Created attachment 180284 [details] Time out error because "kde-applicationdata.mirrors.omnilance.com" domain is unavailable. SUMMARY When you create a fresh install, it looks like 8 (eight) "Artificial Intelligence Files" needed for 8.6.0. But when one of the primary host for is offline/unavailable, download process only gives "retry" or "cancel" options. In my case, digiKam tried to download the "efficientnetb7.onnx" and "classification_classes_ILSVRC2012.txt" files from "kde-applicationdata.mirrors.omnilance.com" domain. But that domain was unreachable as of the time i submitted this bug report. Thus downloads from that domain were failed and didn't try alternative servers for them. STEPS TO REPRODUCE 1. Create an environment for a fresh install 2. Try to download extra files for AI detection. If the domain is available when you are testing it, block "kde-applicationdata.mirrors.omnilance.com" manually. OBSERVED RESULT Downloads failed and didn't look for alternative download servers. I've added one of the failed downloads manually, restarted digiKam, recognised the file and skipped to the next file. EXPECTED RESULT digiKam should add and try alternative download servers. SOFTWARE/OS VERSIONS Windows: 10 Home v 19045.5737 ADDITIONAL INFORMATION See the screenshot from the process.
If you run "try again," an alternative server will be attempted. Otherwise, try again at a different time. As an open source program, we have no alternative download options; we rely on the KDE project's infrastructure. Maik
One more thing to note: random servers are generally chosen. However, there aren't many KDE backup servers available in all parts of the world. Maik
(In reply to Maik Qualmann from comment #2) > One more thing to note: random servers are generally chosen. However, there > aren't many KDE backup servers available in all parts of the world. > > Maik Thanks for the quick reply. I'll try later. One little recommendation for these 3rd party files: Hosting those extra files in a dedicated digiKam repo on huggingface.co or GitHub/GitLab may help in the future. They have enough redundancy.