Here is my video:
I started this idea when I thought about bringing GPT to my PC and adapting it to work with my local data, to create a system that delivers data exactly how I want.
Initially named Robert, the AI has evolved into Alina.
I would like to hear your opinion on this. How would you use it? How do you see this evolving in the future?
Currently, I am considering building a new architecture for a mini AGI to connect other assistants to a Central Intelligence and see how it works.
Interesting. So how are you connecting the data source? Is it hosted in OpenAi as an attachment? How dynamic is that?
Some users report still getting hallucinations from assistant with data retrieval, wondering if your implementation remains accurate in all situations?
Yes is accurate, I know I need to shape it better, for example in manage services (I was trying to get him done fast… so I skip some details at this moment → the confirmation (AI know need to confirm the changes, and i cut some processes for the video, to make it shorter). 
I am thinking to change the architecture, I want to handle multiple assistants by Alina (I am very curious about this).
But first I need to make a good interface so I can present this… and to adapt the code for the web.
Do you know if Chat GPT product came with Chat Plugin or something?
The chat they have → have good efect that cover some time for assistant to process, they have the audio option, the quote option, etc. 
I am not lazy, I just think anyone who would think to sell a product that include a GPT model, should have the GPT chat plugin included. And to add some options for customization and that’s all, full package. 
Your interface could be any kind of application or chat widget, just consume the API for your created Assistant.
Wondering about how you are writing the data, is it writing directly to the files? Any latency there?
Thx, i am lazy to build one, i am expecting OpenAI to offer that with Assistant API as full package in the future.
Look how beautiful looks with the chat, off course i will love to use this 
:
Yes is writing in the files directly in my computer.
I speak with Alina and tell her my needs and she do that for me.
The fun part was, i changed the language and she understand what i want and did it, to change the package to Pretty500. 
I am not sure if in this video, i said something like “i want to upgrade my internet services”, she activate them and then i said “i want to change them back” and deactivate them. 
The purpose for that was to show she understand the context and i don’t use any trigger words or something like that.
I am focusing to make them reasoning and let them chose in unique situation so i can sleep better 
In the case you fill the model with data and expect your answer to be generated for a certain question, i will worry a lot.
We should not use those like a storage, those vectors should be dynamic values all the time because he get smarter with new quality information → those should be like references.
Maybe i am wrong, i am thinking we should always compare GPT to a kid that is learning and take reference all the time (ex: kid hit a table with his head and he will get values: trigger cry, feel pain, brain release endorphins, adjust the speed for next time to not hit the table, adjust height, etc… if he grows will adapt the values). Something like that 
I was in Windows 11, for project used WSL (Ubuntu), on Wi-fi, I set bigger time values in my code. Can be improved.
Thanks, bu still not clear yet how it is reading from the modified CSV file in your computer, have you created a web service to which the assistant is calling? And how do you instruct the assistant to write?
I can not give you more info about the build and mechanisms I made, sorry.
Ok I now understand you only wanted to show off your creation without any contribution. Anyway I can only advice you lower your confidence levels as soon you will see your creation will start to behave unexpected, and you will go crazy and/or mad about it, and you might come here to ask for guidance, and you will find people telling you that they are sorry but cannot tell you more. Welcome to the AI world
I already gave you the solution for your problem.
I am ready for that from the start to encounter fluctuations in values (will be always like that → future is context), that’s why I don’t fine-tune the model (this should be only to make him intelligent not to storage documents).
After I will get the results I want, my direction is to collaborate with OpenAI to make it better and stable for everyone.
Here is a video that will help you understand why i gave you that answer and understand what you need to fix.
At minute 4 pay attention:
I’d use to code.
Having said that, promising but disappointing attitude. If you’re not willing to share knowledge, don’t post things here and expect some kind of advice/feedback. It ends up looking like a scam.
@jandieg.b If you’re interested in an open-source solution, please take a look at GitHub - OpenInterpreter/open-interpreter: A natural language interface for computers
Advice ≠feedback, how you try to shape it.
I didn’t said I come for advice or to share code.
Why so aggressive if I don’t share my code, what is your problem?
That Open Interpreter is like a copy of GPT Engineer, is not even close to what I am doing and what I am targeting.
Right now is not the moment to share anything, I am still working on it because is part of my mini-AGI architecture. When I will share it I am not sure, and in what conditions will be.
