Website content chatbot with Pinecone, Airtable & OpenAI for RAG applications

This n8n workflow shows how to extract website content, index it in Pinecone, and leverage Airtable to power a chat agent for customer Q&A.

Use cases include:

• Building a knowledge base from your website.
• Creating a chatbot that answers customer queries using your own site content.
• Powering RAG workflows for FAQs, support docs, or product knowledge.

How it works

• Workflow starts with a manual trigger or chat message.
• Website content is fetched via HTTP Request.
• The HTML body is extracted and converted into clean Markdown.
• Text is split into chunks (~500 chars with 50 overlap) using the Character Text Splitter.
• OpenAI embeddings are generated for each chunk.
• Content and embeddings are stored in Pinecone with namespace separation.
• A Chat Agent (powered by OpenAI or OpenRouter) retrieves answers from Pinecone and Airtable.
• Memory buffer allows multi-turn conversations.
• A billing tool (Airtable) provides dynamic billing-related answers when needed.

How to use

• Replace the sample website URL in the HTTP Request node with your own domain or content source.
• Update Normalize code based on markdown content output to remove noise.
• Adjust chunk size in the Text Splitter for your website markdown output.
  • In this example, the Character Text Splitter with separator ###### worked really well.
  • Always check the Markdown output to fine-tune your splitting logic.
• Update Pinecone namespace to match your project.
• Customize the Chat Agent system prompt to fit your brand voice and response rules.
• Connect to your own Airtable schema if you want live billing/payment data access.

Requirements

• OpenAI account (for embeddings + chat model).
• Pinecone account (vector DB for semantic search).
• Airtable account (if using the billing tool).
• (Optional) OpenRouter account (alternative chat model provider).
• n8n self-hosted or cloud.

Need Help?

Ask in the n8n Forum!

Happy Automating! 🚀