RAG Pipeline & Chatbot

Turn a knowledge base into grounded answers, with relevant context retrieved before every response.

n8nOpenAIPinecone
RAG Pipeline & Chatbot workflow in the n8n editor

The problem

Answers buried in scattered documentation are hard to find. A general-purpose chatbot may answer without reliable business context.

The solution

The workflow combines semantic retrieval with a language model, using source material to ground an answer rather than relying on model memory.

How it works

  1. 1

    Pick up new documents

    A Google Drive trigger fires when a file is added to the knowledge base folder, and the file is downloaded.

  2. 2

    Store them in Pinecone

    A data loader splits the document into passages, OpenAI embeddings turn each one into a vector, and Pinecone stores them for search.

  3. 3

    Receive a question

    A chat trigger starts the second flow whenever someone sends a message.

  4. 4

    Answer from the documents

    An AI agent on an OpenRouter chat model searches Pinecone for the matching passages and answers from them, not from memory.