Get the Source Documents from the Q&A Chain Agent

Custom n8n Node for Question-Answering with Source Document Output

Overview

n8n-nodes-chain-retrieval-qa is a custom n8n node that leverages the LangChain library to perform question-answering tasks using retrieved documents. This node includes an optional feature to output the source documents that contributed to the response, providing enhanced traceability and transparency.

Features

• Question Answering: Ask questions based on retrieved documents.
• Source Document Output: Enable optional output of source documents for better traceability.
• Integration with LangChain: Uses LangChain's Retrieval QA Chain for powerful retrieval and response generation.

Add to Your n8n Setup

1. Place the installed package into your n8n custom node directory.
2. Restart your n8n instance to load the new node.

Usage

Input

This node requires two inputs:

1. Model: A language model for generating responses (e.g., OpenAI, HuggingFace).
2. Retriever: A retriever to fetch relevant documents from a data source.

Parameters

• Query: The question or input for the QA system.
• Return Source Documents: Enable this option to include the documents that contributed to the response in the output.
• Options:
  • System Prompt Template: Customize the system prompt for context-aware responses.

Output

The output includes:

1. Response: The generated answer to the query.
2. Source Documents (optional): An array of documents used to generate the response (if enabled).

Examples

Example Workflow

1. Add the Question and Answer Chain with Source Documents node to your workflow.
2. Connect it to:
   • A language model node (e.g., OpenAI or HuggingFace).
   • A retriever node for fetching relevant documents.
3. Configure the parameters as needed.

Example Output

{
	"response": "n8n is a powerful workflow automation tool.",
	"documents": [
		{ "content": "n8n is an open-source workflow automation tool." },
		{ "content": "It allows users to automate tasks across various services." }
	]
}

n8n Node example:

{
	"nodes": [
		{
			"parameters": {
				"options": {}
			},
			"type": "@n8n/n8n-nodes-langchain.lmChatOpenAi",
			"typeVersion": 1,
			"position": [80, -140],
			"id": "fe2997f6-f581-419d-be40-ff85fac3b6e8",
			"name": "OpenAI Chat Model",
			"credentials": {
				"openAiApi": {
					"id": "7KA1OCyTLjMmwqUx",
					"name": "OpenAi account"
				}
			}
		},
		{
			"parameters": {},
			"type": "@n8n/n8n-nodes-langchain.retrieverVectorStore",
			"typeVersion": 1,
			"position": [260, -120],
			"id": "cc94a29c-8b8f-4fbc-b7bf-0f3df0fc8b02",
			"name": "Vector Store Retriever"
		},
		{
			"parameters": {
				"options": {}
			},
			"type": "@n8n/n8n-nodes-langchain.embeddingsOpenAi",
			"typeVersion": 1.1,
			"position": [480, 100],
			"id": "77a18ee1-8586-48a0-b4b9-f7bd0b9582b6",
			"name": "Embeddings OpenAI",
			"credentials": {
				"openAiApi": {
					"id": "7KA1OCyTLjMmwqUx",
					"name": "OpenAi account"
				}
			}
		},
		{
			"parameters": {},
			"type": "n8n-nodes-base.executeWorkflowTrigger",
			"typeVersion": 1,
			"position": [-220, -520],
			"id": "92dae8a6-d6b7-4ae5-b5d6-4cc58d743208",
			"name": "Execute Workflow Trigger"
		},
		{
			"parameters": {
				"options": {}
			},
			"type": "CUSTOM.chainRetrievalQaWithSourceDocuments",
			"typeVersion": 1.4,
			"position": [120, -520],
			"id": "097833c9-30cf-4786-b50a-32f3ecc9a651",
			"name": "Question and Answer Chain With Source Documents"
		}
	],
	"connections": {
		"OpenAI Chat Model": {
			"ai_languageModel": [
				[
					{
						"node": "Question and Answer Chain With Source Documents",
						"type": "ai_languageModel",
						"index": 0
					}
				]
			]
		},
		"Vector Store Retriever": {
			"ai_retriever": [
				[
					{
						"node": "Question and Answer Chain With Source Documents",
						"type": "ai_retriever",
						"index": 0
					}
				]
			]
		},
		"Embeddings OpenAI": {
			"ai_embedding": [[]]
		},
		"Execute Workflow Trigger": {
			"main": [
				[
					{
						"node": "Question and Answer Chain With Source Documents",
						"type": "main",
						"index": 0
					}
				]
			]
		}
	},
	"pinData": {
		"Execute Workflow Trigger": [
			{
				"query": {
					"message": "Are there any one bedroom apartment?"
				}
			}
		]
	}
}

Development

Prerequisites

• Node.js >= 18.10
• pnpm >= 9.1

Clone and Build

1. Clone the repository:

   git clone https://github.com/leonardogrig/n8n-metadata-chain-retrieval-qa.git
   cd n8n-metadata-chain-retrieval-qa

2. Install dependencies:

   pnpm install

3. Build the project:

   pnpm build

License

This project is licensed under the MIT License. See the LICENSE file for details.

Author

Leonardo Grigorio Araujo
Email: [email protected]
GitHub: leonardogrig