      }\n    }\n    const docs = await this.docstore.mget(ids);\n    return docs\n      .filter((doc) => doc !== undefined)\n      .slice(0, this.parentK) as Document[];\n  }\n}\n"],"mappings":";;;;;;;;;;;;;;;;;;;;;;;;;AAwCA,IAAa,uBAAb,cAA0C,cAAc;CACtD,OAAO,UAAU;AACf,SAAO;CACR;CAED,eAAe;EAAC;EAAa;EAAc;CAAe;CAE1D,AAAO;CAEP,AAAO;CAEP,AAAU;CAEV,AAAU;CAEV,AAAU;CAEV,YAAYA,MAAiC;EAC3C,MAAM,KAAK;EACX,KAAK,cAAc,KAAK;AACxB,MAAI,KAAK,WACP,KAAK,WAAW,iCAAiC,KAAK,UAAU;WACvD,KAAK,UACd,KAAK,WAAW,KAAK;MAErB,OAAM,IAAI,MACR;EAGJ,KAAK,QAAQ,KAAK,SAAS;EAC3B,KAAK,SAAS,KAAK;EACnB,KAAK,UAAU,KAAK;CACrB;CAED,MAAM,sBAAsBC,OAAoC;EAC9D,MAAM,UAAU,MAAM,KAAK,YAAY,iBAAiB,OAAO,KAAK,OAAO;EAC3E,MAAMC,MAAgB,CAAE;AACxB,OAAK,MAAM,OAAO,QAChB,KAAI,IAAI,SAAS,KAAK,UAAU,CAAC,IAAI,SAAS,IAAI,SAAS,KAAK,OAAO,EACrE,IAAI,KAAK,IAAI,SAAS,KAAK,OAAO;EAGtC,MAAM,OAAO,MAAM,KAAK,SAAS,KAAK,IAAI;AAC1C,SAAO,KACJ,OAAO,CAAC,QAAQ,QAAQ,OAAU,CAClC,MAAM,GAAG,KAAK,QAAQ;CAC1B;AACF"}

package/dist/retrievers/parent_document.cjs

@@ -0,0 +1,137 @@
+const require_rolldown_runtime = require('../_virtual/rolldown_runtime.cjs');
+const require_retrievers_multi_vector = require('./multi_vector.cjs');
+const __langchain_core_documents = require_rolldown_runtime.__toESM(require("@langchain/core/documents"));
+const uuid = require_rolldown_runtime.__toESM(require("uuid"));
+
+//#region src/retrievers/parent_document.ts
+var parent_document_exports = {};
+require_rolldown_runtime.__export(parent_document_exports, { ParentDocumentRetriever: () => ParentDocumentRetriever });
+/**
+* A type of document retriever that splits input documents into smaller chunks
+* while separately storing and preserving the original documents.
+* The small chunks are embedded, then on retrieval, the original
+* "parent" documents are retrieved.
+*
+* This strikes a balance between better targeted retrieval with small documents
+* and the more context-rich larger documents.
+* @example
+* ```typescript
+* const retriever = new ParentDocumentRetriever({
+*   vectorstore: new MemoryVectorStore(new OpenAIEmbeddings()),
+*   byteStore: new InMemoryStore<Uint8Array>(),
+*   parentSplitter: new RecursiveCharacterTextSplitter({
+*     chunkOverlap: 0,
+*     chunkSize: 500,
+*   }),
+*   childSplitter: new RecursiveCharacterTextSplitter({
+*     chunkOverlap: 0,
+*     chunkSize: 50,
+*   }),
+*   childK: 20,
+*   parentK: 5,
+* });
+*
+* const parentDocuments = await getDocuments();
+* await retriever.addDocuments(parentDocuments);
+* const retrievedDocs = await retriever.invoke("justice breyer");
+* ```
+*/
+var ParentDocumentRetriever = class extends require_retrievers_multi_vector.MultiVectorRetriever {
+	static lc_name() {
+		return "ParentDocumentRetriever";
+	}
+	lc_namespace = [
+		"langchain",
+		"retrievers",
+		"parent_document"
+	];
+	vectorstore;
+	childSplitter;
+	parentSplitter;
+	idKey = "doc_id";
+	childK;
+	parentK;
+	childDocumentRetriever;
+	documentCompressor;
+	documentCompressorFilteringFn;
+	constructor(fields) {
+		super(fields);
+		this.vectorstore = fields.vectorstore;
+		this.childSplitter = fields.childSplitter;
+		this.parentSplitter = fields.parentSplitter;
+		this.idKey = fields.idKey ?? this.idKey;
+		this.childK = fields.childK;
+		this.parentK = fields.parentK;
+		this.childDocumentRetriever = fields.childDocumentRetriever;
+		this.documentCompressor = fields.documentCompressor;
+		this.documentCompressorFilteringFn = fields.documentCompressorFilteringFn;
+	}
+	async _getRelevantDocuments(query) {
+		let subDocs = [];
+		if (this.childDocumentRetriever) subDocs = await this.childDocumentRetriever.invoke(query);
+		else subDocs = await this.vectorstore.similaritySearch(query, this.childK);
+		if (this.documentCompressor && subDocs.length) {
+			subDocs = await this.documentCompressor.compressDocuments(subDocs, query);
+			if (this.documentCompressorFilteringFn) subDocs = this.documentCompressorFilteringFn(subDocs);
+		}
+		const parentDocIds = [];
+		for (const doc of subDocs) if (!parentDocIds.includes(doc.metadata[this.idKey])) parentDocIds.push(doc.metadata[this.idKey]);
+		const parentDocs = [];
+		const storedParentDocs = await this.docstore.mget(parentDocIds);
+		const retrievedDocs = storedParentDocs.filter((doc) => doc !== void 0);
+		parentDocs.push(...retrievedDocs);
+		return parentDocs.slice(0, this.parentK);
+	}
+	async _storeDocuments(parentDoc, childDocs, addToDocstore) {
+		if (this.childDocumentRetriever) await this.childDocumentRetriever.addDocuments(childDocs);
+		else await this.vectorstore.addDocuments(childDocs);
+		if (addToDocstore) await this.docstore.mset(Object.entries(parentDoc));
+	}
+	/**
+	* Adds documents to the docstore and vectorstores.
+	* If a retriever is provided, it will be used to add documents instead of the vectorstore.
+	* @param docs The documents to add
+	* @param config.ids Optional list of ids for documents. If provided should be the same
+	*   length as the list of documents. Can provided if parent documents
+	*   are already in the document store and you don't want to re-add
+	*   to the docstore. If not provided, random UUIDs will be used as ids.
+	* @param config.addToDocstore Boolean of whether to add documents to docstore.
+	* This can be false if and only if `ids` are provided. You may want
+	*   to set this to False if the documents are already in the docstore
+	*   and you don't want to re-add them.
+	* @param config.chunkHeaderOptions Object with options for adding Contextual chunk headers
+	*/
+	async addDocuments(docs, config) {
+		const { ids, addToDocstore = true, childDocChunkHeaderOptions = {} } = config ?? {};
+		const parentDocs = this.parentSplitter ? await this.parentSplitter.splitDocuments(docs) : docs;
+		let parentDocIds;
+		if (ids === void 0) {
+			if (!addToDocstore) throw new Error(`If ids are not passed in, "config.addToDocstore" MUST be true`);
+			parentDocIds = parentDocs.map((_doc) => uuid.v4());
+		} else parentDocIds = ids;
+		if (parentDocs.length !== parentDocIds.length) throw new Error(`Got uneven list of documents and ids.\nIf "ids" is provided, should be same length as "documents".`);
+		for (let i = 0; i < parentDocs.length; i += 1) {
+			const parentDoc = parentDocs[i];
+			const parentDocId = parentDocIds[i];
+			const subDocs = await this.childSplitter.splitDocuments([parentDoc], childDocChunkHeaderOptions);
+			const taggedSubDocs = subDocs.map((subDoc) => new __langchain_core_documents.Document({
+				pageContent: subDoc.pageContent,
+				metadata: {
+					...subDoc.metadata,
+					[this.idKey]: parentDocId
+				}
+			}));
+			await this._storeDocuments({ [parentDocId]: parentDoc }, taggedSubDocs, addToDocstore);
+		}
+	}
+};
+
+//#endregion
+exports.ParentDocumentRetriever = ParentDocumentRetriever;
+Object.defineProperty(exports, 'parent_document_exports', {
+  enumerable: true,
+  get: function () {
+    return parent_document_exports;
+  }
+});
+//# sourceMappingURL=parent_document.cjs.map

package/dist/retrievers/parent_document.cjs.map

@@ -0,0 +1 @@
+{"version":3,"file":"parent_document.cjs","names":["MultiVectorRetriever","fields: ParentDocumentRetrieverFields","query: string","subDocs: SubDocs","parentDocIds: string[]","parentDocs: Document[]","retrievedDocs: Document[]","doc?: Document","parentDoc: Record<string, Document>","childDocs: Document[]","addToDocstore: boolean","docs: Document[]","config?: {\n      ids?: string[];\n      addToDocstore?: boolean;\n      childDocChunkHeaderOptions?: TextSplitterChunkHeaderOptions;\n    }","_doc: Document","subDoc: Document","Document"],"sources":["../../src/retrievers/parent_document.ts"],"sourcesContent":["import * as uuid from \"uuid\";\n\nimport {\n  type VectorStoreInterface,\n  type VectorStoreRetrieverInterface,\n} from \"@langchain/core/vectorstores\";\nimport { Document } from \"@langchain/core/documents\";\nimport {\n  TextSplitter,\n  TextSplitterChunkHeaderOptions,\n} from \"@langchain/textsplitters\";\nimport type { BaseDocumentCompressor } from \"./document_compressors/index.js\";\nimport {\n  MultiVectorRetriever,\n  type MultiVectorRetrieverInput,\n} from \"./multi_vector.js\";\n\n// eslint-disable-next-line @typescript-eslint/no-explicit-any\nexport type SubDocs = Document<Record<string, any>>[];\n\n/**\n * Interface for the fields required to initialize a\n * ParentDocumentRetriever instance.\n */\nexport type ParentDocumentRetrieverFields = MultiVectorRetrieverInput & {\n  childSplitter: TextSplitter;\n  parentSplitter?: TextSplitter;\n  /**\n   * A custom retriever to use when retrieving instead of\n   * the `.similaritySearch` method of the vectorstore.\n   */\n  childDocumentRetriever?: VectorStoreRetrieverInterface<VectorStoreInterface>;\n  documentCompressor?: BaseDocumentCompressor | undefined;\n  documentCompressorFilteringFn?: (docs: SubDocs) => SubDocs;\n};\n\n/**\n * A type of document retriever that splits input documents into smaller chunks\n * while separately storing and preserving the original documents.\n * The small chunks are embedded, then on retrieval, the original\n * \"parent\" documents are retrieved.\n *\n * This strikes a balance between better targeted retrieval with small documents\n * and the more context-rich larger documents.\n * @example\n * ```typescript\n * const retriever = new ParentDocumentRetriever({\n *   vectorstore: new MemoryVectorStore(new OpenAIEmbeddings()),\n *   byteStore: new InMemoryStore<Uint8Array>(),\n *   parentSplitter: new RecursiveCharacterTextSplitter({\n *     chunkOverlap: 0,\n *     chunkSize: 500,\n *   }),\n *   childSplitter: new RecursiveCharacterTextSplitter({\n *     chunkOverlap: 0,\n *     chunkSize: 50,\n *   }),\n *   childK: 20,\n *   parentK: 5,\n * });\n *\n * const parentDocuments = await getDocuments();\n * await retriever.addDocuments(parentDocuments);\n * const retrievedDocs = await retriever.invoke(\"justice breyer\");\n * ```\n */\nexport class ParentDocumentRetriever extends MultiVectorRetriever {\n  static lc_name() {\n    return \"ParentDocumentRetriever\";\n  }\n\n  lc_namespace = [\"langchain\", \"retrievers\", \"parent_document\"];\n\n  vectorstore: VectorStoreInterface;\n\n  protected childSplitter: TextSplitter;\n\n  protected parentSplitter?: TextSplitter;\n\n  protected idKey = \"doc_id\";\n\n  protected childK?: number;\n\n  protected parentK?: number;\n\n  childDocumentRetriever:\n    | VectorStoreRetrieverInterface<VectorStoreInterface>\n    | undefined;\n\n  documentCompressor: BaseDocumentCompressor | undefined;\n\n  documentCompressorFilteringFn?: ParentDocumentRetrieverFields[\"documentCompressorFilteringFn\"];\n\n  constructor(fields: ParentDocumentRetrieverFields) {\n    super(fields);\n    this.vectorstore = fields.vectorstore;\n    this.childSplitter = fields.childSplitter;\n    this.parentSplitter = fields.parentSplitter;\n    this.idKey = fields.idKey ?? this.idKey;\n    this.childK = fields.childK;\n    this.parentK = fields.parentK;\n    this.childDocumentRetriever = fields.childDocumentRetriever;\n    this.documentCompressor = fields.documentCompressor;\n    this.documentCompressorFilteringFn = fields.documentCompressorFilteringFn;\n  }\n\n  async _getRelevantDocuments(query: string): Promise<Document[]> {\n    let subDocs: SubDocs = [];\n    if (this.childDocumentRetriever) {\n      subDocs = await this.childDocumentRetriever.invoke(query);\n    } else {\n      subDocs = await this.vectorstore.similaritySearch(query, this.childK);\n    }\n\n    if (this.documentCompressor && subDocs.length) {\n      subDocs = await this.documentCompressor.compressDocuments(subDocs, query);\n      if (this.documentCompressorFilteringFn) {\n        subDocs = this.documentCompressorFilteringFn(subDocs);\n      }\n    }\n\n    // Maintain order\n    const parentDocIds: string[] = [];\n    for (const doc of subDocs) {\n      if (!parentDocIds.includes(doc.metadata[this.idKey])) {\n        parentDocIds.push(doc.metadata[this.idKey]);\n      }\n    }\n    const parentDocs: Document[] = [];\n    const storedParentDocs = await this.docstore.mget(parentDocIds);\n    const retrievedDocs: Document[] = storedParentDocs.filter(\n      (doc?: Document): doc is Document => doc !== undefined\n    );\n    parentDocs.push(...retrievedDocs);\n    return parentDocs.slice(0, this.parentK);\n  }\n\n  async _storeDocuments(\n    parentDoc: Record<string, Document>,\n    childDocs: Document[],\n    addToDocstore: boolean\n  ) {\n    if (this.childDocumentRetriever) {\n      await this.childDocumentRetriever.addDocuments(childDocs);\n    } else {\n      await this.vectorstore.addDocuments(childDocs);\n    }\n    if (addToDocstore) {\n      await this.docstore.mset(Object.entries(parentDoc));\n    }\n  }\n\n  /**\n   * Adds documents to the docstore and vectorstores.\n   * If a retriever is provided, it will be used to add documents instead of the vectorstore.\n   * @param docs The documents to add\n   * @param config.ids Optional list of ids for documents. If provided should be the same\n   *   length as the list of documents. Can provided if parent documents\n   *   are already in the document store and you don't want to re-add\n   *   to the docstore. If not provided, random UUIDs will be used as ids.\n   * @param config.addToDocstore Boolean of whether to add documents to docstore.\n   * This can be false if and only if `ids` are provided. You may want\n   *   to set this to False if the documents are already in the docstore\n   *   and you don't want to re-add them.\n   * @param config.chunkHeaderOptions Object with options for adding Contextual chunk headers\n   */\n  async addDocuments(\n    docs: Document[],\n    config?: {\n      ids?: string[];\n      addToDocstore?: boolean;\n      childDocChunkHeaderOptions?: TextSplitterChunkHeaderOptions;\n    }\n  ): Promise<void> {\n    const {\n      ids,\n      addToDocstore = true,\n      childDocChunkHeaderOptions = {},\n    } = config ?? {};\n    const parentDocs = this.parentSplitter\n      ? await this.parentSplitter.splitDocuments(docs)\n      : docs;\n    let parentDocIds;\n    if (ids === undefined) {\n      if (!addToDocstore) {\n        throw new Error(\n          `If ids are not passed in, \"config.addToDocstore\" MUST be true`\n        );\n      }\n      parentDocIds = parentDocs.map((_doc: Document) => uuid.v4());\n    } else {\n      parentDocIds = ids;\n    }\n    if (parentDocs.length !== parentDocIds.length) {\n      throw new Error(\n        `Got uneven list of documents and ids.\\nIf \"ids\" is provided, should be same length as \"documents\".`\n      );\n    }\n    for (let i = 0; i < parentDocs.length; i += 1) {\n      const parentDoc = parentDocs[i];\n      const parentDocId = parentDocIds[i];\n      const subDocs = await this.childSplitter.splitDocuments(\n        [parentDoc],\n        childDocChunkHeaderOptions\n      );\n      const taggedSubDocs = subDocs.map(\n        (subDoc: Document) =>\n          new Document({\n            pageContent: subDoc.pageContent,\n            metadata: { ...subDoc.metadata, [this.idKey]: parentDocId },\n          })\n      );\n      await this._storeDocuments(\n        { [parentDocId]: parentDoc },\n        taggedSubDocs,\n        addToDocstore\n      );\n    }\n  }\n}\n"],"mappings":";;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;AAkEA,IAAa,0BAAb,cAA6CA,qDAAqB;CAChE,OAAO,UAAU;AACf,SAAO;CACR;CAED,eAAe;EAAC;EAAa;EAAc;CAAkB;CAE7D;CAEA,AAAU;CAEV,AAAU;CAEV,AAAU,QAAQ;CAElB,AAAU;CAEV,AAAU;CAEV;CAIA;CAEA;CAEA,YAAYC,QAAuC;EACjD,MAAM,OAAO;EACb,KAAK,cAAc,OAAO;EAC1B,KAAK,gBAAgB,OAAO;EAC5B,KAAK,iBAAiB,OAAO;EAC7B,KAAK,QAAQ,OAAO,SAAS,KAAK;EAClC,KAAK,SAAS,OAAO;EACrB,KAAK,UAAU,OAAO;EACtB,KAAK,yBAAyB,OAAO;EACrC,KAAK,qBAAqB,OAAO;EACjC,KAAK,gCAAgC,OAAO;CAC7C;CAED,MAAM,sBAAsBC,OAAoC;EAC9D,IAAIC,UAAmB,CAAE;AACzB,MAAI,KAAK,wBACP,UAAU,MAAM,KAAK,uBAAuB,OAAO,MAAM;OAEzD,UAAU,MAAM,KAAK,YAAY,iBAAiB,OAAO,KAAK,OAAO;AAGvE,MAAI,KAAK,sBAAsB,QAAQ,QAAQ;GAC7C,UAAU,MAAM,KAAK,mBAAmB,kBAAkB,SAAS,MAAM;AACzE,OAAI,KAAK,+BACP,UAAU,KAAK,8BAA8B,QAAQ;EAExD;EAGD,MAAMC,eAAyB,CAAE;AACjC,OAAK,MAAM,OAAO,QAChB,KAAI,CAAC,aAAa,SAAS,IAAI,SAAS,KAAK,OAAO,EAClD,aAAa,KAAK,IAAI,SAAS,KAAK,OAAO;EAG/C,MAAMC,aAAyB,CAAE;EACjC,MAAM,mBAAmB,MAAM,KAAK,SAAS,KAAK,aAAa;EAC/D,MAAMC,gBAA4B,iBAAiB,OACjD,CAACC,QAAoC,QAAQ,OAC9C;EACD,WAAW,KAAK,GAAG,cAAc;AACjC,SAAO,WAAW,MAAM,GAAG,KAAK,QAAQ;CACzC;CAED,MAAM,gBACJC,WACAC,WACAC,eACA;AACA,MAAI,KAAK,wBACP,MAAM,KAAK,uBAAuB,aAAa,UAAU;OAEzD,MAAM,KAAK,YAAY,aAAa,UAAU;AAEhD,MAAI,eACF,MAAM,KAAK,SAAS,KAAK,OAAO,QAAQ,UAAU,CAAC;CAEtD;;;;;;;;;;;;;;;CAgBD,MAAM,aACJC,MACAC,QAKe;EACf,MAAM,EACJ,KACA,gBAAgB,MAChB,6BAA6B,CAAE,GAChC,GAAG,UAAU,CAAE;EAChB,MAAM,aAAa,KAAK,iBACpB,MAAM,KAAK,eAAe,eAAe,KAAK,GAC9C;EACJ,IAAI;AACJ,MAAI,QAAQ,QAAW;AACrB,OAAI,CAAC,cACH,OAAM,IAAI,MACR,CAAC,6DAA6D,CAAC;GAGnE,eAAe,WAAW,IAAI,CAACC,SAAmB,KAAK,IAAI,CAAC;EAC7D,OACC,eAAe;AAEjB,MAAI,WAAW,WAAW,aAAa,OACrC,OAAM,IAAI,MACR,CAAC,kGAAkG,CAAC;AAGxG,OAAK,IAAI,IAAI,GAAG,IAAI,WAAW,QAAQ,KAAK,GAAG;GAC7C,MAAM,YAAY,WAAW;GAC7B,MAAM,cAAc,aAAa;GACjC,MAAM,UAAU,MAAM,KAAK,cAAc,eACvC,CAAC,SAAU,GACX,2BACD;GACD,MAAM,gBAAgB,QAAQ,IAC5B,CAACC,WACC,IAAIC,oCAAS;IACX,aAAa,OAAO;IACpB,UAAU;KAAE,GAAG,OAAO;MAAW,KAAK,QAAQ;IAAa;GAC5D,GACJ;GACD,MAAM,KAAK,gBACT,GAAG,cAAc,UAAW,GAC5B,eACA,cACD;EACF;CACF;AACF"}

package/dist/retrievers/parent_document.d.cts

@@ -0,0 +1,92 @@
+import { BaseDocumentCompressor } from "./document_compressors/index.cjs";
+import { MultiVectorRetriever, MultiVectorRetrieverInput } from "./multi_vector.cjs";
+import { VectorStoreInterface, VectorStoreRetrieverInterface } from "@langchain/core/vectorstores";
+import { Document } from "@langchain/core/documents";
+import { TextSplitter, TextSplitterChunkHeaderOptions } from "@langchain/textsplitters";
+
+//#region src/retrievers/parent_document.d.ts
+// eslint-disable-next-line @typescript-eslint/no-explicit-any
+type SubDocs = Document<Record<string, any>>[];
+/**
+ * Interface for the fields required to initialize a
+ * ParentDocumentRetriever instance.
+ */
+type ParentDocumentRetrieverFields = MultiVectorRetrieverInput & {
+  childSplitter: TextSplitter;
+  parentSplitter?: TextSplitter;
+  /**
+   * A custom retriever to use when retrieving instead of
+   * the `.similaritySearch` method of the vectorstore.
+   */
+  childDocumentRetriever?: VectorStoreRetrieverInterface<VectorStoreInterface>;
+  documentCompressor?: BaseDocumentCompressor | undefined;
+  documentCompressorFilteringFn?: (docs: SubDocs) => SubDocs;
+};
+/**
+ * A type of document retriever that splits input documents into smaller chunks
+ * while separately storing and preserving the original documents.
+ * The small chunks are embedded, then on retrieval, the original
+ * "parent" documents are retrieved.
+ *
+ * This strikes a balance between better targeted retrieval with small documents
+ * and the more context-rich larger documents.
+ * @example
+ * ```typescript
+ * const retriever = new ParentDocumentRetriever({
+ *   vectorstore: new MemoryVectorStore(new OpenAIEmbeddings()),
+ *   byteStore: new InMemoryStore<Uint8Array>(),
+ *   parentSplitter: new RecursiveCharacterTextSplitter({
+ *     chunkOverlap: 0,
+ *     chunkSize: 500,
+ *   }),
+ *   childSplitter: new RecursiveCharacterTextSplitter({
+ *     chunkOverlap: 0,
+ *     chunkSize: 50,
+ *   }),
+ *   childK: 20,
+ *   parentK: 5,
+ * });
+ *
+ * const parentDocuments = await getDocuments();
+ * await retriever.addDocuments(parentDocuments);
+ * const retrievedDocs = await retriever.invoke("justice breyer");
+ * ```
+ */
+declare class ParentDocumentRetriever extends MultiVectorRetriever {
+  static lc_name(): string;
+  lc_namespace: string[];
+  vectorstore: VectorStoreInterface;
+  protected childSplitter: TextSplitter;
+  protected parentSplitter?: TextSplitter;
+  protected idKey: string;
+  protected childK?: number;
+  protected parentK?: number;
+  childDocumentRetriever: VectorStoreRetrieverInterface<VectorStoreInterface> | undefined;
+  documentCompressor: BaseDocumentCompressor | undefined;
+  documentCompressorFilteringFn?: ParentDocumentRetrieverFields["documentCompressorFilteringFn"];
+  constructor(fields: ParentDocumentRetrieverFields);
+  _getRelevantDocuments(query: string): Promise<Document[]>;
+  _storeDocuments(parentDoc: Record<string, Document>, childDocs: Document[], addToDocstore: boolean): Promise<void>;
+  /**
+   * Adds documents to the docstore and vectorstores.
+   * If a retriever is provided, it will be used to add documents instead of the vectorstore.
+   * @param docs The documents to add
+   * @param config.ids Optional list of ids for documents. If provided should be the same
+   *   length as the list of documents. Can provided if parent documents
+   *   are already in the document store and you don't want to re-add
+   *   to the docstore. If not provided, random UUIDs will be used as ids.
+   * @param config.addToDocstore Boolean of whether to add documents to docstore.
+   * This can be false if and only if `ids` are provided. You may want
+   *   to set this to False if the documents are already in the docstore
+   *   and you don't want to re-add them.
+   * @param config.chunkHeaderOptions Object with options for adding Contextual chunk headers
+   */
+  addDocuments(docs: Document[], config?: {
+    ids?: string[];
+    addToDocstore?: boolean;
+    childDocChunkHeaderOptions?: TextSplitterChunkHeaderOptions;
+  }): Promise<void>;
+}
+//#endregion
+export { ParentDocumentRetriever, ParentDocumentRetrieverFields, SubDocs };
+//# sourceMappingURL=parent_document.d.cts.map

package/dist/retrievers/parent_document.d.cts.map

@@ -0,0 +1 @@
+{"version":3,"file":"parent_document.d.cts","names":["VectorStoreInterface","VectorStoreRetrieverInterface","Document","TextSplitter","TextSplitterChunkHeaderOptions","BaseDocumentCompressor","MultiVectorRetriever","MultiVectorRetrieverInput","SubDocs","Record","ParentDocumentRetrieverFields","ParentDocumentRetriever","Promise"],"sources":["../../src/retrievers/parent_document.d.ts"],"sourcesContent":["import { type VectorStoreInterface, type VectorStoreRetrieverInterface } from \"@langchain/core/vectorstores\";\nimport { Document } from \"@langchain/core/documents\";\nimport { TextSplitter, TextSplitterChunkHeaderOptions } from \"@langchain/textsplitters\";\nimport type { BaseDocumentCompressor } from \"./document_compressors/index.js\";\nimport { MultiVectorRetriever, type MultiVectorRetrieverInput } from \"./multi_vector.js\";\n// eslint-disable-next-line @typescript-eslint/no-explicit-any\nexport type SubDocs = Document<Record<string, any>>[];\n/**\n * Interface for the fields required to initialize a\n * ParentDocumentRetriever instance.\n */\nexport type ParentDocumentRetrieverFields = MultiVectorRetrieverInput & {\n    childSplitter: TextSplitter;\n    parentSplitter?: TextSplitter;\n    /**\n     * A custom retriever to use when retrieving instead of\n     * the `.similaritySearch` method of the vectorstore.\n     */\n    childDocumentRetriever?: VectorStoreRetrieverInterface<VectorStoreInterface>;\n    documentCompressor?: BaseDocumentCompressor | undefined;\n    documentCompressorFilteringFn?: (docs: SubDocs) => SubDocs;\n};\n/**\n * A type of document retriever that splits input documents into smaller chunks\n * while separately storing and preserving the original documents.\n * The small chunks are embedded, then on retrieval, the original\n * \"parent\" documents are retrieved.\n *\n * This strikes a balance between better targeted retrieval with small documents\n * and the more context-rich larger documents.\n * @example\n * ```typescript\n * const retriever = new ParentDocumentRetriever({\n *   vectorstore: new MemoryVectorStore(new OpenAIEmbeddings()),\n *   byteStore: new InMemoryStore<Uint8Array>(),\n *   parentSplitter: new RecursiveCharacterTextSplitter({\n *     chunkOverlap: 0,\n *     chunkSize: 500,\n *   }),\n *   childSplitter: new RecursiveCharacterTextSplitter({\n *     chunkOverlap: 0,\n *     chunkSize: 50,\n *   }),\n *   childK: 20,\n *   parentK: 5,\n * });\n *\n * const parentDocuments = await getDocuments();\n * await retriever.addDocuments(parentDocuments);\n * const retrievedDocs = await retriever.invoke(\"justice breyer\");\n * ```\n */\nexport declare class ParentDocumentRetriever extends MultiVectorRetriever {\n    static lc_name(): string;\n    lc_namespace: string[];\n    vectorstore: VectorStoreInterface;\n    protected childSplitter: TextSplitter;\n    protected parentSplitter?: TextSplitter;\n    protected idKey: string;\n    protected childK?: number;\n    protected parentK?: number;\n    childDocumentRetriever: VectorStoreRetrieverInterface<VectorStoreInterface> | undefined;\n    documentCompressor: BaseDocumentCompressor | undefined;\n    documentCompressorFilteringFn?: ParentDocumentRetrieverFields[\"documentCompressorFilteringFn\"];\n    constructor(fields: ParentDocumentRetrieverFields);\n    _getRelevantDocuments(query: string): Promise<Document[]>;\n    _storeDocuments(parentDoc: Record<string, Document>, childDocs: Document[], addToDocstore: boolean): Promise<void>;\n    /**\n     * Adds documents to the docstore and vectorstores.\n     * If a retriever is provided, it will be used to add documents instead of the vectorstore.\n     * @param docs The documents to add\n     * @param config.ids Optional list of ids for documents. If provided should be the same\n     *   length as the list of documents. Can provided if parent documents\n     *   are already in the document store and you don't want to re-add\n     *   to the docstore. If not provided, random UUIDs will be used as ids.\n     * @param config.addToDocstore Boolean of whether to add documents to docstore.\n     * This can be false if and only if `ids` are provided. You may want\n     *   to set this to False if the documents are already in the docstore\n     *   and you don't want to re-add them.\n     * @param config.chunkHeaderOptions Object with options for adding Contextual chunk headers\n     */\n    addDocuments(docs: Document[], config?: {\n        ids?: string[];\n        addToDocstore?: boolean;\n        childDocChunkHeaderOptions?: TextSplitterChunkHeaderOptions;\n    }): Promise<void>;\n}\n"],"mappings":";;;;;;;;KAMYQ,OAAAA,GAAUN,SAASO;AAA/B;;;;AAA8B,KAKlBC,6BAAAA,GAAgCH,yBALd,GAAA;EAKlBG,aAAAA,EACOP,YADPO;EAA6B,cAAA,CAAA,EAEpBP,YAFoB;EAAA;;;;EAOsC,sBAAlDF,CAAAA,EAAAA,6BAAAA,CAA8BD,oBAA9BC,CAAAA;EAA6B,kBACjCI,CAAAA,EAAAA,sBAAAA,GAAAA,SAAAA;EAAsB,6BACJG,CAAAA,EAAAA,CAAAA,IAAAA,EAAAA,OAAAA,EAAAA,GAAYA,OAAZA;CAAO;AAAY;AAgC9D;;;;;;;;;;;;;;;;;;;;AAAyE;;;;;;;;;cAApDG,uBAAAA,SAAgCL,oBAAAA;;;eAGpCN;2BACYG;6BACEA;;;;0BAIHF,8BAA8BD;sBAClCK;kCACYK;sBACZA;wCACkBE,QAAQV;6BACnBO,eAAeP,sBAAsBA,qCAAqCU;;;;;;;;;;;;;;;qBAelFV;;;iCAGcE;MAC7BQ"}

package/dist/retrievers/parent_document.d.ts

@@ -0,0 +1,92 @@
+import { BaseDocumentCompressor } from "./document_compressors/index.js";
+import { MultiVectorRetriever, MultiVectorRetrieverInput } from "./multi_vector.js";
+import { Document } from "@langchain/core/documents";
+import { TextSplitter, TextSplitterChunkHeaderOptions } from "@langchain/textsplitters";
+import { VectorStoreInterface, VectorStoreRetrieverInterface } from "@langchain/core/vectorstores";
+
+//#region src/retrievers/parent_document.d.ts
+// eslint-disable-next-line @typescript-eslint/no-explicit-any
+type SubDocs = Document<Record<string, any>>[];
+/**
+ * Interface for the fields required to initialize a
+ * ParentDocumentRetriever instance.
+ */
+type ParentDocumentRetrieverFields = MultiVectorRetrieverInput & {
+  childSplitter: TextSplitter;
+  parentSplitter?: TextSplitter;
+  /**
+   * A custom retriever to use when retrieving instead of
+   * the `.similaritySearch` method of the vectorstore.
+   */
+  childDocumentRetriever?: VectorStoreRetrieverInterface<VectorStoreInterface>;
+  documentCompressor?: BaseDocumentCompressor | undefined;
+  documentCompressorFilteringFn?: (docs: SubDocs) => SubDocs;
+};
+/**
+ * A type of document retriever that splits input documents into smaller chunks
+ * while separately storing and preserving the original documents.
+ * The small chunks are embedded, then on retrieval, the original
+ * "parent" documents are retrieved.
+ *
+ * This strikes a balance between better targeted retrieval with small documents
+ * and the more context-rich larger documents.
+ * @example
+ * ```typescript
+ * const retriever = new ParentDocumentRetriever({
+ *   vectorstore: new MemoryVectorStore(new OpenAIEmbeddings()),
+ *   byteStore: new InMemoryStore<Uint8Array>(),
+ *   parentSplitter: new RecursiveCharacterTextSplitter({
+ *     chunkOverlap: 0,
+ *     chunkSize: 500,
+ *   }),
+ *   childSplitter: new RecursiveCharacterTextSplitter({
+ *     chunkOverlap: 0,
+ *     chunkSize: 50,
+ *   }),
+ *   childK: 20,
+ *   parentK: 5,
+ * });
+ *
+ * const parentDocuments = await getDocuments();
+ * await retriever.addDocuments(parentDocuments);
+ * const retrievedDocs = await retriever.invoke("justice breyer");
+ * ```
+ */
+declare class ParentDocumentRetriever extends MultiVectorRetriever {
+  static lc_name(): string;
+  lc_namespace: string[];
+  vectorstore: VectorStoreInterface;
+  protected childSplitter: TextSplitter;
+  protected parentSplitter?: TextSplitter;
+  protected idKey: string;
+  protected childK?: number;
+  protected parentK?: number;
+  childDocumentRetriever: VectorStoreRetrieverInterface<VectorStoreInterface> | undefined;
+  documentCompressor: BaseDocumentCompressor | undefined;
+  documentCompressorFilteringFn?: ParentDocumentRetrieverFields["documentCompressorFilteringFn"];
+  constructor(fields: ParentDocumentRetrieverFields);
+  _getRelevantDocuments(query: string): Promise<Document[]>;
+  _storeDocuments(parentDoc: Record<string, Document>, childDocs: Document[], addToDocstore: boolean): Promise<void>;
+  /**
+   * Adds documents to the docstore and vectorstores.
+   * If a retriever is provided, it will be used to add documents instead of the vectorstore.
+   * @param docs The documents to add
+   * @param config.ids Optional list of ids for documents. If provided should be the same
+   *   length as the list of documents. Can provided if parent documents
+   *   are already in the document store and you don't want to re-add
+   *   to the docstore. If not provided, random UUIDs will be used as ids.
+   * @param config.addToDocstore Boolean of whether to add documents to docstore.
+   * This can be false if and only if `ids` are provided. You may want
+   *   to set this to False if the documents are already in the docstore
+   *   and you don't want to re-add them.
+   * @param config.chunkHeaderOptions Object with options for adding Contextual chunk headers
+   */
+  addDocuments(docs: Document[], config?: {
+    ids?: string[];
+    addToDocstore?: boolean;
+    childDocChunkHeaderOptions?: TextSplitterChunkHeaderOptions;
+  }): Promise<void>;
+}
+//#endregion
+export { ParentDocumentRetriever, ParentDocumentRetrieverFields, SubDocs };
+//# sourceMappingURL=parent_document.d.ts.map

package/dist/retrievers/parent_document.d.ts.map

@@ -0,0 +1 @@
+{"version":3,"file":"parent_document.d.ts","names":["VectorStoreInterface","VectorStoreRetrieverInterface","Document","TextSplitter","TextSplitterChunkHeaderOptions","BaseDocumentCompressor","MultiVectorRetriever","MultiVectorRetrieverInput","SubDocs","Record","ParentDocumentRetrieverFields","ParentDocumentRetriever","Promise"],"sources":["../../src/retrievers/parent_document.d.ts"],"sourcesContent":["import { type VectorStoreInterface, type VectorStoreRetrieverInterface } from \"@langchain/core/vectorstores\";\nimport { Document } from \"@langchain/core/documents\";\nimport { TextSplitter, TextSplitterChunkHeaderOptions } from \"@langchain/textsplitters\";\nimport type { BaseDocumentCompressor } from \"./document_compressors/index.js\";\nimport { MultiVectorRetriever, type MultiVectorRetrieverInput } from \"./multi_vector.js\";\n// eslint-disable-next-line @typescript-eslint/no-explicit-any\nexport type SubDocs = Document<Record<string, any>>[];\n/**\n * Interface for the fields required to initialize a\n * ParentDocumentRetriever instance.\n */\nexport type ParentDocumentRetrieverFields = MultiVectorRetrieverInput & {\n    childSplitter: TextSplitter;\n    parentSplitter?: TextSplitter;\n    /**\n     * A custom retriever to use when retrieving instead of\n     * the `.similaritySearch` method of the vectorstore.\n     */\n    childDocumentRetriever?: VectorStoreRetrieverInterface<VectorStoreInterface>;\n    documentCompressor?: BaseDocumentCompressor | undefined;\n    documentCompressorFilteringFn?: (docs: SubDocs) => SubDocs;\n};\n/**\n * A type of document retriever that splits input documents into smaller chunks\n * while separately storing and preserving the original documents.\n * The small chunks are embedded, then on retrieval, the original\n * \"parent\" documents are retrieved.\n *\n * This strikes a balance between better targeted retrieval with small documents\n * and the more context-rich larger documents.\n * @example\n * ```typescript\n * const retriever = new ParentDocumentRetriever({\n *   vectorstore: new MemoryVectorStore(new OpenAIEmbeddings()),\n *   byteStore: new InMemoryStore<Uint8Array>(),\n *   parentSplitter: new RecursiveCharacterTextSplitter({\n *     chunkOverlap: 0,\n *     chunkSize: 500,\n *   }),\n *   childSplitter: new RecursiveCharacterTextSplitter({\n *     chunkOverlap: 0,\n *     chunkSize: 50,\n *   }),\n *   childK: 20,\n *   parentK: 5,\n * });\n *\n * const parentDocuments = await getDocuments();\n * await retriever.addDocuments(parentDocuments);\n * const retrievedDocs = await retriever.invoke(\"justice breyer\");\n * ```\n */\nexport declare class ParentDocumentRetriever extends MultiVectorRetriever {\n    static lc_name(): string;\n    lc_namespace: string[];\n    vectorstore: VectorStoreInterface;\n    protected childSplitter: TextSplitter;\n    protected parentSplitter?: TextSplitter;\n    protected idKey: string;\n    protected childK?: number;\n    protected parentK?: number;\n    childDocumentRetriever: VectorStoreRetrieverInterface<VectorStoreInterface> | undefined;\n    documentCompressor: BaseDocumentCompressor | undefined;\n    documentCompressorFilteringFn?: ParentDocumentRetrieverFields[\"documentCompressorFilteringFn\"];\n    constructor(fields: ParentDocumentRetrieverFields);\n    _getRelevantDocuments(query: string): Promise<Document[]>;\n    _storeDocuments(parentDoc: Record<string, Document>, childDocs: Document[], addToDocstore: boolean): Promise<void>;\n    /**\n     * Adds documents to the docstore and vectorstores.\n     * If a retriever is provided, it will be used to add documents instead of the vectorstore.\n     * @param docs The documents to add\n     * @param config.ids Optional list of ids for documents. If provided should be the same\n     *   length as the list of documents. Can provided if parent documents\n     *   are already in the document store and you don't want to re-add\n     *   to the docstore. If not provided, random UUIDs will be used as ids.\n     * @param config.addToDocstore Boolean of whether to add documents to docstore.\n     * This can be false if and only if `ids` are provided. You may want\n     *   to set this to False if the documents are already in the docstore\n     *   and you don't want to re-add them.\n     * @param config.chunkHeaderOptions Object with options for adding Contextual chunk headers\n     */\n    addDocuments(docs: Document[], config?: {\n        ids?: string[];\n        addToDocstore?: boolean;\n        childDocChunkHeaderOptions?: TextSplitterChunkHeaderOptions;\n    }): Promise<void>;\n}\n"],"mappings":";;;;;;;;KAMYQ,OAAAA,GAAUN,SAASO;AAA/B;;;;AAA8B,KAKlBC,6BAAAA,GAAgCH,yBALd,GAAA;EAKlBG,aAAAA,EACOP,YADPO;EAA6B,cAAA,CAAA,EAEpBP,YAFoB;EAAA;;;;EAOsC,sBAAlDF,CAAAA,EAAAA,6BAAAA,CAA8BD,oBAA9BC,CAAAA;EAA6B,kBACjCI,CAAAA,EAAAA,sBAAAA,GAAAA,SAAAA;EAAsB,6BACJG,CAAAA,EAAAA,CAAAA,IAAAA,EAAAA,OAAAA,EAAAA,GAAYA,OAAZA;CAAO;AAAY;AAgC9D;;;;;;;;;;;;;;;;;;;;AAAyE;;;;;;;;;cAApDG,uBAAAA,SAAgCL,oBAAAA;;;eAGpCN;2BACYG;6BACEA;;;;0BAIHF,8BAA8BD;sBAClCK;kCACYK;sBACZA;wCACkBE,QAAQV;6BACnBO,eAAeP,sBAAsBA,qCAAqCU;;;;;;;;;;;;;;;qBAelFV;;;iCAGcE;MAC7BQ"}