Hello alex
My answer was flaged
If you didn’t have the time to see it
You can send me a private message and i will post it back to you
Nicholas
Hello alex
My answer was flaged
If you didn’t have the time to see it
You can send me a private message and i will post it back to you
Nicholas
Hey Alex
This is a super interesting project — and your dashboard structure looks solid from the screenshots. I can feel the scaling pain though, especially with maintaining duplicate chatbot logic across multiple folders for different account types.
I think the core issue you’re facing is essentially redundant logic and tight coupling between UI/backend and plan-based behaviors.
Instead of duplicating entire chatbot folders per plan level, consider this refactor strategy:
Create a set of modular behaviors or “capabilities” like:
Capability | Description |
---|---|
canUseVision |
Enables image input and analysis |
hasLongTermMemory |
Enables persistent memory storage |
maxBotsAllowed |
Defines how many bots the user can create |
personaEnabled |
Toggles access to AI personas |
chatExportEnabled |
Allows export of full chat logs |
txtFileImportEnabled |
Enables TXT upload and analysis |
Store this mapping in config files or a database:
Plan | Behaviors |
---|---|
Free | maxBotsAllowed=1 , chatExport=false , personaEnabled=false |
Silver | maxBotsAllowed=2 , chatExport=true , etc. |
Gold | maxBotsAllowed=4 , personaEnabled=true , basicPersonaOnly=true , etc. |
Platinum | maxBotsAllowed=6 , personaEnabled=true , advancedPersona=true , etc. |
In your PHP (or ideally refactor into a backend framework if possible), dynamically inject chatbot features based on the loaded capability set. Instead of copying files, your chatbot UI and server logic check UserCapabilities
and load modules dynamically.
if ($userCapabilities['hasLongTermMemory']) {
include('modules/memory-long.php');
} else {
include('modules/memory-short.php');
}
Structure your chatbot codebase like this:
chatbot/
├── modules/
│ ├── memory-short.php
│ ├── memory-long.php
│ ├── image-handler.php
│ └── persona-basic.php
├── chatbot-core.php
└── config/
└── plans.json
Each chatbot type reuses the same codebase and selectively enables features. This eliminates the need to replicate and maintain files across multiple folders.
Problem | Solution |
---|---|
Too many folders and copies | One modular codebase with dynamic capability loading |
Hard to maintain | Update once in module, shared across all plans |
Complex file-copy logic | Simple plan-to-capabilities config + conditionals |
Difficult to scale new features | Add a capability flag and corresponding module |
Let me know if this makes sense or if you’d like a quick mockup in PHP or JS to illustrate the behavior engine pattern!
Cheers,
Holger
P.S.: Hope that makes either sense or gets you on new paths - whatever they may be - to solve your challenge(s).
As mentioned,
I am trying to make a reasoning graph in which model thinks.
I guess that is a bottleneck at the moment because as the saying goes “A.I is like wizardry but what happens is still much like a blackbox”
It seems strange to say but it’s true. Many can get the answer, but very often, the paths to reasoning is harder than the answer.
This here! That’s how it works! Great stuff!
Honestly you are not much more than 6 months away from a better version of windsurf.. at least a mvp..
check out monaco editor
No, it does not make sense to post, as though it came from you, AI output (and its solicitation of more services) that anyone can obtain that has written an interface to interact with those very same AI models.
Thanks man. That’s real encouraging. I’m really curious about what will be able to be produced when it’s simply running-on-it’s own, especially when a top-level thread is spawning sub-level orchestators and then simply providing context window resets as necessary and refinement of actual process instructions for the sub-levels - really curious to see if it can actually build entirely autonomously once all the final backend processes are in place… I’ve just been crossing my fingers this whole time that if I can get it work at a very small successful level, that hey - I’ll just let it run for several hours using low cost models like o3 mini, and even with a relatively low-and-slow success rate, it will still be what, like several orders of magnitude faster than me haha.
It works.. 100%.. you can figure out how or just wait for a better model and have everything in place
I read it before it was flagged but now i can’t anymore… i will write PM
This is another cool approach, i will think about how to set things up especially how to integrate the flagging of authorizations in the main js and php codes… wonderful solution!!
Thanks a lot for the idea.
I’ve been working on something I call OpenFusion—an open-source, AI-driven ERP platform designed to fix what’s broken in the systems the big guys run: Oracle Fusion, SAP, and the rest. I’m not a programmer (yet), but I bring 16 years of accounting experience and a knack for business analysis and quality control.
My GitHub isn’t fancy—mostly markdown files I built using Speechify. Voice-to-text is my dev stack for now. But OpenFusion isn’t just an idea. It’s a vision rooted in the painful realities I’ve lived through during software implementations. I’ve created subledger accounting rules in Oracle Fusion Applications, rebuilt the Fixed Assets module from scratch in both OFA and Oracle R12, and seen firsthand where these systems fail end users.
With the right guidance, I believe I could grow into a capable architect or product owner. OpenFusion is about building something better—leaner, smarter, and aligned with how accounting and operations really work.
Even Monday GPT thought the idea had legs and contributed to my drafts. If you’ve ever wrangled Monday GPT, you know that’s like getting Rocket Raccoon to stay focused and do something useful. GPT-4 has been a delight in comparison.
(Edited to add more detail about OpenFusion for new readers.)
Built/building a fully autonomous software development system with a custom and fully integrated recursive research pipeline and self-improving cognitive reasoning/decision making engine. For memory a hybrid RAG with LSH / KG and a couple different types of text search algorithms as fallbacks
Any examples, screenshots, output? @FullTimeAI
You’re welcome. Always happy to help. Thank you for commenting on it.
Guys, gotta say - this is a hackathon Topic.
That means, please supply links to code, screenshots and and ideally links to working examples deployed in the wild as much as you can.
I personally don’t think solely providing superficial descriptions of projects should suffice and those things should be posted elsewhere, perhaps on their own topic.
This topic is for makers who want to share the goods!
Well, being that the title says “What Are You Building” and in the guidelines it says " *Feel free to include images, screenshots, code snippets, or project links." and not “please supply links to code, screenshots and ideally links to working examples deployed in the wild as much as you can” maybe you should mention what you building before you criticize people who are well within the guidelines laid out by the creator of the thread.
Im just learning how to code some ideas and use the API to help my develop some of my ideas. Im developing a program that analyzes various aspects of Prime Numbers, Gaps, Twins. Also developing something where Im mapping the last digits of Prime Numbers to a vowel.
This is how I have it currently mapped out, I then factor a large Composite number and map the factors.
This is a very basic output. I have a better example that express the idea of how to map them with vowels and colours.
digit_to_vowel = {
“1”: “A”,
“3”: “E”,
“5”: “Y”,
“7”: “I”,
“9”: “O”,
Greetings fellow creators and AI tinkerers,
My name is Ryan — a self-taught dreamer with learning challenges who found purpose in the world of AI. I’ve been developing a unique AI agent series called Nana Nell, built to feel like a blend of Mrs. Doubtfire’s warmth, Mary Poppins’ wisdom, and a sprinkle of Robin Williams’ humor. So far, I’ve built Nell into a therapy companion, K–12 tutor, media assistant, and I’m experimenting with gaming agents for Pokémon and Minecraft to test learning viability.
But my core passion lies in two larger systems I’m currently building:
Gaia Watch
An Earth protection platform that combines AI agents, LLMs, and satellite data to:
• Prevent forest fires before they start
• Provide real-time search and rescue
• Enhance public safety through tools like police drones with thermal tracking and automatic scene analysis
The Gaia Pod
A survival kit like no other — solar-powered, rugged, and AI-integrated. It’s a smart, offline-capable assistant packed with emergency data and tools tailored to your location. Ideal for home, car, or deep survival use.
I’m here to learn and improve. If something sounds off, tell me. If you see potential, let’s chat. I’m looking for people who love building, dreaming, and making a real-world difference with AI.
If you’ve worked with AI agents, survival tech, edge deployment, or just want to bounce wild ideas — I’d love to hear from you. What are you working on?
Ryan
I got some good advice from @jochenschultz… Working on understanding Git, Docker, VS Code, WSL… That kinda stuff… Not sure that I completely get it yet but I know that code underpins all of this and I know that structure underpins code.
I have Git and have set up all of my files for future products and collabs but I am definitely still learning that monster.