|
Gianluca Administrator Posts: 1368
6 hours ago
|
Hello everyone,
I'm going a bit off-topic here to tell you about a new, exciting madness I've been diving into lately: it's called GANI.
GANI is a very different program from SyMenu, yet it draws many of its core principles from it: absolute precision, total control, and zero compromises.
So, what exactly is GANI?
In technical terms, it is a RAG system assisted by conversational AI.
Specifically, a RAG (Retrieval-Augmented Generation) is a system usually based on an LLM (Large Language Model), making it similar to ChatGPT, Gemini, or Copilot. Unlike those, however, a RAG concentrates its analytical power primarily on the user's own data. In practice, it's an intelligent assistant that reads your documents, understands them, and answers your questions based predominantly on the materials it has "absorbed".
For those who know a bit about the subject: I should clarify that it isn't exactly an "agent". It won't take initiative, it won't control your accounts, it won't go grocery shopping for you, and it won't clean the cat's litter box. It limits itself to reading and digesting only the documents you actively provide. Likewise, it will erase from its memory any documents you remove.
Is this the "total control" I mentioned earlier?
No. There are various software tools today that offer more or less the same thing. Even AIs like ChatGPT can do it: you give them a document, they read it, and they process it.
The real revolution here is something else: 100% local privacy, with no smoke and mirrors.
Today we are surrounded by cloud-based AI. Even 99% of so-called "local agents" are nothing more than interfaces that ultimately rely on online services for the actual processing.
But for me, privacy is not: "send me your text, I'll analyze it, I'll reply, and then... I promise... trust me... I'll delete it."
To me, privacy means not sending anything out at all, in the most absolute and total way.
GANI's AI runs entirely on your PC.
The whole indexing and analysis process happens right there, on your hardware, completely offline.
Your data never leaves your hard drive.
GANI is an AI with a pure and irreproachable soul that respects your privacy and doesn't exploit your content simply because... technically, it can't.
Clearly, running everything locally comes at a price: the hardware requirements are heavy. This isn't "lightweight" software for every PC but it requires computing power to run the models smoothly. You need a GPU, RAM, and VRAM. If you want to use the most powerful LLM models (GANI lets you choose from several), you'll need to increase that power... in the magic world of AI, power is everything. And unfortunately, power has a cost. And if you think we have all this computing power for free on the cloud today... well, reflect on the fact that you are paying for it with your data.
Consistent with my long-standing tradition, GANI will be completely free for the vast majority of users. However, to make it sustainable, I have included a limit on the number of indexable documents in the Free version (the Pro version will be dedicated to those who need to manage large volumes or want to use it commercially). SyMenu, in its own way, taught me that this is the only way to ensure continuity and remove the boundaries of development.
Why this post?
Well, I need some of you, enthusiastic and technically skilled people, who want to jump into this new adventure.
At this stage of development, I am looking for:
Beta Testers: to measure the AI engine performance on different hardware configurations and to test the user interface.
Translators: to make GANI multilingual.
But I've also opened this thread to gather your suggestions, doubts, technical insights, or critiques.
I want to know what you think and if the idea of a truly local AI seems as sensible to you as it does to me.
If you'd like to give a hand with testing or translations, please write to me in private.
You can learn more and download the beta version of the program on the GANI website here:
https://ganisoft.com
|