Dear KNIME Community,
If you are running KNIME Analytics Platform on macOS 15.4 and have installed any extension that relies on Python—whether or not you ever touch a Python Script node—you may encounter library‑loading errors. The issue affects the bundled Python environments shipped with KNIME and any standalone Conda environments.
Who is affected?
You are affected if your KNIME installation contains any of the following extensions:
- KNIME Python Integration
- KNIME AI Extension
- KNIME Giskard Extension
- KNIME Presidio Extension
- KNIME Web Interaction (Labs)
- KNIME Nodes for Google Ads (Labs)
- KNIME Timeseries (Labs)
- KNIME Nodes for Scikit‑Learn (sklearn) Algorithms
Typical symptoms include:
- The Python Script node failing immediately with the message “Execute failed: Could not connect to the Python process.”
- Missing nodes from the extensions above.
- KNIME log entries like:
ImportError: dlopen(/Applications/KNIME 5.4.2.app/Contents/Eclipse/bundling/envs/org_knime_pythonscripting/lib/python3.11/site-packages/numpy/core/_multiarray_umath.cpython-311-darwin.so, 0x0002):
Library not loaded: @rpath/libgfortran.5.dylib
Root Cause
Some Conda libraries—particularly libgfortran (used by numpy)—contain duplicate RPATH entries, causing them to fail to load on macOS 15.4. This is not limited to KNIME; it stems from a known problem in the numpy conda-forge feedstock (https://github.com/conda-forge/numpy-feedstock/issues/347), and a permanent fix is not yet available. In the meantime, we have prepared a script to remove duplicate RPATH entries and re-sign libraries as needed.
Temporary Workaround
If you do not urgently need to upgrade to macOS 15.4, we recommend waiting until a permanent fix is available. For those who have already updated—or must update—to macOS 15.4, please follow the steps below to modify your KNIME application’s bundled Python environments and any standalone Conda environments. This ensures these libraries can be loaded successfully.
Instructions
- Download the script: Obtain the attached file
fix_knime_python_envs.sh (5.7 KB). - Open a terminal: Launch the Terminal app on macOS. In the terminal, type
cd ~/Downloads(or navigate to the folder where you saved the script) to switch to that directory. - Make the script executable:
chmod +x fix_knime_python_envs.sh - Run the script: Provide the path to your KNIME installation or a specific Conda environment:
./fix_knime_python_envs.sh "/Applications/KNIME 5.4.3.app" ./fix_knime_python_envs.sh /path/to/my_conda_env
We will provide updated Python environments in bugfix releases of KNIME Analytics Platform once conda-forge and numpy offer a general fix for these duplicate RPATH entries. In the meantime, please apply this script if you encounter any Python library loading issues on macOS 15.4.
Thank you for your understanding and patience.
