[GH-ISSUE #4180] account for common config/data locations when OLLAMA_MODELS is unset #49111

Closed
opened 2026-04-28 10:45:41 -05:00 by GiteaMirror · 1 comment
Owner

Originally created by @asmrtfm on GitHub (May 5, 2024).
Original GitHub issue: https://github.com/ollama/ollama/issues/4180

Many, probably most, projects out there which interface with ollama - such as open-webui and privateGPT end up setting the OLLAMA_MODELS variable thus saving models in an alternate location - usually within the users home directory.

This can become problematic considering the chronic instability of these projects where a single variance in one's environment can lead to edge cases that are undocumented.
I encountered one such issue and resolved to just use ollama itself in the terminal and not worry about a gui on top of it.
I pulled down like a hundred gigabytes of models with ollama pull <MODEL>.
I used them in the terminal for a while (ollama run <MODEL>)
After restarting my machine the models are no longer showing up.
I check the docs, they say they will be under /usr/share/ollama/.ollama/models/registry.ollama.ai/...

I freaked out, but then realized what had happened.
However, I thought it was worth pointing out that there is a long-standing convention for programs to look for configs/data/whatever in additional common locations, such as ~/.ollama

Just a thought. Not a big deal.

OS

Linux

GPU

Nvidia

CPU

Intel

Ollama version

0.1.33

Originally created by @asmrtfm on GitHub (May 5, 2024). Original GitHub issue: https://github.com/ollama/ollama/issues/4180 Many, probably most, projects out there which interface with ollama - such as open-webui and privateGPT end up setting the OLLAMA_MODELS variable thus saving models in an alternate location - usually within the users home directory. This can become problematic considering the chronic instability of these projects where a single variance in one's environment can lead to edge cases that are undocumented. I encountered one such issue and resolved to just use ollama itself in the terminal and not worry about a gui on top of it. I pulled down like a hundred gigabytes of models with `ollama pull <MODEL>`. I used them in the terminal for a while (`ollama run <MODEL>`) After restarting my machine the models are no longer showing up. I check the docs, they say they will be under /usr/share/ollama/.ollama/models/registry.ollama.ai/... I freaked out, but then realized what had happened. However, I thought it was worth pointing out that there is a long-standing convention for programs to look for configs/data/whatever in additional common locations, such as ~/.ollama Just a thought. Not a big deal. ### OS Linux ### GPU Nvidia ### CPU Intel ### Ollama version 0.1.33
GiteaMirror added the feature request label 2026-04-28 10:45:41 -05:00
Author
Owner

@pdevine commented on GitHub (May 14, 2024):

Given most users on Linux aren't using it on the desktop, we opted to install Ollama into a centralized place (unlike in Windows and MacOS), which is why it uses systemd by default. It is possible to build/run it yourself though on Linux and have it store everything in ~/.ollama.

It just seemed like most people wouldn't want to do that? I'll go ahead and close the issue, but feel free to keep commenting.

<!-- gh-comment-id:2110972903 --> @pdevine commented on GitHub (May 14, 2024): Given _most_ users on Linux aren't using it on the desktop, we opted to install Ollama into a centralized place (unlike in Windows and MacOS), which is why it uses `systemd` by default. It is possible to build/run it yourself though on Linux and have it store everything in `~/.ollama`. It just seemed like most people wouldn't want to do that? I'll go ahead and close the issue, but feel free to keep commenting.
Sign in to join this conversation.
1 Participants
Notifications
Due Date
No due date set.
Dependencies

No dependencies set.

Reference: github-starred/ollama#49111