n8n AI Chatbot with Google Gemini, Pinecone, and RAG

This n8n workflow demonstrates a powerful AI chatbot that leverages Google Gemini for conversational AI, Pinecone for vector storage and Retrieval Augmented Generation (RAG), and a custom data source from Google Sheets. It allows users to interact with an AI agent that can answer questions based on a pre-loaded knowledge base and maintain conversational memory.

What it does

This workflow automates the following steps:

1. Listens for Chat Messages: It starts by listening for incoming chat messages via the n8n Chat Trigger.
2. Checks for Initial Setup: It uses an "If" node to check if a specific "setup" command is received.
3. Initial Data Loading and Vectorization (Setup Branch):
   • If the "setup" command is received, it retrieves data from a Google Sheet.
   • It converts the Google Sheet data into a file format.
   • It then loads this data using a "Default Data Loader".
   • The loaded text is split into smaller chunks using a "Recursive Character Text Splitter".
   • These text chunks are then converted into embeddings using "Embeddings Google Gemini".
   • Finally, these embeddings are stored in a "Pinecone Vector Store" to create the knowledge base for RAG.
   • A "Wait" node is included, likely to ensure the vector store is fully populated before further interactions.
   • An HTTP Request is made, potentially to confirm the setup completion.
4. AI Chat Interaction (Main Branch):
   • If it's not the "setup" command, the workflow proceeds to the AI chat interaction.
   • It uses a "Simple Memory" node to maintain conversational context.
   • An "AI Agent" node, configured with the "Google Gemini Chat Model" and the "Pinecone Vector Store", processes the user's query. This agent uses RAG to retrieve relevant information from Pinecone based on the user's input and the conversational history, and then uses Gemini to formulate a response.
   • The AI Agent's response is then sent back to the user via the chat interface (implied by the Chat Trigger).

Prerequisites/Requirements

To use this workflow, you will need:

• n8n Instance: A running n8n instance.
• Google Account: For Google Sheets and Google Gemini API access.
• Google Sheets Credential: Configured in n8n to access your spreadsheet.
• Google Gemini API Key: For the "Embeddings Google Gemini" and "Google Gemini Chat Model" nodes.
• Pinecone Account: For vector database storage.
• Pinecone Credential: Configured in n8n to connect to your Pinecone index.
• A Google Sheet: Containing the data you want your chatbot to answer questions about. The structure of this sheet will influence the "Default Data Loader" and subsequent nodes.

Setup/Usage

1. Import the Workflow: Download the workflow JSON and import it into your n8n instance.
2. Configure Credentials:
   • Set up your Google Sheets credential to connect to the Google Sheet containing your knowledge base.
   • Configure your Google Gemini API Key for the embedding and chat model nodes.
   • Set up your Pinecone credential to connect to your Pinecone index. Ensure you have an index created in Pinecone.
3. Update Google Sheets Node: Specify the Spreadsheet ID and Sheet Name in the "Google Sheets" node (Node ID: 18) to point to your knowledge base.
4. Activate the Workflow: Enable the workflow in n8n.
5. Initial Setup: Send a chat message containing the configured "setup" command (e.g., "setup") to trigger the data loading and vectorization process. This will populate your Pinecone index.
6. Start Chatting: Once the setup is complete, you can start sending chat messages to the workflow, and the AI agent will respond based on the loaded knowledge and conversational memory.