n8n AI Agent Workflow for Document-Based Q&A

This n8n workflow demonstrates a powerful AI agent that can answer questions based on documents stored in Google Drive. It leverages LangChain components to process documents, create embeddings, store them in a Pinecone vector database, and then use an OpenAI chat model to answer user queries.

This workflow is ideal for building a knowledge base Q&A system from your existing documents, enabling users to get answers without needing to manually search through files.

What it does

This workflow orchestrates several AI and data management tasks:

1. Triggers on Google Drive Changes: It listens for new or updated files in a specified Google Drive folder.
2. Loads Documents: When a change is detected, it loads the content of the relevant document using a default data loader.
3. Splits Text: The document content is then split into smaller, manageable chunks using a Recursive Character Text Splitter to optimize for embedding and retrieval.
4. Generates Embeddings: OpenAI Embeddings are generated for each text chunk, converting them into numerical representations.
5. Stores in Vector Database: These embeddings are then stored in a Pinecone Vector Store, making them searchable for semantic similarity.
6. Responds to Webhook Queries: A separate webhook endpoint allows users to send questions.
7. Retrieves Relevant Information: The AI Agent uses the Pinecone Vector Store to find the most relevant document chunks based on the user's query.
8. Generates Answers: An OpenAI Chat Model, augmented with a simple memory, synthesizes an answer using the retrieved information.
9. Returns the Answer: The AI-generated answer is sent back as the webhook response.

Prerequisites/Requirements

To use this workflow, you will need:

• n8n Instance: A running n8n instance.
• Google Drive Account: With access to the folder containing your documents.
• OpenAI API Key: For generating embeddings and chat model responses.
• Pinecone Account: To set up and use a vector database for document embeddings.
• LangChain Nodes: Ensure the @n8n/n8n-nodes-langchain package is installed in your n8n instance.

Setup/Usage

1. Import the Workflow: Download the provided JSON and import it into your n8n instance.
2. Configure Credentials:
   • Google Drive Trigger: Set up your Google credentials for the Google Drive Trigger node to monitor your desired folder.
   • OpenAI Embeddings: Configure your OpenAI API key for the Embeddings OpenAI node.
   • OpenAI Chat Model: Configure your OpenAI API key for the OpenAI Chat Model node.
   • Pinecone Vector Store: Set up your Pinecone credentials (API Key and Environment) for the Pinecone Vector Store node.
3. Activate the Workflow: Ensure the Google Drive Trigger is set to "Active" to start monitoring for document changes.
4. Populate Pinecone: The first time the workflow runs (or when new documents are added to Google Drive), it will process your documents and populate the Pinecone vector store.
5. Test the Webhook: Once the vector store is populated, you can send POST requests to the Webhook node's URL with your questions. The workflow will process the query and respond with an AI-generated answer.

This workflow provides a robust foundation for building intelligent Q&A systems directly from your document repositories.