langchain 0.0.196 → 0.0.197-rc.0

package/dist/vectorstores/momento_vector_index.d.ts

@@ -1,7 +1,7 @@
 import { IVectorIndexClient } from "@gomomento/sdk-core";
 import { Document } from "../document.js";
 import { Embeddings } from "../embeddings/base.js";
-import { VectorStore } from "./base.js";
+import { MaxMarginalRelevanceSearchOptions, VectorStore } from "./base.js";
 export interface DocumentProps {
     ids: string[];
 }
@@ -103,6 +103,22 @@ export declare class MomentoVectorIndex extends VectorStore {
      * to the query vector.
      */
     similaritySearchVectorWithScore(query: number[], k: number): Promise<[Document<Record<string, any>>, number][]>;
+    /**
+     * Return documents selected using the maximal marginal relevance.
+     * Maximal marginal relevance optimizes for similarity to the query AND diversity
+     * among selected documents.
+     *
+     * @param {string} query - Text to look up documents similar to.
+     * @param {number} options.k - Number of documents to return.
+     * @param {number} options.fetchK - Number of documents to fetch before passing to the MMR algorithm.
+     * @param {number} options.lambda - Number between 0 and 1 that determines the degree of diversity among the results,
+     *                 where 0 corresponds to maximum diversity and 1 to minimum diversity.
+     * @param {this["FilterType"]} options.filter - Optional filter
+     * @param _callbacks
+     *
+     * @returns {Promise<Document[]>} - List of documents selected by maximal marginal relevance.
+     */
+    maxMarginalRelevanceSearch(query: string, options: MaxMarginalRelevanceSearchOptions<this["FilterType"]>): Promise<Document[]>;
     /**
      * Stores the documents in the index.
      *

package/dist/vectorstores/momento_vector_index.js

@@ -1,9 +1,10 @@
 /* eslint-disable no-instanceof/no-instanceof */
 /* eslint-disable @typescript-eslint/no-explicit-any */
-import { ALL_VECTOR_METADATA, CreateVectorIndex, VectorUpsertItemBatch, VectorDeleteItemBatch, VectorSearch, } from "@gomomento/sdk-core";
+import { ALL_VECTOR_METADATA, CreateVectorIndex, VectorUpsertItemBatch, VectorDeleteItemBatch, VectorSearch, VectorSearchAndFetchVectors, } from "@gomomento/sdk-core";
 import * as uuid from "uuid";
 import { Document } from "../document.js";
 import { VectorStore } from "./base.js";
+import { maximalMarginalRelevance } from "../util/math.js";
 /**
  * A vector store that uses the Momento Vector Index.
  *
@@ -208,6 +209,44 @@ export class MomentoVectorIndex extends VectorStore {
             throw new Error(`Unknown response type: ${response.toString()}`);
         }
     }
+    /**
+     * Return documents selected using the maximal marginal relevance.
+     * Maximal marginal relevance optimizes for similarity to the query AND diversity
+     * among selected documents.
+     *
+     * @param {string} query - Text to look up documents similar to.
+     * @param {number} options.k - Number of documents to return.
+     * @param {number} options.fetchK - Number of documents to fetch before passing to the MMR algorithm.
+     * @param {number} options.lambda - Number between 0 and 1 that determines the degree of diversity among the results,
+     *                 where 0 corresponds to maximum diversity and 1 to minimum diversity.
+     * @param {this["FilterType"]} options.filter - Optional filter
+     * @param _callbacks
+     *
+     * @returns {Promise<Document[]>} - List of documents selected by maximal marginal relevance.
+     */
+    async maxMarginalRelevanceSearch(query, options) {
+        const queryEmbedding = await this.embeddings.embedQuery(query);
+        const response = await this.client.searchAndFetchVectors(this.indexName, queryEmbedding, { topK: options.fetchK ?? 20, metadataFields: ALL_VECTOR_METADATA });
+        if (response instanceof VectorSearchAndFetchVectors.Success) {
+            const hits = response.hits();
+            // Gather the embeddings of the search results
+            const embeddingList = hits.map((hit) => hit.vector);
+            // Gather the ids of the most relevant results when applying MMR
+            const mmrIndexes = maximalMarginalRelevance(queryEmbedding, embeddingList, options.lambda, options.k);
+            const finalResult = mmrIndexes.map((index) => {
+                const hit = hits[index];
+                const { [this.textField]: pageContent, ...metadata } = hit.metadata;
+                return new Document({ metadata, pageContent: pageContent });
+            });
+            return finalResult;
+        }
+        else if (response instanceof VectorSearchAndFetchVectors.Error) {
+            throw new Error(response.toString());
+        }
+        else {
+            throw new Error(`Unknown response type: ${response.toString()}`);
+        }
+    }
     /**
      * Stores the documents in the index.
      *

package/dist/vectorstores/mongodb_atlas.cjs

@@ -99,7 +99,7 @@ class MongoDBAtlasVectorSearch extends base_js_1.VectorStore {
         const pipeline = [
             {
                 $vectorSearch: {
-                    queryVector: query,
+                    queryVector: MongoDBAtlasVectorSearch.fixArrayPrecision(query),
                     index: this.indexName,
                     path: this.embeddingKey,
                     limit: k,
@@ -148,7 +148,7 @@ class MongoDBAtlasVectorSearch extends base_js_1.VectorStore {
             ...filter,
             includeEmbeddings: true,
         };
-        const resultDocs = await this.similaritySearchVectorWithScore(queryEmbedding, fetchK, includeEmbeddingsFilter);
+        const resultDocs = await this.similaritySearchVectorWithScore(MongoDBAtlasVectorSearch.fixArrayPrecision(queryEmbedding), fetchK, includeEmbeddingsFilter);
         const embeddingList = resultDocs.map((doc) => doc[0].metadata[this.embeddingKey]);
         const mmrIndexes = (0, math_js_1.maximalMarginalRelevance)(queryEmbedding, embeddingList, lambda, k);
         return mmrIndexes.map((idx) => {
@@ -196,5 +196,25 @@ class MongoDBAtlasVectorSearch extends base_js_1.VectorStore {
         await instance.addDocuments(docs);
         return instance;
     }
+    /**
+     * Static method to fix the precision of the array that ensures that
+     * every number in this array is always float when casted to other types.
+     * This is needed since MongoDB Atlas Vector Search does not cast integer
+     * inside vector search to float automatically.
+     * This method shall introduce a hint of error but should be safe to use
+     * since introduced error is very small, only applies to integer numbers
+     * returned by embeddings, and most embeddings shall not have precision
+     * as high as 15 decimal places.
+     * @param array Array of number to be fixed.
+     * @returns
+     */
+    static fixArrayPrecision(array) {
+        return array.map((value) => {
+            if (Number.isInteger(value)) {
+                return value + 0.000000000000001;
+            }
+            return value;
+        });
+    }
 }
 exports.MongoDBAtlasVectorSearch = MongoDBAtlasVectorSearch;

package/dist/vectorstores/mongodb_atlas.d.ts

@@ -101,5 +101,18 @@ export declare class MongoDBAtlasVectorSearch extends VectorStore {
      * @returns Promise that resolves to a new instance of MongoDBAtlasVectorSearch.
      */
     static fromDocuments(docs: Document[], embeddings: Embeddings, dbConfig: MongoDBAtlasVectorSearchLibArgs): Promise<MongoDBAtlasVectorSearch>;
+    /**
+     * Static method to fix the precision of the array that ensures that
+     * every number in this array is always float when casted to other types.
+     * This is needed since MongoDB Atlas Vector Search does not cast integer
+     * inside vector search to float automatically.
+     * This method shall introduce a hint of error but should be safe to use
+     * since introduced error is very small, only applies to integer numbers
+     * returned by embeddings, and most embeddings shall not have precision
+     * as high as 15 decimal places.
+     * @param array Array of number to be fixed.
+     * @returns
+     */
+    static fixArrayPrecision(array: number[]): number[];
 }
 export {};

package/dist/vectorstores/mongodb_atlas.js

@@ -96,7 +96,7 @@ export class MongoDBAtlasVectorSearch extends VectorStore {
         const pipeline = [
             {
                 $vectorSearch: {
-                    queryVector: query,
+                    queryVector: MongoDBAtlasVectorSearch.fixArrayPrecision(query),
                     index: this.indexName,
                     path: this.embeddingKey,
                     limit: k,
@@ -145,7 +145,7 @@ export class MongoDBAtlasVectorSearch extends VectorStore {
             ...filter,
             includeEmbeddings: true,
         };
-        const resultDocs = await this.similaritySearchVectorWithScore(queryEmbedding, fetchK, includeEmbeddingsFilter);
+        const resultDocs = await this.similaritySearchVectorWithScore(MongoDBAtlasVectorSearch.fixArrayPrecision(queryEmbedding), fetchK, includeEmbeddingsFilter);
         const embeddingList = resultDocs.map((doc) => doc[0].metadata[this.embeddingKey]);
         const mmrIndexes = maximalMarginalRelevance(queryEmbedding, embeddingList, lambda, k);
         return mmrIndexes.map((idx) => {
@@ -193,4 +193,24 @@ export class MongoDBAtlasVectorSearch extends VectorStore {
         await instance.addDocuments(docs);
         return instance;
     }
+    /**
+     * Static method to fix the precision of the array that ensures that
+     * every number in this array is always float when casted to other types.
+     * This is needed since MongoDB Atlas Vector Search does not cast integer
+     * inside vector search to float automatically.
+     * This method shall introduce a hint of error but should be safe to use
+     * since introduced error is very small, only applies to integer numbers
+     * returned by embeddings, and most embeddings shall not have precision
+     * as high as 15 decimal places.
+     * @param array Array of number to be fixed.
+     * @returns
+     */
+    static fixArrayPrecision(array) {
+        return array.map((value) => {
+            if (Number.isInteger(value)) {
+                return value + 0.000000000000001;
+            }
+            return value;
+        });
+    }
 }

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "langchain",
-  "version": "0.0.196",
+  "version": "0.0.197-rc.0",
   "description": "Typescript bindings for langchain",
   "type": "module",
   "engines": {
@@ -859,12 +859,13 @@
     "@faker-js/faker": "^7.6.0",
     "@getmetal/metal-sdk": "^4.0.0",
     "@getzep/zep-js": "^0.9.0",
-    "@gomomento/sdk": "^1.47.1",
-    "@gomomento/sdk-core": "^1.47.1",
+    "@gomomento/sdk": "^1.51.1",
+    "@gomomento/sdk-core": "^1.51.1",
     "@google-ai/generativelanguage": "^0.2.1",
     "@google-cloud/storage": "^6.10.1",
     "@huggingface/inference": "^2.6.4",
     "@jest/globals": "^29.5.0",
+    "@langchain/anthropic": "workspace:*",
     "@mozilla/readability": "^0.4.4",
     "@notionhq/client": "^2.2.10",
     "@opensearch-project/opensearch": "^2.2.0",
@@ -997,9 +998,9 @@
     "@elastic/elasticsearch": "^8.4.0",
     "@getmetal/metal-sdk": "*",
     "@getzep/zep-js": "^0.9.0",
-    "@gomomento/sdk": "^1.47.1",
-    "@gomomento/sdk-core": "^1.47.1",
-    "@gomomento/sdk-web": "^1.47.1",
+    "@gomomento/sdk": "^1.51.1",
+    "@gomomento/sdk-core": "^1.51.1",
+    "@gomomento/sdk-web": "^1.51.1",
     "@google-ai/generativelanguage": "^0.2.1",
     "@google-cloud/storage": "^6.10.1",
     "@huggingface/inference": "^2.6.4",
@@ -1378,14 +1379,14 @@
     }
   },
   "dependencies": {
-    "@anthropic-ai/sdk": "^0.9.1",
+    "@langchain/anthropic": "^0.0.2",
     "binary-extensions": "^2.2.0",
     "expr-eval": "^2.0.2",
     "flat": "^5.0.2",
     "js-tiktoken": "^1.0.7",
     "js-yaml": "^4.1.0",
     "jsonpointer": "^5.0.1",
-    "langchain-core": "^0.0.1",
+    "langchain-core": "^0.0.4",
     "langchainhub": "~0.0.6",
     "langsmith": "~0.0.48",
     "ml-distance": "^4.0.0",