Safe YOLO Mode: Running LLM agents in vms with Libvirt and Virsh (opens in new tab)

(metachris.dev)

31 pointsmetachris1mo ago9 comments

9 comments

Whilst it is safer to run inside a Vm/container, it doesn't make it safe.

Yes, having your entire filesystem deleted is much less likely now (bonus points for zfs snapshots of the image for each operation) Your context is still vulnerable, as anything the VM has access too.

metachrisOP1mo ago

Good point! Running in isolation does reduce the amount of sensitive things an LLM has access to though, which typically can be quite a lot (SSH keys, Cloud credentials, communication tools, etc.)

dk89961mo ago

Interesting. Im looking for solution to run multiple OpenClaw bots in the cloud, with security and isolation in mind.

ews1mo ago

I ended up using guix shells (container mode) for my agents and sharing just the directories they need

nsonha1mo ago

The comparison between these and lima makes me think that it's AI generated

metachrisOP1mo ago

I've had LLMs assist me in putting together the comparison. I did edit it, and found the comparison good and wanted to keep it. The rest of the post is all hand written and thoroughly manually tested.

xyzsparetimexyz1mo ago

Oh my GOD just STOP with the awful kitschy sepia toned slop images for blog posts like these. I can't take it any more!!!

metachrisOP1mo ago

Gave your comment a bit more thought and I kind of agree, and removed the image! Thanks for your feedback.

nkko1mo ago

For no special reason, beside I could, I’ve slop coded this AI agents ephemeral VM orchestrator which I use inside any agent to manipulate and maintain my coding VMs on Proxmox. Probably it could make sense to simplify it further and move from Proxmox to something like this. Link: https://github.com/nibzard/agentlab

j / k navigate · click thread line to collapse

9 comments

KaiserPro1mo ago

Whilst it is safer to run inside a Vm/container, it doesn't make it safe.

metachrisOP1mo ago

Good point! Running in isolation does reduce the amount of sensitive things an LLM has access to though, which typically can be quite a lot (SSH keys, Cloud credentials, communication tools, etc.)

dk89961mo ago

Interesting. Im looking for solution to run multiple OpenClaw bots in the cloud, with security and isolation in mind.

ews1mo ago

I ended up using guix shells (container mode) for my agents and sharing just the directories they need

nsonha1mo ago

The comparison between these and lima makes me think that it's AI generated

metachrisOP1mo ago

xyzsparetimexyz1mo ago

Oh my GOD just STOP with the awful kitschy sepia toned slop images for blog posts like these. I can't take it any more!!!

metachrisOP1mo ago

Gave your comment a bit more thought and I kind of agree, and removed the image! Thanks for your feedback.

nkko1mo ago

j / k navigate · click thread line to collapse