We are happy to announce that KNIME Analytics Platform 5.4 is now available to download.
Here are the release’s highlights:
Build workflows faster and more efficiently with K-AI
Build workflows node by node with K-AI directly in the canvas
Understand and double check what K-AI did with the documentation it produces
Find and configure nodes faster with UI enhancements
Configure nodes quickly using a dialog in the side panel, enabled in preferences
Discover nodes by category in a Tree view in the node repository
Expand individual cells in the table view to see detailed cell content
Expanded LLM support, system messages to guide LLM tasks, and faster prompt processing
Get access to a wider range of LLMs including Databricks models and GPT4All models on IBM Granite and Microsoft Phi 3.5
Set system messages to provide context for LLM tasks and process more prompts in the same amount of time in the LLM prompter node
Improved GenAI governance with stronger evaluation capabilities and easier model maintenance
Check GenAI workflows for bias, hallucinations, and more with new Giskard nodes to create more reliable applications
Simplify maintenance of retrieval augmented generation (RAG) knowledge bases with a new node to migrate and update vector stores
Filter tables, handle variables, and reuse results in expressions
Reuse results from earlier expressions to perform related custom data manipulations quickly with the enhanced Expressions node
Use the new Expression Row Filter node to filter your table with custom expressions and the Variable Expression node for variable handling
Get AI assistance to manipulate data in all expression nodes
Simplify work with other tools such as Tableau, Box, and Databricks
Get easier access to Tableau data with a new Tableau Reader node
Access Box data with faster authentication
Connect with models and SQL warehouses in Databricks using new nodes
Explore many other major updates and find more details on the ones mentioned above on our blog, update your existing installation, or download the new version of KNIME Analytics Platform below.
KNIME Analytics Platform 5.4 must not be used as an executor on KNIME Server 4.17.x or before. KNIME Server 4.18.0 with full compatibility will be released soon.
Thanks to the KNIME team.
Really enjoying seeing the updates.
Especially, the multi-expression nodes (Expressions and Variable Expressions) and embedded configuration view will be a big help.
When copying/pasting nodes or collections of nodes, can the upstream connections be preserved? I have to do this a lot when creating custom analytics where I need to make tweaks to intermediate nodes but need to keep the same general structure, and looping will not work due to structural changes in the data.
When inside a Metanode, I am unable to drag from the in-port of the metanode start to create another node. It’s just inconsistent with other functions.
I suggest Adding the Node search popup to the “Right Click” menu. I add nodes more often than I Execute All, so it would be more useful in my opinion.
Is it possible to make the new UI dialogue boxes expand when I expand the dialogue box? These boxes expand in the older-style UI boxes.
Sorry to hear that you’re seeing this error. Have you tried closing all workflows in KNIME (so that it doesn’t load a workflow when it is started), restarting it, and then running the update?
The giskard extension is based on Python, and in some edge cases it can happen that KNIME cannot delete the old giskard plugin folder (because Windows slowly releases file locks that Python holds).
There are some fantastic things in this update…well done to all who have worked on it.
Can I ask for some clarification of how K-AI works when we use it to help guide workflows. The info says:
Any information you enter into the chat, as well as information about the workflow (being edited), may be shared with OpenAI and KNIME in order to provide and improve this service.
Would any actual data or metadata read back to the model or just information about the workflow construction?
My company has a one-way access to CoPilot and no access to ChatGpt, so I want to ensure that by using K-AI I’m not crossing any lines with our policy.
Currently, there are two points to keep in mind when it comes to what information K-AI and, consequently, OpenAI have access to (besides what you explicitly provide via messages to K-AI):
Table metadata: column names and types.
Expressions and code: our coding assistant variants of K-AI necessarily share the code with OpenAI.
We are working on expanding K-AI’s workflow-building capabilities, which is likely to involve sampling data in order for it to make more informed decisions, but that would most definitely be configurable and disabled by default.
Thanks for the update. I have feedback about the new release.
I had created an “Error Handling Guide” for my workflow users and when they get an error with the “node path” + “message” in the console, they can search it in the guide and find a solution. However after 5.4 release, they can’t see the node path but only the error message which is not helping as they have the same message for multiple issues. Please note that the workflow they’re using is triggering and executing another workflow which they don’t have permission to import and see it. So, they can go to the node that raising the error. I’m attaching the examples.
Old:
New:
I could hide warnings and see only errors in the console but now I don’t have that option and I see all warnings.
One feedback about the embedded node configuration mode:
On startup, can the embedded node configuration panel be minimized by default?
Then, is it possible to keep the little Settings button that will appear above the node if you hover on it available with the embedded node configuration mode?
Finally, I wish that the node config panel will auto expand if the Settings button is clicked or if the node is double clicked. Thanks!! I’m so excited about the updates and also have had great time using KNIME 5.4!
