Swift-first tooling • on-device & edge AI • pragmatic DevOps
We open-source the things we need every day: lightweight Swift services, self-hosted LLM wrappers, game-dev utilities, AI-powered engineering agents, and home-lab recipes that help indie developers run modern apps and ML workloads on their own hardware.
| Domain | Current Focus |
|---|---|
| Server-Side Swift | • Micro-services with Vapor & Hummingbird (e.g. 🌨️ SnowCover API powered by Sentinel data) • SRM – a Swift replacement for pm2 with auto-restart & startup helpers |
| Local & Edge AI | • Jetson-ready fork of llama.cpp (CUDA 10.2) • SwiftLlama – idiomatic Swift wrapper for llama.cpp • Trae Swift Agent – type-safe Swift port of Bytedance’s Trae Agent for AI-driven software-engineering automation |
| Mobile & Cross-Platform | • iOS sports-tracking, map solutions, all kinds of navigation & AR prototypes • VR prototypes with SwiftGodot on Apple Vision / Meta Quest • React-Native + Python back-ends for LLM micro-services |
| Game Tech | • Godot 4 + SwiftGodot R&D for 2D RPG/strategy games & terrain generation • Unique Unity as a Framework integration to embed Unity into any iOS view • Home-IoT experiments, including hardware setup and selection |
| Home-Lab / Dev-Ops | • Raspberry Pi / Jetson clusters for distributed LLM inference • NGINX, Docker, systemd & PM2 playbooks • Automated deployment scripts and Telegram bot monitors |
| Repo | What it does | Tech |
|---|---|---|
| Trae Swift Agent | Swift implementation of Bytedance’s Trae Agent with 100 % feature parity & compile-time safety | Swift |
| SwiftLlama | Thin, Swifty wrapper around llama.cpp for on-device inference | Swift · C++ |
| llama.cpp (Jetson fork) | CUDA 10.2-compatible build for Jetson Nano 4 GB | C++ |
| PythonKit (fork) | Smooth Swift ↔︎ Python interop for data-science glue code | Swift |
| SRM | CLI process-manager for Swift executables (pm2-style) | Swift |
More experiments land in the Repositories tab — drop us a ⭐ if something helps your workflow.
Got an idea, bug, or question about Swift micro-services, self-hosted LLMs, AI agents, or Godot tooling?
Open an issue or reach out to @Maxim-Lanskoy via LinkedIn or email.