The workflow first populates a Pinecone index with vectors from a Bitcoin whitepaper.

Then, it waits for a manual chat message. When received, the chat message is turned into a vector and compared to the vectors in Pinecone.

The most similar vectors are retrieved and passed to OpenAI for generating a chat response.

Note: To use this template, you need to be on n8n version 1.19.4 or later.