@langchain/classic 1.0.0-alpha.1

package/dist/document_transformers/openai_functions.d.cts

@@ -0,0 +1,29 @@
+import { BaseChain } from "../chains/base.cjs";
+import { TaggingChainOptions } from "../chains/openai_functions/tagging.cjs";
+import { InteropZodObject } from "@langchain/core/utils/types";
+import { ChatOpenAI } from "@langchain/openai";
+import { Document, MappingDocumentTransformer } from "@langchain/core/documents";
+import { JsonSchema7ObjectType } from "@langchain/core/utils/json_schema";
+
+//#region src/document_transformers/openai_functions.d.ts
+
+/**
+ * A transformer that tags metadata to a document using a tagging chain.
+ */
+declare class MetadataTagger extends MappingDocumentTransformer {
+  static lc_name(): string;
+  protected taggingChain: BaseChain;
+  constructor(fields: {
+    taggingChain: BaseChain;
+  });
+  _transformDocument(document: Document): Promise<Document>;
+}
+declare function createMetadataTagger(schema: JsonSchema7ObjectType, options: TaggingChainOptions & {
+  llm?: ChatOpenAI;
+}): MetadataTagger;
+declare function createMetadataTaggerFromZod(schema: InteropZodObject, options: TaggingChainOptions & {
+  llm?: ChatOpenAI;
+}): MetadataTagger;
+//#endregion
+export { MetadataTagger, createMetadataTagger, createMetadataTaggerFromZod };
+//# sourceMappingURL=openai_functions.d.cts.map

package/dist/document_transformers/openai_functions.d.cts.map

@@ -0,0 +1 @@
+{"version":3,"file":"openai_functions.d.cts","names":["Document","MappingDocumentTransformer","ChatOpenAI","InteropZodObject","JsonSchema7ObjectType","BaseChain","TaggingChainOptions","MetadataTagger","Promise","createMetadataTagger","createMetadataTaggerFromZod"],"sources":["../../src/document_transformers/openai_functions.d.ts"],"sourcesContent":["import { Document, MappingDocumentTransformer } from \"@langchain/core/documents\";\nimport { ChatOpenAI } from \"@langchain/openai\";\nimport { InteropZodObject } from \"@langchain/core/utils/types\";\nimport { type JsonSchema7ObjectType } from \"@langchain/core/utils/json_schema\";\nimport { BaseChain } from \"../chains/base.js\";\nimport { TaggingChainOptions } from \"../chains/openai_functions/index.js\";\n/**\n * A transformer that tags metadata to a document using a tagging chain.\n */\nexport declare class MetadataTagger extends MappingDocumentTransformer {\n    static lc_name(): string;\n    protected taggingChain: BaseChain;\n    constructor(fields: {\n        taggingChain: BaseChain;\n    });\n    _transformDocument(document: Document): Promise<Document>;\n}\nexport declare function createMetadataTagger(schema: JsonSchema7ObjectType, options: TaggingChainOptions & {\n    llm?: ChatOpenAI;\n}): MetadataTagger;\nexport declare function createMetadataTaggerFromZod(schema: InteropZodObject, options: TaggingChainOptions & {\n    llm?: ChatOpenAI;\n}): MetadataTagger;\n"],"mappings":";;;;;;;;;;;AASA;AAAmC,cAAdO,cAAAA,SAAuBN,0BAAAA,CAAT;EAAA,OAEPI,OAAAA,CAAAA,CAAAA,EAAAA,MAAAA;EAAS,UAEfA,YAAAA,EAFMA,SAENA;EAAS,WAEEL,CAAAA,MAAAA,EAAAA;IAAmBA,YAAAA,EAF9BK,SAE8BL;EAAQ,CAAA;EAAT,kBANPC,CAAAA,QAAAA,EAMXD,QANWC,CAAAA,EAMAO,OANAP,CAMQD,QANRC,CAAAA;AAA0B;AAQ9CQ,iBAAAA,oBAAAA,CAAoB,MAAA,EAASL,qBAAT,EAAA,OAAA,EAAyCE,mBAAzC,GAAA;EAAA,GAAA,CAAA,EAClCJ,UADkC;CAAA,CAAA,EAExCK,cAFiDH;AAAgCE,iBAG7DI,2BAAAA,CAH6DJ,MAAAA,EAGzBH,gBAHyBG,EAAAA,OAAAA,EAGEA,mBAHFA,GAAAA;EAAmB,GAC9FJ,CAAAA,EAGAA,UAHAA;CAAU,CAAA,EAIhBK,cAHAA"}

package/dist/document_transformers/openai_functions.d.ts

@@ -0,0 +1,29 @@
+import { BaseChain } from "../chains/base.js";
+import { TaggingChainOptions } from "../chains/openai_functions/tagging.js";
+import { Document, MappingDocumentTransformer } from "@langchain/core/documents";
+import { InteropZodObject } from "@langchain/core/utils/types";
+import { JsonSchema7ObjectType } from "@langchain/core/utils/json_schema";
+import { ChatOpenAI } from "@langchain/openai";
+
+//#region src/document_transformers/openai_functions.d.ts
+
+/**
+ * A transformer that tags metadata to a document using a tagging chain.
+ */
+declare class MetadataTagger extends MappingDocumentTransformer {
+  static lc_name(): string;
+  protected taggingChain: BaseChain;
+  constructor(fields: {
+    taggingChain: BaseChain;
+  });
+  _transformDocument(document: Document): Promise<Document>;
+}
+declare function createMetadataTagger(schema: JsonSchema7ObjectType, options: TaggingChainOptions & {
+  llm?: ChatOpenAI;
+}): MetadataTagger;
+declare function createMetadataTaggerFromZod(schema: InteropZodObject, options: TaggingChainOptions & {
+  llm?: ChatOpenAI;
+}): MetadataTagger;
+//#endregion
+export { MetadataTagger, createMetadataTagger, createMetadataTaggerFromZod };
+//# sourceMappingURL=openai_functions.d.ts.map

package/dist/document_transformers/openai_functions.d.ts.map

@@ -0,0 +1 @@
+{"version":3,"file":"openai_functions.d.ts","names":["Document","MappingDocumentTransformer","ChatOpenAI","InteropZodObject","JsonSchema7ObjectType","BaseChain","TaggingChainOptions","MetadataTagger","Promise","createMetadataTagger","createMetadataTaggerFromZod"],"sources":["../../src/document_transformers/openai_functions.d.ts"],"sourcesContent":["import { Document, MappingDocumentTransformer } from \"@langchain/core/documents\";\nimport { ChatOpenAI } from \"@langchain/openai\";\nimport { InteropZodObject } from \"@langchain/core/utils/types\";\nimport { type JsonSchema7ObjectType } from \"@langchain/core/utils/json_schema\";\nimport { BaseChain } from \"../chains/base.js\";\nimport { TaggingChainOptions } from \"../chains/openai_functions/index.js\";\n/**\n * A transformer that tags metadata to a document using a tagging chain.\n */\nexport declare class MetadataTagger extends MappingDocumentTransformer {\n    static lc_name(): string;\n    protected taggingChain: BaseChain;\n    constructor(fields: {\n        taggingChain: BaseChain;\n    });\n    _transformDocument(document: Document): Promise<Document>;\n}\nexport declare function createMetadataTagger(schema: JsonSchema7ObjectType, options: TaggingChainOptions & {\n    llm?: ChatOpenAI;\n}): MetadataTagger;\nexport declare function createMetadataTaggerFromZod(schema: InteropZodObject, options: TaggingChainOptions & {\n    llm?: ChatOpenAI;\n}): MetadataTagger;\n"],"mappings":";;;;;;;;;;;AASA;AAAmC,cAAdO,cAAAA,SAAuBN,0BAAAA,CAAT;EAAA,OAEPI,OAAAA,CAAAA,CAAAA,EAAAA,MAAAA;EAAS,UAEfA,YAAAA,EAFMA,SAENA;EAAS,WAEEL,CAAAA,MAAAA,EAAAA;IAAmBA,YAAAA,EAF9BK,SAE8BL;EAAQ,CAAA;EAAT,kBANPC,CAAAA,QAAAA,EAMXD,QANWC,CAAAA,EAMAO,OANAP,CAMQD,QANRC,CAAAA;AAA0B;AAQ9CQ,iBAAAA,oBAAAA,CAAoB,MAAA,EAASL,qBAAT,EAAA,OAAA,EAAyCE,mBAAzC,GAAA;EAAA,GAAA,CAAA,EAClCJ,UADkC;CAAA,CAAA,EAExCK,cAFiDH;AAAgCE,iBAG7DI,2BAAAA,CAH6DJ,MAAAA,EAGzBH,gBAHyBG,EAAAA,OAAAA,EAGEA,mBAHFA,GAAAA;EAAmB,GAC9FJ,CAAAA,EAGAA,UAHAA;CAAU,CAAA,EAIhBK,cAHAA"}

package/dist/document_transformers/openai_functions.js

@@ -0,0 +1,52 @@
+import { __export } from "../_virtual/rolldown_runtime.js";
+import { createTaggingChain } from "../chains/openai_functions/tagging.js";
+import "../chains/openai_functions/index.js";
+import { Document, MappingDocumentTransformer } from "@langchain/core/documents";
+import { toJsonSchema } from "@langchain/core/utils/json_schema";
+import { ChatOpenAI } from "@langchain/openai";
+
+//#region src/document_transformers/openai_functions.ts
+var openai_functions_exports = {};
+__export(openai_functions_exports, {
+	MetadataTagger: () => MetadataTagger,
+	createMetadataTagger: () => createMetadataTagger,
+	createMetadataTaggerFromZod: () => createMetadataTaggerFromZod
+});
+/**
+* A transformer that tags metadata to a document using a tagging chain.
+*/
+var MetadataTagger = class extends MappingDocumentTransformer {
+	static lc_name() {
+		return "MetadataTagger";
+	}
+	taggingChain;
+	constructor(fields) {
+		super();
+		this.taggingChain = fields.taggingChain;
+		if (this.taggingChain.inputKeys.length !== 1) throw new Error("Invalid input chain. The input chain must have exactly one input.");
+		if (this.taggingChain.outputKeys.length !== 1) throw new Error("Invalid input chain. The input chain must have exactly one output.");
+	}
+	async _transformDocument(document) {
+		const taggingChainResponse = await this.taggingChain.call({ [this.taggingChain.inputKeys[0]]: document.pageContent });
+		const extractedMetadata = taggingChainResponse[this.taggingChain.outputKeys[0]];
+		return new Document({
+			pageContent: document.pageContent,
+			metadata: {
+				...extractedMetadata,
+				...document.metadata
+			}
+		});
+	}
+};
+function createMetadataTagger(schema, options) {
+	const { llm = new ChatOpenAI({ model: "gpt-3.5-turbo-0613" }),...rest } = options;
+	const taggingChain = createTaggingChain(schema, llm, rest);
+	return new MetadataTagger({ taggingChain });
+}
+function createMetadataTaggerFromZod(schema, options) {
+	return createMetadataTagger(toJsonSchema(schema), options);
+}
+
+//#endregion
+export { MetadataTagger, createMetadataTagger, createMetadataTaggerFromZod, openai_functions_exports };
+//# sourceMappingURL=openai_functions.js.map

package/dist/document_transformers/openai_functions.js.map

@@ -0,0 +1 @@
+{"version":3,"file":"openai_functions.js","names":["fields: { taggingChain: BaseChain }","document: Document","schema: JsonSchema7ObjectType","options: TaggingChainOptions & { llm?: ChatOpenAI }","schema: InteropZodObject"],"sources":["../../src/document_transformers/openai_functions.ts"],"sourcesContent":["import {\n  Document,\n  MappingDocumentTransformer,\n} from \"@langchain/core/documents\";\nimport { ChatOpenAI } from \"@langchain/openai\";\nimport { InteropZodObject } from \"@langchain/core/utils/types\";\nimport {\n  type JsonSchema7ObjectType,\n  toJsonSchema,\n} from \"@langchain/core/utils/json_schema\";\nimport { BaseChain } from \"../chains/base.js\";\nimport {\n  TaggingChainOptions,\n  createTaggingChain,\n} from \"../chains/openai_functions/index.js\";\n\n/**\n * A transformer that tags metadata to a document using a tagging chain.\n */\nexport class MetadataTagger extends MappingDocumentTransformer {\n  static lc_name() {\n    return \"MetadataTagger\";\n  }\n\n  protected taggingChain: BaseChain;\n\n  constructor(fields: { taggingChain: BaseChain }) {\n    super();\n    this.taggingChain = fields.taggingChain;\n    if (this.taggingChain.inputKeys.length !== 1) {\n      throw new Error(\n        \"Invalid input chain. The input chain must have exactly one input.\"\n      );\n    }\n    if (this.taggingChain.outputKeys.length !== 1) {\n      throw new Error(\n        \"Invalid input chain. The input chain must have exactly one output.\"\n      );\n    }\n  }\n\n  async _transformDocument(document: Document): Promise<Document> {\n    const taggingChainResponse = await this.taggingChain.call({\n      [this.taggingChain.inputKeys[0]]: document.pageContent,\n    });\n    const extractedMetadata =\n      taggingChainResponse[this.taggingChain.outputKeys[0]];\n    return new Document({\n      pageContent: document.pageContent,\n      metadata: { ...extractedMetadata, ...document.metadata },\n    });\n  }\n}\n\nexport function createMetadataTagger(\n  schema: JsonSchema7ObjectType,\n  options: TaggingChainOptions & { llm?: ChatOpenAI }\n) {\n  const { llm = new ChatOpenAI({ model: \"gpt-3.5-turbo-0613\" }), ...rest } =\n    options;\n  const taggingChain = createTaggingChain(schema, llm, rest);\n  return new MetadataTagger({ taggingChain });\n}\n\nexport function createMetadataTaggerFromZod(\n  schema: InteropZodObject,\n  options: TaggingChainOptions & { llm?: ChatOpenAI }\n) {\n  return createMetadataTagger(\n    toJsonSchema(schema) as JsonSchema7ObjectType,\n    options\n  );\n}\n"],"mappings":";;;;;;;;;;;;;;;;;AAmBA,IAAa,iBAAb,cAAoC,2BAA2B;CAC7D,OAAO,UAAU;AACf,SAAO;CACR;CAED,AAAU;CAEV,YAAYA,QAAqC;EAC/C,OAAO;EACP,KAAK,eAAe,OAAO;AAC3B,MAAI,KAAK,aAAa,UAAU,WAAW,EACzC,OAAM,IAAI,MACR;AAGJ,MAAI,KAAK,aAAa,WAAW,WAAW,EAC1C,OAAM,IAAI,MACR;CAGL;CAED,MAAM,mBAAmBC,UAAuC;EAC9D,MAAM,uBAAuB,MAAM,KAAK,aAAa,KAAK,GACvD,KAAK,aAAa,UAAU,KAAK,SAAS,YAC5C,EAAC;EACF,MAAM,oBACJ,qBAAqB,KAAK,aAAa,WAAW;AACpD,SAAO,IAAI,SAAS;GAClB,aAAa,SAAS;GACtB,UAAU;IAAE,GAAG;IAAmB,GAAG,SAAS;GAAU;EACzD;CACF;AACF;AAED,SAAgB,qBACdC,QACAC,SACA;CACA,MAAM,EAAE,MAAM,IAAI,WAAW,EAAE,OAAO,qBAAsB,GAAG,GAAG,MAAM,GACtE;CACF,MAAM,eAAe,mBAAmB,QAAQ,KAAK,KAAK;AAC1D,QAAO,IAAI,eAAe,EAAE,aAAc;AAC3C;AAED,SAAgB,4BACdC,QACAD,SACA;AACA,QAAO,qBACL,aAAa,OAAO,EACpB,QACD;AACF"}

package/dist/embeddings/cache_backed.cjs

@@ -0,0 +1,140 @@
+const require_rolldown_runtime = require('../_virtual/rolldown_runtime.cjs');
+const require_storage_encoder_backed = require('../storage/encoder_backed.cjs');
+const __langchain_core_utils_hash = require_rolldown_runtime.__toESM(require("@langchain/core/utils/hash"));
+const __langchain_core_embeddings = require_rolldown_runtime.__toESM(require("@langchain/core/embeddings"));
+
+//#region src/embeddings/cache_backed.ts
+var cache_backed_exports = {};
+require_rolldown_runtime.__export(cache_backed_exports, { CacheBackedEmbeddings: () => CacheBackedEmbeddings });
+/**
+* Interface for caching results from embedding models.
+*
+* The interface allows works with any store that implements
+* the abstract store interface accepting keys of type str and values of list of
+* floats.
+*
+* If need be, the interface can be extended to accept other implementations
+* of the value serializer and deserializer, as well as the key encoder.
+* @example
+* ```typescript
+* const underlyingEmbeddings = new OpenAIEmbeddings();
+*
+* const cacheBackedEmbeddings = CacheBackedEmbeddings.fromBytesStore(
+*   underlyingEmbeddings,
+*   new ConvexKVStore({ ctx }),
+*   {
+*     namespace: underlyingEmbeddings.modelName,
+*   },
+* );
+*
+* const loader = new TextLoader("./state_of_the_union.txt");
+* const rawDocuments = await loader.load();
+* const splitter = new RecursiveCharacterTextSplitter({
+*   chunkSize: 1000,
+*   chunkOverlap: 0,
+* });
+* const documents = await splitter.splitDocuments(rawDocuments);
+*
+* let time = Date.now();
+* const vectorstore = await ConvexVectorStore.fromDocuments(
+*   documents,
+*   cacheBackedEmbeddings,
+*   { ctx },
+* );
+* console.log(`Initial creation time: ${Date.now() - time}ms`);
+*
+* time = Date.now();
+* const vectorstore2 = await ConvexVectorStore.fromDocuments(
+*   documents,
+*   cacheBackedEmbeddings,
+*   { ctx },
+* );
+* console.log(`Cached creation time: ${Date.now() - time}ms`);
+*
+* ```
+*/
+var CacheBackedEmbeddings = class extends __langchain_core_embeddings.Embeddings {
+	underlyingEmbeddings;
+	documentEmbeddingStore;
+	constructor(fields) {
+		super(fields);
+		this.underlyingEmbeddings = fields.underlyingEmbeddings;
+		this.documentEmbeddingStore = fields.documentEmbeddingStore;
+	}
+	/**
+	* Embed query text.
+	*
+	* This method does not support caching at the moment.
+	*
+	* Support for caching queries is easy to implement, but might make
+	* sense to hold off to see the most common patterns.
+	*
+	* If the cache has an eviction policy, we may need to be a bit more careful
+	* about sharing the cache between documents and queries. Generally,
+	* one is OK evicting query caches, but document caches should be kept.
+	*
+	* @param document The text to embed.
+	* @returns The embedding for the given text.
+	*/
+	async embedQuery(document) {
+		return this.underlyingEmbeddings.embedQuery(document);
+	}
+	/**
+	* Embed a list of texts.
+	*
+	* The method first checks the cache for the embeddings.
+	* If the embeddings are not found, the method uses the underlying embedder
+	* to embed the documents and stores the results in the cache.
+	*
+	* @param documents
+	* @returns A list of embeddings for the given texts.
+	*/
+	async embedDocuments(documents) {
+		const vectors = await this.documentEmbeddingStore.mget(documents);
+		const missingIndicies = [];
+		const missingDocuments = [];
+		for (let i = 0; i < vectors.length; i += 1) if (vectors[i] === void 0) {
+			missingIndicies.push(i);
+			missingDocuments.push(documents[i]);
+		}
+		if (missingDocuments.length) {
+			const missingVectors = await this.underlyingEmbeddings.embedDocuments(missingDocuments);
+			const keyValuePairs = missingDocuments.map((document, i) => [document, missingVectors[i]]);
+			await this.documentEmbeddingStore.mset(keyValuePairs);
+			for (let i = 0; i < missingIndicies.length; i += 1) vectors[missingIndicies[i]] = missingVectors[i];
+		}
+		return vectors;
+	}
+	/**
+	* Create a new CacheBackedEmbeddings instance from another embeddings instance
+	* and a storage instance.
+	* @param underlyingEmbeddings Embeddings used to populate the cache for new documents.
+	* @param documentEmbeddingStore Stores raw document embedding values. Keys are hashes of the document content.
+	* @param options.namespace Optional namespace for store keys.
+	* @returns A new CacheBackedEmbeddings instance.
+	*/
+	static fromBytesStore(underlyingEmbeddings, documentEmbeddingStore, options) {
+		const encoder = new TextEncoder();
+		const decoder = new TextDecoder();
+		const encoderBackedStore = new require_storage_encoder_backed.EncoderBackedStore({
+			store: documentEmbeddingStore,
+			keyEncoder: (key) => (options?.namespace ?? "") + (0, __langchain_core_utils_hash.sha256)(key),
+			valueSerializer: (value) => encoder.encode(JSON.stringify(value)),
+			valueDeserializer: (serializedValue) => JSON.parse(decoder.decode(serializedValue))
+		});
+		return new this({
+			underlyingEmbeddings,
+			documentEmbeddingStore: encoderBackedStore
+		});
+	}
+};
+
+//#endregion
+exports.CacheBackedEmbeddings = CacheBackedEmbeddings;
+Object.defineProperty(exports, 'cache_backed_exports', {
+  enumerable: true,
+  get: function () {
+    return cache_backed_exports;
+  }
+});
+//# sourceMappingURL=cache_backed.cjs.map

package/dist/embeddings/cache_backed.cjs.map

@@ -0,0 +1 @@
+{"version":3,"file":"cache_backed.cjs","names":["Embeddings","fields: CacheBackedEmbeddingsFields","document: string","documents: string[]","keyValuePairs: [string, number[]][]","underlyingEmbeddings: EmbeddingsInterface","documentEmbeddingStore: BaseStore<string, Uint8Array>","options?: {\n      namespace?: string;\n    }","EncoderBackedStore"],"sources":["../../src/embeddings/cache_backed.ts"],"sourcesContent":["import { sha256 } from \"@langchain/core/utils/hash\";\nimport {\n  type EmbeddingsInterface,\n  Embeddings,\n} from \"@langchain/core/embeddings\";\nimport { BaseStore } from \"@langchain/core/stores\";\n\nimport { AsyncCallerParams } from \"@langchain/core/utils/async_caller\";\nimport { EncoderBackedStore } from \"../storage/encoder_backed.js\";\n\n/**\n * Interface for the fields required to initialize an instance of the\n * CacheBackedEmbeddings class.\n */\nexport interface CacheBackedEmbeddingsFields extends AsyncCallerParams {\n  underlyingEmbeddings: EmbeddingsInterface;\n  documentEmbeddingStore: BaseStore<string, number[]>;\n}\n\n/**\n * Interface for caching results from embedding models.\n *\n * The interface allows works with any store that implements\n * the abstract store interface accepting keys of type str and values of list of\n * floats.\n *\n * If need be, the interface can be extended to accept other implementations\n * of the value serializer and deserializer, as well as the key encoder.\n * @example\n * ```typescript\n * const underlyingEmbeddings = new OpenAIEmbeddings();\n *\n * const cacheBackedEmbeddings = CacheBackedEmbeddings.fromBytesStore(\n *   underlyingEmbeddings,\n *   new ConvexKVStore({ ctx }),\n *   {\n *     namespace: underlyingEmbeddings.modelName,\n *   },\n * );\n *\n * const loader = new TextLoader(\"./state_of_the_union.txt\");\n * const rawDocuments = await loader.load();\n * const splitter = new RecursiveCharacterTextSplitter({\n *   chunkSize: 1000,\n *   chunkOverlap: 0,\n * });\n * const documents = await splitter.splitDocuments(rawDocuments);\n *\n * let time = Date.now();\n * const vectorstore = await ConvexVectorStore.fromDocuments(\n *   documents,\n *   cacheBackedEmbeddings,\n *   { ctx },\n * );\n * console.log(`Initial creation time: ${Date.now() - time}ms`);\n *\n * time = Date.now();\n * const vectorstore2 = await ConvexVectorStore.fromDocuments(\n *   documents,\n *   cacheBackedEmbeddings,\n *   { ctx },\n * );\n * console.log(`Cached creation time: ${Date.now() - time}ms`);\n *\n * ```\n */\nexport class CacheBackedEmbeddings extends Embeddings {\n  protected underlyingEmbeddings: EmbeddingsInterface;\n\n  protected documentEmbeddingStore: BaseStore<string, number[]>;\n\n  constructor(fields: CacheBackedEmbeddingsFields) {\n    super(fields);\n    this.underlyingEmbeddings = fields.underlyingEmbeddings;\n    this.documentEmbeddingStore = fields.documentEmbeddingStore;\n  }\n\n  /**\n   * Embed query text.\n   *\n   * This method does not support caching at the moment.\n   *\n   * Support for caching queries is easy to implement, but might make\n   * sense to hold off to see the most common patterns.\n   *\n   * If the cache has an eviction policy, we may need to be a bit more careful\n   * about sharing the cache between documents and queries. Generally,\n   * one is OK evicting query caches, but document caches should be kept.\n   *\n   * @param document The text to embed.\n   * @returns The embedding for the given text.\n   */\n  async embedQuery(document: string): Promise<number[]> {\n    return this.underlyingEmbeddings.embedQuery(document);\n  }\n\n  /**\n   * Embed a list of texts.\n   *\n   * The method first checks the cache for the embeddings.\n   * If the embeddings are not found, the method uses the underlying embedder\n   * to embed the documents and stores the results in the cache.\n   *\n   * @param documents\n   * @returns A list of embeddings for the given texts.\n   */\n  async embedDocuments(documents: string[]): Promise<number[][]> {\n    const vectors = await this.documentEmbeddingStore.mget(documents);\n    const missingIndicies = [];\n    const missingDocuments = [];\n    for (let i = 0; i < vectors.length; i += 1) {\n      if (vectors[i] === undefined) {\n        missingIndicies.push(i);\n        missingDocuments.push(documents[i]);\n      }\n    }\n    if (missingDocuments.length) {\n      const missingVectors = await this.underlyingEmbeddings.embedDocuments(\n        missingDocuments\n      );\n      const keyValuePairs: [string, number[]][] = missingDocuments.map(\n        (document, i) => [document, missingVectors[i]]\n      );\n      await this.documentEmbeddingStore.mset(keyValuePairs);\n      for (let i = 0; i < missingIndicies.length; i += 1) {\n        vectors[missingIndicies[i]] = missingVectors[i];\n      }\n    }\n    return vectors as number[][];\n  }\n\n  /**\n   * Create a new CacheBackedEmbeddings instance from another embeddings instance\n   * and a storage instance.\n   * @param underlyingEmbeddings Embeddings used to populate the cache for new documents.\n   * @param documentEmbeddingStore Stores raw document embedding values. Keys are hashes of the document content.\n   * @param options.namespace Optional namespace for store keys.\n   * @returns A new CacheBackedEmbeddings instance.\n   */\n  static fromBytesStore(\n    underlyingEmbeddings: EmbeddingsInterface,\n    documentEmbeddingStore: BaseStore<string, Uint8Array>,\n    options?: {\n      namespace?: string;\n    }\n  ) {\n    const encoder = new TextEncoder();\n    const decoder = new TextDecoder();\n    const encoderBackedStore = new EncoderBackedStore<\n      string,\n      number[],\n      Uint8Array\n    >({\n      store: documentEmbeddingStore,\n      keyEncoder: (key) => (options?.namespace ?? \"\") + sha256(key),\n      valueSerializer: (value) => encoder.encode(JSON.stringify(value)),\n      valueDeserializer: (serializedValue) =>\n        JSON.parse(decoder.decode(serializedValue)),\n    });\n    return new this({\n      underlyingEmbeddings,\n      documentEmbeddingStore: encoderBackedStore,\n    });\n  }\n}\n"],"mappings":";;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;AAkEA,IAAa,wBAAb,cAA2CA,uCAAW;CACpD,AAAU;CAEV,AAAU;CAEV,YAAYC,QAAqC;EAC/C,MAAM,OAAO;EACb,KAAK,uBAAuB,OAAO;EACnC,KAAK,yBAAyB,OAAO;CACtC;;;;;;;;;;;;;;;;CAiBD,MAAM,WAAWC,UAAqC;AACpD,SAAO,KAAK,qBAAqB,WAAW,SAAS;CACtD;;;;;;;;;;;CAYD,MAAM,eAAeC,WAA0C;EAC7D,MAAM,UAAU,MAAM,KAAK,uBAAuB,KAAK,UAAU;EACjE,MAAM,kBAAkB,CAAE;EAC1B,MAAM,mBAAmB,CAAE;AAC3B,OAAK,IAAI,IAAI,GAAG,IAAI,QAAQ,QAAQ,KAAK,EACvC,KAAI,QAAQ,OAAO,QAAW;GAC5B,gBAAgB,KAAK,EAAE;GACvB,iBAAiB,KAAK,UAAU,GAAG;EACpC;AAEH,MAAI,iBAAiB,QAAQ;GAC3B,MAAM,iBAAiB,MAAM,KAAK,qBAAqB,eACrD,iBACD;GACD,MAAMC,gBAAsC,iBAAiB,IAC3D,CAAC,UAAU,MAAM,CAAC,UAAU,eAAe,EAAG,EAC/C;GACD,MAAM,KAAK,uBAAuB,KAAK,cAAc;AACrD,QAAK,IAAI,IAAI,GAAG,IAAI,gBAAgB,QAAQ,KAAK,GAC/C,QAAQ,gBAAgB,MAAM,eAAe;EAEhD;AACD,SAAO;CACR;;;;;;;;;CAUD,OAAO,eACLC,sBACAC,wBACAC,SAGA;EACA,MAAM,UAAU,IAAI;EACpB,MAAM,UAAU,IAAI;EACpB,MAAM,qBAAqB,IAAIC,kDAI7B;GACA,OAAO;GACP,YAAY,CAAC,SAAS,SAAS,aAAa,8CAAa,IAAI;GAC7D,iBAAiB,CAAC,UAAU,QAAQ,OAAO,KAAK,UAAU,MAAM,CAAC;GACjE,mBAAmB,CAAC,oBAClB,KAAK,MAAM,QAAQ,OAAO,gBAAgB,CAAC;EAC9C;AACD,SAAO,IAAI,KAAK;GACd;GACA,wBAAwB;EACzB;CACF;AACF"}

package/dist/embeddings/cache_backed.d.cts

@@ -0,0 +1,107 @@
+import { Embeddings, EmbeddingsInterface } from "@langchain/core/embeddings";
+import { BaseStore } from "@langchain/core/stores";
+import { AsyncCallerParams } from "@langchain/core/utils/async_caller";
+
+//#region src/embeddings/cache_backed.d.ts
+
+/**
+ * Interface for the fields required to initialize an instance of the
+ * CacheBackedEmbeddings class.
+ */
+interface CacheBackedEmbeddingsFields extends AsyncCallerParams {
+  underlyingEmbeddings: EmbeddingsInterface;
+  documentEmbeddingStore: BaseStore<string, number[]>;
+}
+/**
+ * Interface for caching results from embedding models.
+ *
+ * The interface allows works with any store that implements
+ * the abstract store interface accepting keys of type str and values of list of
+ * floats.
+ *
+ * If need be, the interface can be extended to accept other implementations
+ * of the value serializer and deserializer, as well as the key encoder.
+ * @example
+ * ```typescript
+ * const underlyingEmbeddings = new OpenAIEmbeddings();
+ *
+ * const cacheBackedEmbeddings = CacheBackedEmbeddings.fromBytesStore(
+ *   underlyingEmbeddings,
+ *   new ConvexKVStore({ ctx }),
+ *   {
+ *     namespace: underlyingEmbeddings.modelName,
+ *   },
+ * );
+ *
+ * const loader = new TextLoader("./state_of_the_union.txt");
+ * const rawDocuments = await loader.load();
+ * const splitter = new RecursiveCharacterTextSplitter({
+ *   chunkSize: 1000,
+ *   chunkOverlap: 0,
+ * });
+ * const documents = await splitter.splitDocuments(rawDocuments);
+ *
+ * let time = Date.now();
+ * const vectorstore = await ConvexVectorStore.fromDocuments(
+ *   documents,
+ *   cacheBackedEmbeddings,
+ *   { ctx },
+ * );
+ * console.log(`Initial creation time: ${Date.now() - time}ms`);
+ *
+ * time = Date.now();
+ * const vectorstore2 = await ConvexVectorStore.fromDocuments(
+ *   documents,
+ *   cacheBackedEmbeddings,
+ *   { ctx },
+ * );
+ * console.log(`Cached creation time: ${Date.now() - time}ms`);
+ *
+ * ```
+ */
+declare class CacheBackedEmbeddings extends Embeddings {
+  protected underlyingEmbeddings: EmbeddingsInterface;
+  protected documentEmbeddingStore: BaseStore<string, number[]>;
+  constructor(fields: CacheBackedEmbeddingsFields);
+  /**
+   * Embed query text.
+   *
+   * This method does not support caching at the moment.
+   *
+   * Support for caching queries is easy to implement, but might make
+   * sense to hold off to see the most common patterns.
+   *
+   * If the cache has an eviction policy, we may need to be a bit more careful
+   * about sharing the cache between documents and queries. Generally,
+   * one is OK evicting query caches, but document caches should be kept.
+   *
+   * @param document The text to embed.
+   * @returns The embedding for the given text.
+   */
+  embedQuery(document: string): Promise<number[]>;
+  /**
+   * Embed a list of texts.
+   *
+   * The method first checks the cache for the embeddings.
+   * If the embeddings are not found, the method uses the underlying embedder
+   * to embed the documents and stores the results in the cache.
+   *
+   * @param documents
+   * @returns A list of embeddings for the given texts.
+   */
+  embedDocuments(documents: string[]): Promise<number[][]>;
+  /**
+   * Create a new CacheBackedEmbeddings instance from another embeddings instance
+   * and a storage instance.
+   * @param underlyingEmbeddings Embeddings used to populate the cache for new documents.
+   * @param documentEmbeddingStore Stores raw document embedding values. Keys are hashes of the document content.
+   * @param options.namespace Optional namespace for store keys.
+   * @returns A new CacheBackedEmbeddings instance.
+   */
+  static fromBytesStore(underlyingEmbeddings: EmbeddingsInterface, documentEmbeddingStore: BaseStore<string, Uint8Array>, options?: {
+    namespace?: string;
+  }): CacheBackedEmbeddings;
+}
+//#endregion
+export { CacheBackedEmbeddings, CacheBackedEmbeddingsFields };
+//# sourceMappingURL=cache_backed.d.cts.map

package/dist/embeddings/cache_backed.d.cts.map

@@ -0,0 +1 @@
+{"version":3,"file":"cache_backed.d.cts","names":["EmbeddingsInterface","Embeddings","BaseStore","AsyncCallerParams","CacheBackedEmbeddingsFields","CacheBackedEmbeddings","Promise","Uint8Array"],"sources":["../../src/embeddings/cache_backed.d.ts"],"sourcesContent":["import { type EmbeddingsInterface, Embeddings } from \"@langchain/core/embeddings\";\nimport { BaseStore } from \"@langchain/core/stores\";\nimport { AsyncCallerParams } from \"@langchain/core/utils/async_caller\";\n/**\n * Interface for the fields required to initialize an instance of the\n * CacheBackedEmbeddings class.\n */\nexport interface CacheBackedEmbeddingsFields extends AsyncCallerParams {\n    underlyingEmbeddings: EmbeddingsInterface;\n    documentEmbeddingStore: BaseStore<string, number[]>;\n}\n/**\n * Interface for caching results from embedding models.\n *\n * The interface allows works with any store that implements\n * the abstract store interface accepting keys of type str and values of list of\n * floats.\n *\n * If need be, the interface can be extended to accept other implementations\n * of the value serializer and deserializer, as well as the key encoder.\n * @example\n * ```typescript\n * const underlyingEmbeddings = new OpenAIEmbeddings();\n *\n * const cacheBackedEmbeddings = CacheBackedEmbeddings.fromBytesStore(\n *   underlyingEmbeddings,\n *   new ConvexKVStore({ ctx }),\n *   {\n *     namespace: underlyingEmbeddings.modelName,\n *   },\n * );\n *\n * const loader = new TextLoader(\"./state_of_the_union.txt\");\n * const rawDocuments = await loader.load();\n * const splitter = new RecursiveCharacterTextSplitter({\n *   chunkSize: 1000,\n *   chunkOverlap: 0,\n * });\n * const documents = await splitter.splitDocuments(rawDocuments);\n *\n * let time = Date.now();\n * const vectorstore = await ConvexVectorStore.fromDocuments(\n *   documents,\n *   cacheBackedEmbeddings,\n *   { ctx },\n * );\n * console.log(`Initial creation time: ${Date.now() - time}ms`);\n *\n * time = Date.now();\n * const vectorstore2 = await ConvexVectorStore.fromDocuments(\n *   documents,\n *   cacheBackedEmbeddings,\n *   { ctx },\n * );\n * console.log(`Cached creation time: ${Date.now() - time}ms`);\n *\n * ```\n */\nexport declare class CacheBackedEmbeddings extends Embeddings {\n    protected underlyingEmbeddings: EmbeddingsInterface;\n    protected documentEmbeddingStore: BaseStore<string, number[]>;\n    constructor(fields: CacheBackedEmbeddingsFields);\n    /**\n     * Embed query text.\n     *\n     * This method does not support caching at the moment.\n     *\n     * Support for caching queries is easy to implement, but might make\n     * sense to hold off to see the most common patterns.\n     *\n     * If the cache has an eviction policy, we may need to be a bit more careful\n     * about sharing the cache between documents and queries. Generally,\n     * one is OK evicting query caches, but document caches should be kept.\n     *\n     * @param document The text to embed.\n     * @returns The embedding for the given text.\n     */\n    embedQuery(document: string): Promise<number[]>;\n    /**\n     * Embed a list of texts.\n     *\n     * The method first checks the cache for the embeddings.\n     * If the embeddings are not found, the method uses the underlying embedder\n     * to embed the documents and stores the results in the cache.\n     *\n     * @param documents\n     * @returns A list of embeddings for the given texts.\n     */\n    embedDocuments(documents: string[]): Promise<number[][]>;\n    /**\n     * Create a new CacheBackedEmbeddings instance from another embeddings instance\n     * and a storage instance.\n     * @param underlyingEmbeddings Embeddings used to populate the cache for new documents.\n     * @param documentEmbeddingStore Stores raw document embedding values. Keys are hashes of the document content.\n     * @param options.namespace Optional namespace for store keys.\n     * @returns A new CacheBackedEmbeddings instance.\n     */\n    static fromBytesStore(underlyingEmbeddings: EmbeddingsInterface, documentEmbeddingStore: BaseStore<string, Uint8Array>, options?: {\n        namespace?: string;\n    }): CacheBackedEmbeddings;\n}\n"],"mappings":";;;;;;;;AAOA;;AAC0BA,UADTI,2BAAAA,SAAoCD,iBAC3BH,CAAAA;EAAmB,oBACjBE,EADFF,mBACEE;EAAS,sBAFgBC,EAEzBD,SAFyBC,CAAAA,MAAAA,EAAAA,MAAAA,EAAAA,CAAAA;AAAiB;AAmDtE;;;;;;;;;;;;AAA6D;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;cAAxCE,qBAAAA,SAA8BJ,UAAAA;kCACfD;oCACEE;sBACdE;;;;;;;;;;;;;;;;gCAgBUE;;;;;;;;;;;uCAWOA;;;;;;;;;8CASON,6CAA6CE,kBAAkBK;;MAEvGF"}

package/dist/embeddings/cache_backed.d.ts

@@ -0,0 +1,107 @@
+import { Embeddings, EmbeddingsInterface } from "@langchain/core/embeddings";
+import { BaseStore } from "@langchain/core/stores";
+import { AsyncCallerParams } from "@langchain/core/utils/async_caller";
+
+//#region src/embeddings/cache_backed.d.ts
+
+/**
+ * Interface for the fields required to initialize an instance of the
+ * CacheBackedEmbeddings class.
+ */
+interface CacheBackedEmbeddingsFields extends AsyncCallerParams {
+  underlyingEmbeddings: EmbeddingsInterface;
+  documentEmbeddingStore: BaseStore<string, number[]>;
+}
+/**
+ * Interface for caching results from embedding models.
+ *
+ * The interface allows works with any store that implements
+ * the abstract store interface accepting keys of type str and values of list of
+ * floats.
+ *
+ * If need be, the interface can be extended to accept other implementations
+ * of the value serializer and deserializer, as well as the key encoder.
+ * @example
+ * ```typescript
+ * const underlyingEmbeddings = new OpenAIEmbeddings();
+ *
+ * const cacheBackedEmbeddings = CacheBackedEmbeddings.fromBytesStore(
+ *   underlyingEmbeddings,
+ *   new ConvexKVStore({ ctx }),
+ *   {
+ *     namespace: underlyingEmbeddings.modelName,
+ *   },
+ * );
+ *
+ * const loader = new TextLoader("./state_of_the_union.txt");
+ * const rawDocuments = await loader.load();
+ * const splitter = new RecursiveCharacterTextSplitter({
+ *   chunkSize: 1000,
+ *   chunkOverlap: 0,
+ * });
+ * const documents = await splitter.splitDocuments(rawDocuments);
+ *
+ * let time = Date.now();
+ * const vectorstore = await ConvexVectorStore.fromDocuments(
+ *   documents,
+ *   cacheBackedEmbeddings,
+ *   { ctx },
+ * );
+ * console.log(`Initial creation time: ${Date.now() - time}ms`);
+ *
+ * time = Date.now();
+ * const vectorstore2 = await ConvexVectorStore.fromDocuments(
+ *   documents,
+ *   cacheBackedEmbeddings,
+ *   { ctx },
+ * );
+ * console.log(`Cached creation time: ${Date.now() - time}ms`);
+ *
+ * ```
+ */
+declare class CacheBackedEmbeddings extends Embeddings {
+  protected underlyingEmbeddings: EmbeddingsInterface;
+  protected documentEmbeddingStore: BaseStore<string, number[]>;
+  constructor(fields: CacheBackedEmbeddingsFields);
+  /**
+   * Embed query text.
+   *
+   * This method does not support caching at the moment.
+   *
+   * Support for caching queries is easy to implement, but might make
+   * sense to hold off to see the most common patterns.
+   *
+   * If the cache has an eviction policy, we may need to be a bit more careful
+   * about sharing the cache between documents and queries. Generally,
+   * one is OK evicting query caches, but document caches should be kept.
+   *
+   * @param document The text to embed.
+   * @returns The embedding for the given text.
+   */
+  embedQuery(document: string): Promise<number[]>;
+  /**
+   * Embed a list of texts.
+   *
+   * The method first checks the cache for the embeddings.
+   * If the embeddings are not found, the method uses the underlying embedder
+   * to embed the documents and stores the results in the cache.
+   *
+   * @param documents
+   * @returns A list of embeddings for the given texts.
+   */
+  embedDocuments(documents: string[]): Promise<number[][]>;
+  /**
+   * Create a new CacheBackedEmbeddings instance from another embeddings instance
+   * and a storage instance.
+   * @param underlyingEmbeddings Embeddings used to populate the cache for new documents.
+   * @param documentEmbeddingStore Stores raw document embedding values. Keys are hashes of the document content.
+   * @param options.namespace Optional namespace for store keys.
+   * @returns A new CacheBackedEmbeddings instance.
+   */
+  static fromBytesStore(underlyingEmbeddings: EmbeddingsInterface, documentEmbeddingStore: BaseStore<string, Uint8Array>, options?: {
+    namespace?: string;
+  }): CacheBackedEmbeddings;
+}
+//#endregion
+export { CacheBackedEmbeddings, CacheBackedEmbeddingsFields };
+//# sourceMappingURL=cache_backed.d.ts.map

package/dist/embeddings/cache_backed.d.ts.map

@@ -0,0 +1 @@
+{"version":3,"file":"cache_backed.d.ts","names":["EmbeddingsInterface","Embeddings","BaseStore","AsyncCallerParams","CacheBackedEmbeddingsFields","CacheBackedEmbeddings","Promise","Uint8Array"],"sources":["../../src/embeddings/cache_backed.d.ts"],"sourcesContent":["import { type EmbeddingsInterface, Embeddings } from \"@langchain/core/embeddings\";\nimport { BaseStore } from \"@langchain/core/stores\";\nimport { AsyncCallerParams } from \"@langchain/core/utils/async_caller\";\n/**\n * Interface for the fields required to initialize an instance of the\n * CacheBackedEmbeddings class.\n */\nexport interface CacheBackedEmbeddingsFields extends AsyncCallerParams {\n    underlyingEmbeddings: EmbeddingsInterface;\n    documentEmbeddingStore: BaseStore<string, number[]>;\n}\n/**\n * Interface for caching results from embedding models.\n *\n * The interface allows works with any store that implements\n * the abstract store interface accepting keys of type str and values of list of\n * floats.\n *\n * If need be, the interface can be extended to accept other implementations\n * of the value serializer and deserializer, as well as the key encoder.\n * @example\n * ```typescript\n * const underlyingEmbeddings = new OpenAIEmbeddings();\n *\n * const cacheBackedEmbeddings = CacheBackedEmbeddings.fromBytesStore(\n *   underlyingEmbeddings,\n *   new ConvexKVStore({ ctx }),\n *   {\n *     namespace: underlyingEmbeddings.modelName,\n *   },\n * );\n *\n * const loader = new TextLoader(\"./state_of_the_union.txt\");\n * const rawDocuments = await loader.load();\n * const splitter = new RecursiveCharacterTextSplitter({\n *   chunkSize: 1000,\n *   chunkOverlap: 0,\n * });\n * const documents = await splitter.splitDocuments(rawDocuments);\n *\n * let time = Date.now();\n * const vectorstore = await ConvexVectorStore.fromDocuments(\n *   documents,\n *   cacheBackedEmbeddings,\n *   { ctx },\n * );\n * console.log(`Initial creation time: ${Date.now() - time}ms`);\n *\n * time = Date.now();\n * const vectorstore2 = await ConvexVectorStore.fromDocuments(\n *   documents,\n *   cacheBackedEmbeddings,\n *   { ctx },\n * );\n * console.log(`Cached creation time: ${Date.now() - time}ms`);\n *\n * ```\n */\nexport declare class CacheBackedEmbeddings extends Embeddings {\n    protected underlyingEmbeddings: EmbeddingsInterface;\n    protected documentEmbeddingStore: BaseStore<string, number[]>;\n    constructor(fields: CacheBackedEmbeddingsFields);\n    /**\n     * Embed query text.\n     *\n     * This method does not support caching at the moment.\n     *\n     * Support for caching queries is easy to implement, but might make\n     * sense to hold off to see the most common patterns.\n     *\n     * If the cache has an eviction policy, we may need to be a bit more careful\n     * about sharing the cache between documents and queries. Generally,\n     * one is OK evicting query caches, but document caches should be kept.\n     *\n     * @param document The text to embed.\n     * @returns The embedding for the given text.\n     */\n    embedQuery(document: string): Promise<number[]>;\n    /**\n     * Embed a list of texts.\n     *\n     * The method first checks the cache for the embeddings.\n     * If the embeddings are not found, the method uses the underlying embedder\n     * to embed the documents and stores the results in the cache.\n     *\n     * @param documents\n     * @returns A list of embeddings for the given texts.\n     */\n    embedDocuments(documents: string[]): Promise<number[][]>;\n    /**\n     * Create a new CacheBackedEmbeddings instance from another embeddings instance\n     * and a storage instance.\n     * @param underlyingEmbeddings Embeddings used to populate the cache for new documents.\n     * @param documentEmbeddingStore Stores raw document embedding values. Keys are hashes of the document content.\n     * @param options.namespace Optional namespace for store keys.\n     * @returns A new CacheBackedEmbeddings instance.\n     */\n    static fromBytesStore(underlyingEmbeddings: EmbeddingsInterface, documentEmbeddingStore: BaseStore<string, Uint8Array>, options?: {\n        namespace?: string;\n    }): CacheBackedEmbeddings;\n}\n"],"mappings":";;;;;;;;AAOA;;AAC0BA,UADTI,2BAAAA,SAAoCD,iBAC3BH,CAAAA;EAAmB,oBACjBE,EADFF,mBACEE;EAAS,sBAFgBC,EAEzBD,SAFyBC,CAAAA,MAAAA,EAAAA,MAAAA,EAAAA,CAAAA;AAAiB;AAmDtE;;;;;;;;;;;;AAA6D;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;cAAxCE,qBAAAA,SAA8BJ,UAAAA;kCACfD;oCACEE;sBACdE;;;;;;;;;;;;;;;;gCAgBUE;;;;;;;;;;;uCAWOA;;;;;;;;;8CASON,6CAA6CE,kBAAkBK;;MAEvGF"}

package/dist/embeddings/cache_backed.js

@@ -0,0 +1,134 @@
+import { __export } from "../_virtual/rolldown_runtime.js";
+import { EncoderBackedStore } from "../storage/encoder_backed.js";
+import { sha256 } from "@langchain/core/utils/hash";
+import { Embeddings } from "@langchain/core/embeddings";
+
+//#region src/embeddings/cache_backed.ts
+var cache_backed_exports = {};
+__export(cache_backed_exports, { CacheBackedEmbeddings: () => CacheBackedEmbeddings });
+/**
+* Interface for caching results from embedding models.
+*
+* The interface allows works with any store that implements
+* the abstract store interface accepting keys of type str and values of list of
+* floats.
+*
+* If need be, the interface can be extended to accept other implementations
+* of the value serializer and deserializer, as well as the key encoder.
+* @example
+* ```typescript
+* const underlyingEmbeddings = new OpenAIEmbeddings();
+*
+* const cacheBackedEmbeddings = CacheBackedEmbeddings.fromBytesStore(
+*   underlyingEmbeddings,
+*   new ConvexKVStore({ ctx }),
+*   {
+*     namespace: underlyingEmbeddings.modelName,
+*   },
+* );
+*
+* const loader = new TextLoader("./state_of_the_union.txt");
+* const rawDocuments = await loader.load();
+* const splitter = new RecursiveCharacterTextSplitter({
+*   chunkSize: 1000,
+*   chunkOverlap: 0,
+* });
+* const documents = await splitter.splitDocuments(rawDocuments);
+*
+* let time = Date.now();
+* const vectorstore = await ConvexVectorStore.fromDocuments(
+*   documents,
+*   cacheBackedEmbeddings,
+*   { ctx },
+* );
+* console.log(`Initial creation time: ${Date.now() - time}ms`);
+*
+* time = Date.now();
+* const vectorstore2 = await ConvexVectorStore.fromDocuments(
+*   documents,
+*   cacheBackedEmbeddings,
+*   { ctx },
+* );
+* console.log(`Cached creation time: ${Date.now() - time}ms`);
+*
+* ```
+*/
+var CacheBackedEmbeddings = class extends Embeddings {
+	underlyingEmbeddings;
+	documentEmbeddingStore;
+	constructor(fields) {
+		super(fields);
+		this.underlyingEmbeddings = fields.underlyingEmbeddings;
+		this.documentEmbeddingStore = fields.documentEmbeddingStore;
+	}
+	/**
+	* Embed query text.
+	*
+	* This method does not support caching at the moment.
+	*
+	* Support for caching queries is easy to implement, but might make
+	* sense to hold off to see the most common patterns.
+	*
+	* If the cache has an eviction policy, we may need to be a bit more careful
+	* about sharing the cache between documents and queries. Generally,
+	* one is OK evicting query caches, but document caches should be kept.
+	*
+	* @param document The text to embed.
+	* @returns The embedding for the given text.
+	*/
+	async embedQuery(document) {
+		return this.underlyingEmbeddings.embedQuery(document);
+	}
+	/**
+	* Embed a list of texts.
+	*
+	* The method first checks the cache for the embeddings.
+	* If the embeddings are not found, the method uses the underlying embedder
+	* to embed the documents and stores the results in the cache.
+	*
+	* @param documents
+	* @returns A list of embeddings for the given texts.
+	*/
+	async embedDocuments(documents) {
+		const vectors = await this.documentEmbeddingStore.mget(documents);
+		const missingIndicies = [];
+		const missingDocuments = [];
+		for (let i = 0; i < vectors.length; i += 1) if (vectors[i] === void 0) {
+			missingIndicies.push(i);
+			missingDocuments.push(documents[i]);
+		}
+		if (missingDocuments.length) {
+			const missingVectors = await this.underlyingEmbeddings.embedDocuments(missingDocuments);
+			const keyValuePairs = missingDocuments.map((document, i) => [document, missingVectors[i]]);
+			await this.documentEmbeddingStore.mset(keyValuePairs);
+			for (let i = 0; i < missingIndicies.length; i += 1) vectors[missingIndicies[i]] = missingVectors[i];
+		}
+		return vectors;
+	}
+	/**
+	* Create a new CacheBackedEmbeddings instance from another embeddings instance
+	* and a storage instance.
+	* @param underlyingEmbeddings Embeddings used to populate the cache for new documents.
+	* @param documentEmbeddingStore Stores raw document embedding values. Keys are hashes of the document content.
+	* @param options.namespace Optional namespace for store keys.
+	* @returns A new CacheBackedEmbeddings instance.
+	*/
+	static fromBytesStore(underlyingEmbeddings, documentEmbeddingStore, options) {
+		const encoder = new TextEncoder();
+		const decoder = new TextDecoder();
+		const encoderBackedStore = new EncoderBackedStore({
+			store: documentEmbeddingStore,
+			keyEncoder: (key) => (options?.namespace ?? "") + sha256(key),
+			valueSerializer: (value) => encoder.encode(JSON.stringify(value)),
+			valueDeserializer: (serializedValue) => JSON.parse(decoder.decode(serializedValue))
+		});
+		return new this({
+			underlyingEmbeddings,
+			documentEmbeddingStore: encoderBackedStore
+		});
+	}
+};
+
+//#endregion
+export { CacheBackedEmbeddings, cache_backed_exports };
+//# sourceMappingURL=cache_backed.js.map