Ecosystem WebUI Updates

EDIT
Repo is here: https://github.com/UselessToys/Ecosystem_WebUI

Hey all just a short update here nothing huge, just letting you know on a lot of the upcoming structure changes to the ecosystem.

So first off the FUTURE of this is always in the hanging balance, you guys know that we suffer from on/off extreme burnout from endeavours in AI. This tool is a LOT beyond our usual comprehension, and we're extremely proud but at the same time - extremely unproud. It's less the agentic assistance, and more the "it's just a wrapper" thing that my brain loves to remind me about.

For now though, until we can get this into a little more of a stable state, it's of course going to be there for the long haul. (Unless someone wants to take it off our hands, and go nuts and multiply?)

Right! Changes!

Generation UI:

We now vendor ComfyUI with pre-determined templates based on Guy90s Anima workflow, and my own SDXL version that was a fork of his. This is a little stagnant because trying to get an agentic tool to understand what is "NICHE" and what is "USED" is like pulling teeth. It's working otherwise, especially for ANIMA, and overall it's a good start. Despite my current diving opinions on generative AI, i'm extremely big on accessible interfaces and I hate Gradio.

Merge Tool:

We swapped from a pure Supermerger clone to using Chatti... Chattiori? The tool is EXTREMELY basic still because we're not exposing even REMOTLEY everything that it does. Works really well with the basic merging logic, the lora bake works especially well for ANIMA.

Trainer Tool:

SD-scripts with Derrian Distro and custom optimizers, things are a little twitchy at times. We're vendoring extremely new forked code from the community, so if something's wonky it might be a patch we pulled.

Tagging System:

Customized tagging, the downside is Blip and GIT might not be working entirely, but i'm not sure.

Installation & Starting

Git Clone via: https://github.com/UselessToys/Ecosystem_WebUI

I tried to add the current release, but it's 42mb not 30mb LOL.

Any questions just ask!

Train, tag, and generate — a full LoRA & checkpoint training suite with ComfyUI built in.

A web interface built on Kohya SS. Prepare datasets, auto-tag, train across many model types (SDXL, Flux, SD3, Lumina, Anima…), then generate and test results in a bundled ComfyUI workspace.

Requirements

Hardware

GPU: NVIDIA with CUDA 12.1+
Disk: 50 GB+ free space

GPUStatusNVIDIA (CUDA 12.1+)Fully supportedAMD (ROCm)Community supported — see AMD ROCm docsNVIDIA via ZLUDACommunity supported — see ZLUDACPU-onlyNot recommended for training; tagging/captioning will work

While baseline VRAM requirements can be high, the community has figured out plenty of workarounds to train on constrained hardware.

Software

Python: 3.10, 3.11, or 3.12
Node.js: 20.19+ (22.x recommended)
Git

Full platform requirements are in documentation/installation/INSTALLATION.md.

Installation

Windows Users: Install to a path you own (e.g. C:\Users\YourName\Projects\). Restricted system directories (C:\, Program Files, OneDrive folders) will cause permission errors.

Windows:

git clone https://github.com/Ktiseos-Nyx/Ktiseos-Nyx-Trainer.git
cd Ktiseos-Nyx-Trainer
install.bat

Linux:

git clone https://github.com/Ktiseos-Nyx/Ktiseos-Nyx-Trainer.git
cd Ktiseos-Nyx-Trainer
./install.sh

Both scripts will prompt about creating a virtual environment. Pass --venv or --no-venv to skip the prompt; --auto for fully unattended install.

ComfyUI (image generation) installs by default — including on cloud. To skip it, pass --no-comfyui (e.g. install.bat --no-comfyui, ./install.sh --no-comfyui, or python installer.py --no-comfyui on a remote instance).

Cloud: Use the VastAI or RunPod deploy buttons. Both auto-configure on launch.

Starting the App

# Default: frontend on 3000, backend on 8000
start_services_local.bat    # Windows
./start_services_local.sh   # Linux

# Custom ports
start_services_local.bat --port 4000 --backend-port 9000
./start_services_local.sh  --port 4000 --backend-port 9000

# Quick restart without reinstall
restart.bat     # Windows
./restart.sh    # Linux

# Restart with Reinstall + Restart supervisory process (Remote only)

bash fetch-restart.sh
./fetch-restart.sh

URLs:

New Features:

Crop tool

UNDER HEAVY CONSTRUCTION

Inspired by BirME, isn't working currently but we're going to TRY and figure it out - the damn LLM kept telling me a DIALOG POP UP WAS REQUIRED )(it wasn't) and I dont know how to do Node by hand, and I'm running low on funds so I can't re-sub to Claude to fix it.

CONVERSION TOOL

Works 100% and includes whatever SHARP supports for file types - Windows users are gonna like this because Pinterest defaults to JFIF.

Dataset Tools Integration

Dataset tools's NodeJS is integrated, it's a LITTLE JANKY if you're using remote, but single image drop from local totally works.

Future Paths

Trainer

Future of this is up in the air because i'm struggling to understand a lot of the backend details, and i'm a solo + LLM developer for this, and I'm becoming far overwhelmed with it. Not only that but I am still highly burned out from previous issues in AI, and there may come a time that when it's more stable that it'll be "FORKABLE" but archived.

Training Models

Same issue sort of as above: Due to not being employed, and wanting to pursue a lot more things that CAN help me gain employment, I've not really been as interested in training more content as often. I have been toying with things because of the trainer UI/ecosystem, and it's neat sure! But uh, yea I've really shifted in how things work, because I'm unable to really pick up commissions for training as nobody really wants to pay for things that aren't NSFW.

Commissions

We'll leave our commissions open for the time being on this, but there is still a potent warning that we may eventually cease activities finally. The cost outweighs the "vibe chase", and we're incredibly grateful for the help everyone's given us AND the referral tickets have afforded us the ability to likely HOPEFULLY take out money we'd deposited MONTHS ago in the buzz system..

But that'll only likely amount to less than 60 bucks, and most of the "buzz" we're 'earning' is blue or otherwise. Which is valid, but we rarely generate images on site at this stage, and we really REALLY appreciate people enjoy our content, but with the economical situation we're in - I'm not employed and I'm unable to keep paying into doing things long term.

NSFW?

Not only that but of course, y'all REALLY ONLY like the NSFW loras we've trained, and we've tried to stipulate several times that we don't WANT to train NSFW or XXX content, but people paid into this before reading rules and skirted around it. This makes us incredibly uncomfortable, and we can't get around refunding anyone because it costs us twice the amount after fees etc.

Mental Health & Family

As always a lot of this is due to mental health AND other things going on, we've been dealing with our stepdad recovering from his stroke last year, and his hand surgery this year. My mother directly has COPD and severe Rheumatoid + Psioratic Arthritis plus Osteo, and currently is losing a lot of weight due to unforseen circumstances (She doesnt' care cause it's weight loss but pfft). I'm also responsible for long term duties with our new floof Baxter, and i'm walking him twice a day and babysitting him when my parents are away. We're also dealing with supporting our partners and their family has had medical issue after medical issue..

At this stage we're just incredibly exhausted, and never expected to get this far but that's another thing entirely.

We don't know when our "EXPIRATION" date literally is this time around, but we're sort of feeling it out to see what our future holds.

Edit: Likely early august is our finality: We've decided that whatever holds y'all can reupload we don't care AND mind anymore -- it's sort of an ethos of open source, we know Civitai can't do the transfers and we dont want to force OUR name on nayhting so when we delete and you scrape through our HF repos- DO NOT credit us, we're trying to detach and go back to what we did before AI.

Ecosystem WebUI Updates + Installation Tutorial