Inline execution Quarto in Positron with ssh connection possible?

[smokyJ07]

Hi,
I started using Positron recently and was interested to use the new inline execution feature for Quarto files. I use Positron with ssh to a remote machine. After switching on inline execution it allows me to select a kernel specific to the Quarto file (similar to Jupyter notebooks) but then fails to load the kernel with an unknown error:
Failed to start kernel for 'test_r_env.qmd': Unknown error
Is this a known limitation on the current patch or should it work?
Thanks!

Positron Version: 2026.08.1 (user setup) build 2
Code - OSS Version: 1.124.0
Commit: 48dc4ce
Date: 2026-08-09T13:35:16-06:00
Electron: 42.2.0
Chromium: 148.0.7778.97
Node.js: 24.15.0
V8: 14.8.178.14-electron.0
OS: Windows_NT x64 10.0.26100

[jmcphers]

This should absolutely work. Can you attach a runtime startup diagnostics report after observing the failure? From the Command Palette, run Developer: Runtime Startup Diagnostics.

[smokyJ07]

Sure! Thanks for checking
Runtime Startup Diagnostics.txt

[juliasilge]

Thanks for the diagnostics! They show where the startup stops at least. Looks like the problem is not specific to Quarto. Your .qmd file starts an R session in notebook mode, and that session never completes its handshake with the kernel supervisor:

[session r-notebook-57c53c6a] Waiting for JEP 66 handshake
[session r-notebook-57c53c6a] Timeout waiting for handshake

The kernel process stays alive for the full 30 seconds and sends no output but it does not crash. Your R console session started correctly 27 seconds earlier, with the same interpreter, the same environment, and the same conda activate command.

One difference could maybe explain this, related to the working directory. A console session starts in the workspace root (/home/rauchhau) while a notebook session starts in the folder of the document (/home/rauchhau/scratch_rauchhau/visium). Your interpreter needs conda activate, so Positron starts the kernel through bash --login -i. Your full shell configuration then runs in that second folder. If a line in your profile or in .bashrc reads the current directory, and that directory is on a slow scratch filesystem, the shell can stop there but not in your home directory.

Can you try these four steps?

1. Set notebook.workingDirectory to ${workspaceFolder}. Then open the .qmd file again. If the kernel starts then, the directory is the cause.
2. Open a normal .ipynb notebook with R in the same folder. If it fails in the same way, the problem is not related to Quarto.
3. Set positron.r.kernel.logLevel to trace. Do the same operation again, then attach the "R Kernel" output channel. At the default level this log stays empty, so it does not show us the failure.
4. On the remote machine, compare the time of bash --login -i -c 'echo hi' in your home directory and in /home/rauchhau/scratch_rauchhau/visium.

Your report also shows two problems on our side about errors and retries. I'll get those logged!

[juliasilge]

I opened that issue about the retries and errors here: #15570