Hello, and welcome to my latest project

About a year ago I decided to make a tool to fill some gaps and see what I could do. The primary frustrations I was having centered around me being new to AI, and most tools being either chat focused or coding focused, but rarely both. Also, most things seemed to encourage using a paid model most of the time. It very much seemed like the "easy button" was, and sometimes still is, to go to x provider, download their tool, and pay for their API token usage.

Now, in fairness, as the year has gone on, lots of people have been working on this. Harnesses have gotten better. Web and desktop tools have gotten better. There are more people working on more flexibility, more local open-weight models, better tools, etc. All that being said... over the last year I embarked on a project inspired by a few people. [So I am going to credit them here]

Theo

NetworkChuck

ThePrimeagen and the team at TheStandupPod

The Big Project

So what did I do?

Well, I set out to make a thing that would help anyone starting their AI journey have an easier time learning, testing, and pushing capabilities, without being immediately intimidating.

The general idea is that as long as you have WSL with Ubuntu or Debian Linux, a Linux laptop or host, or a Linux VPS, you can install this using a single command that pulls an installer, asks some questions, and sets the whole thing up for you. It installs Docker (if you don't have it), installs the containers, sets them up with defaults (which only requires you to have an OpenAI-compatible API endpoint for the initial setup; from there you can adjust models to your heart's content), and gives you an endpoint you can access with a clickable link.

The other thing I didn't want to do was reinvent the wheel, so this leans on several tools that already exist: Hermes for the orchestration agent and assistant. Opencode as the built-in coder (and powering the worker agents). Git for version control of projects, Monaco for editing, a simple viewer, and diff checkers. Excalidraw for the spatial board (co-work whiteboard). A wiki for long-term "memory" and files and concepts you want the agent to keep and reference. And the normal tools/skills/memory, endpoint configuration, kanban board for agents, CLI and TUIs, and agent permission controls. Also a project server space and preview space to allow you to test and deploy projects locally before committing them.

It's a lot, I know.

So to help, I included a help guide, and you can always ask it what it can do. I also built in a free AI tracker that pulls from the ClawLabsAI Tracker on GitHub, so you can use paid models, free models, or local models.

It is not something I would call a proper project, as in when I update it, it'll be because something bothers me or I added a feature I wanted. But that's the other thing about this; it's something I do use. I wanted to solve a problem I was finding, and also provide something open source and at no cost to the community, since I have received so much help from everyone in the open-source community.

You can find it here: AI Workbench. And if you just want to grab the installer command, here it is, as well as several screenshots so you can see if it's something you might want to try.

curl -fsSLo install.sh https://forgejo.hybridcloudhq.com/jsvoros/spatial-workbench/raw/branch/main/deploy/install.sh && bash install.sh

Images to come and github repo instead of my forgejo, but for now...