2026.09.17 Franklin Ezeji – SchedNext: Building a Lightweight Agent Runtime for Cloud, Edge, and…

SchedNext is a lightweight agent runtime for cloud, edge, and embedded Linux. Learn how it schedules, executes, and exposes workload state without a database, always-on web API, or heavy orchestration stack. An "agent" here is a deployable workload, such as a shell script or native binary. The runtime handles execution; the workload owns its business logic. TL;DR - Trace the CLI/daemon architecture: Unix-socket IPC, cron scheduling, persistent state, runtime locking, and atomic configuration updates. - Inspect live execution state with StateLens, a FUSE-based filesystem inspired by Linux /proc and /sys. - Follow a scheduled snapshot workload to a remote ESP32-S3 camera and verify the result in a browser. - See experimental SchedNext OS boot in QEMU, with SSH access and the runtime built into a minimal Linux image. The discussion focuses on where a small execution layer fits beneath or outside systems such as Kubernetes and systemd, rather than replacing them everywhere. Tech stack Runtime and interfaces: SchedNext, StateLens, Go, Linux, FUSE, Unix sockets, JSON configuration, cron expressions. Examples and demos: shell scripts, Python, C++, ESP32-S3, ngrok, QEMU, SSH, OpenCV, Telegram. Chapters 00:00 Introduction to SchedNext 00:47 Automation on constrained edge devices 02:27 Agents as deployable workloads 03:18 Architecture: CLI, IPC, scheduler, and executor 05:22 Minimal dependencies, persistent state, and locking 06:09 StateLens: filesystem-native observability 07:37 ESP32-S3 snapshot demo overview 09:08 Starting the daemon and exploring the CLI 11:51 JSON configuration and workload control 14:18 Inspecting live execution state with StateLens 16:49 Camera snapshots in the browser 20:42 Design lessons: reliability and edge constraints 22:10 SchedNext OS: a minimal Linux environment 24:08 QEMU boot, SSH access, and runtime inspection 31:33 Roadmap and boundaries: cloud, edge, and device fleets 35:15 Runtime source and invitation to contribute 36:14 Q&A: AI workloads, OpenCV, and fall detection 39:47 Q&A: logging, connectivity, and workload boundaries 43:26 Q&A: alternatives and minimal Linux design Speaker Franklin Ezeji, Senior Software Engineer at Eesti Energia LinkedIn: https://www.linkedin.com/in/franklin-ezeji-a68068160/ SchedNext runtime source https://github.com/Ezeji/schednext Try the runtime on a Linux VM or…