KNIME Documentation

Add-on feature

Bring your own model is a paid add-on available for the Business and Enterprise plans. For all available add-ons, see Add-ons.

Bring your own model for K-AI ​

K-AI is the Knime AI Assistant. It helps users efficiently answer questions about Knime and build customized workflows, simplifying their data analysis tasks.

By default, K-AI uses a model provided by OpenAI with a default configuration. You can't change the K-AI configuration yourself as a global admin, but you can request changes by contacting your Knime account manager.

This guide describes the default model and the settings you can ask to change.

Which models K-AI uses ​

By default, K-AI uses the following models provided by OpenAI:

  • gpt-6.1-sol to generate responses
  • text-embedding-3-small to create embeddings

To learn how K-AI handles your data, see Privacy and security.

What you can customize ​

You can request changes to the following K-AI settings:

  • LLM backend
  • Welcome message and disclaimer
  • Workflows and components in K-AI answers

To request a change, contact your Knime account manager.

LLM backend ​

Connect K-AI to your own OpenAI or Azure OpenAI account. K-AI then sends requests to your provider instead of the Knime-managed models. You pay your provider directly for tokens, and Knime charges a reduced AI credit rate. For details, see Understanding AI credits.

To configure OpenAI as LLM backend:

  1. Create an OpenAI account.
  2. Generate an API key on the API keys page. It is recommended to create a fresh API key that is used exclusively by K-AI.
  3. Contact your Knime account manager to complete the setup.

To configure Azure OpenAI as LLM backend:

  1. Create an Azure account with access to Azure OpenAI.
  2. Create an Azure OpenAI resource to be used by K-AI. It is recommended to create a separate resource that is used exclusively by K-AI to avoid any interference with other applications.
  3. Deploy a model to generate responses and a model to create embeddings with API version 2024-08-01-preview or newer.
  4. Contact your Knime account manager to complete the setup.

If your model supports reasoning, you can also request to set a reasoning effort level.

Welcome message and disclaimer ​

K-AI conversations start with a default welcome message. You can request a custom message, or remove it.

The first time a user opens K-AI, or uses K-AI to generate code, K-AI shows a disclaimer. You can request a custom disclaimer.

Workflows and components in K-AI answers ​

K-AI can answer questions about the public workflows and components on your Knime instance. When it searches them, K-AI shares their descriptions with your LLM provider. This option is on by default. You can request to turn it off.

Access the K-AI chat history ​

Global admins can retrieve an audit record of the conversations users had with K-AI. The record doesn't include message content. Each user message and model response is replaced by a marker that shows its category and size.

To retrieve the record, send a GET request to https://api.<base-url>/chat-history. Include an Authorization header with a valid bearer token. The endpoint returns records in pages, and records from the same conversation appear together. Group them by conversation ID to reconstruct each conversation.