Stopping point
I am pushing forward with my systems programming and distributed systems journey for agent infrastructure with another project. I recently finished task-orchestrator. That project is a container orchestration platform that uses manager and worker nodes. I built a CLI to maintain the platform and run the commands to send tasks, add a managers, and add workers, and the other actions needed to use the platform.
Side note: I have recently watched several Google Borg Youtube videos by engineers on the project because I am fascinated about the scale Borg operates at. My mind explodes thinking of how much software is ran by Borg daily.
Next Project: Quickspin
The next quest I will take on in my journey to become great at agent infrastructure is a project to build my first agent sandbox platform. It is called Quickspin. The pitch for this platform is "Quickspin is a platform that allows agents to dynamically create sandboxes to work in. Your agents can use Quickspin's Python and Typescript SDKs to interact with their sandboxes. Quickspin can pause and snapshot sandboxes to give your agents more flexibility when they do their work."
Quickspin will be written in Go, use Lima to setup Linux VM development using a Mac, and it will use Docker containers as its first sandbox environment. The platform will have both TypeScript and Python SDKs to help agents create and delete sandboxes. The final phase will deploy Quickspin onto cloud hosts and support Firecracker Microvm for better isolation.
Current Progress
So far I completed the first step of:
- installing Lima
- creating a Lima template that sets up Docker o n Ubuntu LTS vm
- Forward the internal docker socket to a local socket on my host machine
- configured my Makefile to create lima vm and create a docker context against the forwarded docker socket so that my hosts docker cli works against the vm's Docker
- create linux amd64 binary and validate it runs in my vm