just for info:
on my machine, i had no problems and able to install the extension today, although it was a bit slow on my low internet connection. running knime 5.1 on linux mint 21.2,
Hi @nemad, yes I’m on WIN 10 on a company PC (thus some limitations), but the registry field LongPathsEnabled is set to ‘1’.
I just (again) freshly installed/unzipped KNIME 5.1.0 (via zip-archive) and tried to install the Knime AI Extension (see above) as the only additional package. The result is, that after some seconds the installation process gets stuck at 65% when installing the package ‘org.knime.python.llm.channel.bin.win32.x86_64’ (see image). After some time (=minutes) the Error Box pops up showing the message below as the detail:
I am sorry that it was not the long paths.
A failed installation like this should leave an entry in the error log (you can open it via View->Error Log):
If you double-click on the entry that I selected, you should see the log output of the installation process .
Should the error log come up empty, you can also find the logs of the installation create_env.err and create_env.out in the plugin folder located under <knime-installation-dir>\plugins\org.knime.python.llm.channel.bin.win32.x86_64_5.1.0.202307290225\bin.
Hopefully, these logs will give us some more clues.
Hi Adrian,
attached please find the 'Message’ content copied from the Event Detail window - following your instruction.
As said: Fresh unzipped Knime_5.1.0 installation only installing the ‘Knime AI Extension’ as described above!
Thx
Erich
@yupliu Thanks for verifying that long path is not the issue. Could you please also provide the three logs as described by Adrian further above? The error log and also the content of create_env.err and create_env.out (see Adrians second post above for where to find these two files). We are trying our best but we need as much information as we can get here
@Erich_Gstrein Thanks for sharing! Do you have a proxy? Is there a way to double-check whether outside of your company network you have the same issue accessing pip?
@marijakil Thanks for confirming the issue. Could you please verify that you have long paths enabled as Adrian outlined in his first post above? If you confirm that it exists, could you please help us pinning this issue down by also providing create_env.err and create_env.out (see Adrians second post above for where to find these two files)?
We continue looking into this. But as we cannot reproduce it, we need your help guys!
I am not sure whether you did something to the installation package or not. After restarting my PC multiple times, I tried to install those plugins again and it worked this time, thanks!
Hi Steffen,
yes, in my company’s environment there is a proxy in place - however, this proxy was set in ‘preferences’! Furthermore, working directly with Python/Notebooks on my machine I do not have these pip issues!
And yes, on my private machine - with no proxy - I do not have these issues.
@Erich_Gstrein, @steffen_KNIME I think there are currently some problems with Python based nodes and Proxies. They are not always working. There seem to be this workaround of creating a local collection of necessary packages in a KNIME version without a proxy and then transferring this. I have not tried it.
The zipped update packages are of no use. It would be great if the Python based nodes can use the Proxies set. I think I already initiated a ticket (which does not have a number as of yet it seems).
Also for some time now KNIME seems to struggle with recognizing which extensions are already installed and which are not. I have mainly seen this on Apple Silicon but with 5.1 it starts to show on Windows machines too.
@rsalencar I get the same error message even when I set the “KNIME_PYTHON_PACKAGE_REPO_URL” variable and point it to the collected Python packages from another system. From the logfile I can see that the installation sees the local Python packages and then uses them but then stops at the end with some more error messages.
Is there some pattern like
“it does not work on os xyz if python version is abc or if certain dependency incompatibilites show up” or is it completely random?
AI nodes worked for me so I can’t reproduce errors, but I am not completely sure whether i tried them on latest 5.x version
br
@rsalencar I tried a few more things and was able to get it to work. Had to remove some external references from a pip config file. Details in the grey box in this workflow
