Fly Labs
Ask FlyBot

Holos – QEMU/KVM with a compose-style YAML, GPUs and health checks

I got tired of libvirt XML and Vagrant's Ruby/reload dance for single-host VM stacks, so I built a compose-style runtime directly on QEMU/KVM.What's there: GPU passthrough as a first-class primitive (VFIO, OVMF, per-instance EFI vars), healthchecks that gate depends_on over SSH, socket-multicast L2 between VMs with no root and no bridge config, cloud-init wired through the YAML, Dockerfile support for provisioning.What it's not: Kubernetes. No clustering, no live migration, no control plane. Single host. Prototype, but I'm running it on real hardware. Curious what breaks for people.

Hacker NewsToolMedicine & HealthSource
0Sign in to voteCopy link

FL score

62

out of 100

Verdict

VALIDATE

medium confidence

Competition

No competitor data yet

Trend

No signal yet

A YAML-first VM orchestration tool for single-host GPU workloads that replaces libvirt XML complexity with Docker Compose-like syntax.

The pain

Engineers managing GPU-accelerated VMs on single hosts waste time wrestling with libvirt XML verbosity, Vagrant's reload cycles, and manual GPU passthrough configuration. Health checks don't integrate with dependency ordering. Setting up inter-VM networking requires root access and bridge manipulation.

The gap

No tool bridges the gap between Docker Compose's simplicity and QEMU/KVM's raw power for GPU workloads. Kubernetes is overkill for single-host ML dev, simulation, or medical imaging pipelines. Vagrant doesn't expose GPU primitives cleanly.

Build angle

Start by validating with 3-5 specific user segments: (1) ML researchers doing local GPU experiments before cloud, (2) medical imaging software teams needing reproducible multi-VM test environments, (3) game dev studios testing on multiple GPU configs. Build a tight feedback loop around GPU passthrough reliability and health check semantics. Don't chase clustering, own the single-host niche deeply first.

Strengths

  • Solves a real friction point (libvirt XML is genuinely painful)
  • GPU passthrough as first-class citizen is rare and valuable
  • Working prototype on real hardware reduces technical risk
  • YAML syntax lowers barrier vs. XML or Ruby DSLs
  • Health checks gating depends_on is a smart primitive missing elsewhere
  • Socket-multicast L2 networking without root is technically elegant

Risks

  • Market size unclear, how many people actually need this vs. using cloud VMs or containers?
  • GPU passthrough is hardware-dependent; support matrix could explode
  • No clustering means it dies at scale; limits TAM to single-host use cases
  • Competing against free libvirt + Vagrant + community knowledge
  • Health checks over SSH are fragile; what happens with network partitions?
  • Prototype status means stability and edge cases unknown; early adopters will hit bugs
  • Medical/health category claim seems forced, no evidence this solves healthcare problems specifically

Questions about this idea?

FlyBot reads the scoring and gives you a second opinion on “Holos – QEMU/KVM with a compose-style YAML, GPUs and health checks”.

Open FlyBot