@langchain/anthropic
Version:
Anthropic integrations for LangChain.js
50 lines (49 loc) • 1.54 kB
JavaScript
;
Object.defineProperty(exports, "__esModule", { value: true });
exports.convertPromptToAnthropic = void 0;
const message_inputs_js_1 = require("./message_inputs.cjs");
/**
* Convert a formatted LangChain prompt (e.g. pulled from the hub) into
* a format expected by Anthropic's JS SDK.
*
* Requires the "@langchain/anthropic" package to be installed in addition
* to the Anthropic SDK.
*
* @example
* ```ts
* import { convertPromptToAnthropic } from "langsmith/utils/hub/anthropic";
* import { pull } from "langchain/hub";
*
* import Anthropic from '@anthropic-ai/sdk';
*
* const prompt = await pull("jacob/joke-generator");
* const formattedPrompt = await prompt.invoke({
* topic: "cats",
* });
*
* const { system, messages } = convertPromptToAnthropic(formattedPrompt);
*
* const anthropicClient = new Anthropic({
* apiKey: 'your_api_key',
* });
*
* const anthropicResponse = await anthropicClient.messages.create({
* model: "claude-3-5-sonnet-20240620",
* max_tokens: 1024,
* stream: false,
* system,
* messages,
* });
* ```
* @param formattedPrompt
* @returns A partial Anthropic payload.
*/
function convertPromptToAnthropic(formattedPrompt) {
const messages = formattedPrompt.toChatMessages();
const anthropicBody = (0, message_inputs_js_1._convertMessagesToAnthropicPayload)(messages);
if (anthropicBody.messages === undefined) {
anthropicBody.messages = [];
}
return anthropicBody;
}
exports.convertPromptToAnthropic = convertPromptToAnthropic;