---
layout: home
hero:
name: "gflow"
text: "Single-Node Job Scheduler"
tagline: "Run, queue, inspect, and control GPU or CPU jobs on one Linux machine without deploying a cluster"
actions:
- theme: brand
text: Quick Start
link: /getting-started/quick-start
- theme: alt
text: Installation
link: /getting-started/installation
- theme: alt
text: Command Reference
link: /reference/quick-reference
image:
src: /logo.svg
alt: gflow
features:
- icon: 🚀
title: Lightweight by Design
details: Bring queueing and scheduling to a single machine without deploying a full cluster stack.
- icon: 📋
title: Flexible Submission
details: Submit scripts or commands with GPU requests, priorities, time limits, Conda environments, and arrays.
- icon: 🔗
title: Dependencies and Arrays
details: Chain jobs together, fan out workloads, and build small research pipelines with familiar CLI commands.
- icon: 📊
title: Queue Visibility
details: Inspect active or completed jobs, filter by user or project, and switch between table, tree, JSON, CSV, or YAML output.
- icon: 🖥️
title: tmux-backed Execution
details: Every job runs in its own tmux session so you can attach, follow logs, and recover long-running work.
- icon: 🎛️
title: GPU-aware Control
details: Restrict visible GPUs, manage shared scheduling with VRAM limits, and apply runtime controls without editing scripts.
---