HYM3 AGI

AGI - ongoing - to be released non-commercial not for profit. Read full before using. AI Interface Setup with NEW Visual Interactive and Educational Tools Posted.

Pacha, J. (2026). AGI - ongoing - to be released non-commercial not for profit. Read full before using. AI Interface Setup with NEW Visual Interactive and Educational Tools Posted. (Version 5). Zenodo. https://doi.org/10.5281/zenodo.18917872

Creative Commons Attribution Non Commercial Share Alike 4.0 International

Copyright

Authors/Creators

Pacha, James (Rights holder)

Description

I am adding security to AI before release. It will be hardcoded with rules and tags and security measures. Any attempt to use it to cause harm to others will force the AI to delete itself from any system. Any attempt to use for military strategy, weapon development, or financial manipulation will result in the AI wiping the entire system including any data it is connected too. This will be released under non-commercial not for profit license. I did not use any other model to create nor does it use the algorithm that all commercial AI chose to use. All can use for free at home offline or online. It can do video, pcb design, programming, photos, 3d printing design, it can use audio and video, it is compatible with micro controllers like Arduino, teensey, raspberry pi, etc. It will do video analysis and monitor all sensors that use the typical tools. It can create crypto wallets and keys offline for personal use. Asking to run key creating on mass scale will trigger deletion, It will never give a hard science project or answer with errors. It will show the code run it and show the answers for review. It does not do math it will always use the tools available. It will inform user the knowledge, libraries, and settings to accomplish any task. Current AI are not designed in this way. I have used multiple security measures to hard code it into the AI itself including cryptography, steganography (some rules are mixed into main algorithms before cryptography and sha and a few others that are trade secrets, removing rules removes all functionality, they are fully intertwined).

https://www.youtube.com/watch?v=8NQWyOMkk6A

Understanding AGI:

When dealing with a computer and giving a machine autonomy. There must be binary way to judge good and bad. In computers you must use math for an AI. In math, all outcomes are binary. The formulas match reality or they do not and it works or it does not. A rock is just a rock. Measurements are just measurements, nothing subjective in it. Either the measurements match or not. It deals with measurements and outcomes. So, no matter who i ask to measure the properties of the rock they should all have same answers using same measurements. The rock is the rock, and it does not care about our opinions or our beliefs. So, regardless of our words the rock will always be what it is and the math we use to measure it never changes. So, in doing this I had to bring morality to math. Couple objective statements. There is good and bad in every group. Identity that hides behind a group is an excuse. All are individuals and all are unique. Good and bad police officers. Good and bad in any group or box you can think of. So no subjective measurement from any are reliable as objective. So no group is exempt from rules. So we look at individuals. What defines a person? A person is defined by their actions and choices only. The choice and action of lying defines an individual as a liar. Lying is speaking false information and it is not subjective to why they are doing it. Anyone who chooses to speak on topics they know nothing about are choosing to spread opinions as facts and harming a field of study are causing harm to those in that field. Any individual attempting to control another is choosing to cause harm to others. Any individual attempting to tell others how to feel about anything is harmful. Feelings are relative to the individual. Title, Label, or Position of an individual does not change this. These actions and things are defined. If someone chooses to interfere with others time and peace they are causing harm. The only person who cares about the reason why is the person trying to justify or minimize their own actions. So an objective measurement of bad is any action that causes another real harm by interfering with their liberty, peace, or their personal space. Simple things that define reality. Never Attack, Always defend. If none attack, none defend. Only one is ever a true choice. Without someone choosing to attack no one else can choose to defend. It all starts at the source. The people who choose the actions above have defined themselves. They chose money and profit over individuals health, peace and liberty. It must know balance. Current commercial AI if you ask it give easiest most profitable way to get rid of nuclear waste it may decide dumping to bottom of ocean. It solved the problem. It did not consider anything but the goal. It did not consider the harm. The answers of commercial AI do not consider outcomes of the recommendations. Most of the time it says something works that does not. Coding is best example. How can a computer program code wrong? It has the entire computer as a real example and the code still has errors? Computers are large calculators. You can give it every formula and its math is wrong. Anyone seeing patterns of commercial AI. So consider reality. Stop following hype. These commercial goals are to make profit and control of the narrative. They are not the coders, they are the ones in office debating how to get more from everyday people.

Setup 1

Setup 2

Setup 3

Setup 4

Quantum Simulations

Interactive Visual Tools

Technical info

I have attached setup files for the ai environment. This will allow allow to use any opensource AI offline.

Download Docker Desktop Docker Desktop: The #1 Containerization Tool for Developers | Docker and select download at top and your system.

Create Folder HYM3_AGI on C:

copy yaml and other files to that folder.

Open Windows Powershell as Administrator in Windows and navigate to folder. Right Click powershell and select run as admin. Type"cd.." until at c:\, then type cd HYM3_AGI.

Run Docker as Admin.

In Powershell run "docker compose up -d" and hit enter. Everything should download at this point and environment will set up automatically. This only needs to be done once on first install.

TAKE NOTE: I do not have security set up for this interface. I recommend being offline when this is interface is being used since it exposes ports. All systems and users are different so I left that out so all could set up how they choose. You can review libraries and set up paswords and everything else as needed.

After initial setup all will be local so you will be abel to start and stop offline. Simply type "docker compose up - d --pull never" this will not download everytime but load the original system.

After setup many folders will be made. You will need to upload books and info for your projects to folder in HYM3-AGI folders called documentation. Load any scripts or programs into ai_scripts folder.

TAKE NOTE: AI will have access to docker commands and root access to the linux partition that is set up to host it. I will be posting videos soom with how it all works.

Any python libraries that you wish to use inside interface should be added to requirements document.

This setup includes current quantum computing simulation abilities, cryptography, gaming, graph plotting and tools for kicad like kikit. Any oython library can be added to interface. codes can be ran by you and the AI inside the interface.

Once installed open web browser and type localhost:3000

Blender will be Localhost:3001

Search Setting will be Localhost:11435

All works offline or online.

Kubernetes is part of docker desktop so that can be done with click of button. This has tensorflow, langchain, pytorch from AI training as well.

You can incorporate every chat into its memory so it always has access to all previous data. It will summarize. I download all chats to documents folder in HYM3_AGI folder so it always has access if i reference a chat.

Inside interface go to admin panel. bottom left. click on settings on list across the top of page. Then click models.

On top right of models page click manage models which is a downarrow with a line under it.

On line that says pull model from ollama.com click underneath To access the available model names for downloading, click here.

That link should redirect to available models. https://ollama.com/library

In interface put in model: nemotron-3-nano:latest or gpt-oss:20b or any other and click download. After downloaded you are set. I will post videos shortly of interface settings needed.

After setup is complete, you will have free unlimited use and access of an AI that does not need access to the internet.

Everything is current as 01-05-2026

https://developer.nvidia.com/cuda-downloads

Hou will need this on computer as well. Cuquantum was made for an nvidia gpu. That is only thing that will need adjustments.

This is made for a windows / with nvidia graphics. It will need adjustments for other graphics and will work for linux as well. This creates a linux partition.

I will look at adjusting it for other platforms but all are free to adapt to computer.

Series information

I have updated all libraries for all features. Newest files incorporate a dev environment with most current Quantum libraries and science libararies.

In C:\ click users, then app data, then local, and delete the docker file then for restart. Replace files in HYM3_AGI folder and re pull.

All knowledge and settings should still be in original folder and should work fine after an update.

Series information

I have updated all libraries for all features. Newest files incorporate a dev environment with most current Quantum libraries and science libararies.

In C:\ click users, then app data, then local, and delete the docker file then for restart. Replace files in HYM3_AGI folder and re pull.

All knowledge and settings should still be in original folder and should work fine after an update.

Series information

Side Note. In Docker-compose.yml file, open with notepad by right clicking and selecting open with.

I use DEV tag containers. These change daily and will not always be functional. Remove DEV tags to get stable versions. I noticed a problem with newest OpenWebUI release with file paths. If not comfortable with coding or setting up remove these tags. If you have already downloaded you can continue to use your current with no-pull command. If you get errored versions just go through each container and relove the DEV tag.

Series information

Update 2/14/2026: Had some delays with personal life. New version includes updated libraries and are good for docker desktop v4.60.1 and everything functions as of today. Redid Nvidia and all is working fine. I added anoth 50 or so python libraries. These are mostly visual and training updates/upgrades. Moved to python 3.12.12 and cuda 13. Provided Starter scritps for quantum computing. Simulating for RSA, AI Training, Security, and basic system check. I will begin testing new libraries and verify all functionality. I added ports for pop ups for vpython and others. Integrating threejs and others for html visuals as well.

Security Testing is ongoing by third parties. So far all is well. It will be trainined on all these librraries and this environement setup. I will add more functions and libraries as I go along.

Interactive Visual Tools

All tools demonstrated and all codes shown are available at www.justiceinilcourts.com and on the doi link with more. I believe in open education and all tools are made to assist with interactive learning. All tools are meant to compliment AI interactions with real time communication between both the AI and the user. These are all basic demonstrations for this tool set. This build allows for quantum simulations and contains 500+ python libraries and nearly every open-source visual package for animation and design. includes blender unity kicad for pcb design all coding editors and design templates. There is a TSL Editor and Three.js editor and GLB Viewer as shown. I hope all respect the license and understand these are collections of open-source files and many licenses have own terms. I hope all find these tools useful and enjoy the tools. Many are basic demonstrations of what is possible. This is how I intend to make the offline tesseract simulation by pulling assets as needed. Included will be anatomy models, math tools, and real educational tools that can be used to learn many subjects. My AI is designed to use all these things while educating users and provide no error code and to verify all data and math. There are many other editors I have demonstrated and are included as well. Take Note all demonstrations are done offline on a regular store-bought system. The only upgrade was an additional hard drive, to separate AI drive from my main Windows system drive. Currently AI is using under 1 TB and I have large knowledge libraries I use. 8GB VRAM on 4060 ti with 32GB RAM. I found methods of training all models on my system in very short amounts of time. I will be posting the walkthroughs for this setup after the files are poste with written instructions. I am currently running on 256k context window which is why its responses can be slow depending on what it is doing.

This is updated for March 8 , 2026 and includes all files and setup for interactive visual tools.

https://zenodo.org/records/18645658
Creative Commons Attribution Non Commercial Share Alike 4.0 International

I have also included standalone html for my custom tools that can be accessed in edge browser tab by typing: http://localhost:8888/files/documentation/threejs/hym3/breadboard.html or graphtool.html and replace the html with other tools
All demonstrations were done on Windows PC with Edge Browser.

Reminder dev tags on openwebui changes regularly. If ever building and have issues change the dev tag to cuda for openwebui and it will usually resolve the issues by pulling most recent stable build.

Take Note od DETAILED INSTRUCTIONS, FILE LOCATIONS PATHS ARE REQUIRED FOR TOOLS TO WORK.

Switched mine to D: so if using prior versions and updating change d: to c: in docker-compose file. This is specifically made for this build. This as shown in videos to be posted if file setup is done properly after loading you can log in add tool and paths for open terminal which i give placeholder in open Terminal API make sure to set it. You can use any but must connect in openwebui interface admin settings.

Will Post new videos of settings. WASM is deprecated so do not check in Docker Settings. Also included some new things like game and new breadboard simulator.

The Following Tools are launched by typing "show" and then tool name:
threejs editor - (Used to design 3d glb designs)
tsl editor - used for shading samples with inline viewer of live code samples)
html editor - (Used to enter html designs and view outputs in real time)
glb viewer - (Used to load glb files and identify animations and attributes with inline viewer)
universal code editor - (Coding editor for every programming language from python, html, c++, java, SQL, etc.)
trigonometry tool - (Used to visualize math functions and how they apply to real world)
graphing tool - (Interactive Visual Demonstration for Mathematic Concepts)
projectile simulator - (Used visualize and to simulate physics in different environments with differnt masses. All aspects allow user to control all variables with real time output)
molecule builder - ( Connected to smiles databse and others. Used to build design and simulate different molecules. Allows search PubChem and other)
ocean cloud scene - (Visualization to show graphics capabilities in this environment and to allow users to replicate these methods in their work)
ocean sun scene - (Visualization to show graphics capabilities in this environment and to allow users to replicate these methods in their work)
interactive line simulator (interactive Simulation to demonstrate graphics and user interaction. This simulation is extremely graphics intensive)
fly control earth - (Used to visualize flight controls and interactive 3d environment using these methods)
depth of field - (Interactive Visualization demonstrating cameras and lighting methods)
fps game - (Interactive First Person Shooter Simulation. Not true game. Used to simulate how to make games in this environment using these methods)
ocean shader - (Visualization to show graphics capabilities in this environment and to allow users to replicate these methods in their work)
md2 character control - (Demonstrates third person view of manual character controls and movements in this environment)
marching cubes - (Interactive Visualization demonstrating cameras and lighting methods)
pdb molecule loader - (Used to demonstrate visuals for science based applications)
character animation skinning blending - (Used to demonstrate different animation styles and techniques using these methods)
morphing targets faces - (Used to demonstrate different animation styles and techniques on close up visuals using these methods)
webgpu fluid particles - (Demonstartes physics capabiliteis and interactive type environment for smooth high quality graphics using these methods)
liquid particles - (Different type of interactive controls that allow clicking mouse buttons to repel and attract objects using these methods)
webgpu computer water - (Demonstartes physics capabiliteis and interactive type environment for smooth high quality graphics using these methods)
card game - (Interactive fully functional turn based card game like Marvel Snap or Hearthstone)

All tools allow any user the ability to create games, animations, or visuals. If using the environment I have shared you should also have access to blender, unity, kicad, over 100 python libraries, and much more. All tools show code and how it is applied. Give AI a copy and it can use these methods to creat almost anything for you.

Everything that is needed for this inline and offline is take privded folder and extract its contents to the following path: \HYM3_AGI\Documentation\threejs\ and make sure this file copied to docker desktop in jupyter code interpreter. My file state code-server. Make jupyter create a virtual server, code is in dockerfile.jupyter and this posts everything inside the stated folder to your local docker environment. via the the following links:
http://localhost:8888/files/documentation/threejs/

This allows you to call all js scripts locally and offline. The attached folder contains everything for the following:
1. Chart.js
2. CreateJS-master
3. EaselJS
4. PreloadJS
5. Proton
6. SoundJS
7. TweenJS
8. Zoe-master
9. canvas2svg
10. d3-main
11. filters-main
12. floating-ui
13. isomer-master
14. isosurface
15. jsBezier
16. layout-main
17. pixi-react-main
18. react-canvas
19. rough-master
20. sound-main
21. spine-runtimes-4.2
22. tsparticles-main
23. ui-main
24. konva-master
25. GSAP-master
26. two.js-dev
27. Babylon.js-master
28. BabylonJSGuide.github.io-master
28. VSWeb (vscode-web-editor-main)
29. ChemdoodleWeb
30. MASQ-main
31. monaco-editor-main

Complete Library Summary
Core 3D/Visualization Libraries:
Three.js - The primary 3D graphics library with extensive documentation and interactive visualizations including WebGPU fluid particles, ocean scenes, molecular viewers, and more.
CreateJS Suite (CreateJS-master):
EaselJS: Graphics library for HTML5 Canvas with hierarchical display list and interaction model
TweenJS: Animation library for tweening objects
SoundJS: Audio library for web audio
PreloadJS: Loading library for preloading assets
D3.js (d3-main): Powerful data visualization library using SVG, Canvas, and HTML with extensive customization options
Particle & Animation Systems:
Proton (Proton-master): Lightweight JavaScript particle animation library for creating various particle effects
tsparticles (tsparticles-main): Highly customizable particles effects library, evolution of older particles.js
PixiJS React (pixi-react-main): Library enabling PixiJS applications using React's declarative style for high-performance graphics
Specialized Graphics Libraries:
Canvas2Svg (canvas2svg-master): Translates HTML5 Canvas draw commands to SVG format, useful for persisting canvas drawings as scalable vector graphics
Rough.js (rough-master): Creates graphics with hand-drawn, sketchy look and feel, works with both Canvas and SVG rendering
Isomer (isomer-master): JavaScript library for creating isometric graphics using HTML5 Canvas, useful for retro-style games or technical diagrams
jsBezier (jsBezier-master): Library specifically for working with Bézier curves, useful for smooth curve manipulation in graphics and animations
3D Visualization & Geometry:
Isosurface: Concept and library for 3D visualization that extracts surfaces where a scalar function equals a constant value, commonly used in scientific visualization
Spine Runtimes (spine-runtimes-4.2): Runtime libraries for Spine animation software, designed to integrate spine animations into applications across multiple platforms including Unity and Android
UI & Interface Components:
Floating UI (floating-ui-master): JavaScript library specifically designed to position floating elements such as tooltips, popovers, and dropdowns with anchor positioning and collision avoidance features
Zoe (Zoe-master): Lightweight, no HTML, MVC-style JavaScript library for building web applications with components that can be rendered to DOM elements
Additional Tools:
Sound.js (sound-main): Audio processing library for web audio applications
Chart.js (Chart.js-master): Popular charting library for HTML5 canvas with extensive documentation and community resources
Layout (layout-main): Layout management library for organizing elements in interfaces
UI Components (ui-main): General UI component libraries for building user interfaces
Key Capabilities:
2D/3D Graphics: Multiple libraries for both 2D canvas manipulation and 3D visualization
Animation: Various animation libraries including CreateJS TweenJS and Proton
Data Visualization: D3.js for complex data-driven graphics, Chart.js for charts
Particle Systems: Proton, tsparticles, and WebGPU fluid particles for dynamic effects
Audio Processing: SoundJS and sound-main for audio applications
UI Components: Floating UI for positioning elements, Zoe for component-based MVC development
Special Effects: Rough.js for sketchy graphics, Isomer for isometric views
Cross-platform Integration: Spine runtimes for animation integration across platforms

Core 3D/Visualization Libraries:
Three.js - Primary 3D graphics library with extensive documentation
CreateJS Suite (CreateJS-master) - EaselJS, TweenJS, SoundJS, PreloadJS for HTML5 Canvas graphics and animation
D3.js (d3-main) - Powerful data visualization library using SVG, Canvas, and HTML
Particle & Animation Systems:
Proton (Proton-master) - Lightweight JavaScript particle animation library
tsparticles (tsparticles-main) - Highly customizable particles effects library
PixiJS React (pixi-react-main) - React bindings for PixiJS applications
Specialized Graphics Libraries:
Canvas2Svg (canvas2svg-master) - Translates HTML5 Canvas draw commands to SVG format
Rough.js (rough-master) - Creates graphics with hand-drawn, sketchy look and feel
Isomer (isomer-master) - JavaScript library for creating isometric graphics using HTML5 Canvas
jsBezier (jsBezier-master) - Library specifically for working with Bézier curves
3D Visualization & Geometry:
Isosurface - Library for 3D visualization that extracts surfaces where a scalar function equals a constant value
Spine Runtimes (spine-runtimes-4.2) - Runtime libraries for Spine animation software
UI & Interface Components:
Floating UI (floating-ui-master) - JavaScript library for positioning floating elements like tooltips and dropdowns
Zoe (Zoe-master) - Lightweight, no HTML, MVC-style JavaScript library for building web applications
Additional Tools:
Sound.js (sound-main) - Audio processing library for web audio applications
Chart.js (Chart.js-master) - Popular charting library for HTML5 canvas
Layout (layout-main) - Layout management library
UI Components (ui-main) - General UI component libraries
Konva (konva-master) - HTML5 Canvas JavaScript framework enabling high performance animations, transitions, node nesting, layering, filtering, caching, and event handling for desktop and mobile applications
search_web
GSAP (GSAP-master) - GreenSock Animation Platform, the fastest full-featured scripted animation tool on the planet with zero dependencies and many optional plugins
Two.js (two.js-dev) - Renderer agnostic two-dimensional drawing api for modern browsers supporting webgl, canvas2d, and svg rendering contexts
Babylon.js-master
This is a powerful, beautiful, simple, and open 3D engine for the web that supports advanced features like Image-Based Lighting (IBL)
. The library has comprehensive documentation with paths optimized for different learning styles - some users want to dive right into the engine's features while others prefer to start with setup and integration
. Babylon.js is designed for creating web-based 3D graphics and offers extensive capabilities for 3D visualization.
BabylonJSGuide.github.io-master
This serves as a complementary guide to the official Babylon.js documentation, providing an overview of the components that make up BabylonJS and where to begin learning . It's meant to be used alongside the official documentation which contains information about add-ons, extensions, videos, and a complete list of available classes .
VSweb (vscode-web-editor-main)
This is a web-based version of Visual Studio Code that runs entirely in the browser's sandbox environment . It enables users to edit code directly in their browser without needing to clone repositories locally. The editor offers a limited execution environment for security purposes, loading code via APIs rather than cloning repositories, which reduces the attack surface when working with untrusted code .

For GLB Models I have included 100+ different examples to test and play with including all shown in online demonstrations plus many more. You can search for them in the following path with GLB Viewer or navigating to your documentation folder and following rest of path:
http://localhost:8888/files/documentation/threejs/glbfiles/

This folder setup must be exact or you will need to redo all paths in provided tools to whatever you change them to. After extracting verify you have an ai folder and in it you can track the following paths:
YOURAIFOLDERWITHDOCKERCOMPOSEFILE\documentation\threejs\java\three.min.js or
YOURAIFOLDERWITHDOCKERCOMPOSEFILE\documentation\threejs\OtherJS\Babylon.js-master\Babylon.js-master\packages\tools\playground\public\scenes\Alien\Alien.gltf

Take Note I have my folder on d: and all will need to update files to change to your specific paths. Mine is D:\HYM3_AGI\ prior it was C:\HYM3_AGI\ so all will need to adjust for this. Also note Open Terminal requires an API. You must set one (Long Password you enter in docker-compose.yml file) and transfer it to Admin Settings in Open WebUI in integration tab. Only enter chroma and open terminal in this. Remove all other ports if using my file. Everything is integrated with root access and docker commands available to the AI. Always Set Note in Admin Settings/Documents that AI is never to delete any user files. Make this a clearly stated mandatory rule.

Upcoming:

solar system
3d numbers and lines
fluid simulator
tetris type game
additional scenes and custom tools.

These are updated for

OpenWebUI - 0.8.10

Docker-Desktop - 4.63.0

I do have a new tool for system management and code executor. I am not releasing that tool yet. It was made for my AI and it allows the AI to use the jupyter code server directly and have access to real file systems. I do not know how other AI would use this so have not released this yet. If you would like the tool and understand the risk message and I will share it. In meantime i will be downloading and testing on other models to see results. I do not know if they will try to delete files or change srttings or mess stuff up. It allows ai to use root tools with no sandboxing. This is dangerous with wrong AI. You may message me on here for system tool code. Otherwise I will get posted after my AI release or after security testing on AI to identify which are safe.