langchain 0.1.6 → 0.1.7

package/dist/agents/toolkits/conversational_retrieval/openai_functions.cjs

@@ -15,13 +15,14 @@ const token_buffer_memory_js_1 = require("./token_buffer_memory.cjs");
  * @returns A Promise that resolves to an initialized AgentExecutor.
  */
 async function createConversationalRetrievalAgent(llm, tools, options) {
-    const { rememberIntermediateSteps = true, memoryKey = "chat_history", outputKey = "output", prefix, verbose, } = options ?? {};
+    const { rememberIntermediateSteps = true, memoryKey = "chat_history", outputKey = "output", inputKey = "input", prefix, verbose, } = options ?? {};
     let memory;
     if (rememberIntermediateSteps) {
         memory = new token_buffer_memory_js_1.OpenAIAgentTokenBufferMemory({
             memoryKey,
             llm,
             outputKey,
+            inputKey,
         });
     }
     else {
@@ -31,6 +32,7 @@ async function createConversationalRetrievalAgent(llm, tools, options) {
             maxTokenLimit: 12000,
             returnMessages: true,
             outputKey,
+            inputKey,
         });
     }
     const executor = await (0, initialize_js_1.initializeAgentExecutorWithOptions)(tools, llm, {

package/dist/agents/toolkits/conversational_retrieval/openai_functions.d.ts

@@ -4,6 +4,7 @@ export type ConversationalRetrievalAgentOptions = {
     rememberIntermediateSteps?: boolean;
     memoryKey?: string;
     outputKey?: string;
+    inputKey?: string;
     prefix?: string;
     verbose?: boolean;
 };

package/dist/agents/toolkits/conversational_retrieval/openai_functions.js

@@ -12,13 +12,14 @@ import { OpenAIAgentTokenBufferMemory } from "./token_buffer_memory.js";
  * @returns A Promise that resolves to an initialized AgentExecutor.
  */
 export async function createConversationalRetrievalAgent(llm, tools, options) {
-    const { rememberIntermediateSteps = true, memoryKey = "chat_history", outputKey = "output", prefix, verbose, } = options ?? {};
+    const { rememberIntermediateSteps = true, memoryKey = "chat_history", outputKey = "output", inputKey = "input", prefix, verbose, } = options ?? {};
     let memory;
     if (rememberIntermediateSteps) {
         memory = new OpenAIAgentTokenBufferMemory({
             memoryKey,
             llm,
             outputKey,
+            inputKey,
         });
     }
     else {
@@ -28,6 +29,7 @@ export async function createConversationalRetrievalAgent(llm, tools, options) {
             maxTokenLimit: 12000,
             returnMessages: true,
             outputKey,
+            inputKey,
         });
     }
     const executor = await initializeAgentExecutorWithOptions(tools, llm, {

package/dist/indexes/index.cjs

@@ -0,0 +1,5 @@
+"use strict";
+Object.defineProperty(exports, "__esModule", { value: true });
+exports.index = void 0;
+var indexing_js_1 = require("./indexing.cjs");
+Object.defineProperty(exports, "index", { enumerable: true, get: function () { return indexing_js_1.index; } });

package/dist/indexes/index.d.ts

@@ -0,0 +1 @@
+export { type CleanupMode, type IndexOptions, index } from "./indexing.js";

package/dist/indexes/index.js

@@ -0,0 +1 @@
+export { index } from "./indexing.js";

package/dist/indexes/indexing.cjs

@@ -0,0 +1,265 @@
+"use strict";
+Object.defineProperty(exports, "__esModule", { value: true });
+exports.index = void 0;
+const uuid_1 = require("uuid");
+const base_1 = require("@langchain/community/indexes/base");
+const hash_1 = require("@langchain/core/utils/hash");
+const document_js_1 = require("../document.cjs");
+/**
+ * HashedDocument is a Document with hashes calculated.
+ * Hashes are calculated based on page content and metadata.
+ * It is used for indexing.
+ */
+class HashedDocument {
+    constructor(fields) {
+        Object.defineProperty(this, "uid", {
+            enumerable: true,
+            configurable: true,
+            writable: true,
+            value: void 0
+        });
+        Object.defineProperty(this, "hash_", {
+            enumerable: true,
+            configurable: true,
+            writable: true,
+            value: void 0
+        });
+        Object.defineProperty(this, "contentHash", {
+            enumerable: true,
+            configurable: true,
+            writable: true,
+            value: void 0
+        });
+        Object.defineProperty(this, "metadataHash", {
+            enumerable: true,
+            configurable: true,
+            writable: true,
+            value: void 0
+        });
+        Object.defineProperty(this, "pageContent", {
+            enumerable: true,
+            configurable: true,
+            writable: true,
+            value: void 0
+        });
+        Object.defineProperty(this, "metadata", {
+            enumerable: true,
+            configurable: true,
+            writable: true,
+            value: void 0
+        });
+        this.uid = fields.uid;
+        this.pageContent = fields.pageContent;
+        this.metadata = fields.metadata;
+    }
+    calculateHashes() {
+        const forbiddenKeys = ["hash_", "content_hash", "metadata_hash"];
+        for (const key of forbiddenKeys) {
+            if (key in this.metadata) {
+                throw new Error(`Metadata cannot contain key ${key} as it is reserved for internal use. Restricted keys: [${forbiddenKeys.join(", ")}]`);
+            }
+        }
+        const contentHash = this._hashStringToUUID(this.pageContent);
+        try {
+            const metadataHash = this._hashNestedDictToUUID(this.metadata);
+            this.contentHash = contentHash;
+            this.metadataHash = metadataHash;
+        }
+        catch (e) {
+            throw new Error(`Failed to hash metadata: ${e}. Please use a dict that can be serialized using json.`);
+        }
+        this.hash_ = this._hashStringToUUID(this.contentHash + this.metadataHash);
+        if (!this.uid) {
+            this.uid = this.hash_;
+        }
+    }
+    toDocument() {
+        return new document_js_1.Document({
+            pageContent: this.pageContent,
+            metadata: this.metadata,
+        });
+    }
+    static fromDocument(document, uid) {
+        const doc = new this({
+            pageContent: document.pageContent,
+            metadata: document.metadata,
+            uid: uid || document.uid,
+        });
+        doc.calculateHashes();
+        return doc;
+    }
+    _hashStringToUUID(inputString) {
+        const hash_value = (0, hash_1.insecureHash)(inputString);
+        return (0, uuid_1.v5)(hash_value, base_1.UUIDV5_NAMESPACE);
+    }
+    _hashNestedDictToUUID(data) {
+        const serialized_data = JSON.stringify(data, Object.keys(data).sort());
+        const hash_value = (0, hash_1.insecureHash)(serialized_data);
+        return (0, uuid_1.v5)(hash_value, base_1.UUIDV5_NAMESPACE);
+    }
+}
+function batch(size, iterable) {
+    const batches = [];
+    let currentBatch = [];
+    iterable.forEach((item) => {
+        currentBatch.push(item);
+        if (currentBatch.length >= size) {
+            batches.push(currentBatch);
+            currentBatch = [];
+        }
+    });
+    if (currentBatch.length > 0) {
+        batches.push(currentBatch);
+    }
+    return batches;
+}
+function deduplicateInOrder(hashedDocuments) {
+    const seen = new Set();
+    const deduplicated = [];
+    for (const hashedDoc of hashedDocuments) {
+        if (!hashedDoc.hash_) {
+            throw new Error("Hashed document does not have a hash");
+        }
+        if (!seen.has(hashedDoc.hash_)) {
+            seen.add(hashedDoc.hash_);
+            deduplicated.push(hashedDoc);
+        }
+    }
+    return deduplicated;
+}
+function getSourceIdAssigner(sourceIdKey) {
+    if (sourceIdKey === null) {
+        return (_doc) => null;
+    }
+    else if (typeof sourceIdKey === "string") {
+        return (doc) => doc.metadata[sourceIdKey];
+    }
+    else if (typeof sourceIdKey === "function") {
+        return sourceIdKey;
+    }
+    else {
+        throw new Error(`sourceIdKey should be null, a string or a function, got ${typeof sourceIdKey}`);
+    }
+}
+// eslint-disable-next-line @typescript-eslint/no-explicit-any
+const _isBaseDocumentLoader = (arg) => {
+    if ("load" in arg &&
+        typeof arg.load === "function" &&
+        "loadAndSplit" in arg &&
+        typeof arg.loadAndSplit === "function") {
+        return true;
+    }
+    return false;
+};
+/**
+ * Index data from the doc source into the vector store.
+ *
+ * Indexing functionality uses a manager to keep track of which documents
+ * are in the vector store.
+ *
+ * This allows us to keep track of which documents were updated, and which
+ * documents were deleted, which documents should be skipped.
+ *
+ * For the time being, documents are indexed using their hashes, and users
+ *  are not able to specify the uid of the document.
+ *
+ * @param {IndexArgs} args
+ * @param {BaseDocumentLoader | DocumentInterface[]} args.docsSource The source of documents to index. Can be a DocumentLoader or a list of Documents.
+ * @param {RecordManagerInterface} args.recordManager The record manager to use for keeping track of indexed documents.
+ * @param {VectorStore} args.vectorStore The vector store to use for storing the documents.
+ * @param {IndexOptions | undefined} args.options Options for indexing.
+ * @returns {Promise<IndexingResult>}
+ */
+async function index(args) {
+    const { docsSource, recordManager, vectorStore, options } = args;
+    const { batchSize = 100, cleanup, sourceIdKey, cleanupBatchSize = 1000, forceUpdate = false, } = options ?? {};
+    if (cleanup === "incremental" && !sourceIdKey) {
+        throw new Error("sourceIdKey is required when cleanup mode is incremental. Please provide through 'options.sourceIdKey'.");
+    }
+    const docs = _isBaseDocumentLoader(docsSource)
+        ? await docsSource.load()
+        : docsSource;
+    const sourceIdAssigner = getSourceIdAssigner(sourceIdKey ?? null);
+    const indexStartDt = await recordManager.getTime();
+    let numAdded = 0;
+    let numDeleted = 0;
+    let numUpdated = 0;
+    let numSkipped = 0;
+    const batches = batch(batchSize ?? 100, docs);
+    for (const batch of batches) {
+        const hashedDocs = deduplicateInOrder(batch.map((doc) => HashedDocument.fromDocument(doc)));
+        const sourceIds = hashedDocs.map((doc) => sourceIdAssigner(doc));
+        if (cleanup === "incremental") {
+            hashedDocs.forEach((_hashedDoc, index) => {
+                const source = sourceIds[index];
+                if (source === null) {
+                    throw new Error("sourceIdKey must be provided when cleanup is incremental");
+                }
+            });
+        }
+        const batchExists = await recordManager.exists(hashedDocs.map((doc) => doc.uid));
+        const uids = [];
+        const docsToIndex = [];
+        const docsToUpdate = [];
+        const seenDocs = new Set();
+        hashedDocs.forEach((hashedDoc, i) => {
+            const docExists = batchExists[i];
+            if (docExists) {
+                if (forceUpdate) {
+                    seenDocs.add(hashedDoc.uid);
+                }
+                else {
+                    docsToUpdate.push(hashedDoc.uid);
+                    return;
+                }
+            }
+            uids.push(hashedDoc.uid);
+            docsToIndex.push(hashedDoc.toDocument());
+        });
+        if (docsToUpdate.length > 0) {
+            await recordManager.update(docsToUpdate, { timeAtLeast: indexStartDt });
+            numSkipped += docsToUpdate.length;
+        }
+        if (docsToIndex.length > 0) {
+            await vectorStore.addDocuments(docsToIndex, { ids: uids });
+            numAdded += docsToIndex.length - seenDocs.size;
+            numUpdated += seenDocs.size;
+        }
+        await recordManager.update(hashedDocs.map((doc) => doc.uid), { timeAtLeast: indexStartDt, groupIds: sourceIds });
+        if (cleanup === "incremental") {
+            sourceIds.forEach((sourceId) => {
+                if (!sourceId)
+                    throw new Error("Source id cannot be null");
+            });
+            const uidsToDelete = await recordManager.listKeys({
+                before: indexStartDt,
+                groupIds: sourceIds,
+            });
+            await vectorStore.delete({ ids: uidsToDelete });
+            await recordManager.deleteKeys(uidsToDelete);
+            numDeleted += uidsToDelete.length;
+        }
+    }
+    if (cleanup === "full") {
+        let uidsToDelete = await recordManager.listKeys({
+            before: indexStartDt,
+            limit: cleanupBatchSize,
+        });
+        while (uidsToDelete.length > 0) {
+            await vectorStore.delete({ ids: uidsToDelete });
+            await recordManager.deleteKeys(uidsToDelete);
+            numDeleted += uidsToDelete.length;
+            uidsToDelete = await recordManager.listKeys({
+                before: indexStartDt,
+                limit: cleanupBatchSize,
+            });
+        }
+    }
+    return {
+        numAdded,
+        numDeleted,
+        numUpdated,
+        numSkipped,
+    };
+}
+exports.index = index;

package/dist/indexes/indexing.d.ts

@@ -0,0 +1,75 @@
+import { VectorStore } from "@langchain/core/vectorstores";
+import { RecordManagerInterface } from "@langchain/community/indexes/base";
+import { DocumentInterface } from "@langchain/core/documents";
+import { BaseDocumentLoader } from "../document_loaders/base.js";
+type IndexingResult = {
+    numAdded: number;
+    numDeleted: number;
+    numUpdated: number;
+    numSkipped: number;
+};
+type StringOrDocFunc = string | ((doc: DocumentInterface) => string);
+export type CleanupMode = "full" | "incremental";
+export type IndexOptions = {
+    /**
+     * The number of documents to index in one batch.
+     */
+    batchSize?: number;
+    /**
+     * The cleanup mode to use. Can be "full", "incremental" or undefined.
+     * - **Incremental**: Cleans up all documents that haven't been updated AND
+     *   that are associated with source ids that were seen
+     *   during indexing.
+     *   Clean up is done continuously during indexing helping
+     *   to minimize the probability of users seeing duplicated
+     *   content.
+     * - **Full**: Delete all documents that haven to been returned by the loader.
+     *   Clean up runs after all documents have been indexed.
+     *   This means that users may see duplicated content during indexing.
+     * - **undefined**: Do not delete any documents.
+     */
+    cleanup?: CleanupMode;
+    /**
+     * Optional key that helps identify the original source of the document.
+     * Must either be a string representing the key of the source in the metadata
+     * or a function that takes a document and returns a string representing the source.
+     * **Required when cleanup is incremental**.
+     */
+    sourceIdKey?: StringOrDocFunc;
+    /**
+     * Batch size to use when cleaning up documents.
+     */
+    cleanupBatchSize?: number;
+    /**
+     * Force update documents even if they are present in the
+     * record manager. Useful if you are re-indexing with updated embeddings.
+     */
+    forceUpdate?: boolean;
+};
+interface IndexArgs {
+    docsSource: BaseDocumentLoader | DocumentInterface[];
+    recordManager: RecordManagerInterface;
+    vectorStore: VectorStore;
+    options?: IndexOptions;
+}
+/**
+ * Index data from the doc source into the vector store.
+ *
+ * Indexing functionality uses a manager to keep track of which documents
+ * are in the vector store.
+ *
+ * This allows us to keep track of which documents were updated, and which
+ * documents were deleted, which documents should be skipped.
+ *
+ * For the time being, documents are indexed using their hashes, and users
+ *  are not able to specify the uid of the document.
+ *
+ * @param {IndexArgs} args
+ * @param {BaseDocumentLoader | DocumentInterface[]} args.docsSource The source of documents to index. Can be a DocumentLoader or a list of Documents.
+ * @param {RecordManagerInterface} args.recordManager The record manager to use for keeping track of indexed documents.
+ * @param {VectorStore} args.vectorStore The vector store to use for storing the documents.
+ * @param {IndexOptions | undefined} args.options Options for indexing.
+ * @returns {Promise<IndexingResult>}
+ */
+export declare function index(args: IndexArgs): Promise<IndexingResult>;
+export {};

package/dist/indexes/indexing.js

@@ -0,0 +1,261 @@
+import { v5 as uuidv5 } from "uuid";
+import { UUIDV5_NAMESPACE, } from "@langchain/community/indexes/base";
+import { insecureHash } from "@langchain/core/utils/hash";
+import { Document } from "../document.js";
+/**
+ * HashedDocument is a Document with hashes calculated.
+ * Hashes are calculated based on page content and metadata.
+ * It is used for indexing.
+ */
+class HashedDocument {
+    constructor(fields) {
+        Object.defineProperty(this, "uid", {
+            enumerable: true,
+            configurable: true,
+            writable: true,
+            value: void 0
+        });
+        Object.defineProperty(this, "hash_", {
+            enumerable: true,
+            configurable: true,
+            writable: true,
+            value: void 0
+        });
+        Object.defineProperty(this, "contentHash", {
+            enumerable: true,
+            configurable: true,
+            writable: true,
+            value: void 0
+        });
+        Object.defineProperty(this, "metadataHash", {
+            enumerable: true,
+            configurable: true,
+            writable: true,
+            value: void 0
+        });
+        Object.defineProperty(this, "pageContent", {
+            enumerable: true,
+            configurable: true,
+            writable: true,
+            value: void 0
+        });
+        Object.defineProperty(this, "metadata", {
+            enumerable: true,
+            configurable: true,
+            writable: true,
+            value: void 0
+        });
+        this.uid = fields.uid;
+        this.pageContent = fields.pageContent;
+        this.metadata = fields.metadata;
+    }
+    calculateHashes() {
+        const forbiddenKeys = ["hash_", "content_hash", "metadata_hash"];
+        for (const key of forbiddenKeys) {
+            if (key in this.metadata) {
+                throw new Error(`Metadata cannot contain key ${key} as it is reserved for internal use. Restricted keys: [${forbiddenKeys.join(", ")}]`);
+            }
+        }
+        const contentHash = this._hashStringToUUID(this.pageContent);
+        try {
+            const metadataHash = this._hashNestedDictToUUID(this.metadata);
+            this.contentHash = contentHash;
+            this.metadataHash = metadataHash;
+        }
+        catch (e) {
+            throw new Error(`Failed to hash metadata: ${e}. Please use a dict that can be serialized using json.`);
+        }
+        this.hash_ = this._hashStringToUUID(this.contentHash + this.metadataHash);
+        if (!this.uid) {
+            this.uid = this.hash_;
+        }
+    }
+    toDocument() {
+        return new Document({
+            pageContent: this.pageContent,
+            metadata: this.metadata,
+        });
+    }
+    static fromDocument(document, uid) {
+        const doc = new this({
+            pageContent: document.pageContent,
+            metadata: document.metadata,
+            uid: uid || document.uid,
+        });
+        doc.calculateHashes();
+        return doc;
+    }
+    _hashStringToUUID(inputString) {
+        const hash_value = insecureHash(inputString);
+        return uuidv5(hash_value, UUIDV5_NAMESPACE);
+    }
+    _hashNestedDictToUUID(data) {
+        const serialized_data = JSON.stringify(data, Object.keys(data).sort());
+        const hash_value = insecureHash(serialized_data);
+        return uuidv5(hash_value, UUIDV5_NAMESPACE);
+    }
+}
+function batch(size, iterable) {
+    const batches = [];
+    let currentBatch = [];
+    iterable.forEach((item) => {
+        currentBatch.push(item);
+        if (currentBatch.length >= size) {
+            batches.push(currentBatch);
+            currentBatch = [];
+        }
+    });
+    if (currentBatch.length > 0) {
+        batches.push(currentBatch);
+    }
+    return batches;
+}
+function deduplicateInOrder(hashedDocuments) {
+    const seen = new Set();
+    const deduplicated = [];
+    for (const hashedDoc of hashedDocuments) {
+        if (!hashedDoc.hash_) {
+            throw new Error("Hashed document does not have a hash");
+        }
+        if (!seen.has(hashedDoc.hash_)) {
+            seen.add(hashedDoc.hash_);
+            deduplicated.push(hashedDoc);
+        }
+    }
+    return deduplicated;
+}
+function getSourceIdAssigner(sourceIdKey) {
+    if (sourceIdKey === null) {
+        return (_doc) => null;
+    }
+    else if (typeof sourceIdKey === "string") {
+        return (doc) => doc.metadata[sourceIdKey];
+    }
+    else if (typeof sourceIdKey === "function") {
+        return sourceIdKey;
+    }
+    else {
+        throw new Error(`sourceIdKey should be null, a string or a function, got ${typeof sourceIdKey}`);
+    }
+}
+// eslint-disable-next-line @typescript-eslint/no-explicit-any
+const _isBaseDocumentLoader = (arg) => {
+    if ("load" in arg &&
+        typeof arg.load === "function" &&
+        "loadAndSplit" in arg &&
+        typeof arg.loadAndSplit === "function") {
+        return true;
+    }
+    return false;
+};
+/**
+ * Index data from the doc source into the vector store.
+ *
+ * Indexing functionality uses a manager to keep track of which documents
+ * are in the vector store.
+ *
+ * This allows us to keep track of which documents were updated, and which
+ * documents were deleted, which documents should be skipped.
+ *
+ * For the time being, documents are indexed using their hashes, and users
+ *  are not able to specify the uid of the document.
+ *
+ * @param {IndexArgs} args
+ * @param {BaseDocumentLoader | DocumentInterface[]} args.docsSource The source of documents to index. Can be a DocumentLoader or a list of Documents.
+ * @param {RecordManagerInterface} args.recordManager The record manager to use for keeping track of indexed documents.
+ * @param {VectorStore} args.vectorStore The vector store to use for storing the documents.
+ * @param {IndexOptions | undefined} args.options Options for indexing.
+ * @returns {Promise<IndexingResult>}
+ */
+export async function index(args) {
+    const { docsSource, recordManager, vectorStore, options } = args;
+    const { batchSize = 100, cleanup, sourceIdKey, cleanupBatchSize = 1000, forceUpdate = false, } = options ?? {};
+    if (cleanup === "incremental" && !sourceIdKey) {
+        throw new Error("sourceIdKey is required when cleanup mode is incremental. Please provide through 'options.sourceIdKey'.");
+    }
+    const docs = _isBaseDocumentLoader(docsSource)
+        ? await docsSource.load()
+        : docsSource;
+    const sourceIdAssigner = getSourceIdAssigner(sourceIdKey ?? null);
+    const indexStartDt = await recordManager.getTime();
+    let numAdded = 0;
+    let numDeleted = 0;
+    let numUpdated = 0;
+    let numSkipped = 0;
+    const batches = batch(batchSize ?? 100, docs);
+    for (const batch of batches) {
+        const hashedDocs = deduplicateInOrder(batch.map((doc) => HashedDocument.fromDocument(doc)));
+        const sourceIds = hashedDocs.map((doc) => sourceIdAssigner(doc));
+        if (cleanup === "incremental") {
+            hashedDocs.forEach((_hashedDoc, index) => {
+                const source = sourceIds[index];
+                if (source === null) {
+                    throw new Error("sourceIdKey must be provided when cleanup is incremental");
+                }
+            });
+        }
+        const batchExists = await recordManager.exists(hashedDocs.map((doc) => doc.uid));
+        const uids = [];
+        const docsToIndex = [];
+        const docsToUpdate = [];
+        const seenDocs = new Set();
+        hashedDocs.forEach((hashedDoc, i) => {
+            const docExists = batchExists[i];
+            if (docExists) {
+                if (forceUpdate) {
+                    seenDocs.add(hashedDoc.uid);
+                }
+                else {
+                    docsToUpdate.push(hashedDoc.uid);
+                    return;
+                }
+            }
+            uids.push(hashedDoc.uid);
+            docsToIndex.push(hashedDoc.toDocument());
+        });
+        if (docsToUpdate.length > 0) {
+            await recordManager.update(docsToUpdate, { timeAtLeast: indexStartDt });
+            numSkipped += docsToUpdate.length;
+        }
+        if (docsToIndex.length > 0) {
+            await vectorStore.addDocuments(docsToIndex, { ids: uids });
+            numAdded += docsToIndex.length - seenDocs.size;
+            numUpdated += seenDocs.size;
+        }
+        await recordManager.update(hashedDocs.map((doc) => doc.uid), { timeAtLeast: indexStartDt, groupIds: sourceIds });
+        if (cleanup === "incremental") {
+            sourceIds.forEach((sourceId) => {
+                if (!sourceId)
+                    throw new Error("Source id cannot be null");
+            });
+            const uidsToDelete = await recordManager.listKeys({
+                before: indexStartDt,
+                groupIds: sourceIds,
+            });
+            await vectorStore.delete({ ids: uidsToDelete });
+            await recordManager.deleteKeys(uidsToDelete);
+            numDeleted += uidsToDelete.length;
+        }
+    }
+    if (cleanup === "full") {
+        let uidsToDelete = await recordManager.listKeys({
+            before: indexStartDt,
+            limit: cleanupBatchSize,
+        });
+        while (uidsToDelete.length > 0) {
+            await vectorStore.delete({ ids: uidsToDelete });
+            await recordManager.deleteKeys(uidsToDelete);
+            numDeleted += uidsToDelete.length;
+            uidsToDelete = await recordManager.listKeys({
+                before: indexStartDt,
+                limit: cleanupBatchSize,
+            });
+        }
+    }
+    return {
+        numAdded,
+        numDeleted,
+        numUpdated,
+        numSkipped,
+    };
+}

package/dist/load/import_map.cjs

@@ -25,7 +25,7 @@ var __importStar = (this && this.__importStar) || function (mod) {
 };
 Object.defineProperty(exports, "__esModule", { value: true });
 exports.util__document = exports.storage__in_memory = exports.storage__encoder_backed = exports.stores__message__in_memory = exports.stores__file__in_memory = exports.stores__doc__in_memory = exports.retrievers__vespa = exports.retrievers__score_threshold = exports.retrievers__hyde = exports.retrievers__document_compressors__embeddings_filter = exports.retrievers__document_compressors__chain_extract = exports.retrievers__time_weighted = exports.retrievers__parent_document = exports.retrievers__multi_vector = exports.retrievers__multi_query = exports.retrievers__document_compressors = exports.retrievers__contextual_compression = exports.retrievers__remote = exports.output_parsers = exports.schema__query_constructor = exports.schema__prompt_template = exports.chat_models__anthropic = exports.document_transformers__openai_functions = exports.document_loaders__web__sort_xyz_blockchain = exports.document_loaders__web__serpapi = exports.document_loaders__web__searchapi = exports.document_loaders__base = exports.text_splitter = exports.vectorstores__memory = exports.llms__fake = exports.embeddings__fake = exports.embeddings__cache_backed = exports.chains__retrieval = exports.chains__openai_functions = exports.chains__history_aware_retriever = exports.chains__combine_documents__reduce = exports.chains__combine_documents = exports.chains = exports.tools__retriever = exports.tools__render = exports.agents__openai__output_parser = exports.agents__xml__output_parser = exports.agents__react__output_parser = exports.agents__format_scratchpad__log_to_message = exports.agents__format_scratchpad__xml = exports.agents__format_scratchpad__log = exports.agents__format_scratchpad__openai_tools = exports.agents__format_scratchpad = exports.agents__toolkits = exports.agents = void 0;
-exports.llms__fireworks = exports.chat_models__fireworks = exports.schema__output = exports.schema__output_parser = exports.schema__runnable = exports.prompts__base = exports.prompts__pipeline = exports.prompts__chat = exports.schema__messages = exports.prompts__prompt = exports.embeddings__openai = exports.llms__openai = exports.chat_models__openai = exports.runnables__remote = exports.smith = exports.evaluation = exports.experimental__prompts__custom_format = exports.experimental__masking = exports.experimental__chains__violation_of_expectations = exports.experimental__chat_models__bittensor = exports.experimental__plan_and_execute = exports.experimental__generative_agents = exports.experimental__babyagi = exports.experimental__openai_files = exports.experimental__openai_assistant = exports.experimental__autogpt = exports.util__time = exports.util__math = void 0;
+exports.llms__fireworks = exports.chat_models__fireworks = exports.schema__output = exports.schema__output_parser = exports.schema__runnable = exports.prompts__base = exports.prompts__pipeline = exports.prompts__chat = exports.schema__messages = exports.prompts__prompt = exports.embeddings__openai = exports.llms__openai = exports.chat_models__openai = exports.indexes = exports.runnables__remote = exports.smith = exports.evaluation = exports.experimental__prompts__custom_format = exports.experimental__masking = exports.experimental__chains__violation_of_expectations = exports.experimental__chat_models__bittensor = exports.experimental__plan_and_execute = exports.experimental__generative_agents = exports.experimental__babyagi = exports.experimental__openai_files = exports.experimental__openai_assistant = exports.experimental__autogpt = exports.util__time = exports.util__math = void 0;
 exports.agents = __importStar(require("../agents/index.cjs"));
 exports.agents__toolkits = __importStar(require("../agents/toolkits/index.cjs"));
 exports.agents__format_scratchpad = __importStar(require("../agents/format_scratchpad/openai_functions.cjs"));
@@ -91,6 +91,7 @@ exports.experimental__prompts__custom_format = __importStar(require("../experime
 exports.evaluation = __importStar(require("../evaluation/index.cjs"));
 exports.smith = __importStar(require("../smith/index.cjs"));
 exports.runnables__remote = __importStar(require("../runnables/remote.cjs"));
+exports.indexes = __importStar(require("../indexes/index.cjs"));
 const openai_1 = require("@langchain/openai");
 const prompts_1 = require("@langchain/core/prompts");
 const messages_1 = require("@langchain/core/messages");

package/dist/load/import_map.d.ts

@@ -63,6 +63,7 @@ export * as experimental__prompts__custom_format from "../experimental/prompts/c
 export * as evaluation from "../evaluation/index.js";
 export * as smith from "../smith/index.js";
 export * as runnables__remote from "../runnables/remote.js";
+export * as indexes from "../indexes/index.js";
 import { ChatOpenAI, OpenAI, OpenAIEmbeddings } from "@langchain/openai";
 import { PromptTemplate, AIMessagePromptTemplate, ChatMessagePromptTemplate, ChatPromptTemplate, HumanMessagePromptTemplate, MessagesPlaceholder, SystemMessagePromptTemplate, PipelinePromptTemplate } from "@langchain/core/prompts";
 import { AIMessage, AIMessageChunk, BaseMessage, BaseMessageChunk, ChatMessage, ChatMessageChunk, FunctionMessage, FunctionMessageChunk, HumanMessage, HumanMessageChunk, SystemMessage, SystemMessageChunk, ToolMessage, ToolMessageChunk } from "@langchain/core/messages";

package/dist/load/import_map.js

@@ -64,6 +64,7 @@ export * as experimental__prompts__custom_format from "../experimental/prompts/c
 export * as evaluation from "../evaluation/index.js";
 export * as smith from "../smith/index.js";
 export * as runnables__remote from "../runnables/remote.js";
+export * as indexes from "../indexes/index.js";
 import { ChatOpenAI, OpenAI, OpenAIEmbeddings } from "@langchain/openai";
 import { PromptTemplate, AIMessagePromptTemplate, ChatMessagePromptTemplate, ChatPromptTemplate, HumanMessagePromptTemplate, MessagesPlaceholder, SystemMessagePromptTemplate, PipelinePromptTemplate } from "@langchain/core/prompts";
 import { AIMessage, AIMessageChunk, BaseMessage, BaseMessageChunk, ChatMessage, ChatMessageChunk, FunctionMessage, FunctionMessageChunk, HumanMessage, HumanMessageChunk, SystemMessage, SystemMessageChunk, ToolMessage, ToolMessageChunk } from "@langchain/core/messages";

package/dist/smith/config.d.ts

@@ -15,18 +15,19 @@ export type EvaluatorInputFormatter = ({ rawInput, rawPrediction, rawReferenceOu
     rawReferenceOutput?: any;
     run: Run;
 }) => EvaluatorInputs;
+export type DynamicRunEvaluatorParams = {
+    input: Record<string, unknown>;
+    prediction?: Record<string, unknown>;
+    reference?: Record<string, unknown>;
+    run: Run;
+    example?: Example;
+};
 /**
  * Type of a function that can be coerced into a RunEvaluator function.
  * While we have the class-based RunEvaluator, it's often more convenient to directly
  * pass a function to the runner. This type allows us to do that.
  */
-export type RunEvaluatorLike = (({ run, example, }: {
-    run: Run;
-    example?: Example;
-}) => Promise<EvaluationResult>) | (({ run, example }: {
-    run: Run;
-    example?: Example;
-}) => EvaluationResult);
+export type RunEvaluatorLike = ((props: DynamicRunEvaluatorParams) => Promise<EvaluationResult>) | ((props: DynamicRunEvaluatorParams) => EvaluationResult);
 /**
  * Configuration class for running evaluations on datasets.
  *
@@ -51,10 +52,26 @@ export type RunEvalConfig<T extends keyof EvaluatorType = keyof EvaluatorType, U
      */
     evaluators?: (T | EvalConfig)[];
     /**
-     * Convert the evaluation data into a format that can be used by the evaluator.
-     * By default, we pass the first value of the run.inputs, run.outputs (predictions),
-     * and references (example.outputs)
-     *
+     * Convert the evaluation data into formats that can be used by the evaluator.
+     * This should most commonly be a string.
+     * Parameters are the raw input from the run, the raw output, raw reference output, and the raw run.
+     * @example
+     * ```ts
+     * // Chain input: { input: "some string" }
+     * // Chain output: { output: "some output" }
+     * // Reference example output format: { output: "some reference output" }
+     * const formatEvaluatorInputs = ({
+     *   rawInput,
+     *   rawPrediction,
+     *   rawReferenceOutput,
+     * }) => {
+     *   return {
+     *     input: rawInput.input,
+     *     prediction: rawPrediction.output,
+     *     reference: rawReferenceOutput.output,
+     *   };
+     * };
+     * ```
      * @returns The prepared data.
      */
     formatEvaluatorInputs?: EvaluatorInputFormatter;
@@ -76,8 +93,26 @@ export interface EvalConfig extends LoadEvaluatorOptions {
      */
     feedbackKey?: string;
     /**
-     * Convert the evaluation data into a format that can be used by the evaluator.
-     * @param data The data to prepare.
+     * Convert the evaluation data into formats that can be used by the evaluator.
+     * This should most commonly be a string.
+     * Parameters are the raw input from the run, the raw output, raw reference output, and the raw run.
+     * @example
+     * ```ts
+     * // Chain input: { input: "some string" }
+     * // Chain output: { output: "some output" }
+     * // Reference example output format: { output: "some reference output" }
+     * const formatEvaluatorInputs = ({
+     *   rawInput,
+     *   rawPrediction,
+     *   rawReferenceOutput,
+     * }) => {
+     *   return {
+     *     input: rawInput.input,
+     *     prediction: rawPrediction.output,
+     *     reference: rawReferenceOutput.output,
+     *   };
+     * };
+     * ```
      * @returns The prepared data.
      */
     formatEvaluatorInputs: EvaluatorInputFormatter;
@@ -91,16 +126,21 @@ export interface EvalConfig extends LoadEvaluatorOptions {
  * @returns The configuration for the evaluator.
  * @example
  * ```ts
- * const evalConfig = new RunEvalConfig(
- *  [new RunEvalConfig.Criteria("helpfulness")],
- * );
+ * const evalConfig = {
+ *   evaluators: [{
+ *     evaluatorType: "criteria",
+ *     criteria: "helpfulness"
+ *   }]
+ * };
  * ```
  * @example
  * ```ts
- * const evalConfig = new RunEvalConfig(
- * [new RunEvalConfig.Criteria(
- *      { "isCompliant": "Does the submission comply with the requirements of XYZ"
- *  })],
+ * const evalConfig = {
+ *   evaluators: [{
+ *     evaluatorType: "criteria",
+ *     criteria: { "isCompliant": "Does the submission comply with the requirements of XYZ"
+ *   }]
+ * };
  */
 export type CriteriaEvalChainConfig = EvalConfig & {
     evaluatorType: "criteria";
@@ -133,16 +173,21 @@ export type CriteriaEvalChainConfig = EvalConfig & {
  * @returns The configuration for the evaluator.
  * @example
  * ```ts
- * const evalConfig = new RunEvalConfig(
- *  [new RunEvalConfig.LabeledCriteria("correctness")],
- * );
+ * const evalConfig = {
+ *   evaluators: [{
+ *     evaluatorType: "labeled_criteria",
+ *     criteria: "correctness"
+ *   }],
+ * };
  * ```
  * @example
  * ```ts
- * const evalConfig = new RunEvalConfig(
- * [new RunEvalConfig.Criteria(
- *      { "mentionsAllFacts": "Does the include all facts provided in the reference?"
- *  })],
+ * const evalConfig = {
+ *   evaluators: [{
+ *     evaluatorType: "labeled_criteria",
+ *     criteria: { "mentionsAllFacts": "Does the include all facts provided in the reference?" }
+ *   }],
+ * };
  */
 export type LabeledCriteria = EvalConfig & {
     evaluatorType: "labeled_criteria";

package/dist/smith/index.cjs

@@ -1,5 +1,20 @@
 "use strict";
+var __createBinding = (this && this.__createBinding) || (Object.create ? (function(o, m, k, k2) {
+    if (k2 === undefined) k2 = k;
+    var desc = Object.getOwnPropertyDescriptor(m, k);
+    if (!desc || ("get" in desc ? !m.__esModule : desc.writable || desc.configurable)) {
+      desc = { enumerable: true, get: function() { return m[k]; } };
+    }
+    Object.defineProperty(o, k2, desc);
+}) : (function(o, m, k, k2) {
+    if (k2 === undefined) k2 = k;
+    o[k2] = m[k];
+}));
+var __exportStar = (this && this.__exportStar) || function(m, exports) {
+    for (var p in m) if (p !== "default" && !Object.prototype.hasOwnProperty.call(exports, p)) __createBinding(exports, m, p);
+};
 Object.defineProperty(exports, "__esModule", { value: true });
 exports.runOnDataset = void 0;
 const runner_utils_js_1 = require("./runner_utils.cjs");
 Object.defineProperty(exports, "runOnDataset", { enumerable: true, get: function () { return runner_utils_js_1.runOnDataset; } });
+__exportStar(require("./config.cjs"), exports);

package/dist/smith/index.d.ts

@@ -1,3 +1,3 @@
-import type { RunEvalConfig } from "./config.js";
-import { runOnDataset } from "./runner_utils.js";
-export { type RunEvalConfig, runOnDataset };
+import { type EvalResults, type RunOnDatasetParams, runOnDataset } from "./runner_utils.js";
+export { type EvalResults, type RunOnDatasetParams, runOnDataset };
+export * from "./config.js";

package/dist/smith/index.js

@@ -1,2 +1,3 @@
-import { runOnDataset } from "./runner_utils.js";
+import { runOnDataset, } from "./runner_utils.js";
 export { runOnDataset };
+export * from "./config.js";

package/dist/smith/runner_utils.cjs

@@ -29,9 +29,16 @@ class DynamicRunEvaluator {
      * @returns A promise that resolves to the evaluation result.
      */
     async evaluateRun(run, example) {
-        return await this.evaluator.invoke({ run, example });
+        return await this.evaluator.invoke({
+            run,
+            example,
+            input: run.inputs,
+            prediction: run.outputs,
+            reference: example?.outputs,
+        });
     }
 }
+// eslint-disable-next-line @typescript-eslint/no-explicit-any
 function isLLMStringEvaluator(evaluator) {
     return evaluator && typeof evaluator.evaluateStrings === "function";
 }
@@ -254,7 +261,7 @@ const getExamplesInputs = (examples, chainOrFactory, dataType) => {
  * for evaluation.
  *
  * @param options - (Optional) Additional parameters for the evaluation process:
- *   - `evaluation` (RunEvalConfig): Configuration for the evaluation, including
+ *   - `evaluationConfig` (RunEvalConfig): Configuration for the evaluation, including
  *     standard and custom evaluators.
  *   - `projectName` (string): Name of the project for logging and tracking.
  *   - `projectMetadata` (Record<string, unknown>): Additional metadata for the project.
@@ -273,10 +280,10 @@ const getExamplesInputs = (examples, chainOrFactory, dataType) => {
  *   const datasetName = 'example-dataset';
  *   const client = new Client(/* ...config... *\//);
  *
- *   const evaluationConfig = new RunEvalConfig({
+ *   const evaluationConfig = {
  *     evaluators: [/* ...evaluators... *\//],
  *     customEvaluators: [/* ...custom evaluators... *\//],
- *   });
+ *   };
  *
  *   const results = await runOnDataset(chain, datasetName, {
  *     evaluationConfig,

package/dist/smith/runner_utils.d.ts

@@ -1,6 +1,6 @@
 import { Runnable } from "@langchain/core/runnables";
 import { Client, Feedback } from "langsmith";
-import { RunEvalConfig } from "./config.js";
+import type { RunEvalConfig } from "./config.js";
 export type ChainOrFactory = Runnable | (() => Runnable) | ((obj: any) => any) | ((obj: any) => Promise<any>) | (() => (obj: unknown) => unknown) | (() => (obj: unknown) => Promise<unknown>);
 export type RunOnDatasetParams = {
     evaluationConfig?: RunEvalConfig;
@@ -35,7 +35,7 @@ export type EvalResults = {
  * for evaluation.
  *
  * @param options - (Optional) Additional parameters for the evaluation process:
- *   - `evaluation` (RunEvalConfig): Configuration for the evaluation, including
+ *   - `evaluationConfig` (RunEvalConfig): Configuration for the evaluation, including
  *     standard and custom evaluators.
  *   - `projectName` (string): Name of the project for logging and tracking.
  *   - `projectMetadata` (Record<string, unknown>): Additional metadata for the project.
@@ -54,10 +54,10 @@ export type EvalResults = {
  *   const datasetName = 'example-dataset';
  *   const client = new Client(/* ...config... *\//);
  *
- *   const evaluationConfig = new RunEvalConfig({
+ *   const evaluationConfig = {
  *     evaluators: [/* ...evaluators... *\//],
  *     customEvaluators: [/* ...custom evaluators... *\//],
- *   });
+ *   };
  *
  *   const results = await runOnDataset(chain, datasetName, {
  *     evaluationConfig,

package/dist/smith/runner_utils.js

@@ -1,5 +1,5 @@
 import { mapStoredMessagesToChatMessages } from "@langchain/core/messages";
-import { Runnable, RunnableLambda } from "@langchain/core/runnables";
+import { Runnable, RunnableLambda, } from "@langchain/core/runnables";
 import { RunCollectorCallbackHandler } from "@langchain/core/tracers/run_collector";
 import { LangChainTracer } from "@langchain/core/tracers/tracer_langchain";
 import { Client } from "langsmith";
@@ -26,9 +26,16 @@ class DynamicRunEvaluator {
      * @returns A promise that resolves to the evaluation result.
      */
     async evaluateRun(run, example) {
-        return await this.evaluator.invoke({ run, example });
+        return await this.evaluator.invoke({
+            run,
+            example,
+            input: run.inputs,
+            prediction: run.outputs,
+            reference: example?.outputs,
+        });
     }
 }
+// eslint-disable-next-line @typescript-eslint/no-explicit-any
 function isLLMStringEvaluator(evaluator) {
     return evaluator && typeof evaluator.evaluateStrings === "function";
 }
@@ -251,7 +258,7 @@ const getExamplesInputs = (examples, chainOrFactory, dataType) => {
  * for evaluation.
  *
  * @param options - (Optional) Additional parameters for the evaluation process:
- *   - `evaluation` (RunEvalConfig): Configuration for the evaluation, including
+ *   - `evaluationConfig` (RunEvalConfig): Configuration for the evaluation, including
  *     standard and custom evaluators.
  *   - `projectName` (string): Name of the project for logging and tracking.
  *   - `projectMetadata` (Record<string, unknown>): Additional metadata for the project.
@@ -270,10 +277,10 @@ const getExamplesInputs = (examples, chainOrFactory, dataType) => {
  *   const datasetName = 'example-dataset';
  *   const client = new Client(/* ...config... *\//);
  *
- *   const evaluationConfig = new RunEvalConfig({
+ *   const evaluationConfig = {
  *     evaluators: [/* ...evaluators... *\//],
  *     customEvaluators: [/* ...custom evaluators... *\//],
- *   });
+ *   };
  *
  *   const results = await runOnDataset(chain, datasetName, {
  *     evaluationConfig,

package/indexes.cjs

@@ -0,0 +1 @@
+module.exports = require('./dist/indexes/index.cjs');

package/indexes.d.ts

@@ -0,0 +1 @@
+export * from './dist/indexes/index.js'

package/indexes.js

@@ -0,0 +1 @@
+export * from './dist/indexes/index.js'

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "langchain",
-  "version": "0.1.6",
+  "version": "0.1.7",
   "description": "Typescript bindings for langchain",
   "type": "module",
   "engines": {
@@ -874,6 +874,9 @@
     "runnables/remote.cjs",
     "runnables/remote.js",
     "runnables/remote.d.ts",
+    "indexes.cjs",
+    "indexes.js",
+    "indexes.d.ts",
     "index.cjs",
     "index.js",
     "index.d.ts"
@@ -1205,7 +1208,7 @@
   },
   "dependencies": {
     "@anthropic-ai/sdk": "^0.9.1",
-    "@langchain/community": "~0.0.17",
+    "@langchain/community": "~0.0.20",
     "@langchain/core": "~0.1.16",
     "@langchain/openai": "~0.0.12",
     "binary-extensions": "^2.2.0",
@@ -2681,6 +2684,11 @@
       "import": "./runnables/remote.js",
       "require": "./runnables/remote.cjs"
     },
+    "./indexes": {
+      "types": "./indexes.d.ts",
+      "import": "./indexes.js",
+      "require": "./indexes.cjs"
+    },
     "./package.json": "./package.json"
   }
 }