@langchain/classic 1.0.0-alpha.1

package/dist/chains/conversational_retrieval_chain.d.ts

@@ -0,0 +1,149 @@
+import { SerializedChatVectorDBQAChain } from "./serde.js";
+import { BaseChain, ChainInputs } from "./base.js";
+import { LLMChain } from "./llm_chain.js";
+import { QAChainParams } from "./question_answering/load.js";
+import { BaseLanguageModelInterface } from "@langchain/core/language_models/base";
+import { CallbackManagerForChainRun } from "@langchain/core/callbacks/manager";
+import { BaseMessage } from "@langchain/core/messages";
+import { ChainValues } from "@langchain/core/utils/types";
+import { BaseRetrieverInterface } from "@langchain/core/retrievers";
+
+//#region src/chains/conversational_retrieval_chain.d.ts
+// eslint-disable-next-line @typescript-eslint/no-explicit-any
+type LoadValues = Record<string, any>;
+/**
+ * Interface for the input parameters of the
+ * ConversationalRetrievalQAChain class.
+ */
+interface ConversationalRetrievalQAChainInput extends ChainInputs {
+  retriever: BaseRetrieverInterface;
+  combineDocumentsChain: BaseChain;
+  questionGeneratorChain: LLMChain;
+  returnSourceDocuments?: boolean;
+  returnGeneratedQuestion?: boolean;
+  inputKey?: string;
+}
+/**
+ * Class for conducting conversational question-answering tasks with a
+ * retrieval component. Extends the BaseChain class and implements the
+ * ConversationalRetrievalQAChainInput interface.
+ * @example
+ * ```typescript
+ * import { ChatAnthropic } from "@langchain/anthropic";
+ * import {
+ *   ChatPromptTemplate,
+ *   MessagesPlaceholder,
+ * } from "@langchain/core/prompts";
+ * import { BaseMessage } from "@langchain/core/messages";
+ * import { createStuffDocumentsChain } from "@langchain/classic/chains/combine_documents";
+ * import { createHistoryAwareRetriever } from "@langchain/classic/chains/history_aware_retriever";
+ * import { createRetrievalChain } from "@langchain/classic/chains/retrieval";
+ *
+ * const retriever = ...your retriever;
+ * const llm = new ChatAnthropic();
+ *
+ * // Contextualize question
+ * const contextualizeQSystemPrompt = `
+ * Given a chat history and the latest user question
+ * which might reference context in the chat history,
+ * formulate a standalone question which can be understood
+ * without the chat history. Do NOT answer the question, just
+ * reformulate it if needed and otherwise return it as is.`;
+ * const contextualizeQPrompt = ChatPromptTemplate.fromMessages([
+ *   ["system", contextualizeQSystemPrompt],
+ *   new MessagesPlaceholder("chat_history"),
+ *   ["human", "{input}"],
+ * ]);
+ * const historyAwareRetriever = await createHistoryAwareRetriever({
+ *   llm,
+ *   retriever,
+ *   rephrasePrompt: contextualizeQPrompt,
+ * });
+ *
+ * // Answer question
+ * const qaSystemPrompt = `
+ * You are an assistant for question-answering tasks. Use
+ * the following pieces of retrieved context to answer the
+ * question. If you don't know the answer, just say that you
+ * don't know. Use three sentences maximum and keep the answer
+ * concise.
+ * \n\n
+ * {context}`;
+ * const qaPrompt = ChatPromptTemplate.fromMessages([
+ *   ["system", qaSystemPrompt],
+ *   new MessagesPlaceholder("chat_history"),
+ *   ["human", "{input}"],
+ * ]);
+ *
+ * // Below we use createStuffDocuments_chain to feed all retrieved context
+ * // into the LLM. Note that we can also use StuffDocumentsChain and other
+ * // instances of BaseCombineDocumentsChain.
+ * const questionAnswerChain = await createStuffDocumentsChain({
+ *   llm,
+ *   prompt: qaPrompt,
+ * });
+ *
+ * const ragChain = await createRetrievalChain({
+ *   retriever: historyAwareRetriever,
+ *   combineDocsChain: questionAnswerChain,
+ * });
+ *
+ * // Usage:
+ * const chat_history: BaseMessage[] = [];
+ * const response = await ragChain.invoke({
+ *   chat_history,
+ *   input: "...",
+ * });
+ * ```
+ */
+declare class ConversationalRetrievalQAChain extends BaseChain implements ConversationalRetrievalQAChainInput {
+  static lc_name(): string;
+  inputKey: string;
+  chatHistoryKey: string;
+  get inputKeys(): string[];
+  get outputKeys(): string[];
+  retriever: BaseRetrieverInterface;
+  combineDocumentsChain: BaseChain;
+  questionGeneratorChain: LLMChain;
+  returnSourceDocuments: boolean;
+  returnGeneratedQuestion: boolean;
+  constructor(fields: ConversationalRetrievalQAChainInput);
+  /**
+   * Static method to convert the chat history input into a formatted
+   * string.
+   * @param chatHistory Chat history input which can be a string, an array of BaseMessage instances, or an array of string arrays.
+   * @returns A formatted string representing the chat history.
+   */
+  static getChatHistoryString(chatHistory: string | BaseMessage[] | string[][]): string;
+  /** @ignore */
+  _call(values: ChainValues, runManager?: CallbackManagerForChainRun): Promise<ChainValues>;
+  _chainType(): string;
+  static deserialize(_data: SerializedChatVectorDBQAChain, _values: LoadValues): Promise<ConversationalRetrievalQAChain>;
+  serialize(): SerializedChatVectorDBQAChain;
+  /**
+   * Static method to create a new ConversationalRetrievalQAChain from a
+   * BaseLanguageModel and a BaseRetriever.
+   * @param llm {@link BaseLanguageModelInterface} instance used to generate a new question.
+   * @param retriever {@link BaseRetrieverInterface} instance used to retrieve relevant documents.
+   * @param options.returnSourceDocuments Whether to return source documents in the final output
+   * @param options.questionGeneratorChainOptions Options to initialize the standalone question generation chain used as the first internal step
+   * @param options.qaChainOptions {@link QAChainParams} used to initialize the QA chain used as the second internal step
+   * @returns A new instance of ConversationalRetrievalQAChain.
+   */
+  static fromLLM(llm: BaseLanguageModelInterface, retriever: BaseRetrieverInterface, options?: {
+    outputKey?: string; // not used
+    returnSourceDocuments?: boolean;
+    /** @deprecated Pass in questionGeneratorChainOptions.template instead */
+    questionGeneratorTemplate?: string;
+    /** @deprecated Pass in qaChainOptions.prompt instead */
+    qaTemplate?: string;
+    questionGeneratorChainOptions?: {
+      llm?: BaseLanguageModelInterface;
+      template?: string;
+    };
+    qaChainOptions?: QAChainParams;
+  } & Omit<ConversationalRetrievalQAChainInput, "retriever" | "combineDocumentsChain" | "questionGeneratorChain">): ConversationalRetrievalQAChain;
+}
+//#endregion
+export { ConversationalRetrievalQAChain, ConversationalRetrievalQAChainInput };
+//# sourceMappingURL=conversational_retrieval_chain.d.ts.map

package/dist/chains/conversational_retrieval_chain.d.ts.map

@@ -0,0 +1 @@
+{"version":3,"file":"conversational_retrieval_chain.d.ts","names":["BaseLanguageModelInterface","BaseRetrieverInterface","BaseMessage","ChainValues","CallbackManagerForChainRun","SerializedChatVectorDBQAChain","BaseChain","ChainInputs","LLMChain","QAChainParams","LoadValues","Record","ConversationalRetrievalQAChainInput","ConversationalRetrievalQAChain","Promise","Omit"],"sources":["../../src/chains/conversational_retrieval_chain.d.ts"],"sourcesContent":["import type { BaseLanguageModelInterface } from \"@langchain/core/language_models/base\";\nimport type { BaseRetrieverInterface } from \"@langchain/core/retrievers\";\nimport { BaseMessage } from \"@langchain/core/messages\";\nimport { ChainValues } from \"@langchain/core/utils/types\";\nimport { CallbackManagerForChainRun } from \"@langchain/core/callbacks/manager\";\nimport { SerializedChatVectorDBQAChain } from \"./serde.js\";\nimport { BaseChain, ChainInputs } from \"./base.js\";\nimport { LLMChain } from \"./llm_chain.js\";\nimport { QAChainParams } from \"./question_answering/load.js\";\n// eslint-disable-next-line @typescript-eslint/no-explicit-any\nexport type LoadValues = Record<string, any>;\n/**\n * Interface for the input parameters of the\n * ConversationalRetrievalQAChain class.\n */\nexport interface ConversationalRetrievalQAChainInput extends ChainInputs {\n    retriever: BaseRetrieverInterface;\n    combineDocumentsChain: BaseChain;\n    questionGeneratorChain: LLMChain;\n    returnSourceDocuments?: boolean;\n    returnGeneratedQuestion?: boolean;\n    inputKey?: string;\n}\n/**\n * Class for conducting conversational question-answering tasks with a\n * retrieval component. Extends the BaseChain class and implements the\n * ConversationalRetrievalQAChainInput interface.\n * @example\n * ```typescript\n * import { ChatAnthropic } from \"@langchain/anthropic\";\n * import {\n *   ChatPromptTemplate,\n *   MessagesPlaceholder,\n * } from \"@langchain/core/prompts\";\n * import { BaseMessage } from \"@langchain/core/messages\";\n * import { createStuffDocumentsChain } from \"@langchain/classic/chains/combine_documents\";\n * import { createHistoryAwareRetriever } from \"@langchain/classic/chains/history_aware_retriever\";\n * import { createRetrievalChain } from \"@langchain/classic/chains/retrieval\";\n *\n * const retriever = ...your retriever;\n * const llm = new ChatAnthropic();\n *\n * // Contextualize question\n * const contextualizeQSystemPrompt = `\n * Given a chat history and the latest user question\n * which might reference context in the chat history,\n * formulate a standalone question which can be understood\n * without the chat history. Do NOT answer the question, just\n * reformulate it if needed and otherwise return it as is.`;\n * const contextualizeQPrompt = ChatPromptTemplate.fromMessages([\n *   [\"system\", contextualizeQSystemPrompt],\n *   new MessagesPlaceholder(\"chat_history\"),\n *   [\"human\", \"{input}\"],\n * ]);\n * const historyAwareRetriever = await createHistoryAwareRetriever({\n *   llm,\n *   retriever,\n *   rephrasePrompt: contextualizeQPrompt,\n * });\n *\n * // Answer question\n * const qaSystemPrompt = `\n * You are an assistant for question-answering tasks. Use\n * the following pieces of retrieved context to answer the\n * question. If you don't know the answer, just say that you\n * don't know. Use three sentences maximum and keep the answer\n * concise.\n * \\n\\n\n * {context}`;\n * const qaPrompt = ChatPromptTemplate.fromMessages([\n *   [\"system\", qaSystemPrompt],\n *   new MessagesPlaceholder(\"chat_history\"),\n *   [\"human\", \"{input}\"],\n * ]);\n *\n * // Below we use createStuffDocuments_chain to feed all retrieved context\n * // into the LLM. Note that we can also use StuffDocumentsChain and other\n * // instances of BaseCombineDocumentsChain.\n * const questionAnswerChain = await createStuffDocumentsChain({\n *   llm,\n *   prompt: qaPrompt,\n * });\n *\n * const ragChain = await createRetrievalChain({\n *   retriever: historyAwareRetriever,\n *   combineDocsChain: questionAnswerChain,\n * });\n *\n * // Usage:\n * const chat_history: BaseMessage[] = [];\n * const response = await ragChain.invoke({\n *   chat_history,\n *   input: \"...\",\n * });\n * ```\n */\nexport declare class ConversationalRetrievalQAChain extends BaseChain implements ConversationalRetrievalQAChainInput {\n    static lc_name(): string;\n    inputKey: string;\n    chatHistoryKey: string;\n    get inputKeys(): string[];\n    get outputKeys(): string[];\n    retriever: BaseRetrieverInterface;\n    combineDocumentsChain: BaseChain;\n    questionGeneratorChain: LLMChain;\n    returnSourceDocuments: boolean;\n    returnGeneratedQuestion: boolean;\n    constructor(fields: ConversationalRetrievalQAChainInput);\n    /**\n     * Static method to convert the chat history input into a formatted\n     * string.\n     * @param chatHistory Chat history input which can be a string, an array of BaseMessage instances, or an array of string arrays.\n     * @returns A formatted string representing the chat history.\n     */\n    static getChatHistoryString(chatHistory: string | BaseMessage[] | string[][]): string;\n    /** @ignore */\n    _call(values: ChainValues, runManager?: CallbackManagerForChainRun): Promise<ChainValues>;\n    _chainType(): string;\n    static deserialize(_data: SerializedChatVectorDBQAChain, _values: LoadValues): Promise<ConversationalRetrievalQAChain>;\n    serialize(): SerializedChatVectorDBQAChain;\n    /**\n     * Static method to create a new ConversationalRetrievalQAChain from a\n     * BaseLanguageModel and a BaseRetriever.\n     * @param llm {@link BaseLanguageModelInterface} instance used to generate a new question.\n     * @param retriever {@link BaseRetrieverInterface} instance used to retrieve relevant documents.\n     * @param options.returnSourceDocuments Whether to return source documents in the final output\n     * @param options.questionGeneratorChainOptions Options to initialize the standalone question generation chain used as the first internal step\n     * @param options.qaChainOptions {@link QAChainParams} used to initialize the QA chain used as the second internal step\n     * @returns A new instance of ConversationalRetrievalQAChain.\n     */\n    static fromLLM(llm: BaseLanguageModelInterface, retriever: BaseRetrieverInterface, options?: {\n        outputKey?: string; // not used\n        returnSourceDocuments?: boolean;\n        /** @deprecated Pass in questionGeneratorChainOptions.template instead */\n        questionGeneratorTemplate?: string;\n        /** @deprecated Pass in qaChainOptions.prompt instead */\n        qaTemplate?: string;\n        questionGeneratorChainOptions?: {\n            llm?: BaseLanguageModelInterface;\n            template?: string;\n        };\n        qaChainOptions?: QAChainParams;\n    } & Omit<ConversationalRetrievalQAChainInput, \"retriever\" | \"combineDocumentsChain\" | \"questionGeneratorChain\">): ConversationalRetrievalQAChain;\n}\n"],"mappings":";;;;;;;;;;;;KAUYU,UAAAA,GAAaC;AAAzB;AAKA;;;AAE2BL,UAFVM,mCAAAA,SAA4CL,WAElCD,CAAAA;EAAS,SACRE,EAFbP,sBAEaO;EAAQ,qBAHyBD,EAElCD,SAFkCC;EAAW,sBAAA,EAG5CC,QAH4C;EAiFnDK,qBAAAA,CAAAA,EAAAA,OAAAA;EAA8B,uBAAA,CAAA,EAAA,OAAA;EAAA,QAMpCZ,CAAAA,EAAAA,MAAAA;;;;;;;;;;;;;;;;;;;;;;;AANqG;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;cAA/FY,8BAAAA,SAAuCP,SAAAA,YAAqBM;;;;;;aAMlEX;yBACYK;0BACCE;;;sBAGJI;;;;;;;oDAO8BV;;gBAEpCC,0BAA0BC,6BAA6BU,QAAQX;;4BAEnDE,wCAAwCK,aAAaI,QAAQD;eAC1ER;;;;;;;;;;;sBAWOL,uCAAuCC;;;;;;;;YAQ7CD;;;qBAGOS;MACjBM,KAAKH,yGAAyGC"}

package/dist/chains/conversational_retrieval_chain.js

@@ -0,0 +1,214 @@
+import { BaseChain } from "./base.js";
+import { LLMChain } from "./llm_chain.js";
+import { loadQAChain } from "./question_answering/load.js";
+import { PromptTemplate } from "@langchain/core/prompts";
+import { AIMessage, HumanMessage } from "@langchain/core/messages";
+
+//#region src/chains/conversational_retrieval_chain.ts
+const question_generator_template = `Given the following conversation and a follow up question, rephrase the follow up question to be a standalone question.
+
+Chat History:
+{chat_history}
+Follow Up Input: {question}
+Standalone question:`;
+/**
+* Class for conducting conversational question-answering tasks with a
+* retrieval component. Extends the BaseChain class and implements the
+* ConversationalRetrievalQAChainInput interface.
+* @example
+* ```typescript
+* import { ChatAnthropic } from "@langchain/anthropic";
+* import {
+*   ChatPromptTemplate,
+*   MessagesPlaceholder,
+* } from "@langchain/core/prompts";
+* import { BaseMessage } from "@langchain/core/messages";
+* import { createStuffDocumentsChain } from "@langchain/classic/chains/combine_documents";
+* import { createHistoryAwareRetriever } from "@langchain/classic/chains/history_aware_retriever";
+* import { createRetrievalChain } from "@langchain/classic/chains/retrieval";
+*
+* const retriever = ...your retriever;
+* const llm = new ChatAnthropic();
+*
+* // Contextualize question
+* const contextualizeQSystemPrompt = `
+* Given a chat history and the latest user question
+* which might reference context in the chat history,
+* formulate a standalone question which can be understood
+* without the chat history. Do NOT answer the question, just
+* reformulate it if needed and otherwise return it as is.`;
+* const contextualizeQPrompt = ChatPromptTemplate.fromMessages([
+*   ["system", contextualizeQSystemPrompt],
+*   new MessagesPlaceholder("chat_history"),
+*   ["human", "{input}"],
+* ]);
+* const historyAwareRetriever = await createHistoryAwareRetriever({
+*   llm,
+*   retriever,
+*   rephrasePrompt: contextualizeQPrompt,
+* });
+*
+* // Answer question
+* const qaSystemPrompt = `
+* You are an assistant for question-answering tasks. Use
+* the following pieces of retrieved context to answer the
+* question. If you don't know the answer, just say that you
+* don't know. Use three sentences maximum and keep the answer
+* concise.
+* \n\n
+* {context}`;
+* const qaPrompt = ChatPromptTemplate.fromMessages([
+*   ["system", qaSystemPrompt],
+*   new MessagesPlaceholder("chat_history"),
+*   ["human", "{input}"],
+* ]);
+*
+* // Below we use createStuffDocuments_chain to feed all retrieved context
+* // into the LLM. Note that we can also use StuffDocumentsChain and other
+* // instances of BaseCombineDocumentsChain.
+* const questionAnswerChain = await createStuffDocumentsChain({
+*   llm,
+*   prompt: qaPrompt,
+* });
+*
+* const ragChain = await createRetrievalChain({
+*   retriever: historyAwareRetriever,
+*   combineDocsChain: questionAnswerChain,
+* });
+*
+* // Usage:
+* const chat_history: BaseMessage[] = [];
+* const response = await ragChain.invoke({
+*   chat_history,
+*   input: "...",
+* });
+* ```
+*/
+var ConversationalRetrievalQAChain = class ConversationalRetrievalQAChain extends BaseChain {
+	static lc_name() {
+		return "ConversationalRetrievalQAChain";
+	}
+	inputKey = "question";
+	chatHistoryKey = "chat_history";
+	get inputKeys() {
+		return [this.inputKey, this.chatHistoryKey];
+	}
+	get outputKeys() {
+		return this.combineDocumentsChain.outputKeys.concat(this.returnSourceDocuments ? ["sourceDocuments"] : []);
+	}
+	retriever;
+	combineDocumentsChain;
+	questionGeneratorChain;
+	returnSourceDocuments = false;
+	returnGeneratedQuestion = false;
+	constructor(fields) {
+		super(fields);
+		this.retriever = fields.retriever;
+		this.combineDocumentsChain = fields.combineDocumentsChain;
+		this.questionGeneratorChain = fields.questionGeneratorChain;
+		this.inputKey = fields.inputKey ?? this.inputKey;
+		this.returnSourceDocuments = fields.returnSourceDocuments ?? this.returnSourceDocuments;
+		this.returnGeneratedQuestion = fields.returnGeneratedQuestion ?? this.returnGeneratedQuestion;
+	}
+	/**
+	* Static method to convert the chat history input into a formatted
+	* string.
+	* @param chatHistory Chat history input which can be a string, an array of BaseMessage instances, or an array of string arrays.
+	* @returns A formatted string representing the chat history.
+	*/
+	static getChatHistoryString(chatHistory) {
+		let historyMessages;
+		if (Array.isArray(chatHistory)) {
+			if (Array.isArray(chatHistory[0]) && typeof chatHistory[0][0] === "string") {
+				console.warn("Passing chat history as an array of strings is deprecated.\nPlease see https://js.langchain.com/docs/modules/chains/popular/chat_vector_db#externally-managed-memory for more information.");
+				historyMessages = chatHistory.flat().map((stringMessage, i) => {
+					if (i % 2 === 0) return new HumanMessage(stringMessage);
+					else return new AIMessage(stringMessage);
+				});
+			} else historyMessages = chatHistory;
+			return historyMessages.map((chatMessage) => {
+				if (chatMessage._getType() === "human") return `Human: ${chatMessage.content}`;
+				else if (chatMessage._getType() === "ai") return `Assistant: ${chatMessage.content}`;
+				else return `${chatMessage.content}`;
+			}).join("\n");
+		}
+		return chatHistory;
+	}
+	/** @ignore */
+	async _call(values, runManager) {
+		if (!(this.inputKey in values)) throw new Error(`Question key ${this.inputKey} not found.`);
+		if (!(this.chatHistoryKey in values)) throw new Error(`Chat history key ${this.chatHistoryKey} not found.`);
+		const question = values[this.inputKey];
+		const chatHistory = ConversationalRetrievalQAChain.getChatHistoryString(values[this.chatHistoryKey]);
+		let newQuestion = question;
+		if (chatHistory.length > 0) {
+			const result$1 = await this.questionGeneratorChain.call({
+				question,
+				chat_history: chatHistory
+			}, runManager?.getChild("question_generator"));
+			const keys = Object.keys(result$1);
+			if (keys.length === 1) newQuestion = result$1[keys[0]];
+			else throw new Error("Return from llm chain has multiple values, only single values supported.");
+		}
+		const docs = await this.retriever.invoke(newQuestion, runManager?.getChild("retriever"));
+		const inputs = {
+			question: newQuestion,
+			input_documents: docs,
+			chat_history: chatHistory
+		};
+		let result = await this.combineDocumentsChain.call(inputs, runManager?.getChild("combine_documents"));
+		if (this.returnSourceDocuments) result = {
+			...result,
+			sourceDocuments: docs
+		};
+		if (this.returnGeneratedQuestion) result = {
+			...result,
+			generatedQuestion: newQuestion
+		};
+		return result;
+	}
+	_chainType() {
+		return "conversational_retrieval_chain";
+	}
+	static async deserialize(_data, _values) {
+		throw new Error("Not implemented.");
+	}
+	serialize() {
+		throw new Error("Not implemented.");
+	}
+	/**
+	* Static method to create a new ConversationalRetrievalQAChain from a
+	* BaseLanguageModel and a BaseRetriever.
+	* @param llm {@link BaseLanguageModelInterface} instance used to generate a new question.
+	* @param retriever {@link BaseRetrieverInterface} instance used to retrieve relevant documents.
+	* @param options.returnSourceDocuments Whether to return source documents in the final output
+	* @param options.questionGeneratorChainOptions Options to initialize the standalone question generation chain used as the first internal step
+	* @param options.qaChainOptions {@link QAChainParams} used to initialize the QA chain used as the second internal step
+	* @returns A new instance of ConversationalRetrievalQAChain.
+	*/
+	static fromLLM(llm, retriever, options = {}) {
+		const { questionGeneratorTemplate, qaTemplate, qaChainOptions = {
+			type: "stuff",
+			prompt: qaTemplate ? PromptTemplate.fromTemplate(qaTemplate) : void 0
+		}, questionGeneratorChainOptions, verbose,...rest } = options;
+		const qaChain = loadQAChain(llm, qaChainOptions);
+		const questionGeneratorChainPrompt = PromptTemplate.fromTemplate(questionGeneratorChainOptions?.template ?? questionGeneratorTemplate ?? question_generator_template);
+		const questionGeneratorChain = new LLMChain({
+			prompt: questionGeneratorChainPrompt,
+			llm: questionGeneratorChainOptions?.llm ?? llm,
+			verbose
+		});
+		const instance = new this({
+			retriever,
+			combineDocumentsChain: qaChain,
+			questionGeneratorChain,
+			verbose,
+			...rest
+		});
+		return instance;
+	}
+};
+
+//#endregion
+export { ConversationalRetrievalQAChain };
+//# sourceMappingURL=conversational_retrieval_chain.js.map

package/dist/chains/conversational_retrieval_chain.js.map

@@ -0,0 +1 @@
+{"version":3,"file":"conversational_retrieval_chain.js","names":["fields: ConversationalRetrievalQAChainInput","chatHistory: string | BaseMessage[] | string[][]","historyMessages: BaseMessage[]","values: ChainValues","runManager?: CallbackManagerForChainRun","question: string","chatHistory: string","result","_data: SerializedChatVectorDBQAChain","_values: LoadValues","llm: BaseLanguageModelInterface","retriever: BaseRetrieverInterface","options: {\n      outputKey?: string; // not used\n      returnSourceDocuments?: boolean;\n      /** @deprecated Pass in questionGeneratorChainOptions.template instead */\n      questionGeneratorTemplate?: string;\n      /** @deprecated Pass in qaChainOptions.prompt instead */\n      qaTemplate?: string;\n      questionGeneratorChainOptions?: {\n        llm?: BaseLanguageModelInterface;\n        template?: string;\n      };\n      qaChainOptions?: QAChainParams;\n    } & Omit<\n      ConversationalRetrievalQAChainInput,\n      \"retriever\" | \"combineDocumentsChain\" | \"questionGeneratorChain\"\n    >"],"sources":["../../src/chains/conversational_retrieval_chain.ts"],"sourcesContent":["import type { BaseLanguageModelInterface } from \"@langchain/core/language_models/base\";\nimport type { BaseRetrieverInterface } from \"@langchain/core/retrievers\";\nimport { PromptTemplate } from \"@langchain/core/prompts\";\nimport { BaseMessage, HumanMessage, AIMessage } from \"@langchain/core/messages\";\nimport { ChainValues } from \"@langchain/core/utils/types\";\nimport { CallbackManagerForChainRun } from \"@langchain/core/callbacks/manager\";\nimport { SerializedChatVectorDBQAChain } from \"./serde.js\";\nimport { BaseChain, ChainInputs } from \"./base.js\";\nimport { LLMChain } from \"./llm_chain.js\";\nimport { QAChainParams, loadQAChain } from \"./question_answering/load.js\";\n\n// eslint-disable-next-line @typescript-eslint/no-explicit-any\nexport type LoadValues = Record<string, any>;\n\nconst question_generator_template = `Given the following conversation and a follow up question, rephrase the follow up question to be a standalone question.\n\nChat History:\n{chat_history}\nFollow Up Input: {question}\nStandalone question:`;\n\n/**\n * Interface for the input parameters of the\n * ConversationalRetrievalQAChain class.\n */\nexport interface ConversationalRetrievalQAChainInput extends ChainInputs {\n  retriever: BaseRetrieverInterface;\n  combineDocumentsChain: BaseChain;\n  questionGeneratorChain: LLMChain;\n  returnSourceDocuments?: boolean;\n  returnGeneratedQuestion?: boolean;\n  inputKey?: string;\n}\n\n/**\n * Class for conducting conversational question-answering tasks with a\n * retrieval component. Extends the BaseChain class and implements the\n * ConversationalRetrievalQAChainInput interface.\n * @example\n * ```typescript\n * import { ChatAnthropic } from \"@langchain/anthropic\";\n * import {\n *   ChatPromptTemplate,\n *   MessagesPlaceholder,\n * } from \"@langchain/core/prompts\";\n * import { BaseMessage } from \"@langchain/core/messages\";\n * import { createStuffDocumentsChain } from \"@langchain/classic/chains/combine_documents\";\n * import { createHistoryAwareRetriever } from \"@langchain/classic/chains/history_aware_retriever\";\n * import { createRetrievalChain } from \"@langchain/classic/chains/retrieval\";\n *\n * const retriever = ...your retriever;\n * const llm = new ChatAnthropic();\n *\n * // Contextualize question\n * const contextualizeQSystemPrompt = `\n * Given a chat history and the latest user question\n * which might reference context in the chat history,\n * formulate a standalone question which can be understood\n * without the chat history. Do NOT answer the question, just\n * reformulate it if needed and otherwise return it as is.`;\n * const contextualizeQPrompt = ChatPromptTemplate.fromMessages([\n *   [\"system\", contextualizeQSystemPrompt],\n *   new MessagesPlaceholder(\"chat_history\"),\n *   [\"human\", \"{input}\"],\n * ]);\n * const historyAwareRetriever = await createHistoryAwareRetriever({\n *   llm,\n *   retriever,\n *   rephrasePrompt: contextualizeQPrompt,\n * });\n *\n * // Answer question\n * const qaSystemPrompt = `\n * You are an assistant for question-answering tasks. Use\n * the following pieces of retrieved context to answer the\n * question. If you don't know the answer, just say that you\n * don't know. Use three sentences maximum and keep the answer\n * concise.\n * \\n\\n\n * {context}`;\n * const qaPrompt = ChatPromptTemplate.fromMessages([\n *   [\"system\", qaSystemPrompt],\n *   new MessagesPlaceholder(\"chat_history\"),\n *   [\"human\", \"{input}\"],\n * ]);\n *\n * // Below we use createStuffDocuments_chain to feed all retrieved context\n * // into the LLM. Note that we can also use StuffDocumentsChain and other\n * // instances of BaseCombineDocumentsChain.\n * const questionAnswerChain = await createStuffDocumentsChain({\n *   llm,\n *   prompt: qaPrompt,\n * });\n *\n * const ragChain = await createRetrievalChain({\n *   retriever: historyAwareRetriever,\n *   combineDocsChain: questionAnswerChain,\n * });\n *\n * // Usage:\n * const chat_history: BaseMessage[] = [];\n * const response = await ragChain.invoke({\n *   chat_history,\n *   input: \"...\",\n * });\n * ```\n */\nexport class ConversationalRetrievalQAChain\n  extends BaseChain\n  implements ConversationalRetrievalQAChainInput\n{\n  static lc_name() {\n    return \"ConversationalRetrievalQAChain\";\n  }\n\n  inputKey = \"question\";\n\n  chatHistoryKey = \"chat_history\";\n\n  get inputKeys() {\n    return [this.inputKey, this.chatHistoryKey];\n  }\n\n  get outputKeys() {\n    return this.combineDocumentsChain.outputKeys.concat(\n      this.returnSourceDocuments ? [\"sourceDocuments\"] : []\n    );\n  }\n\n  retriever: BaseRetrieverInterface;\n\n  combineDocumentsChain: BaseChain;\n\n  questionGeneratorChain: LLMChain;\n\n  returnSourceDocuments = false;\n\n  returnGeneratedQuestion = false;\n\n  constructor(fields: ConversationalRetrievalQAChainInput) {\n    super(fields);\n    this.retriever = fields.retriever;\n    this.combineDocumentsChain = fields.combineDocumentsChain;\n    this.questionGeneratorChain = fields.questionGeneratorChain;\n    this.inputKey = fields.inputKey ?? this.inputKey;\n    this.returnSourceDocuments =\n      fields.returnSourceDocuments ?? this.returnSourceDocuments;\n    this.returnGeneratedQuestion =\n      fields.returnGeneratedQuestion ?? this.returnGeneratedQuestion;\n  }\n\n  /**\n   * Static method to convert the chat history input into a formatted\n   * string.\n   * @param chatHistory Chat history input which can be a string, an array of BaseMessage instances, or an array of string arrays.\n   * @returns A formatted string representing the chat history.\n   */\n  static getChatHistoryString(\n    chatHistory: string | BaseMessage[] | string[][]\n  ) {\n    let historyMessages: BaseMessage[];\n    if (Array.isArray(chatHistory)) {\n      // TODO: Deprecate on a breaking release\n      if (\n        Array.isArray(chatHistory[0]) &&\n        typeof chatHistory[0][0] === \"string\"\n      ) {\n        console.warn(\n          \"Passing chat history as an array of strings is deprecated.\\nPlease see https://js.langchain.com/docs/modules/chains/popular/chat_vector_db#externally-managed-memory for more information.\"\n        );\n        historyMessages = chatHistory.flat().map((stringMessage, i) => {\n          if (i % 2 === 0) {\n            return new HumanMessage(stringMessage);\n          } else {\n            return new AIMessage(stringMessage);\n          }\n        });\n      } else {\n        historyMessages = chatHistory as BaseMessage[];\n      }\n      return historyMessages\n        .map((chatMessage) => {\n          if (chatMessage._getType() === \"human\") {\n            return `Human: ${chatMessage.content}`;\n          } else if (chatMessage._getType() === \"ai\") {\n            return `Assistant: ${chatMessage.content}`;\n          } else {\n            return `${chatMessage.content}`;\n          }\n        })\n        .join(\"\\n\");\n    }\n    return chatHistory;\n  }\n\n  /** @ignore */\n  async _call(\n    values: ChainValues,\n    runManager?: CallbackManagerForChainRun\n  ): Promise<ChainValues> {\n    if (!(this.inputKey in values)) {\n      throw new Error(`Question key ${this.inputKey} not found.`);\n    }\n    if (!(this.chatHistoryKey in values)) {\n      throw new Error(`Chat history key ${this.chatHistoryKey} not found.`);\n    }\n    const question: string = values[this.inputKey];\n    const chatHistory: string =\n      ConversationalRetrievalQAChain.getChatHistoryString(\n        values[this.chatHistoryKey]\n      );\n    let newQuestion = question;\n    if (chatHistory.length > 0) {\n      const result = await this.questionGeneratorChain.call(\n        {\n          question,\n          chat_history: chatHistory,\n        },\n        runManager?.getChild(\"question_generator\")\n      );\n      const keys = Object.keys(result);\n      if (keys.length === 1) {\n        newQuestion = result[keys[0]];\n      } else {\n        throw new Error(\n          \"Return from llm chain has multiple values, only single values supported.\"\n        );\n      }\n    }\n    const docs = await this.retriever.invoke(\n      newQuestion,\n      runManager?.getChild(\"retriever\")\n    );\n    const inputs = {\n      question: newQuestion,\n      input_documents: docs,\n      chat_history: chatHistory,\n    };\n    let result = await this.combineDocumentsChain.call(\n      inputs,\n      runManager?.getChild(\"combine_documents\")\n    );\n    if (this.returnSourceDocuments) {\n      result = {\n        ...result,\n        sourceDocuments: docs,\n      };\n    }\n    if (this.returnGeneratedQuestion) {\n      result = {\n        ...result,\n        generatedQuestion: newQuestion,\n      };\n    }\n    return result;\n  }\n\n  _chainType(): string {\n    return \"conversational_retrieval_chain\";\n  }\n\n  static async deserialize(\n    _data: SerializedChatVectorDBQAChain,\n    _values: LoadValues\n  ): Promise<ConversationalRetrievalQAChain> {\n    throw new Error(\"Not implemented.\");\n  }\n\n  serialize(): SerializedChatVectorDBQAChain {\n    throw new Error(\"Not implemented.\");\n  }\n\n  /**\n   * Static method to create a new ConversationalRetrievalQAChain from a\n   * BaseLanguageModel and a BaseRetriever.\n   * @param llm {@link BaseLanguageModelInterface} instance used to generate a new question.\n   * @param retriever {@link BaseRetrieverInterface} instance used to retrieve relevant documents.\n   * @param options.returnSourceDocuments Whether to return source documents in the final output\n   * @param options.questionGeneratorChainOptions Options to initialize the standalone question generation chain used as the first internal step\n   * @param options.qaChainOptions {@link QAChainParams} used to initialize the QA chain used as the second internal step\n   * @returns A new instance of ConversationalRetrievalQAChain.\n   */\n  static fromLLM(\n    llm: BaseLanguageModelInterface,\n    retriever: BaseRetrieverInterface,\n    options: {\n      outputKey?: string; // not used\n      returnSourceDocuments?: boolean;\n      /** @deprecated Pass in questionGeneratorChainOptions.template instead */\n      questionGeneratorTemplate?: string;\n      /** @deprecated Pass in qaChainOptions.prompt instead */\n      qaTemplate?: string;\n      questionGeneratorChainOptions?: {\n        llm?: BaseLanguageModelInterface;\n        template?: string;\n      };\n      qaChainOptions?: QAChainParams;\n    } & Omit<\n      ConversationalRetrievalQAChainInput,\n      \"retriever\" | \"combineDocumentsChain\" | \"questionGeneratorChain\"\n    > = {}\n  ): ConversationalRetrievalQAChain {\n    const {\n      questionGeneratorTemplate,\n      qaTemplate,\n      qaChainOptions = {\n        type: \"stuff\",\n        prompt: qaTemplate\n          ? PromptTemplate.fromTemplate(qaTemplate)\n          : undefined,\n      },\n      questionGeneratorChainOptions,\n      verbose,\n      ...rest\n    } = options;\n\n    const qaChain = loadQAChain(llm, qaChainOptions);\n\n    const questionGeneratorChainPrompt = PromptTemplate.fromTemplate(\n      questionGeneratorChainOptions?.template ??\n        questionGeneratorTemplate ??\n        question_generator_template\n    );\n    const questionGeneratorChain = new LLMChain({\n      prompt: questionGeneratorChainPrompt,\n      llm: questionGeneratorChainOptions?.llm ?? llm,\n      verbose,\n    });\n    const instance = new this({\n      retriever,\n      combineDocumentsChain: qaChain,\n      questionGeneratorChain,\n      verbose,\n      ...rest,\n    });\n    return instance;\n  }\n}\n"],"mappings":";;;;;;;AAcA,MAAM,8BAA8B,CAAC;;;;;oBAKjB,CAAC;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;AAwFrB,IAAa,iCAAb,MAAa,uCACH,UAEV;CACE,OAAO,UAAU;AACf,SAAO;CACR;CAED,WAAW;CAEX,iBAAiB;CAEjB,IAAI,YAAY;AACd,SAAO,CAAC,KAAK,UAAU,KAAK,cAAe;CAC5C;CAED,IAAI,aAAa;AACf,SAAO,KAAK,sBAAsB,WAAW,OAC3C,KAAK,wBAAwB,CAAC,iBAAkB,IAAG,CAAE,EACtD;CACF;CAED;CAEA;CAEA;CAEA,wBAAwB;CAExB,0BAA0B;CAE1B,YAAYA,QAA6C;EACvD,MAAM,OAAO;EACb,KAAK,YAAY,OAAO;EACxB,KAAK,wBAAwB,OAAO;EACpC,KAAK,yBAAyB,OAAO;EACrC,KAAK,WAAW,OAAO,YAAY,KAAK;EACxC,KAAK,wBACH,OAAO,yBAAyB,KAAK;EACvC,KAAK,0BACH,OAAO,2BAA2B,KAAK;CAC1C;;;;;;;CAQD,OAAO,qBACLC,aACA;EACA,IAAIC;AACJ,MAAI,MAAM,QAAQ,YAAY,EAAE;AAE9B,OACE,MAAM,QAAQ,YAAY,GAAG,IAC7B,OAAO,YAAY,GAAG,OAAO,UAC7B;IACA,QAAQ,KACN,6LACD;IACD,kBAAkB,YAAY,MAAM,CAAC,IAAI,CAAC,eAAe,MAAM;AAC7D,SAAI,IAAI,MAAM,EACZ,QAAO,IAAI,aAAa;SAExB,QAAO,IAAI,UAAU;IAExB,EAAC;GACH,OACC,kBAAkB;AAEpB,UAAO,gBACJ,IAAI,CAAC,gBAAgB;AACpB,QAAI,YAAY,UAAU,KAAK,QAC7B,QAAO,CAAC,OAAO,EAAE,YAAY,SAAS;aAC7B,YAAY,UAAU,KAAK,KACpC,QAAO,CAAC,WAAW,EAAE,YAAY,SAAS;QAE1C,QAAO,GAAG,YAAY,SAAS;GAElC,EAAC,CACD,KAAK,KAAK;EACd;AACD,SAAO;CACR;;CAGD,MAAM,MACJC,QACAC,YACsB;AACtB,MAAI,EAAE,KAAK,YAAY,QACrB,OAAM,IAAI,MAAM,CAAC,aAAa,EAAE,KAAK,SAAS,WAAW,CAAC;AAE5D,MAAI,EAAE,KAAK,kBAAkB,QAC3B,OAAM,IAAI,MAAM,CAAC,iBAAiB,EAAE,KAAK,eAAe,WAAW,CAAC;EAEtE,MAAMC,WAAmB,OAAO,KAAK;EACrC,MAAMC,cACJ,+BAA+B,qBAC7B,OAAO,KAAK,gBACb;EACH,IAAI,cAAc;AAClB,MAAI,YAAY,SAAS,GAAG;GAC1B,MAAMC,WAAS,MAAM,KAAK,uBAAuB,KAC/C;IACE;IACA,cAAc;GACf,GACD,YAAY,SAAS,qBAAqB,CAC3C;GACD,MAAM,OAAO,OAAO,KAAKA,SAAO;AAChC,OAAI,KAAK,WAAW,GAClB,cAAcA,SAAO,KAAK;OAE1B,OAAM,IAAI,MACR;EAGL;EACD,MAAM,OAAO,MAAM,KAAK,UAAU,OAChC,aACA,YAAY,SAAS,YAAY,CAClC;EACD,MAAM,SAAS;GACb,UAAU;GACV,iBAAiB;GACjB,cAAc;EACf;EACD,IAAI,SAAS,MAAM,KAAK,sBAAsB,KAC5C,QACA,YAAY,SAAS,oBAAoB,CAC1C;AACD,MAAI,KAAK,uBACP,SAAS;GACP,GAAG;GACH,iBAAiB;EAClB;AAEH,MAAI,KAAK,yBACP,SAAS;GACP,GAAG;GACH,mBAAmB;EACpB;AAEH,SAAO;CACR;CAED,aAAqB;AACnB,SAAO;CACR;CAED,aAAa,YACXC,OACAC,SACyC;AACzC,QAAM,IAAI,MAAM;CACjB;CAED,YAA2C;AACzC,QAAM,IAAI,MAAM;CACjB;;;;;;;;;;;CAYD,OAAO,QACLC,KACAC,WACAC,UAeI,CAAE,GAC0B;EAChC,MAAM,EACJ,2BACA,YACA,iBAAiB;GACf,MAAM;GACN,QAAQ,aACJ,eAAe,aAAa,WAAW,GACvC;EACL,GACD,+BACA,QACA,GAAG,MACJ,GAAG;EAEJ,MAAM,UAAU,YAAY,KAAK,eAAe;EAEhD,MAAM,+BAA+B,eAAe,aAClD,+BAA+B,YAC7B,6BACA,4BACH;EACD,MAAM,yBAAyB,IAAI,SAAS;GAC1C,QAAQ;GACR,KAAK,+BAA+B,OAAO;GAC3C;EACD;EACD,MAAM,WAAW,IAAI,KAAK;GACxB;GACA,uBAAuB;GACvB;GACA;GACA,GAAG;EACJ;AACD,SAAO;CACR;AACF"}

package/dist/chains/graph_qa/cypher.cjs

@@ -0,0 +1,133 @@
+const require_rolldown_runtime = require('../../_virtual/rolldown_runtime.cjs');
+const require_base = require('../base.cjs');
+const require_llm_chain = require('../llm_chain.cjs');
+const require_prompts = require('./prompts.cjs');
+require('../../util/entrypoint_deprecation.cjs');
+
+//#region src/chains/graph_qa/cypher.ts
+var cypher_exports = {};
+require_rolldown_runtime.__export(cypher_exports, {
+	GraphCypherQAChain: () => GraphCypherQAChain,
+	INTERMEDIATE_STEPS_KEY: () => INTERMEDIATE_STEPS_KEY
+});
+const INTERMEDIATE_STEPS_KEY = "intermediateSteps";
+/**
+* Chain for question-answering against a graph by generating Cypher statements.
+*
+* @example
+* ```typescript
+* const chain = new GraphCypherQAChain({
+*   llm: new ChatOpenAI({ model: "gpt-4o-mini", temperature: 0 }),
+*   graph: new Neo4jGraph(),
+* });
+* const res = await chain.invoke("Who played in Pulp Fiction?");
+* ```
+*
+* @security
+* This chain will execute Cypher statements against the provided database.
+* Make sure that the database connection uses credentials
+* that are narrowly-scoped to only include necessary permissions.
+* Failure to do so may result in data corruption or loss, since the calling code
+* may attempt commands that would result in deletion, mutation of data
+* if appropriately prompted or reading sensitive data if such data is present in the database.
+* The best way to guard against such negative outcomes is to (as appropriate) limit the
+* permissions granted to the credentials used with this tool.
+*
+* See https://js.langchain.com/docs/security for more information.
+*/
+var GraphCypherQAChain = class GraphCypherQAChain extends require_base.BaseChain {
+	graph;
+	cypherGenerationChain;
+	qaChain;
+	inputKey = "query";
+	outputKey = "result";
+	topK = 10;
+	returnDirect = false;
+	returnIntermediateSteps = false;
+	constructor(props) {
+		super(props);
+		const { graph, cypherGenerationChain, qaChain, inputKey, outputKey, topK, returnIntermediateSteps, returnDirect } = props;
+		this.graph = graph;
+		this.cypherGenerationChain = cypherGenerationChain;
+		this.qaChain = qaChain;
+		if (inputKey) this.inputKey = inputKey;
+		if (outputKey) this.outputKey = outputKey;
+		if (topK) this.topK = topK;
+		if (returnIntermediateSteps) this.returnIntermediateSteps = returnIntermediateSteps;
+		if (returnDirect) this.returnDirect = returnDirect;
+	}
+	_chainType() {
+		return "graph_cypher_chain";
+	}
+	get inputKeys() {
+		return [this.inputKey];
+	}
+	get outputKeys() {
+		return [this.outputKey];
+	}
+	static fromLLM(props) {
+		const { graph, qaPrompt = require_prompts.CYPHER_QA_PROMPT, cypherPrompt = require_prompts.CYPHER_GENERATION_PROMPT, llm, cypherLLM, qaLLM, returnIntermediateSteps = false, returnDirect = false } = props;
+		if (!cypherLLM && !llm) throw new Error("Either 'llm' or 'cypherLLM' parameters must be provided");
+		if (!qaLLM && !llm) throw new Error("Either 'llm' or 'qaLLM' parameters must be provided");
+		if (cypherLLM && qaLLM && llm) throw new Error("You can specify up to two of 'cypherLLM', 'qaLLM', and 'llm', but not all three simultaneously.");
+		const qaChain = new require_llm_chain.LLMChain({
+			llm: qaLLM || llm,
+			prompt: qaPrompt
+		});
+		const cypherGenerationChain = new require_llm_chain.LLMChain({
+			llm: cypherLLM || llm,
+			prompt: cypherPrompt
+		});
+		return new GraphCypherQAChain({
+			cypherGenerationChain,
+			qaChain,
+			graph,
+			returnIntermediateSteps,
+			returnDirect
+		});
+	}
+	extractCypher(text) {
+		const pattern = /```(.*?)```/s;
+		const matches = text.match(pattern);
+		return matches ? matches[1] : text;
+	}
+	async _call(values, runManager) {
+		const callbacks = runManager?.getChild();
+		const question = values[this.inputKey];
+		const intermediateSteps = [];
+		const generatedCypher = await this.cypherGenerationChain.call({
+			question,
+			schema: this.graph.getSchema()
+		}, callbacks);
+		const extractedCypher = this.extractCypher(generatedCypher.text);
+		await runManager?.handleText(`Generated Cypher:\n`);
+		await runManager?.handleText(`${extractedCypher} green\n`);
+		intermediateSteps.push({ query: extractedCypher });
+		let chainResult;
+		const context = await this.graph.query(extractedCypher, { topK: this.topK });
+		if (this.returnDirect) chainResult = { [this.outputKey]: context };
+		else {
+			await runManager?.handleText("Full Context:\n");
+			await runManager?.handleText(`${context} green\n`);
+			intermediateSteps.push({ context });
+			const result = await this.qaChain.call({
+				question,
+				context: JSON.stringify(context)
+			}, callbacks);
+			chainResult = { [this.outputKey]: result[this.qaChain.outputKey] };
+		}
+		if (this.returnIntermediateSteps) chainResult[INTERMEDIATE_STEPS_KEY] = intermediateSteps;
+		return chainResult;
+	}
+};
+
+//#endregion
+exports.GraphCypherQAChain = GraphCypherQAChain;
+exports.INTERMEDIATE_STEPS_KEY = INTERMEDIATE_STEPS_KEY;
+Object.defineProperty(exports, 'cypher_exports', {
+  enumerable: true,
+  get: function () {
+    return cypher_exports;
+  }
+});
+//# sourceMappingURL=cypher.cjs.map

package/dist/chains/graph_qa/cypher.cjs.map

@@ -0,0 +1 @@
+{"version":3,"file":"cypher.cjs","names":["BaseChain","props: GraphCypherQAChainInput","props: FromLLMInput","CYPHER_QA_PROMPT","CYPHER_GENERATION_PROMPT","LLMChain","text: string","values: ChainValues","runManager?: CallbackManagerForChainRun","chainResult: ChainValues"],"sources":["../../../src/chains/graph_qa/cypher.ts"],"sourcesContent":["import type { BaseLanguageModelInterface } from \"@langchain/core/language_models/base\";\nimport { ChainValues } from \"@langchain/core/utils/types\";\nimport { BasePromptTemplate } from \"@langchain/core/prompts\";\nimport { CallbackManagerForChainRun } from \"@langchain/core/callbacks/manager\";\nimport { LLMChain } from \"../llm_chain.js\";\nimport { BaseChain, ChainInputs } from \"../base.js\";\nimport { CYPHER_GENERATION_PROMPT, CYPHER_QA_PROMPT } from \"./prompts.js\";\nimport { logVersion020MigrationWarning } from \"../../util/entrypoint_deprecation.js\";\n\n/* #__PURE__ */ logVersion020MigrationWarning({\n  oldEntrypointName: \"chains/graph_qa/cypher\",\n  newPackageName: \"@langchain/community\",\n});\n\nexport const INTERMEDIATE_STEPS_KEY = \"intermediateSteps\";\n\nexport interface GraphCypherQAChainInput extends ChainInputs {\n  // eslint-disable-next-line @typescript-eslint/no-explicit-any\n  graph: any;\n  cypherGenerationChain: LLMChain;\n  qaChain: LLMChain;\n  inputKey?: string;\n  outputKey?: string;\n  topK?: number;\n  returnIntermediateSteps?: boolean;\n  returnDirect?: boolean;\n}\n\nexport interface FromLLMInput {\n  // eslint-disable-next-line @typescript-eslint/no-explicit-any\n  graph: any;\n  llm?: BaseLanguageModelInterface;\n  cypherLLM?: BaseLanguageModelInterface;\n  qaLLM?: BaseLanguageModelInterface;\n  qaPrompt?: BasePromptTemplate;\n  cypherPrompt?: BasePromptTemplate;\n  returnIntermediateSteps?: boolean;\n  returnDirect?: boolean;\n}\n\n/**\n * Chain for question-answering against a graph by generating Cypher statements.\n *\n * @example\n * ```typescript\n * const chain = new GraphCypherQAChain({\n *   llm: new ChatOpenAI({ model: \"gpt-4o-mini\", temperature: 0 }),\n *   graph: new Neo4jGraph(),\n * });\n * const res = await chain.invoke(\"Who played in Pulp Fiction?\");\n * ```\n *\n * @security\n * This chain will execute Cypher statements against the provided database.\n * Make sure that the database connection uses credentials\n * that are narrowly-scoped to only include necessary permissions.\n * Failure to do so may result in data corruption or loss, since the calling code\n * may attempt commands that would result in deletion, mutation of data\n * if appropriately prompted or reading sensitive data if such data is present in the database.\n * The best way to guard against such negative outcomes is to (as appropriate) limit the\n * permissions granted to the credentials used with this tool.\n *\n * See https://js.langchain.com/docs/security for more information.\n */\nexport class GraphCypherQAChain extends BaseChain {\n  // eslint-disable-next-line @typescript-eslint/no-explicit-any\n  private graph: any;\n\n  private cypherGenerationChain: LLMChain;\n\n  private qaChain: LLMChain;\n\n  private inputKey = \"query\";\n\n  private outputKey = \"result\";\n\n  private topK = 10;\n\n  private returnDirect = false;\n\n  private returnIntermediateSteps = false;\n\n  constructor(props: GraphCypherQAChainInput) {\n    super(props);\n    const {\n      graph,\n      cypherGenerationChain,\n      qaChain,\n      inputKey,\n      outputKey,\n      topK,\n      returnIntermediateSteps,\n      returnDirect,\n    } = props;\n\n    this.graph = graph;\n    this.cypherGenerationChain = cypherGenerationChain;\n    this.qaChain = qaChain;\n\n    if (inputKey) {\n      this.inputKey = inputKey;\n    }\n    if (outputKey) {\n      this.outputKey = outputKey;\n    }\n    if (topK) {\n      this.topK = topK;\n    }\n    if (returnIntermediateSteps) {\n      this.returnIntermediateSteps = returnIntermediateSteps;\n    }\n    if (returnDirect) {\n      this.returnDirect = returnDirect;\n    }\n  }\n\n  _chainType() {\n    return \"graph_cypher_chain\" as const;\n  }\n\n  get inputKeys(): string[] {\n    return [this.inputKey];\n  }\n\n  get outputKeys(): string[] {\n    return [this.outputKey];\n  }\n\n  static fromLLM(props: FromLLMInput): GraphCypherQAChain {\n    const {\n      graph,\n      qaPrompt = CYPHER_QA_PROMPT,\n      cypherPrompt = CYPHER_GENERATION_PROMPT,\n      llm,\n      cypherLLM,\n      qaLLM,\n      returnIntermediateSteps = false,\n      returnDirect = false,\n    } = props;\n\n    if (!cypherLLM && !llm) {\n      throw new Error(\n        \"Either 'llm' or 'cypherLLM' parameters must be provided\"\n      );\n    }\n\n    if (!qaLLM && !llm) {\n      throw new Error(\"Either 'llm' or 'qaLLM' parameters must be provided\");\n    }\n\n    if (cypherLLM && qaLLM && llm) {\n      throw new Error(\n        \"You can specify up to two of 'cypherLLM', 'qaLLM', and 'llm', but not all three simultaneously.\"\n      );\n    }\n\n    const qaChain = new LLMChain({\n      llm: (qaLLM || llm) as BaseLanguageModelInterface,\n      prompt: qaPrompt,\n    });\n\n    const cypherGenerationChain = new LLMChain({\n      llm: (cypherLLM || llm) as BaseLanguageModelInterface,\n      prompt: cypherPrompt,\n    });\n\n    return new GraphCypherQAChain({\n      cypherGenerationChain,\n      qaChain,\n      graph,\n      returnIntermediateSteps,\n      returnDirect,\n    });\n  }\n\n  private extractCypher(text: string): string {\n    const pattern = /```(.*?)```/s;\n    const matches = text.match(pattern);\n    return matches ? matches[1] : text;\n  }\n\n  async _call(\n    values: ChainValues,\n    runManager?: CallbackManagerForChainRun\n  ): Promise<ChainValues> {\n    const callbacks = runManager?.getChild();\n    const question = values[this.inputKey];\n\n    const intermediateSteps = [];\n\n    const generatedCypher = await this.cypherGenerationChain.call(\n      { question, schema: this.graph.getSchema() },\n      callbacks\n    );\n\n    const extractedCypher = this.extractCypher(generatedCypher.text);\n\n    await runManager?.handleText(`Generated Cypher:\\n`);\n    await runManager?.handleText(`${extractedCypher} green\\n`);\n\n    intermediateSteps.push({ query: extractedCypher });\n\n    let chainResult: ChainValues;\n    const context = await this.graph.query(extractedCypher, {\n      topK: this.topK,\n    });\n\n    if (this.returnDirect) {\n      chainResult = { [this.outputKey]: context };\n    } else {\n      await runManager?.handleText(\"Full Context:\\n\");\n      await runManager?.handleText(`${context} green\\n`);\n\n      intermediateSteps.push({ context });\n\n      const result = await this.qaChain.call(\n        { question, context: JSON.stringify(context) },\n        callbacks\n      );\n\n      chainResult = {\n        [this.outputKey]: result[this.qaChain.outputKey],\n      };\n    }\n\n    if (this.returnIntermediateSteps) {\n      chainResult[INTERMEDIATE_STEPS_KEY] = intermediateSteps;\n    }\n\n    return chainResult;\n  }\n}\n"],"mappings":";;;;;;;;;;;;AAcA,MAAa,yBAAyB;;;;;;;;;;;;;;;;;;;;;;;;;AAkDtC,IAAa,qBAAb,MAAa,2BAA2BA,uBAAU;CAEhD,AAAQ;CAER,AAAQ;CAER,AAAQ;CAER,AAAQ,WAAW;CAEnB,AAAQ,YAAY;CAEpB,AAAQ,OAAO;CAEf,AAAQ,eAAe;CAEvB,AAAQ,0BAA0B;CAElC,YAAYC,OAAgC;EAC1C,MAAM,MAAM;EACZ,MAAM,EACJ,OACA,uBACA,SACA,UACA,WACA,MACA,yBACA,cACD,GAAG;EAEJ,KAAK,QAAQ;EACb,KAAK,wBAAwB;EAC7B,KAAK,UAAU;AAEf,MAAI,UACF,KAAK,WAAW;AAElB,MAAI,WACF,KAAK,YAAY;AAEnB,MAAI,MACF,KAAK,OAAO;AAEd,MAAI,yBACF,KAAK,0BAA0B;AAEjC,MAAI,cACF,KAAK,eAAe;CAEvB;CAED,aAAa;AACX,SAAO;CACR;CAED,IAAI,YAAsB;AACxB,SAAO,CAAC,KAAK,QAAS;CACvB;CAED,IAAI,aAAuB;AACzB,SAAO,CAAC,KAAK,SAAU;CACxB;CAED,OAAO,QAAQC,OAAyC;EACtD,MAAM,EACJ,OACA,WAAWC,kCACX,eAAeC,0CACf,KACA,WACA,OACA,0BAA0B,OAC1B,eAAe,OAChB,GAAG;AAEJ,MAAI,CAAC,aAAa,CAAC,IACjB,OAAM,IAAI,MACR;AAIJ,MAAI,CAAC,SAAS,CAAC,IACb,OAAM,IAAI,MAAM;AAGlB,MAAI,aAAa,SAAS,IACxB,OAAM,IAAI,MACR;EAIJ,MAAM,UAAU,IAAIC,2BAAS;GAC3B,KAAM,SAAS;GACf,QAAQ;EACT;EAED,MAAM,wBAAwB,IAAIA,2BAAS;GACzC,KAAM,aAAa;GACnB,QAAQ;EACT;AAED,SAAO,IAAI,mBAAmB;GAC5B;GACA;GACA;GACA;GACA;EACD;CACF;CAED,AAAQ,cAAcC,MAAsB;EAC1C,MAAM,UAAU;EAChB,MAAM,UAAU,KAAK,MAAM,QAAQ;AACnC,SAAO,UAAU,QAAQ,KAAK;CAC/B;CAED,MAAM,MACJC,QACAC,YACsB;EACtB,MAAM,YAAY,YAAY,UAAU;EACxC,MAAM,WAAW,OAAO,KAAK;EAE7B,MAAM,oBAAoB,CAAE;EAE5B,MAAM,kBAAkB,MAAM,KAAK,sBAAsB,KACvD;GAAE;GAAU,QAAQ,KAAK,MAAM,WAAW;EAAE,GAC5C,UACD;EAED,MAAM,kBAAkB,KAAK,cAAc,gBAAgB,KAAK;EAEhE,MAAM,YAAY,WAAW,CAAC,mBAAmB,CAAC,CAAC;EACnD,MAAM,YAAY,WAAW,GAAG,gBAAgB,QAAQ,CAAC,CAAC;EAE1D,kBAAkB,KAAK,EAAE,OAAO,gBAAiB,EAAC;EAElD,IAAIC;EACJ,MAAM,UAAU,MAAM,KAAK,MAAM,MAAM,iBAAiB,EACtD,MAAM,KAAK,KACZ,EAAC;AAEF,MAAI,KAAK,cACP,cAAc,GAAG,KAAK,YAAY,QAAS;OACtC;GACL,MAAM,YAAY,WAAW,kBAAkB;GAC/C,MAAM,YAAY,WAAW,GAAG,QAAQ,QAAQ,CAAC,CAAC;GAElD,kBAAkB,KAAK,EAAE,QAAS,EAAC;GAEnC,MAAM,SAAS,MAAM,KAAK,QAAQ,KAChC;IAAE;IAAU,SAAS,KAAK,UAAU,QAAQ;GAAE,GAC9C,UACD;GAED,cAAc,GACX,KAAK,YAAY,OAAO,KAAK,QAAQ,WACvC;EACF;AAED,MAAI,KAAK,yBACP,YAAY,0BAA0B;AAGxC,SAAO;CACR;AACF"}

package/dist/chains/graph_qa/cypher.d.cts

@@ -0,0 +1,76 @@
+import { BaseChain, ChainInputs } from "../base.cjs";
+import { LLMChain } from "../llm_chain.cjs";
+import { ChainValues } from "@langchain/core/utils/types";
+import { BasePromptTemplate } from "@langchain/core/prompts";
+import { BaseLanguageModelInterface } from "@langchain/core/language_models/base";
+import { CallbackManagerForChainRun } from "@langchain/core/callbacks/manager";
+
+//#region src/chains/graph_qa/cypher.d.ts
+declare const INTERMEDIATE_STEPS_KEY = "intermediateSteps";
+interface GraphCypherQAChainInput extends ChainInputs {
+  // eslint-disable-next-line @typescript-eslint/no-explicit-any
+  graph: any;
+  cypherGenerationChain: LLMChain;
+  qaChain: LLMChain;
+  inputKey?: string;
+  outputKey?: string;
+  topK?: number;
+  returnIntermediateSteps?: boolean;
+  returnDirect?: boolean;
+}
+interface FromLLMInput {
+  // eslint-disable-next-line @typescript-eslint/no-explicit-any
+  graph: any;
+  llm?: BaseLanguageModelInterface;
+  cypherLLM?: BaseLanguageModelInterface;
+  qaLLM?: BaseLanguageModelInterface;
+  qaPrompt?: BasePromptTemplate;
+  cypherPrompt?: BasePromptTemplate;
+  returnIntermediateSteps?: boolean;
+  returnDirect?: boolean;
+}
+/**
+ * Chain for question-answering against a graph by generating Cypher statements.
+ *
+ * @example
+ * ```typescript
+ * const chain = new GraphCypherQAChain({
+ *   llm: new ChatOpenAI({ model: "gpt-4o-mini", temperature: 0 }),
+ *   graph: new Neo4jGraph(),
+ * });
+ * const res = await chain.invoke("Who played in Pulp Fiction?");
+ * ```
+ *
+ * @security
+ * This chain will execute Cypher statements against the provided database.
+ * Make sure that the database connection uses credentials
+ * that are narrowly-scoped to only include necessary permissions.
+ * Failure to do so may result in data corruption or loss, since the calling code
+ * may attempt commands that would result in deletion, mutation of data
+ * if appropriately prompted or reading sensitive data if such data is present in the database.
+ * The best way to guard against such negative outcomes is to (as appropriate) limit the
+ * permissions granted to the credentials used with this tool.
+ *
+ * See https://js.langchain.com/docs/security for more information.
+ */
+declare class GraphCypherQAChain extends BaseChain {
+  // eslint-disable-next-line @typescript-eslint/no-explicit-any
+  private graph;
+  private cypherGenerationChain;
+  private qaChain;
+  private inputKey;
+  private outputKey;
+  private topK;
+  private returnDirect;
+  private returnIntermediateSteps;
+  constructor(props: GraphCypherQAChainInput);
+  _chainType(): "graph_cypher_chain";
+  get inputKeys(): string[];
+  get outputKeys(): string[];
+  static fromLLM(props: FromLLMInput): GraphCypherQAChain;
+  private extractCypher;
+  _call(values: ChainValues, runManager?: CallbackManagerForChainRun): Promise<ChainValues>;
+}
+//#endregion
+export { FromLLMInput, GraphCypherQAChain, GraphCypherQAChainInput, INTERMEDIATE_STEPS_KEY };
+//# sourceMappingURL=cypher.d.cts.map

package/dist/chains/graph_qa/cypher.d.cts.map

@@ -0,0 +1 @@
+{"version":3,"file":"cypher.d.cts","names":["BaseLanguageModelInterface","ChainValues","BasePromptTemplate","CallbackManagerForChainRun","LLMChain","BaseChain","ChainInputs","INTERMEDIATE_STEPS_KEY","GraphCypherQAChainInput","FromLLMInput","GraphCypherQAChain","Promise"],"sources":["../../../src/chains/graph_qa/cypher.d.ts"],"sourcesContent":["import type { BaseLanguageModelInterface } from \"@langchain/core/language_models/base\";\nimport { ChainValues } from \"@langchain/core/utils/types\";\nimport { BasePromptTemplate } from \"@langchain/core/prompts\";\nimport { CallbackManagerForChainRun } from \"@langchain/core/callbacks/manager\";\nimport { LLMChain } from \"../llm_chain.js\";\nimport { BaseChain, ChainInputs } from \"../base.js\";\nexport declare const INTERMEDIATE_STEPS_KEY = \"intermediateSteps\";\nexport interface GraphCypherQAChainInput extends ChainInputs {\n    // eslint-disable-next-line @typescript-eslint/no-explicit-any\n    graph: any;\n    cypherGenerationChain: LLMChain;\n    qaChain: LLMChain;\n    inputKey?: string;\n    outputKey?: string;\n    topK?: number;\n    returnIntermediateSteps?: boolean;\n    returnDirect?: boolean;\n}\nexport interface FromLLMInput {\n    // eslint-disable-next-line @typescript-eslint/no-explicit-any\n    graph: any;\n    llm?: BaseLanguageModelInterface;\n    cypherLLM?: BaseLanguageModelInterface;\n    qaLLM?: BaseLanguageModelInterface;\n    qaPrompt?: BasePromptTemplate;\n    cypherPrompt?: BasePromptTemplate;\n    returnIntermediateSteps?: boolean;\n    returnDirect?: boolean;\n}\n/**\n * Chain for question-answering against a graph by generating Cypher statements.\n *\n * @example\n * ```typescript\n * const chain = new GraphCypherQAChain({\n *   llm: new ChatOpenAI({ model: \"gpt-4o-mini\", temperature: 0 }),\n *   graph: new Neo4jGraph(),\n * });\n * const res = await chain.invoke(\"Who played in Pulp Fiction?\");\n * ```\n *\n * @security\n * This chain will execute Cypher statements against the provided database.\n * Make sure that the database connection uses credentials\n * that are narrowly-scoped to only include necessary permissions.\n * Failure to do so may result in data corruption or loss, since the calling code\n * may attempt commands that would result in deletion, mutation of data\n * if appropriately prompted or reading sensitive data if such data is present in the database.\n * The best way to guard against such negative outcomes is to (as appropriate) limit the\n * permissions granted to the credentials used with this tool.\n *\n * See https://js.langchain.com/docs/security for more information.\n */\nexport declare class GraphCypherQAChain extends BaseChain {\n    // eslint-disable-next-line @typescript-eslint/no-explicit-any\n    private graph;\n    private cypherGenerationChain;\n    private qaChain;\n    private inputKey;\n    private outputKey;\n    private topK;\n    private returnDirect;\n    private returnIntermediateSteps;\n    constructor(props: GraphCypherQAChainInput);\n    _chainType(): \"graph_cypher_chain\";\n    get inputKeys(): string[];\n    get outputKeys(): string[];\n    static fromLLM(props: FromLLMInput): GraphCypherQAChain;\n    private extractCypher;\n    _call(values: ChainValues, runManager?: CallbackManagerForChainRun): Promise<ChainValues>;\n}\n"],"mappings":";;;;;;;;cAMqBO,sBAAAA;UACJC,uBAAAA,SAAgCF;EAD5BC;EACJC,KAAAA,EAAAA,GAAAA;EAAuB,qBAAA,EAGbJ,QAHa;EAAA,OAGbA,EACdA,QADcA;EAAQ,QACtBA,CAAAA,EAAAA,MAAAA;EAAQ,SAJ4BE,CAAAA,EAAAA,MAAAA;EAAW,IAAA,CAAA,EAAA,MAAA;EAW3CG,uBAAY,CAAA,EAAA,OAAA;EAAA,YAAA,CAAA,EAAA,OAAA;;AAIbT,UAJCS,YAAAA,CAIDT;EAA0B;EACJ,KACvBE,EAAAA,GAAAA;EAAkB,GACdA,CAAAA,EAJTF,0BAISE;EAAkB,SAAA,CAAA,EAHrBF,0BAGqB;EA4BhBU,KAAAA,CAAAA,EA9BTV,0BA8B2B;EAAA,QAAA,CAAA,EA7BxBE,kBA6BwB;EAAA,YAUhBM,CAAAA,EAtCJN,kBAsCIM;EAAuB,uBAIpBC,CAAAA,EAAAA,OAAAA;EAAY,YAAGC,CAAAA,EAAAA,OAAAA;;;;;;AAdgB;;;;;;;;;;;;;;;;;;;;cAApCA,kBAAAA,SAA2BL,SAAAA;;;;;;;;;;qBAUzBG;;;;wBAIGC,eAAeC;;gBAEvBT,0BAA0BE,6BAA6BQ,QAAQV"}