Support delete vectorstore operations (#1830)

* Adds optional id field to document base class to support upsert and delete operations

* Rehydrate retrieved Chroma docs with ids

* Allow ids or a filter type to be passed into vector store delete method

* Remove id from TypeORM Document model

* Revert base document class changes and use pinecone precendent to pass delete args

* Add deletion tests

* Update tests

* Change vectorstore add docs methods to accept an options object, update docs

* Update VectorStoreRetriever interface

docs/docs/modules/indexes/vector_stores/index.mdx

@@ -14,9 +14,15 @@ A vector store is a particular type of database optimized for storing documents
 ```typescript
 interface VectorStore {
   /**
-   * Add more documents to an existing VectorStore
+   * Add more documents to an existing VectorStore.
+   * Some providers support additional parameters, e.g. to associate custom ids
+   * with added documents or to change the batch size of bulk inserts.
+   * Returns an array of ids for the documents or nothing.
    */
-  addDocuments(documents: Document[]): Promise<void>;
+  addDocuments(
+    documents: Document[],
+    options?: Record<string, any>
+  ): Promise<string[] | void>;
 
   /**
    * Search for the most similar documents to a query
@@ -42,11 +48,21 @@ interface VectorStore {
    */
   asRetriever(k?: number): BaseRetriever;
 
+  /**
+   * Delete embedded documents from the vector store matching the passed in parameter.
+   * Not supported by every provider.
+   */
+  delete(params?: Record<string, any>): Promise<void>;
+
   /**
    * Advanced: Add more documents to an existing VectorStore,
    * when you already have their embeddings
    */
-  addVectors(vectors: number[][], documents: Document[]): Promise<void>;
+  addVectors(
+    vectors: number[][],
+    documents: Document[],
+    options?: Record<string, any>
+  ): Promise<string[] | void>;
 
   /**
    * Advanced: Search for the most similar documents to a query,

docs/docs/modules/indexes/vector_stores/integrations/chroma.mdx

@@ -74,3 +74,9 @@ import FromTexts from "@examples/indexes/vector_stores/chroma/fromTexts.ts";
 import Search from "@examples/indexes/vector_stores/chroma/search.ts";
 
 <CodeBlock language="typescript">{Search}</CodeBlock>
+
+## Usage, delete docs
+
+import Delete from "@examples/indexes/vector_stores/chroma/delete.ts";
+
+<CodeBlock language="typescript">{Delete}</CodeBlock>

docs/docs/modules/indexes/vector_stores/integrations/pinecone.md

@@ -8,10 +8,10 @@ sidebar_class_name: node-only
 Only available on Node.js.
 :::
 
-Langchain.js accepts [@pinecone-database/pinecone](https://docs.pinecone.io/docs/node-client) as the client for Pinecone vectorstore. Install the library with
+LangChain.js accepts [@pinecone-database/pinecone](https://docs.pinecone.io/docs/node-client) as the client for Pinecone vectorstore. Install the library with:
 
 ```bash npm2yarn
-npm install -S dotenv langchain @pinecone-database/pinecone
+npm install -S dotenv @pinecone-database/pinecone
 ```
 
 ## Index docs
@@ -114,3 +114,76 @@ console.log(response);
 }
 */
 ```
+
+## Delete docs
+
+```typescript
+import { PineconeClient } from "@pinecone-database/pinecone";
+import * as dotenv from "dotenv";
+import { Document } from "langchain/document";
+import { OpenAIEmbeddings } from "langchain/embeddings/openai";
+import { PineconeStore } from "langchain/vectorstores/pinecone";
+
+dotenv.config();
+
+const client = new PineconeClient();
+await client.init({
+  apiKey: process.env.PINECONE_API_KEY,
+  environment: process.env.PINECONE_ENVIRONMENT,
+});
+const pineconeIndex = client.Index(process.env.PINECONE_INDEX);
+const embeddings = new OpenAIEmbeddings();
+const pineconeStore = new PineconeStore(embeddings, { pineconeIndex });
+
+const docs = [
+  new Document({
+    metadata: { foo: "bar" },
+    pageContent: "pinecone is a vector db",
+  }),
+  new Document({
+    metadata: { foo: "bar" },
+    pageContent: "the quick brown fox jumped over the lazy dog",
+  }),
+  new Document({
+    metadata: { baz: "qux" },
+    pageContent: "lorem ipsum dolor sit amet",
+  }),
+  new Document({
+    metadata: { baz: "qux" },
+    pageContent: "pinecones are the woody fruiting body and of a pine tree",
+  }),
+];
+
+const ids = await pineconeStore.addDocuments(docs);
+
+const results = await pineconeStore.similaritySearch(pageContent, 2, {
+  foo: "bar",
+});
+
+console.log(results);
+/*
+[
+  Document {
+    pageContent: 'pinecone is a vector db',
+    metadata: { foo: 'bar' },
+  },
+  Document {
+    pageContent: "the quick brown fox jumped over the lazy dog",
+    metadata: { foo: "bar" },
+  }
+]
+*/
+
+await pineconeStore.delete({
+  ids: [ids[0], ids[1]],
+});
+
+const results2 = await pineconeStore.similaritySearch(pageContent, 2, {
+  foo: "bar",
+});
+
+console.log(results2);
+/*
+  []
+*/
+```

docs/docs/modules/indexes/vector_stores/integrations/supabase.mdx

@@ -61,6 +61,7 @@ import CodeBlock from "@theme/CodeBlock";
 import Example from "@examples/indexes/vector_stores/supabase.ts";
 import MetadataFilterExample from "@examples/indexes/vector_stores/supabase_with_metadata_filter.ts";
 import MetadataQueryBuilderFilterExample from "@examples/indexes/vector_stores/supabase_with_query_builder_metadata_filter.ts";
+import DeletionExample from "@examples/indexes/vector_stores/supabase_deletion.ts";
 
 ### Standard Usage
 
@@ -81,3 +82,7 @@ Given the above `match_documents` Postgres function, you can also pass a filter
 You can also use query builder-style filtering similar to how [the Supabase JavaScript library works](https://supabase.com/docs/reference/javascript/using-filters) instead of passing an object. Note that since most of the filter properties are in the metadata column, you need to use arrow operators (`->` for integer or `->>` for text) as defined in [Postgrest API documentation](https://postgrest.org/en/stable/references/api/tables_views.html?highlight=operators#json-columns) and specify the data type of the property (e.g. the column should look something like `metadata->some_int_value::int`).
 
 <CodeBlock language="typescript">{MetadataQueryBuilderFilterExample}</CodeBlock>
+
+### Document deletion
+
+<CodeBlock language="typescript">{DeletionExample}</CodeBlock>

docs/docs/modules/indexes/vector_stores/integrations/weaviate.mdx

@@ -29,3 +29,9 @@ import InsertExample from "@examples/indexes/vector_stores/weaviate_fromTexts.ts
 import QueryExample from "@examples/indexes/vector_stores/weaviate_search.ts";
 
 <CodeBlock language="typescript">{QueryExample}</CodeBlock>
+
+## Usage delete documents
+
+import DeleteExample from "@examples/indexes/vector_stores/weaviate_delete.ts";
+
+<CodeBlock language="typescript">{DeleteExample}</CodeBlock>

examples/src/indexes/vector_stores/chroma/delete.ts

@@ -0,0 +1,69 @@
+import { Chroma } from "langchain/vectorstores/chroma";
+import { OpenAIEmbeddings } from "langchain/embeddings/openai";
+
+const embeddings = new OpenAIEmbeddings();
+const vectorStore = new Chroma(embeddings, {
+  collectionName: "godel-escher-bach",
+});
+
+const documents = [
+  {
+    pageContent: `Tortoise: Labyrinth? Labyrinth? Could it Are we in the notorious Little
+    Harmonic Labyrinth of the dreaded Majotaur?`,
+    metadata: {
+      speaker: "Tortoise",
+    },
+  },
+  {
+    pageContent: "Achilles: Yiikes! What is that?",
+    metadata: {
+      speaker: "Achilles",
+    },
+  },
+  {
+    pageContent: `Tortoise: They say-although I person never believed it myself-that an I
+    Majotaur has created a tiny labyrinth sits in a pit in the middle of
+    it, waiting innocent victims to get lost in its fears complexity.
+    Then, when they wander and dazed into the center, he laughs and
+    laughs at them-so hard, that he laughs them to death!`,
+    metadata: {
+      speaker: "Tortoise",
+    },
+  },
+  {
+    pageContent: "Achilles: Oh, no!",
+    metadata: {
+      speaker: "Achilles",
+    },
+  },
+  {
+    pageContent: "Tortoise: But it's only a myth. Courage, Achilles.",
+    metadata: {
+      speaker: "Tortoise",
+    },
+  },
+];
+
+const ids = await vectorStore.addDocuments(documents);
+
+const response = await vectorStore.similaritySearch("scared", 2);
+console.log(response);
+/*
+[
+  Document { pageContent: 'Achilles: Oh, no!', metadata: {} },
+  Document {
+    pageContent: 'Achilles: Yiikes! What is that?',
+    metadata: { id: 1 }
+  }
+]
+*/
+
+// You can also pass a "filter" parameter instead
+await vectorStore.delete({ ids });
+
+const response2 = await vectorStore.similaritySearch("scared", 2);
+console.log(response2);
+
+/*
+  []
+*/

examples/src/indexes/vector_stores/elasticsearch/elasticsearch.ts

@@ -54,10 +54,11 @@ export async function run() {
     baseOptions: { temperature: 0 },
   });
 
-  await ElasticVectorSearch.fromDocuments(docs, embeddings, clientArgs);
-
+  // await ElasticVectorSearch.fromDocuments(docs, embeddings, clientArgs);
   const vectorStore = new ElasticVectorSearch(embeddings, clientArgs);
 
+  const ids = await vectorStore.addDocuments(docs);
+
   /* Search the vector DB independently with meta filters */
   const results = await vectorStore.similaritySearch("fox jump", 1);
   console.log(JSON.stringify(results, null, 2));
@@ -93,4 +94,14 @@ export async function run() {
       ]
     }
     */
+
+  await vectorStore.delete({ ids });
+
+  const response2 = await chain.call({ query: "What is Elasticsearch?" });
+
+  console.log(JSON.stringify(response2, null, 2));
+
+  /*
+    []
+  */
 }

examples/src/indexes/vector_stores/supabase_deletion.ts

@@ -0,0 +1,49 @@
+import { SupabaseVectorStore } from "langchain/vectorstores/supabase";
+import { OpenAIEmbeddings } from "langchain/embeddings/openai";
+import { createClient } from "@supabase/supabase-js";
+
+// First, follow set-up instructions at
+// https://js.langchain.com/docs/modules/indexes/vector_stores/integrations/supabase
+
+const privateKey = process.env.SUPABASE_PRIVATE_KEY;
+if (!privateKey) throw new Error(`Expected env var SUPABASE_PRIVATE_KEY`);
+
+const url = process.env.SUPABASE_URL;
+if (!url) throw new Error(`Expected env var SUPABASE_URL`);
+
+export const run = async () => {
+  const client = createClient(url, privateKey);
+
+  const embeddings = new OpenAIEmbeddings();
+
+  const store = new SupabaseVectorStore(embeddings, {
+    client,
+    tableName: "documents",
+  });
+
+  const docs = [
+    { pageContent: "hello", metadata: { b: 1, c: 9, stuff: "right" } },
+    { pageContent: "hello", metadata: { b: 1, c: 9, stuff: "wrong" } },
+  ];
+
+  const ids = await store.addDocuments(docs);
+
+  const resultA = await store.similaritySearch("hello", 2);
+  console.log(resultA);
+
+  /*
+    [
+      Document { pageContent: "hello", metadata: { b: 1, c: 9, stuff: "right" } },
+      Document { pageContent: "hello", metadata: { b: 1, c: 9, stuff: "wrong" } },
+    ]
+  */
+
+  await store.delete({ ids });
+
+  const resultB = await store.similaritySearch("hello", 2);
+  console.log(resultB);
+
+  /*
+    []
+  */
+};

examples/src/indexes/vector_stores/weaviate_delete.ts

@@ -0,0 +1,41 @@
+/* eslint-disable @typescript-eslint/no-explicit-any */
+import weaviate from "weaviate-ts-client";
+import { WeaviateStore } from "langchain/vectorstores/weaviate";
+import { OpenAIEmbeddings } from "langchain/embeddings/openai";
+
+export async function run() {
+  // Something wrong with the weaviate-ts-client types, so we need to disable
+  const client = (weaviate as any).client({
+    scheme: process.env.WEAVIATE_SCHEME || "https",
+    host: process.env.WEAVIATE_HOST || "localhost",
+    apiKey: new (weaviate as any).ApiKey(
+      process.env.WEAVIATE_API_KEY || "default"
+    ),
+  });
+
+  // Create a store for an existing index
+  const store = await WeaviateStore.fromExistingIndex(new OpenAIEmbeddings(), {
+    client,
+    indexName: "Test",
+    metadataKeys: ["foo"],
+  });
+
+  const docs = [{ pageContent: "see ya!", metadata: { foo: "bar" } }];
+
+  const ids = await store.addDocuments(docs);
+
+  // Search the index without any filters
+  const results = await store.similaritySearch("see ya!", 1);
+  console.log(results);
+  /*
+  [ Document { pageContent: 'see ya!', metadata: { foo: 'bar' } } ]
+  */
+
+  await store.delete({ ids });
+
+  const results2 = await store.similaritySearch("see ya!", 1);
+  console.log(results2);
+  /*
+  []
+  */
+}

langchain/package.json

@@ -516,7 +516,7 @@
     "apify-client": "^2.7.1",
     "axios": "^0.26.0",
     "cheerio": "^1.0.0-rc.12",
-    "chromadb": "^1.5.2",
+    "chromadb": "^1.5.3",
     "cohere-ai": "^5.0.2",
     "d3-dsv": "^2.0.0",
     "dotenv": "^16.0.3",
@@ -588,7 +588,7 @@
     "apify-client": "^2.7.1",
     "axios": "*",
     "cheerio": "^1.0.0-rc.12",
-    "chromadb": "^1.5.2",
+    "chromadb": "^1.5.3",
     "cohere-ai": "^5.0.2",
     "d3-dsv": "^2.0.0",
     "epub2": "^3.0.1",