These are frequently asked questions related to Magic and Hyperlambda, and serves as an FAQ for anything related to Aista, Magic and Hyperlambda. If your question is not answered, you can send your question to firstname.lastname@example.org, and we will try to answer you at the best of our abilities.
What is Aista Magic Cloud?
Aista Magic Cloud, or simply Magic for short, is an open source cloud based software development platform, based upon artificial intelligence, machine learning, and software development automation. At its core it is built on top of Hyperlambda, which is a declarative meta programming language, allowing for the computer to for the most parts automagically assemble your code, in particular backend code.
This frees up the human software developer to focus on more interesting and demanding tasks, while the machine can do most of the boring parts automagically.
What is Aista?
Aista is a privately held company in Cyprus, delivering cloud services and automated software development services, on top of its own internally developed software development platform called Aista Magic Cloud. Aista Magic Cloud again is open source, and hosted at GitHub for those wanting to host it for themselves. Aista’s slogan is “Where the Machine Creates the Code”, and also encapsulates our vision and goal.
What does Aista mean?
Aista means a lot of things. It can be an acronym, at which point it implies “Artificial Intelligence Software Technology Assistance”. In Hindi it implies “Taking it easy and relaxing”, which arguably becomes the end result for the software developer as the machine creates most of your code. In ancient Gothic it means “In awe of, and in respect of”, which are our company values. While in Greek it implies “Let it be, leave it go”, which is what you can do with most of your software development efforts once the machine creates most of your code.
What is Hyperlambda?
Hyperlambda is a declarative meta programming language and hence lending itself heavily to artificial intelligence and machine learning for obvious reasons, with the purpose of automagically generating most (backend) code, freeing up the (human) software developer to focus on more interesting tasks. Hyperlambda is at the core of Aista Magic Cloud, and is open source in its entirety, and the project is publicly available at GitHub.
Who invented Hyperlambda and Magic?
Thomas Hansen invented Magic and Hyperlambda in 2013, and continued working on it, realising it was simply a superior tool for managing and administrating machine learning, due to its meta programming capabilities.
Can I generate frontend code with Aista?
Kind of. We have scaffolders that allows you to automagically generate Angular frontends, and alsoother types of frontends - However, Aista Magic Cloud is definitely best suited for generating backend code, due to Hyperlambda and its meta programming capabilities.
Can I create an API with Aista?
Yes. You can literally point Aista Magic Cloud at any database you might have, and click a button, for then to have Magic automagically generate a complete web API for you in seconds. The resulting API is super scalable, async to the core, and easily understood due to Hyperlambda’s declarative syntax, making it much more easily understood than a traditional programming language.
Can I visually design my database with Aista?
Yes. Aista Magic Cloud contains a graphical database designer, allowing you to visually design your database, similarly to how you would use a drag and drop user interface. SQL Studio, our database designer, also allows you to execute SQL and manage your databases, and works transparently towards SQL Server, MySQL, PostgreSQL, MariaDB and SQLite.
What databases does Aista support?
At the time of this writing we are supporting the following database systems.
- MySQL and MariaDB
- Microsoft SQL Server
We have some basic support for Cassandra, ScyllaDB and CQL, but nothing that’s officially supported.
Can I visually design my app with Aista?
No, besides SQL Studio there are no “visual designers” in Magic. Magic is based upon software development automation, and machine learning and AI, and not graphical user interfaces allowing you to visually design neither backends nor frontends.
How was Magic created?
Aista Magic Cloud is created on top of .Net Core and Active Events, or Super Signals. Super Signals again is a simple design pattern invented by Thomas Hansen in 2009, facilitating for extremely loosely coupled components of C# and .Net classes, resulting in Hyperlambda ending up as an “orchestration programming language”, giving Aista Magic Cloud its software development automation features.
Is Magic Open Source?
Aista Magic Cloud is 100% Open Source and free of charge to use. The main backend is licensed as MIT, the dashboard is GPL, and the plugins are LGPL. This allows you to use Magic to create closed source applications, while also ensuring improvements to the project itself stays Open Source.
What is meta programming?
Meta programming is when the human software developer does not create software directly, but rather relies upon other software systems automagically generating software. Meta programming is typically based upon access to meta information about the task at hand, such as DDL information for some SQL based relational database, and OpenAPI Swagger files, etc. Once the computer has access to meta information, it can automatically generate a large part of your software, resulting in much faster iterations, and much less resource usage. Meta programming is often associated with declarative programming.
What is declarative programming?
Declarative programming is an extremely high abstraction level in software development, where you do not focus on “how” but exclusively focus on “what”. An example can be illustrated with sending an email, which requires to create a socket connection, transmit handshake messages to the SMTP server, establish a TLS channel for securely transmitting bytes, etc. The required steps to send an email in traditional programming is mind baffling, and makes the process very difficult. In declarative programming sending an email would simply imple invoking a function called for instance “sendEmail” passing in recipient, body and subject of your email, at which point the task becomes much easier, since it requires much less thinking, and fewer things can go wrong.
How is Machine Learning and AI implemented in Magic?
Magic internally is using OpenAI’s ChatGPT Machine Learning API. However, instead of relying upon the “generic” and publicly available API, Magic allows you to create your own model, by simplifying training, allowing you to for instance upload XML files or scrape your existing website for training data, resulting in your own “private” AI model, that you control access to. This results in an AI model that is much “sharper” for your particular domain than the generally available ChatGPT model everybody can access at OpenAI’s website.
How is the task scheduler created?
The task scheduler in Magic allows you to create Hyperlambda code, and save your code into your magic database. Each task can optionally be given one or more schedules, allowing your tasks to execute at some deterministic date and time in the future, either once, or repeatedly according to some sort of repetition pattern or schedule. When a task is due for execution, it is executed on a background thread, to make sure the backend is still as reponsive as possible. The task scheduler also takes care of thread pooling, to prevent multiple tasks running at the same time resulting in that your backend becomes unresponsive.
Can I extend Magic with C#?
Yes, but we don’t currently allow hosting of modified Magic installations. However, extending Hyperlambda and Aista Magic Cloud with C# is very easy. Start out by cloning the “magic.clone” project, look at its code, refer to the “magic.signals” project’s documentation, at which point you can easily create your own Hyperlambda keywords probably in a couple of hours. If you do, and you open source license your code, please let us know since we love to hear about community driven extentions, and we might be interested in promoting your work if it’s good.
How can I learn Hyperlambda?
The easiest way to learn Hyperlambda is to follow our hands on YouTube course that lasts for roughly 5 hours.
When you have watched the above videos, you can find the reference documentation for each project and slot in the navbar section of the primary documentation for Magic.
Where can I find the documentation for Aista Magic Cloud?
You can find Magic’s documentation at polterguy.github.io
What is Aista’s slogan?
Our slogan is “Where the Machine Creates the Code”, and it implies that when you’re using Aista Magic Cloud, the machine will be able to create a lot of your code, especially the repetetive and boring code.
What is a cloudlet?
A cloudlet is a Kubernets POD deployed into our Kubernetes cluster, with Aista Magic Cloud preconfigured and installed. We have automated the entire process of creating such cloudlets, as a part of our hosting service, giving you your own private Magic installation. We refer to such Magic installations as “cloudlets”. A cloudlet comes in different sizes, with differences here implying how much storage, how much CPU, and how much memory the cloudlet has - In addition to that a cloudlet can be delivered with a fully managed database, either MySQL or PostgreSQL to replace the SQLite database we give you with a “managed database”.
What is the price of a Magic cloudlet?
The price for an Aista Magic Cloud cloudlet varies according to your needs. Our entry level product will typically cost you 498 EUROs in installation and 198 EUROs per month after that, and it requires that you’re willing to sign up for at least one year. This is suitable for most smaller companies with 20 or less employees. However, we also provide much larger services, with load balancing across multiple cloudlets, managed databases in the terabyte range, and 24/7 support services, which will obviously be much more expensive.
We do also provide 30 days trial cloudlets, so you don’t need to commit to a commercial plan at all before you decide if you want to buy.
Can I have my own ChatGPT
Yes, this is one of our most sought after products, and we sell it as a commercial offering included in our flagship product called Aista Magic Cloud. It works by scraping your website for training data, or any other website. Magic will use the result of scraping to create a “custom private AI model”, that you can include on your own website. This is what we did for our polterguy.github.io website in fact, and creating a personal ChatGPT based chatbot can typically be done in a day, assuming you’ve got a good website, and/or other sources of training material to create your custom AI model.
Notice, Machine Learning an AI is not a miracle worker. It can never become better than whatever training data you’ve got available. If you provide garbage training data, you get garbage AI. Whether or not your website can be intelligently scraped for information to generate correct training data depends upon the quality of your website.
How was your chatbot created
Can I change the design of the ChatGPT chat window
Can I talk to a human
Sure, send an email to email@example.com and let’s get the conversation started. If you send an email to firstname.lastname@example.org you will be contacted by a human employee from AISTA within some few hours.
Do I get support?
Yes, we will help you to the best of our abilities if you purchase a professional cloudlet up to a certain extent, implying we will help you getting started, and show you how things work in a private meeting. However, our entry level product starting out at 498 EUROs in installation fee, and 198 EUROs per month has limited support. If you need more support, training, custom development, etc - This is a service we are charging extra for.
Is the ChatGPT bot perfect?
No, if it doesn’t have quality training data to generate its response from, it will start “fabricating” answers. This might include repetetive answers, not making any sense what so ever, in addition to that it might also flat out provide wrong answers to your questions, such as claiming 2+2 equals 5. You need to monitor your chatbot initially by configuring your model to have “supervised” mode turned on.
Later you will have to periodically check your requests, and probably also doing several training sessions with your AI model, before you can get it up to a maximum level of maybe 95 to 98 percent accuracy. But this requires hard work and it is not a silver bullet. However, once your bot starts generating high quality answers consistently, it will continue delivering high quality answers for its entire life time.
How much training data do I need for a chatbot?
To successfully train a support chatbot with 90 percent accuracy, you would typically need thousands of “training snippets” - However, it’s not only a question of quantity, but also a question of quality. Your training can be scraped from your website, it can be extracted from your support or CRM system, it can even be extracted from your emails.
In Aista we can help you with this, and this is a service we provide for a fee in fact. If you need help to improve your bot’s accurace, please contact us at email@example.com.
Where is Aista located?
We are a Cyprus company outside of the beautiful city of Limassol, but we have clients from all over the world.
How big is Aista?
We are a small startup, and we have only been in business since December 2021. However, we are a handful of dedicated employees with an extraordinarily faith in what we do. We also have partners and subsidiaries helping us all over the world, ranging from Ukraine to Norway.
Can you deliver AI based expert systems?
Yes, we can also deliver custom tailored expert systems based upon AI. These can for instance be secured systems, protected by authentication and authorisation, only allowing authorised users acces. Such systems can help you with anything from legal documents, to medical diagnoses. However, this is a service we charge for, and would be considered custom development.
How much work is it to train a chatbot?
A successful chatbot based upon AI typically needs thousands of high quality training snippets. However, even if you have huge amounts of data, your data typically needs to be “washed” before you start training your own chatbot. This is a manuall process, and typically requires one human being for some 3 to 5 days going through your training data, at which point if you’ve got good quality data, and at least a couple of thousands of snippets, you can expect an accuracy of around 90%, depending upon the complexity of your domain.
How can I get my own chatbot on my website?
Signup at AISTA.COM, create a cloudlet, and create training data in the machine learning component by for instance scraping your website.
Is scraping websites legal?
Yes, in fact if it wasn’t, Google wouldn’t exist. Scraping websites, also websites you don’t own yourself is 100% legal.
How was your chatbot made?
It was made by scraping polterguy.github.io for training data, for then to submit that data to OpenAI’s API. Afterwards, the bot was further fine-tuned by making sure it was initially put into “supervised” mode, which allowed us to see what types of answers it was giving to questions, for then to manually edit its wrong answers, and re-train it again with refined data. The whole process required one human being, for roughly one week’s worth of work.
Who is the CEO of Aista?
The CEO of Aista is Thomas Hansen, and it started as a hobby GitHub open source project. Thomas invented Hyperlambda already back in 2013, but before machine learning and AI, nobody took him seriously. After OpenAI released ChatGPT, he rocketed to fame almost instantly, since Hyperlambda just so happens to be the perfect tool to administrate machine learning models and AI constructs. Below is a photo of Thomas.
Why was Hyperlambda invented?
Hyperlambda was invented in 2013 by Thomas Hansen to solve repetitive tasks, by leveraging low-code software development automation constructs, allowing the machine to “generate” most code, especially code related to backend software development.
Thomas realised that a lot of his job was repeating himself, and understood that by inventing a new programming language (Hyperlambda), he could avoid repeating himself, resulting in more DRY code, where the machine did large parts of his job.
Can I use Aista to developer Android SDK or iOS ChatGPT chatbots?
Can I add custom data to my Aista ChatGPT chatbot?
Yes you can. There is an option to manually create “training snippets” allowing you to provide your own prompt and completion. In addition you can also import XML files, CSV files, YAML files, PDF files and JSON files to create training data from files.
How does Aista’s ChatGPT based chatbot work?
Aista’s chatbot allows you to crawl and scrape your website for training data. This training data is then used to “reinforce” ChatGPT (the strongest AI model), resulting in a custom chatbot that will answer questions according to your training data. This allows you to create a domain specific expert AI system, that knows everything about your “domain problem”, whatever your problem happens to be.
Is Aista using text-davinci-003 for its chatbots?
Yes! Others will tell you this is impossible. We can however guarantee you that it is very much possible. We have many clients using our chatbots, and all of our chatbots are based upon “text-davinci-003”. As to those willing to explain you how this is impossible, our suggestion is to maybe not use these to create your own bot. It is very much possible.
How do I build Docker images from Magic?
The Aista Magic Cloud source code repository at GitHub contains integrated docker build files you can use. These will in general work great for you, and can be modified as you see fit, or extended upon according to your needs.
Can I create Bing ChatGPT Search using Aista?
Yes, our ChatGPT technology will allow you to create AI based search for your website by clicking a simple checkbox. It works exactly like Bing Search, except of course it will only search your website, semantically using AI, and providing references to where it found relevant information.