Icon

Hub Documentation RAG

<p>This workflow builds and retrieves vector chunks from the documentation to provide additional context for answering complex user queries. It is used when the information available in the system message is not sufficient, when the user requests more detailed explanations or clarifications, or when additional context is required to support more in depth reasoning.</p>

Tool: Hub Documentation RAG

KNIME Documentation Vector Store creator

This workflow builds a vector store from KNIME documentation to support Retrieval Augmented Generation workflows. It processes the documentation, splits it into semantically meaningful chunks, computes embeddings, and stores them

Before executing this workflow, make sure that an embedding model is configured and available in the KNIME GenAI Gateway. The embedding model is required to convert documentation chunks into numerical vectors suitable for similarity search.

User Query

Join all chunks
GroupBy
String Manipulation
Count section length
String Manipulation
Hub Documentation
Row Filter
Chunk Size 2000
Text Chunker
Column Filter
FAISS Vector Store Creator
title
String Manipulation
At least 200 characters, this removes useless bodies (f.e. indexes)
Row Filter
Column Filter
Read KNIME Docas a table
Table Reader
Create embeddings
Text Embedder
Vector Search Hub Documentation
String Configuration
Retrieve documentation chunks
Vector Store Retriever
Tool Message Output
Use Current Hub to getHub Base URL
KNIME Hub Authenticator
String Manipulation
add title as reference
String Manipulation
Ungroup
Variable to Table Row
KNIME Hub Embedding Model Selector

Nodes

Extensions

Links