import { NodeConnectionType } from 'n8n-workflow'; import type { INodeTypeBaseDescription, IExecuteFunctions, INodeExecutionData, INodeType, INodeTypeDescription, IDataObject, } from 'n8n-workflow'; import { loadSummarizationChain } from 'langchain/chains'; import type { BaseLanguageModel } from 'langchain/dist/base_language'; import type { Document } from 'langchain/document'; import type { TextSplitter } from 'langchain/text_splitter'; import { RecursiveCharacterTextSplitter } from 'langchain/text_splitter'; import { N8nJsonLoader } from '../../../../utils/N8nJsonLoader'; import { N8nBinaryLoader } from '../../../../utils/N8nBinaryLoader'; import { getTemplateNoticeField } from '../../../../utils/sharedFields'; import { REFINE_PROMPT_TEMPLATE, DEFAULT_PROMPT_TEMPLATE } from '../prompt'; import { getChainPromptsArgs } from '../helpers'; function getInputs(parameters: IDataObject) { const chunkingMode = parameters?.chunkingMode; const operationMode = parameters?.operationMode; const inputs = [ { displayName: '', type: NodeConnectionType.Main }, { displayName: 'Model', maxConnections: 1, type: NodeConnectionType.AiLanguageModel, required: true, }, ]; if (operationMode === 'documentLoader') { inputs.push({ displayName: 'Document', type: NodeConnectionType.AiDocument, required: true, maxConnections: 1, }); return inputs; } if (chunkingMode === 'advanced') { inputs.push({ displayName: 'Text Splitter', type: NodeConnectionType.AiTextSplitter, required: false, maxConnections: 1, }); return inputs; } return inputs; } export class ChainSummarizationV2 implements INodeType { description: INodeTypeDescription; constructor(baseDescription: INodeTypeBaseDescription) { this.description = { ...baseDescription, version: [2], defaults: { name: 'Summarization Chain', color: '#909298', }, // eslint-disable-next-line n8n-nodes-base/node-class-description-inputs-wrong-regular-node inputs: `={{ ((parameter) => { ${getInputs.toString()}; return getInputs(parameter) })($parameter) }}`, outputs: [NodeConnectionType.Main], credentials: [], properties: [ getTemplateNoticeField(1951), { displayName: 'Data to Summarize', name: 'operationMode', noDataExpression: true, type: 'options', description: 'How to pass data into the summarization chain', default: 'nodeInputJson', options: [ { name: 'Use Node Input (JSON)', value: 'nodeInputJson', description: 'Summarize the JSON data coming into this node from the previous one', }, { name: 'Use Node Input (Binary)', value: 'nodeInputBinary', description: 'Summarize the binary data coming into this node from the previous one', }, { name: 'Use Document Loader', value: 'documentLoader', description: 'Use a loader sub-node with more configuration options', }, ], }, { displayName: 'Chunking Strategy', name: 'chunkingMode', noDataExpression: true, type: 'options', description: 'Chunk splitting strategy', default: 'simple', options: [ { name: 'Simple (Define Below)', value: 'simple', }, { name: 'Advanced', value: 'advanced', description: 'Use a splitter sub-node with more configuration options', }, ], displayOptions: { show: { '/operationMode': ['nodeInputJson', 'nodeInputBinary'], }, }, }, { displayName: 'Characters Per Chunk', name: 'chunkSize', description: 'Controls the max size (in terms of number of characters) of the final document chunk', type: 'number', default: 1000, displayOptions: { show: { '/chunkingMode': ['simple'], }, }, }, { displayName: 'Chunk Overlap (Characters)', name: 'chunkOverlap', type: 'number', description: 'Specifies how much characters overlap there should be between chunks', default: 200, displayOptions: { show: { '/chunkingMode': ['simple'], }, }, }, { displayName: 'Options', name: 'options', type: 'collection', default: {}, placeholder: 'Add Option', options: [ { displayName: 'Input Data Field Name', name: 'binaryDataKey', type: 'string', default: 'data', description: 'The name of the field in the agent or chain’s input that contains the binary file to be processed', displayOptions: { show: { '/operationMode': ['nodeInputBinary'], }, }, }, { displayName: 'Summarization Method and Prompts', name: 'summarizationMethodAndPrompts', type: 'fixedCollection', default: { values: { summarizationMethod: 'map_reduce', prompt: DEFAULT_PROMPT_TEMPLATE, combineMapPrompt: DEFAULT_PROMPT_TEMPLATE, }, }, placeholder: 'Add Option', typeOptions: {}, options: [ { name: 'values', displayName: 'Values', values: [ { displayName: 'Summarization Method', name: 'summarizationMethod', type: 'options', description: 'The type of summarization to run', default: 'map_reduce', options: [ { name: 'Map Reduce (Recommended)', value: 'map_reduce', description: 'Summarize each document (or chunk) individually, then summarize those summaries', }, { name: 'Refine', value: 'refine', description: 'Summarize the first document (or chunk). Then update that summary based on the next document (or chunk), and repeat.', }, { name: 'Stuff', value: 'stuff', description: 'Pass all documents (or chunks) at once. Ideal for small datasets.', }, ], }, { displayName: 'Final Prompt to Combine', name: 'combineMapPrompt', type: 'string', hint: 'The prompt to combine individual summaries', displayOptions: { hide: { '/options.summarizationMethodAndPrompts.values.summarizationMethod': [ 'stuff', 'refine', ], }, }, default: DEFAULT_PROMPT_TEMPLATE, typeOptions: { rows: 9, }, }, { displayName: 'Individual Summary Prompt', name: 'prompt', type: 'string', default: DEFAULT_PROMPT_TEMPLATE, hint: 'The prompt to summarize an individual document (or chunk)', displayOptions: { hide: { '/options.summarizationMethodAndPrompts.values.summarizationMethod': [ 'stuff', 'refine', ], }, }, typeOptions: { rows: 9, }, }, { displayName: 'Prompt', name: 'prompt', type: 'string', default: DEFAULT_PROMPT_TEMPLATE, displayOptions: { hide: { '/options.summarizationMethodAndPrompts.values.summarizationMethod': [ 'refine', 'map_reduce', ], }, }, typeOptions: { rows: 9, }, }, { displayName: 'Subsequent (Refine) Prompt', name: 'refinePrompt', type: 'string', displayOptions: { hide: { '/options.summarizationMethodAndPrompts.values.summarizationMethod': [ 'stuff', 'map_reduce', ], }, }, default: REFINE_PROMPT_TEMPLATE, hint: 'The prompt to refine the summary based on the next document (or chunk)', typeOptions: { rows: 9, }, }, { displayName: 'Initial Prompt', name: 'refineQuestionPrompt', type: 'string', displayOptions: { hide: { '/options.summarizationMethodAndPrompts.values.summarizationMethod': [ 'stuff', 'map_reduce', ], }, }, default: DEFAULT_PROMPT_TEMPLATE, hint: 'The prompt for the first document (or chunk)', typeOptions: { rows: 9, }, }, ], }, ], }, ], }, ], }; } async execute(this: IExecuteFunctions): Promise { this.logger.verbose('Executing Summarization Chain V2'); const operationMode = this.getNodeParameter('operationMode', 0, 'nodeInputJson') as | 'nodeInputJson' | 'nodeInputBinary' | 'documentLoader'; const chunkingMode = this.getNodeParameter('chunkingMode', 0, 'simple') as | 'simple' | 'advanced'; const model = (await this.getInputConnectionData( NodeConnectionType.AiLanguageModel, 0, )) as BaseLanguageModel; const items = this.getInputData(); const returnData: INodeExecutionData[] = []; for (let itemIndex = 0; itemIndex < items.length; itemIndex++) { const summarizationMethodAndPrompts = this.getNodeParameter( 'options.summarizationMethodAndPrompts.values', itemIndex, {}, ) as { prompt?: string; refineQuestionPrompt?: string; refinePrompt?: string; summarizationMethod: 'map_reduce' | 'stuff' | 'refine'; combineMapPrompt?: string; }; const chainArgs = getChainPromptsArgs( summarizationMethodAndPrompts.summarizationMethod ?? 'map_reduce', summarizationMethodAndPrompts, ); const chain = loadSummarizationChain(model, chainArgs); const item = items[itemIndex]; let processedDocuments: Document[]; // Use dedicated document loader input to load documents if (operationMode === 'documentLoader') { const documentInput = (await this.getInputConnectionData( NodeConnectionType.AiDocument, 0, )) as N8nJsonLoader | Array>>; const isN8nLoader = documentInput instanceof N8nJsonLoader || documentInput instanceof N8nBinaryLoader; processedDocuments = isN8nLoader ? await documentInput.processItem(item, itemIndex) : documentInput; const response = await chain.call({ input_documents: processedDocuments, }); returnData.push({ json: { response } }); } // Take the input and use binary or json loader if (['nodeInputJson', 'nodeInputBinary'].includes(operationMode)) { let textSplitter: TextSplitter | undefined; switch (chunkingMode) { // In simple mode we use recursive character splitter with default settings case 'simple': const chunkSize = this.getNodeParameter('chunkSize', itemIndex, 1000) as number; const chunkOverlap = this.getNodeParameter('chunkOverlap', itemIndex, 200) as number; textSplitter = new RecursiveCharacterTextSplitter({ chunkOverlap, chunkSize }); break; // In advanced mode user can connect text splitter node so we just retrieve it case 'advanced': textSplitter = (await this.getInputConnectionData( NodeConnectionType.AiTextSplitter, 0, )) as TextSplitter | undefined; break; default: break; } let processor: N8nJsonLoader | N8nBinaryLoader; if (operationMode === 'nodeInputBinary') { const binaryDataKey = this.getNodeParameter( 'options.binaryDataKey', itemIndex, 'data', ) as string; processor = new N8nBinaryLoader(this, 'options.', binaryDataKey, textSplitter); } else { processor = new N8nJsonLoader(this, 'options.', textSplitter); } const processedItem = await processor.processItem(item, itemIndex); const response = await chain.call({ input_documents: processedItem, }); returnData.push({ json: { response } }); } } return this.prepareOutputData(returnData); } }