langchain 0.2.9 → 0.2.11

package/dist/retrievers/tests/hyde.int.test.js

@@ -1,44 +0,0 @@
-import { expect, test } from "@jest/globals";
-import { OpenAIEmbeddings, OpenAI } from "@langchain/openai";
-import { Document } from "@langchain/core/documents";
-import { MemoryVectorStore } from "../../vectorstores/memory.js";
-import { HydeRetriever } from "../hyde.js";
-test("Hyde retriever", async () => {
-    const embeddings = new OpenAIEmbeddings();
-    const vectorStore = new MemoryVectorStore(embeddings);
-    const llm = new OpenAI();
-    const retriever = new HydeRetriever({
-        vectorStore,
-        llm,
-        k: 1,
-    });
-    await vectorStore.addDocuments([
-        "My name is John.",
-        "My name is Bob.",
-        "My favourite food is pizza.",
-        "My favourite food is pasta.",
-    ].map((pageContent) => new Document({ pageContent })));
-    const results = await retriever.getRelevantDocuments("What is my favourite food?");
-    expect(results.length).toBe(1);
-    console.log(results);
-});
-test("Hyde retriever with default prompt template", async () => {
-    const embeddings = new OpenAIEmbeddings();
-    const vectorStore = new MemoryVectorStore(embeddings);
-    const llm = new OpenAI();
-    const retriever = new HydeRetriever({
-        vectorStore,
-        llm,
-        k: 1,
-        promptTemplate: "websearch",
-    });
-    await vectorStore.addDocuments([
-        "My name is John.",
-        "My name is Bob.",
-        "My favourite food is pizza.",
-        "My favourite food is pasta.",
-    ].map((pageContent) => new Document({ pageContent })));
-    const results = await retriever.getRelevantDocuments("What is my favourite food?");
-    expect(results.length).toBe(1);
-    console.log(results);
-});

package/dist/retrievers/tests/matryoshka_retriever.int.test.d.ts

@@ -1 +0,0 @@
-export {};

package/dist/retrievers/tests/matryoshka_retriever.int.test.js

@@ -1,113 +0,0 @@
-import { test, expect } from "@jest/globals";
-import { OpenAIEmbeddings } from "@langchain/openai";
-import { faker } from "@faker-js/faker";
-import { Document } from "@langchain/core/documents";
-import { v4 as uuidV4 } from "uuid";
-import { SyntheticEmbeddings } from "@langchain/core/utils/testing";
-import { MatryoshkaRetriever } from "../matryoshka_retriever.js";
-import { MemoryVectorStore } from "../../vectorstores/memory.js";
-test("MatryoshkaRetriever can retrieve", async () => {
-    const smallEmbeddings = new OpenAIEmbeddings({
-        modelName: "text-embedding-3-small",
-        dimensions: 512, // Min num for small
-    });
-    const largeEmbeddings = new OpenAIEmbeddings({
-        modelName: "text-embedding-3-large",
-        dimensions: 3072, // Max num for large
-    });
-    const vectorStore = new MemoryVectorStore(smallEmbeddings);
-    const retriever = new MatryoshkaRetriever({
-        largeEmbeddingModel: largeEmbeddings,
-        vectorStore,
-        largeK: 5,
-    });
-    const irrelevantDocs = Array.from({ length: 250 }).map(() => new Document({
-        pageContent: faker.lorem.words(7),
-        metadata: { id: uuidV4() },
-    }));
-    const relevantDocContents = [
-        "LangChain is an open source github repo",
-        "There are JS and PY versions of the LangChain github repos",
-        "LangGraph is a new open source library by the LangChain team",
-        "LangChain announced GA of LangSmith last week!",
-        "I heart LangChain",
-    ];
-    const relevantDocs = [
-        new Document({
-            pageContent: relevantDocContents[0],
-            metadata: { id: uuidV4() },
-        }),
-        new Document({
-            pageContent: relevantDocContents[1],
-            metadata: { id: uuidV4() },
-        }),
-        new Document({
-            pageContent: relevantDocContents[2],
-            metadata: { id: uuidV4() },
-        }),
-        new Document({
-            pageContent: relevantDocContents[3],
-            metadata: { id: uuidV4() },
-        }),
-        new Document({
-            pageContent: relevantDocContents[4],
-            metadata: { id: uuidV4() },
-        }),
-    ];
-    const allDocs = [...irrelevantDocs, ...relevantDocs];
-    await retriever.addDocuments(allDocs);
-    const query = "What is LangChain?";
-    console.log("Querying documents");
-    const results = await retriever.getRelevantDocuments(query);
-    const retrieverResultContents = new Set(results.map((doc) => doc.pageContent));
-    console.log([...retrieverResultContents]);
-    expect(results.length).toBe(5);
-    expect(retrieverResultContents).toEqual(new Set(relevantDocContents));
-});
-test("Can change number of docs returned (largeK)", async () => {
-    const smallEmbeddings = new SyntheticEmbeddings({
-        vectorSize: 512,
-    });
-    const largeEmbeddings = new SyntheticEmbeddings({
-        vectorSize: 3072,
-    });
-    const vectorStore = new MemoryVectorStore(smallEmbeddings);
-    const docsWithId = Array.from({ length: 10 }).map(() => new Document({
-        pageContent: "hello world",
-        metadata: { id: uuidV4() },
-    }));
-    const retriever = new MatryoshkaRetriever({
-        largeEmbeddingModel: largeEmbeddings,
-        vectorStore,
-        largeK: 10,
-    });
-    await retriever.addDocuments(docsWithId);
-    const query = "hello world";
-    const results = await retriever.getRelevantDocuments(query);
-    expect(results.length).toBe(10);
-});
-test("AddDocunents adds large embeddings metadata field", async () => {
-    const testId = uuidV4();
-    const doc = new Document({
-        pageContent: "hello world",
-        metadata: { id: testId },
-    });
-    const smallEmbeddings = new OpenAIEmbeddings({
-        modelName: "text-embedding-3-small",
-        dimensions: 512, // Min num for small
-    });
-    const largeEmbeddings = new OpenAIEmbeddings({
-        modelName: "text-embedding-3-large",
-        dimensions: 3072, // Max num for large
-    });
-    const vectorStore = new MemoryVectorStore(smallEmbeddings);
-    const retriever = new MatryoshkaRetriever({
-        largeEmbeddingModel: largeEmbeddings,
-        vectorStore,
-    });
-    await retriever.addDocuments([doc]);
-    const relevantDocs = await retriever.getRelevantDocuments("hello world");
-    expect(relevantDocs[0].metadata.id).toBe(testId);
-    expect(relevantDocs[0].metadata[retriever.largeEmbeddingKey]).toBeDefined();
-    expect(JSON.parse(relevantDocs[0].metadata[retriever.largeEmbeddingKey]).length).toBe(3072);
-});

package/dist/retrievers/tests/multi_query.int.test.d.ts

@@ -1 +0,0 @@
-export {};

package/dist/retrievers/tests/multi_query.int.test.js

@@ -1,45 +0,0 @@
-import { expect, test } from "@jest/globals";
-import { CohereEmbeddings } from "@langchain/cohere";
-import { ChatOpenAI } from "@langchain/openai";
-import { MemoryVectorStore } from "../../vectorstores/memory.js";
-import { MultiQueryRetriever } from "../multi_query.js";
-test("Should work with a question input", async () => {
-    const vectorstore = await MemoryVectorStore.fromTexts([
-        "Buildings are made out of brick",
-        "Buildings are made out of wood",
-        "Buildings are made out of stone",
-        "Cars are made out of metal",
-        "Cars are made out of plastic",
-        "mitochondria is the powerhouse of the cell",
-        "mitochondria is made of lipids",
-    ], [{ id: 1 }, { id: 2 }, { id: 3 }, { id: 4 }, { id: 5 }], new CohereEmbeddings({ model: "embed-english-v3.0" }));
-    const model = new ChatOpenAI({});
-    const retriever = MultiQueryRetriever.fromLLM({
-        llm: model,
-        retriever: vectorstore.asRetriever(),
-        verbose: true,
-    });
-    const query = "What are mitochondria made of?";
-    const retrievedDocs = await retriever.getRelevantDocuments(query);
-    expect(retrievedDocs[0].pageContent).toContain("mitochondria");
-});
-test("Should work with a keyword", async () => {
-    const vectorstore = await MemoryVectorStore.fromTexts([
-        "Buildings are made out of brick",
-        "Buildings are made out of wood",
-        "Buildings are made out of stone",
-        "Cars are made out of metal",
-        "Cars are made out of plastic",
-        "mitochondria is the powerhouse of the cell",
-        "mitochondria is made of lipids",
-    ], [{ id: 1 }, { id: 2 }, { id: 3 }, { id: 4 }, { id: 5 }], new CohereEmbeddings({ model: "embed-english-v3.0" }));
-    const model = new ChatOpenAI({});
-    const retriever = MultiQueryRetriever.fromLLM({
-        llm: model,
-        retriever: vectorstore.asRetriever(),
-        verbose: true,
-    });
-    const query = "cars";
-    const retrievedDocs = await retriever.getRelevantDocuments(query);
-    expect(retrievedDocs[0].pageContent).toContain("Cars");
-});

package/dist/retrievers/tests/parent_document.int.test.d.ts

@@ -1 +0,0 @@
-export {};

package/dist/retrievers/tests/parent_document.int.test.js

@@ -1,122 +0,0 @@
-import { expect, test } from "@jest/globals";
-import { OpenAIEmbeddings } from "@langchain/openai";
-import { InMemoryDocstore } from "../../stores/doc/in_memory.js";
-import { TextLoader } from "../../document_loaders/fs/text.js";
-import { InMemoryStore } from "../../storage/in_memory.js";
-import { MemoryVectorStore } from "../../vectorstores/memory.js";
-import { ParentDocumentRetriever } from "../parent_document.js";
-import { RecursiveCharacterTextSplitter } from "../../text_splitter.js";
-import { ScoreThresholdRetriever } from "../score_threshold.js";
-test("Should return the full document if an unsplit parent document has been added", async () => {
-    const vectorstore = new MemoryVectorStore(new OpenAIEmbeddings());
-    const retriever = new ParentDocumentRetriever({
-        vectorstore,
-        docstore: new InMemoryStore(),
-        childSplitter: new RecursiveCharacterTextSplitter({
-            chunkOverlap: 0,
-            chunkSize: 100,
-        }),
-    });
-    const docs = await new TextLoader("../examples/state_of_the_union.txt").load();
-    await retriever.addDocuments(docs);
-    const query = "justice breyer";
-    const retrievedDocs = await retriever.getRelevantDocuments(query);
-    expect(retrievedDocs.length).toEqual(1);
-    expect(retrievedDocs[0].pageContent.length).toBeGreaterThan(1000);
-});
-test("Should return a part of a document if a parent splitter is passed", async () => {
-    const vectorstore = new MemoryVectorStore(new OpenAIEmbeddings());
-    const docstore = new InMemoryStore();
-    const retriever = new ParentDocumentRetriever({
-        vectorstore,
-        docstore,
-        parentSplitter: new RecursiveCharacterTextSplitter({
-            chunkOverlap: 0,
-            chunkSize: 500,
-        }),
-        childSplitter: new RecursiveCharacterTextSplitter({
-            chunkOverlap: 0,
-            chunkSize: 50,
-        }),
-    });
-    const docs = await new TextLoader("../examples/state_of_the_union.txt").load();
-    await retriever.addDocuments(docs);
-    const query = "justice breyer";
-    const retrievedDocs = await retriever.getRelevantDocuments(query);
-    const vectorstoreRetreivedDocs = await vectorstore.similaritySearch("justice breyer");
-    console.log(vectorstoreRetreivedDocs, vectorstoreRetreivedDocs.length);
-    console.log(retrievedDocs);
-    expect(retrievedDocs.length).toBeGreaterThan(1);
-    expect(retrievedDocs[0].pageContent.length).toBeGreaterThan(100);
-});
-test("Should work with a backwards compatible docstore too", async () => {
-    const vectorstore = new MemoryVectorStore(new OpenAIEmbeddings());
-    const retriever = new ParentDocumentRetriever({
-        vectorstore,
-        docstore: new InMemoryDocstore(),
-        childSplitter: new RecursiveCharacterTextSplitter({
-            chunkOverlap: 0,
-            chunkSize: 100,
-        }),
-    });
-    const docs = await new TextLoader("../examples/state_of_the_union.txt").load();
-    await retriever.addDocuments(docs);
-    const query = "justice breyer";
-    const retrievedDocs = await retriever.getRelevantDocuments(query);
-    expect(retrievedDocs.length).toEqual(1);
-    expect(retrievedDocs[0].pageContent.length).toBeGreaterThan(1000);
-});
-test("Should return a part of a document if a parent splitter is passed", async () => {
-    const vectorstore = new MemoryVectorStore(new OpenAIEmbeddings());
-    const docstore = new InMemoryStore();
-    const retriever = new ParentDocumentRetriever({
-        vectorstore,
-        docstore,
-        parentSplitter: new RecursiveCharacterTextSplitter({
-            chunkOverlap: 0,
-            chunkSize: 500,
-        }),
-        childSplitter: new RecursiveCharacterTextSplitter({
-            chunkOverlap: 0,
-            chunkSize: 50,
-        }),
-    });
-    const docs = await new TextLoader("../examples/state_of_the_union.txt").load();
-    await retriever.addDocuments(docs);
-    const query = "justice breyer";
-    const retrievedDocs = await retriever.getRelevantDocuments(query);
-    const vectorstoreRetreivedDocs = await vectorstore.similaritySearch("justice breyer");
-    console.log(vectorstoreRetreivedDocs, vectorstoreRetreivedDocs.length);
-    console.log(retrievedDocs);
-    expect(retrievedDocs.length).toBeGreaterThan(1);
-    expect(retrievedDocs[0].pageContent.length).toBeGreaterThan(100);
-});
-test("Should use a custom retriever to retrieve one doc", async () => {
-    const vectorstore = new MemoryVectorStore(new OpenAIEmbeddings());
-    const docstore = new InMemoryStore();
-    const childDocumentRetriever = ScoreThresholdRetriever.fromVectorStore(vectorstore, {
-        minSimilarityScore: 0.01,
-        maxK: 1, // Only return the top result
-    });
-    const retriever = new ParentDocumentRetriever({
-        vectorstore,
-        docstore,
-        childDocumentRetriever,
-        parentSplitter: new RecursiveCharacterTextSplitter({
-            chunkOverlap: 0,
-            chunkSize: 500,
-        }),
-        childSplitter: new RecursiveCharacterTextSplitter({
-            chunkOverlap: 0,
-            chunkSize: 50,
-        }),
-    });
-    const docs = await new TextLoader("../examples/state_of_the_union.txt").load();
-    await retriever.addDocuments(docs);
-    const query = "justice breyer";
-    const retrievedDocs = await retriever.getRelevantDocuments(query);
-    const vectorstoreRetreivedDocs = await vectorstore.similaritySearch("justice breyer");
-    console.log(vectorstoreRetreivedDocs, vectorstoreRetreivedDocs.length);
-    console.log(retrievedDocs);
-    expect(retrievedDocs).toHaveLength(1);
-});

package/dist/retrievers/tests/score_threshold.int.test.d.ts

@@ -1 +0,0 @@
-export {};

package/dist/retrievers/tests/score_threshold.int.test.js

@@ -1,83 +0,0 @@
-/* eslint-disable no-process-env */
-import { expect, test } from "@jest/globals";
-import { OpenAIEmbeddings, ChatOpenAI } from "@langchain/openai";
-import { ConversationalRetrievalQAChain } from "../../chains/conversational_retrieval_chain.js";
-import { BufferMemory } from "../../memory/buffer_memory.js";
-import { MemoryVectorStore } from "../../vectorstores/memory.js";
-import { ScoreThresholdRetriever } from "../score_threshold.js";
-test("ConversationalRetrievalQAChain.fromLLM should use its vector store recursively until it gets all the similar results with the minimum similarity score provided", async () => {
-    const vectorStore = await MemoryVectorStore.fromTexts([
-        "Buildings are made out of brick",
-        "Buildings are made out of wood",
-        "Buildings are made out of stone",
-        "Cars are made out of metal",
-        "Cars are made out of plastic",
-    ], [{ id: 1 }, { id: 2 }, { id: 3 }, { id: 4 }, { id: 5 }], new OpenAIEmbeddings());
-    const model = new ChatOpenAI({
-        modelName: "gpt-3.5-turbo",
-        temperature: 0,
-    });
-    const chain = ConversationalRetrievalQAChain.fromLLM(model, ScoreThresholdRetriever.fromVectorStore(vectorStore, {
-        minSimilarityScore: 0.9,
-        kIncrement: 1,
-    }), {
-        returnSourceDocuments: true,
-        memory: new BufferMemory({
-            memoryKey: "chat_history",
-            inputKey: "question",
-            outputKey: "text",
-        }),
-    });
-    const res = await chain.invoke({
-        question: "Buildings are made out of what?",
-    });
-    console.log("response:", res);
-    expect(res).toEqual(expect.objectContaining({
-        text: expect.any(String),
-        sourceDocuments: expect.arrayContaining([
-            expect.objectContaining({
-                metadata: expect.objectContaining({
-                    id: 1,
-                }),
-            }),
-            expect.objectContaining({
-                metadata: expect.objectContaining({
-                    id: 2,
-                }),
-            }),
-            expect.objectContaining({
-                metadata: expect.objectContaining({
-                    id: 3,
-                }),
-            }),
-        ]),
-    }));
-});
-test("ConversationalRetrievalQAChain.fromLLM should use its vector store to get up to X results that matches the provided similarity score", async () => {
-    const vectorStore = await MemoryVectorStore.fromTexts([
-        "Buildings are made out of brick",
-        "Buildings are made out of wood",
-        "Buildings are made out of stone",
-        "Cars are made out of metal",
-        "Cars are made out of plastic",
-    ], [{ id: 1 }, { id: 2 }, { id: 3 }, { id: 4 }, { id: 5 }], new OpenAIEmbeddings());
-    const model = new ChatOpenAI({
-        modelName: "gpt-3.5-turbo",
-        temperature: 0,
-    });
-    const chain = ConversationalRetrievalQAChain.fromLLM(model, ScoreThresholdRetriever.fromVectorStore(vectorStore, {
-        minSimilarityScore: 0.9,
-        maxK: 2,
-    }), {
-        returnSourceDocuments: true,
-        memory: new BufferMemory({
-            memoryKey: "chat_history",
-            inputKey: "question",
-            outputKey: "text",
-        }),
-    });
-    const res = await chain.call({
-        question: "Buildings are made out of what?",
-    });
-    expect(res.sourceDocuments).toHaveLength(2);
-});

package/dist/retrievers/tests/time_weighted.test.d.ts

@@ -1 +0,0 @@
-export {};