n8n/packages/@n8n/nodes-langchain/nodes/llms/LMChatOpenAi/LmChatOpenAi.node.ts

/* eslint-disable n8n-nodes-base/node-dirname-against-convention */
import {
	NodeConnectionType,
	type IExecuteFunctions,
	type INodeType,
	type INodeTypeDescription,
	type SupplyData,
	type JsonObject,
	NodeApiError,
} from 'n8n-workflow';

import { ChatOpenAI, type ClientOptions } from '@langchain/openai';
import { getConnectionHintNoticeField } from '../../../utils/sharedFields';
import { N8nLlmTracing } from '../N8nLlmTracing';
import { RateLimitError } from 'openai';
import { getCustomErrorMessage } from '../../vendors/OpenAi/helpers/error-handling';

export class LmChatOpenAi implements INodeType {
	description: INodeTypeDescription = {
		displayName: 'OpenAI Chat Model',
		// eslint-disable-next-line n8n-nodes-base/node-class-description-name-miscased
		name: 'lmChatOpenAi',
		icon: { light: 'file:openAiLight.svg', dark: 'file:openAiLight.dark.svg' },
		group: ['transform'],
		version: 1,
		description: 'For advanced usage with an AI chain',
		defaults: {
			name: 'OpenAI Chat Model',
		},
		codex: {
			categories: ['AI'],
			subcategories: {
				AI: ['Language Models', 'Root Nodes'],
				'Language Models': ['Chat Models (Recommended)'],
			},
			resources: {
				primaryDocumentation: [
					{
						url: 'https://docs.n8n.io/integrations/builtin/cluster-nodes/sub-nodes/n8n-nodes-langchain.lmchatopenai/',
					},
				],
			},
		},
		// eslint-disable-next-line n8n-nodes-base/node-class-description-inputs-wrong-regular-node
		inputs: [],
		// eslint-disable-next-line n8n-nodes-base/node-class-description-outputs-wrong
		outputs: [NodeConnectionType.AiLanguageModel],
		outputNames: ['Model'],
		credentials: [
			{
				name: 'openAiApi',
				required: true,
			},
		],
		requestDefaults: {
			ignoreHttpStatusErrors: true,
			baseURL:
				'={{ $parameter.options?.baseURL?.split("/").slice(0,-1).join("/") || "https://api.openai.com" }}',
		},
		properties: [
			getConnectionHintNoticeField([NodeConnectionType.AiChain, NodeConnectionType.AiAgent]),
			{
				displayName:
					'If using JSON response format, you must include word "json" in the prompt in your chain or agent. Also, make sure to select latest models released post November 2023.',
				name: 'notice',
				type: 'notice',
				default: '',
				displayOptions: {
					show: {
						'/options.responseFormat': ['json_object'],
					},
				},
			},
			{
				displayName: 'Model',
				name: 'model',
				type: 'options',
				description:
					'The model which will generate the completion. <a href="https://beta.openai.com/docs/models/overview">Learn more</a>.',
				typeOptions: {
					loadOptions: {
						routing: {
							request: {
								method: 'GET',
								url: '={{ $parameter.options?.baseURL?.split("/").slice(-1).pop() || "v1"  }}/models',
							},
							output: {
								postReceive: [
									{
										type: 'rootProperty',
										properties: {
											property: 'data',
										},
									},
									{
										type: 'filter',
										properties: {
											// If the baseURL is not set or is set to api.openai.com, include only chat models
											pass: `={{
												($parameter.options?.baseURL && !$parameter.options?.baseURL?.includes('api.openai.com')) ||
												$responseItem.id.startsWith('ft:') ||
												$responseItem.id.startsWith('o1') ||
												($responseItem.id.startsWith('gpt-') && !$responseItem.id.includes('instruct'))
											}}`,
										},
									},
									{
										type: 'setKeyValue',
										properties: {
											name: '={{$responseItem.id}}',
											value: '={{$responseItem.id}}',
										},
									},
									{
										type: 'sort',
										properties: {
											key: 'name',
										},
									},
								],
							},
						},
					},
				},
				routing: {
					send: {
						type: 'body',
						property: 'model',
					},
				},
				default: 'gpt-3.5-turbo',
			},
			{
				displayName:
					'When using non-OpenAI models via "Base URL" override, not all models might be chat-compatible or support other features, like tools calling or JSON response format',
				name: 'notice',
				type: 'notice',
				default: '',
				displayOptions: {
					show: {
						'/options.baseURL': [{ _cnd: { exists: true } }],
					},
				},
			},
			{
				displayName: 'Options',
				name: 'options',
				placeholder: 'Add Option',
				description: 'Additional options to add',
				type: 'collection',
				default: {},
				options: [
					{
						displayName: 'Base URL',
						name: 'baseURL',
						default: 'https://api.openai.com/v1',
						description: 'Override the default base URL for the API',
						type: 'string',
					},
					{
						displayName: 'Frequency Penalty',
						name: 'frequencyPenalty',
						default: 0,
						typeOptions: { maxValue: 2, minValue: -2, numberPrecision: 1 },
						description:
							"Positive values penalize new tokens based on their existing frequency in the text so far, decreasing the model's likelihood to repeat the same line verbatim",
						type: 'number',
					},
					{
						displayName: 'Maximum Number of Tokens',
						name: 'maxTokens',
						default: -1,
						description:
							'The maximum number of tokens to generate in the completion. Most models have a context length of 2048 tokens (except for the newest models, which support 32,768).',
						type: 'number',
						typeOptions: {
							maxValue: 32768,
						},
					},
					{
						displayName: 'Response Format',
						name: 'responseFormat',
						default: 'text',
						type: 'options',
						options: [
							{
								name: 'Text',
								value: 'text',
								description: 'Regular text response',
							},
							{
								name: 'JSON',
								value: 'json_object',
								description:
									'Enables JSON mode, which should guarantee the message the model generates is valid JSON',
							},
						],
					},
					{
						displayName: 'Presence Penalty',
						name: 'presencePenalty',
						default: 0,
						typeOptions: { maxValue: 2, minValue: -2, numberPrecision: 1 },
						description:
							"Positive values penalize new tokens based on whether they appear in the text so far, increasing the model's likelihood to talk about new topics",
						type: 'number',
					},
					{
						displayName: 'Sampling Temperature',
						name: 'temperature',
						default: 0.7,
						typeOptions: { maxValue: 1, minValue: 0, numberPrecision: 1 },
						description:
							'Controls randomness: Lowering results in less random completions. As the temperature approaches zero, the model will become deterministic and repetitive.',
						type: 'number',
					},
					{
						displayName: 'Timeout',
						name: 'timeout',
						default: 60000,
						description: 'Maximum amount of time a request is allowed to take in milliseconds',
						type: 'number',
					},
					{
						displayName: 'Max Retries',
						name: 'maxRetries',
						default: 2,
						description: 'Maximum number of retries to attempt',
						type: 'number',
					},
					{
						displayName: 'Top P',
						name: 'topP',
						default: 1,
						typeOptions: { maxValue: 1, minValue: 0, numberPrecision: 1 },
						description:
							'Controls diversity via nucleus sampling: 0.5 means half of all likelihood-weighted options are considered. We generally recommend altering this or temperature but not both.',
						type: 'number',
					},
				],
			},
		],
	};

	async supplyData(this: IExecuteFunctions, itemIndex: number): Promise<SupplyData> {
		const credentials = await this.getCredentials('openAiApi');

		const modelName = this.getNodeParameter('model', itemIndex) as string;
		const options = this.getNodeParameter('options', itemIndex, {}) as {
			baseURL?: string;
			frequencyPenalty?: number;
			maxTokens?: number;
			maxRetries: number;
			timeout: number;
			presencePenalty?: number;
			temperature?: number;
			topP?: number;
			responseFormat?: 'text' | 'json_object';
		};

		const configuration: ClientOptions = {};
		if (options.baseURL) {
			configuration.baseURL = options.baseURL;
		}

		const model = new ChatOpenAI({
			openAIApiKey: credentials.apiKey as string,
			modelName,
			...options,
			timeout: options.timeout ?? 60000,
			maxRetries: options.maxRetries ?? 2,
			configuration,
			callbacks: [new N8nLlmTracing(this)],
			modelKwargs: options.responseFormat
				? {
						response_format: { type: options.responseFormat },
					}
				: undefined,
			onFailedAttempt: (error: any) => {
				// If the error is a rate limit error, we want to handle it differently
				// because OpenAI has multiple different rate limit errors
				if (error instanceof RateLimitError) {
					const errorCode = error?.code;
					if (errorCode) {
						const customErrorMessage = getCustomErrorMessage(errorCode);

						const apiError = new NodeApiError(this.getNode(), error as unknown as JsonObject);
						if (customErrorMessage) {
							apiError.message = customErrorMessage;
						}

						throw apiError;
					}
				}

				throw error;
			},
		});

		return {
			response: model,
		};
	}
}