n8n Dynamic Website Assistant with DeepSeek AI & Pinecone Vectorstore

This n8n workflow creates a dynamic website assistant that leverages DeepSeek AI, Pinecone for vector storage, and PostgreSQL for chat memory. It's designed to provide intelligent, context-aware responses based on website content, with site-based routing for handling different website contexts.

What it does

This workflow automates the following steps:

1. Receives Webhook Request: It starts by listening for incoming HTTP requests, acting as the entry point for user queries.
2. Routes Requests Based on Site: A Switch node inspects the incoming request (presumably for a site parameter) to route the request to the appropriate AI processing path.
3. Initializes AI Agent: For each site, an AI Agent is initialized, configured with specific tools and models.
4. Configures DeepSeek AI Chat Model: The AI Agent utilizes the OpenRouter Chat Model to access DeepSeek AI for generating responses.
5. Sets up Cohere Embeddings: Embeddings Cohere is used to create vector embeddings of the input and potentially retrieved documents, enabling semantic search.
6. Connects to Pinecone Vector Store: The workflow integrates with a Pinecone Vector Store to retrieve relevant documents or context based on the user's query, ensuring responses are informed by specific website content.
7. Manages Chat History with PostgreSQL: Postgres Chat Memory is employed to maintain conversational context, allowing the AI to remember previous interactions within a session.
8. Responds to Webhook: After processing the query and generating a response, the workflow sends the AI's answer back to the original webhook caller.
9. Provides Documentation/Notes: Sticky notes are included in the workflow for documentation and to explain specific parts of the setup, such as "DeepSeek AI" and "Pinecone Vector Store".

Prerequisites/Requirements

To use this workflow, you will need:

• n8n Instance: A running instance of n8n.
• Webhook Endpoint: An external system or application to send HTTP requests to the n8n webhook.
• DeepSeek AI (via OpenRouter): An OpenRouter API key configured to access DeepSeek AI models.
• Cohere API Key: Credentials for Cohere Embeddings.
• Pinecone Account: An active Pinecone account with a configured index for vector storage.
• PostgreSQL Database: Access to a PostgreSQL database for storing chat memory.
• Langchain Nodes: Ensure the @n8n/n8n-nodes-langchain package is installed in your n8n instance.

Setup/Usage

1. Import the Workflow: Import the provided JSON into your n8n instance.
2. Configure Credentials:
   • Set up your OpenRouter credentials (for DeepSeek AI) in the OpenRouter Chat Model node.
   • Configure your Cohere credentials in the Embeddings Cohere node.
   • Provide your Pinecone API key and environment details in the Pinecone Vector Store node.
   • Set up your PostgreSQL database connection details in the Postgres Chat Memory node.
3. Activate the Webhook: Activate the Webhook trigger node to get its unique URL.
4. Send Requests: Send HTTP POST requests to the webhook URL with your user queries and a site parameter to route the request. The workflow expects the input to be structured in a way that the Switch node can interpret for routing.
5. Customize AI Agent: Adjust the AI Agent node's configuration (e.g., tools, prompt templates) to fit your specific website content and desired assistant behavior.
6. Populate Pinecone: Ensure your Pinecone index is populated with relevant website content, embedded using Cohere, to allow the AI to retrieve accurate information.