New feat: Self Query Retriever (langchain-ai#1266)

* self-query with meriyah

* removed builtin translators so end-user needs to specify their own translator, but a basic translator that can be used with pinecone and chroma is included and can be imported

* fixed class name typo

* moved meriyah to main dependency so expression parser can work in test-exports-*

* added back self_query to entrypoints

* moved meriyah to dev/peer dep, fixed unused parser types, minor fixes

* renaming expression type to reflect meriyah better

* removed unused lint disable on self_query/base.ts

docs/docs/modules/indexes/retrievers/chroma-self-query.mdx

@@ -0,0 +1,14 @@
+# Self Query Chroma Retriever
+
+The Self Query Retriever, which as the name suggests, has the ability to query itself. Specifically, given any natural language query, the retriever uses a query-constructing LLM chain to write a structured query and then applies that structured query to it's underlying vector store. This allows the retriever to not only use the user-input query for semantic similarity comparison with the contents of stored documents, but to also extract filters from the user query on the metadata of stored documents and to execute those filter.
+
+This example uses Chroma vector store.
+
+## Usage
+
+This example shows how to intialize a `SelfQueryRetriever` with a vector store:
+
+import CodeBlock from "@theme/CodeBlock";
+import Example from "@examples/retrievers/chroma_self_query.ts";
+
+<CodeBlock language="typescript">{Example}</CodeBlock>

docs/docs/modules/indexes/retrievers/pinecone-self-query.mdx

@@ -0,0 +1,14 @@
+# Self Query Pinecone Retriever
+
+The Self Query Retriever, which as the name suggests, has the ability to query itself. Specifically, given any natural language query, the retriever uses a query-constructing LLM chain to write a structured query and then applies that structured query to it's underlying vector store. This allows the retriever to not only use the user-input query for semantic similarity comparison with the contents of stored documents, but to also extract filters from the user query on the metadata of stored documents and to execute those filter.
+
+This example uses Pinecone vector store.
+
+## Usage
+
+This example shows how to intialize a `SelfQueryRetriever` with a vector store:
+
+import CodeBlock from "@theme/CodeBlock";
+import Example from "@examples/retrievers/pinecone_self_query.ts";
+
+<CodeBlock language="typescript">{Example}</CodeBlock>

examples/src/retrievers/chroma_self_query.ts

@@ -0,0 +1,134 @@
+import { AttributeInfo } from "langchain/schema/query_constructor";
+import { Document } from "langchain/document";
+import { OpenAIEmbeddings } from "langchain/embeddings/openai";
+import { SelfQueryRetriever } from "langchain/retrievers/self_query/base";
+import { BasicTranslator } from "langchain/retrievers/self_query/translator";
+import { OpenAI } from "langchain/llms/openai";
+import { Chroma } from "langchain/vectorstores/chroma";
+
+const run = async () => {
+  /**
+   * First, we create a bunch of documents. You can load your own documents here instead.
+   * Each document has a pageContent and a metadata field. Make sure your metadata matches the AttributeInfo below.
+   */
+  const docs = [
+    new Document({
+      pageContent:
+        "A bunch of scientists bring back dinosaurs and mayhem breaks loose",
+      metadata: { year: 1993, rating: 7.7, genre: "science fiction" },
+    }),
+    new Document({
+      pageContent:
+        "Leo DiCaprio gets lost in a dream within a dream within a dream within a ...",
+      metadata: { year: 2010, director: "Christopher Nolan", rating: 8.2 },
+    }),
+    new Document({
+      pageContent:
+        "A psychologist / detective gets lost in a series of dreams within dreams within dreams and Inception reused the idea",
+      metadata: { year: 2006, director: "Satoshi Kon", rating: 8.6 },
+    }),
+    new Document({
+      pageContent:
+        "A bunch of normal-sized women are supremely wholesome and some men pine after them",
+      metadata: { year: 2019, director: "Greta Gerwig", rating: 8.3 },
+    }),
+    new Document({
+      pageContent: "Toys come alive and have a blast doing so",
+      metadata: { year: 1995, genre: "animated" },
+    }),
+    new Document({
+      pageContent:
+        "Three men walk into the Zone, three men walk out of the Zone",
+      metadata: {
+        year: 1979,
+        director: "Andrei Tarkovsky",
+        genre: "science fiction",
+        rating: 9.9,
+      },
+    }),
+  ];
+
+  /**
+   * Next, we define the attributes we want to be able to query on.
+   * in this case, we want to be able to query on the genre, year, director, rating, and length of the movie.
+   * We also provide a description of each attribute and the type of the attribute.
+   * This is used to generate the query prompts.
+   */
+  const attributeInfo: AttributeInfo[] = [
+    {
+      name: "genre",
+      description: "The genre of the movie",
+      type: "string or array of strings",
+    },
+    {
+      name: "year",
+      description: "The year the movie was released",
+      type: "number",
+    },
+    {
+      name: "director",
+      description: "The director of the movie",
+      type: "string",
+    },
+    {
+      name: "rating",
+      description: "The rating of the movie (1-10)",
+      type: "number",
+    },
+    {
+      name: "length",
+      description: "The length of the movie in minutes",
+      type: "number",
+    },
+  ];
+
+  /**
+   * Next, we instantiate a vector store. This is where we store the embeddings of the documents.
+   * We use the Pinecone vector store here, but you can use any vector store you want.
+   * At this point we only support Chroma and Pinecone, but we will add more in the future.
+   * We also need to provide an embeddings object. This is used to embed the documents.
+   */
+  const embeddings = new OpenAIEmbeddings();
+  const llm = new OpenAI();
+  const documentContents = "Brief summary of a movie";
+  const vectorStore = await Chroma.fromDocuments(docs, embeddings, {
+    collectionName: "a-movie-collection",
+  });
+  const selfQueryRetriever = await SelfQueryRetriever.fromLLM({
+    llm,
+    vectorStore,
+    documentContents,
+    attributeInfo,
+    /**
+     * We need to create a basic translator that translates the queries into a
+     * filter format that the vector store can understand. We provide a basic translator
+     * translator here (which works for Chroma and Pinecone), but you can create
+     * your own translator by extending BaseTranslator abstract class. Note that the
+     * vector store needs to support filtering on the metadata attributes you want to
+     * query on.
+     */
+    structuredQueryTranslator: new BasicTranslator(),
+  });
+
+  /**
+   * Now we can query the vector store.
+   * We can ask questions like "Which movies are less than 90 minutes?" or "Which movies are rated higher than 8.5?".
+   * We can also ask questions like "Which movies are either comedy or drama and are less than 90 minutes?".
+   * The retriever will automatically convert these questions into queries that can be used to retrieve documents.
+   */
+  const query1 = await selfQueryRetriever.getRelevantDocuments(
+    "Which movies are less than 90 minutes?"
+  );
+  const query2 = await selfQueryRetriever.getRelevantDocuments(
+    "Which movies are rated higher than 8.5?"
+  );
+  const query3 = await selfQueryRetriever.getRelevantDocuments(
+    "Which movies are directed by Greta Gerwig?"
+  );
+  const query4 = await selfQueryRetriever.getRelevantDocuments(
+    "Which movies are either comedy or drama and are less than 90 minutes?"
+  );
+  console.log(query1, query2, query3, query4);
+};
+
+run();

examples/src/retrievers/pinecone_self_query.ts

@@ -0,0 +1,152 @@
+import { PineconeClient } from "@pinecone-database/pinecone";
+import { AttributeInfo } from "langchain/schema/query_constructor";
+import { Document } from "langchain/document";
+import { OpenAIEmbeddings } from "langchain/embeddings/openai";
+import { SelfQueryRetriever } from "langchain/retrievers/self_query/base";
+import { BasicTranslator } from "langchain/retrievers/self_query/translator";
+import { PineconeStore } from "langchain/vectorstores/pinecone";
+import { OpenAI } from "langchain/llms/openai";
+
+const run = async () => {
+  /**
+   * First, we create a bunch of documents. You can load your own documents here instead.
+   * Each document has a pageContent and a metadata field. Make sure your metadata matches the AttributeInfo below.
+   */
+  const docs = [
+    new Document({
+      pageContent:
+        "A bunch of scientists bring back dinosaurs and mayhem breaks loose",
+      metadata: { year: 1993, rating: 7.7, genre: "science fiction" },
+    }),
+    new Document({
+      pageContent:
+        "Leo DiCaprio gets lost in a dream within a dream within a dream within a ...",
+      metadata: { year: 2010, director: "Christopher Nolan", rating: 8.2 },
+    }),
+    new Document({
+      pageContent:
+        "A psychologist / detective gets lost in a series of dreams within dreams within dreams and Inception reused the idea",
+      metadata: { year: 2006, director: "Satoshi Kon", rating: 8.6 },
+    }),
+    new Document({
+      pageContent:
+        "A bunch of normal-sized women are supremely wholesome and some men pine after them",
+      metadata: { year: 2019, director: "Greta Gerwig", rating: 8.3 },
+    }),
+    new Document({
+      pageContent: "Toys come alive and have a blast doing so",
+      metadata: { year: 1995, genre: "animated" },
+    }),
+    new Document({
+      pageContent:
+        "Three men walk into the Zone, three men walk out of the Zone",
+      metadata: {
+        year: 1979,
+        director: "Andrei Tarkovsky",
+        genre: "science fiction",
+        rating: 9.9,
+      },
+    }),
+  ];
+
+  /**
+   * Next, we define the attributes we want to be able to query on.
+   * in this case, we want to be able to query on the genre, year, director, rating, and length of the movie.
+   * We also provide a description of each attribute and the type of the attribute.
+   * This is used to generate the query prompts.
+   */
+  const attributeInfo: AttributeInfo[] = [
+    {
+      name: "genre",
+      description: "The genre of the movie",
+      type: "string or array of strings",
+    },
+    {
+      name: "year",
+      description: "The year the movie was released",
+      type: "number",
+    },
+    {
+      name: "director",
+      description: "The director of the movie",
+      type: "string",
+    },
+    {
+      name: "rating",
+      description: "The rating of the movie (1-10)",
+      type: "number",
+    },
+    {
+      name: "length",
+      description: "The length of the movie in minutes",
+      type: "number",
+    },
+  ];
+
+  /**
+   * Next, we instantiate a vector store. This is where we store the embeddings of the documents.
+   * We use the Pinecone vector store here, but you can use any vector store you want.
+   * At this point we only support Chroma and Pinecone, but we will add more in the future.
+   * We also need to provide an embeddings object. This is used to embed the documents.
+   */
+  if (
+    !process.env.PINECONE_API_KEY ||
+    !process.env.PINECONE_ENVIRONMENT ||
+    !process.env.PINECONE_INDEX
+  ) {
+    throw new Error(
+      "PINECONE_ENVIRONMENT and PINECONE_API_KEY and PINECONE_INDEX must be set"
+    );
+  }
+
+  const client = new PineconeClient();
+  await client.init({
+    apiKey: process.env.PINECONE_API_KEY,
+    environment: process.env.PINECONE_ENVIRONMENT,
+  });
+  const index = client.Index(process.env.PINECONE_INDEX);
+
+  const embeddings = new OpenAIEmbeddings();
+  const llm = new OpenAI();
+  const documentContents = "Brief summary of a movie";
+  const vectorStore = await PineconeStore.fromDocuments(docs, embeddings, {
+    pineconeIndex: index,
+  });
+  const selfQueryRetriever = await SelfQueryRetriever.fromLLM({
+    llm,
+    vectorStore,
+    documentContents,
+    attributeInfo,
+    /**
+     * We need to create a basic translator that translates the queries into a
+     * filter format that the vector store can understand. We provide a basic translator
+     * translator here (which works for Chroma and Pinecone), but you can create
+     * your own translator by extending BaseTranslator abstract class. Note that the
+     * vector store needs to support filtering on the metadata attributes you want to
+     * query on.
+     */
+    structuredQueryTranslator: new BasicTranslator(),
+  });
+
+  /**
+   * Now we can query the vector store.
+   * We can ask questions like "Which movies are less than 90 minutes?" or "Which movies are rated higher than 8.5?".
+   * We can also ask questions like "Which movies are either comedy or drama and are less than 90 minutes?".
+   * The retriever will automatically convert these questions into queries that can be used to retrieve documents.
+   */
+  const query1 = await selfQueryRetriever.getRelevantDocuments(
+    "Which movies are less than 90 minutes?"
+  );
+  const query2 = await selfQueryRetriever.getRelevantDocuments(
+    "Which movies are rated higher than 8.5?"
+  );
+  const query3 = await selfQueryRetriever.getRelevantDocuments(
+    "Which movies are directed by Greta Gerwig?"
+  );
+  const query4 = await selfQueryRetriever.getRelevantDocuments(
+    "Which movies are either comedy or drama and are less than 90 minutes?"
+  );
+  console.log(query1, query2, query3, query4);
+};
+
+run();

langchain/.gitignore

@@ -220,6 +220,9 @@ schema.d.ts
 schema/output_parser.cjs
 schema/output_parser.js
 schema/output_parser.d.ts
+schema/query_constructor.cjs
+schema/query_constructor.js
+schema/query_constructor.d.ts
 sql_db.cjs
 sql_db.js
 sql_db.d.ts
@@ -259,6 +262,24 @@ retrievers/document_compressors/chain_extract.d.ts
 retrievers/hyde.cjs
 retrievers/hyde.js
 retrievers/hyde.d.ts
+retrievers/self_query/base.cjs
+retrievers/self_query/base.js
+retrievers/self_query/base.d.ts
+retrievers/self_query/translator.cjs
+retrievers/self_query/translator.js
+retrievers/self_query/translator.d.ts
+output_parsers/expression.cjs
+output_parsers/expression.js
+output_parsers/expression.d.ts
+chains/query_constructor/base.cjs
+chains/query_constructor/base.js
+chains/query_constructor/base.d.ts
+chains/query_constructor/ir.cjs
+chains/query_constructor/ir.js
+chains/query_constructor/ir.d.ts
+chains/query_constructor/parser.cjs
+chains/query_constructor/parser.js
+chains/query_constructor/parser.d.ts
 cache.cjs
 cache.js
 cache.d.ts