PiFlow-Lite is a lightweight, Python-native MLOps orchestration framework designed for Raspberry Pi clusters and other resource-constrained edge environments.
Inspired by Kubeflow and MLflow, PiFlow-Lite reimagines the machine-learning pipeline stack from the ground up — replacing heavy microservices with simple, asynchronous Python components. It enables users to define workflows in YAML, execute them across multiple Pis via SSH, and track experiments locally using SQLite, all without relying on Kubernetes or cloud infrastructure.
⚡ Think of it as “Kubeflow without the complexity” — built for edge AI, research labs, and makers who want full control over their ML lifecycle.
- 🧩 YAML-based pipeline definitions
- 🔄 Asynchronous workflow controller (pure Python)
- 📊 Lightweight experiment tracker using SQLite
- 🌐 Optional REST API (FastAPI) and Streamlit dashboard
- 🖧 Multi-node execution via SSH on Raspberry Pi clusters
PiFlow-Lite aims to democratize MLOps — bringing experiment tracking, workflow automation, and reproducibility to the edge computing world, where every watt and megabyte counts.