The Future of Aru Ai and the Roadmap

If you open the Aru Ai website or application now, there is a roadmap in the guide section.

Overall, when I wrote this document, I tried to be as concise and informative as possible; if a person uses Aru and has managed to set everything up, they probably don't need a guide at all.

In this post, I will try to elaborate a bit more on the future, modules, refactoring, new releases, and the roadmap.

In the article about the history of Aru Ai development, I talked about what I went through and why many modules that existed before didn't make it into the current version, but you will find out what will appear soon and what will never be implemented.

The Future as an AI Assistant:

Overall, everything will remain as it is now, but many new features will appear. Right now, I'm working on STT and TTS; it will be possible to communicate with Aru by voice. For children, especially those who cannot yet read, this will be excellent functionality.

Of course, thanks to new modules, Aru will become smarter and better. But now the focus is on basic things and convenience. For example, Aru will have tabs and nested settings. It will be possible to work simultaneously with different chats on separate tabs and even assign separate LLM providers to different chats. Aru will not lose all its capabilities, character, semantics, and heuristics because of this.

The project will feature the ability to generate images and videos. Initially, at the level supported by your chosen model. Then, also with the help of Aru Native.

I declare that there will be more translations; currently, everything works with i18n. I will slightly rework the algorithm; ideally, I want to reach a point where anyone can take a JSON and translate it into the language they need. Of course, this will only work in personal versions hosted on one's own hardware or computer, but I guarantee that translations for many languages will appear in the project by default.

Python code execution is not difficult, and in fact, it was even in the latest versions, but I cut the functionality because I didn't like how it looked on the interface. Of course, don't expect pip install everything, but it will be useful for learning and testing simple algorithms.

Ephemeral mode - the ability to create chats that will use Aru's power but will not save anything, use semantics, or heuristics will be added. Aru's age modes and character will remain, but in such chats, she will know nothing about you and will never remember anything; nothing will be saved to the database either.

Speaker mode - I have already started thinking through these implementations. Essentially, it's an easier-to-understand implementation of STT and TTS but for simple tasks, as is the case with most smart speakers from various manufacturers - quickly asking for the weather, checking traffic, getting directions, or simply playing music on demand. Many questions will arise here - where to get music? I am working on making it possible to connect streaming accounts legally and securely within the project.

Full-fledged tasks, a calendar, notes, and even a Kanban board will appear. In the future - integration with Google (possibly).

Of course, there will be more improvements in semantics, heuristics, more emotions, and stickers. Enhanced work with artifacts and a visual editor directly on the canvas.

An iPhone version is already under development.

Budget and Sponsorship:

Despite earning from third-party developments, business automation, Telegram bots, and acting as a visiting manager, Aru remains a non-commercial project. I finance it myself as much as I can, but without external help, the entire roadmap and new modules will be added incredibly slowly, which is why I haven't specified any deadlines in the guide document or this article.

Any investment will be useful and will go towards the project's development. If you wish to help:
You can set up a monthly payment on Ko-Fi (all cards accepted)
You can send a crypto donation if it's more convenient for you via Cryptocloud

I am very grateful to everyone who contributed to the project's development; particularly large donations will be noted directly in the project's information section.

New Modules:

Telegram Module - will appear at the stage when the code is fully open and some other modules are added. It will be possible to create a separate chat and connect it to a Telegram bot to communicate with Aru directly in the messenger without losing her character, emotions, and even semantics. Yes, this will only work on local versions.

"Health" Module - this topic is personally close to me, so I have read many articles about chronic conditions and diseases requiring daily monitoring, and studied all aspects concerning both myself and people living in a controlled regimen. This will be something like a state diary - hypothyroidism, diabetes, insulin resistance, migraine attacks, etc. All of this can be asked for Aru to analyze and provide recommendations, which will be recorded as separate cards so that the user can easily understand what is happening and how to monitor it. It will also include analysis interpretation and dietary assistance based on all the data the user inputs. This is not a substitute for a doctor or medical consultations; it's more about automation than an AI doctor.

"Sport" Module - Aru will start tracking your progress, providing workout recommendations, creating plans, and evaluating the work done. The sport and health modules will be released simultaneously, so they will be data-linked.

Aru Game Studio - I have only just started thinking about this module. Ideally, it would create simple games from pre-made resources based on the simplest voice commands, even in speaker mode. Of course, it won't just be a JS file; I will try to organize work with a simple and accessible JS game engine. In advanced mode, it is planned to add custom resources - textures, graphics, models, tiles. I think it will be interesting. The very first prototypes I made separately from Aru showed interesting results; everything needs to be well thought out and organized.

Aru Connector - this module once existed in previous attempts to refine Aru. The idea is to allow users to connect to SQL databases, such as those of online stores, websites, or services. And with the help of NLQ SQL, enable Aru to safely work with this data, perform analytics, extract summaries, and highlight weak points (in sales, for example). It's not certain that this will appear, but I hope it will work out.

Aru Native - What I Aim For:

I very much hope to achieve this. When most of the modules are completed, the source code is fully open, and the project reaches stable branches, I will start thinking about a separate provider where nothing needs to be connected by users themselves.

This will undoubtedly only work if there is sufficient funding and the project is important. I will choose a sufficiently powerful multimodal model and either host it on dedicated hardware or rent it via cloud APIs with maximum security.

In this mode, Aru will not lose its ideology or principles; nothing will ever be saved except your own database. Users will simply have the option to start communicating with Aru without connecting anything at all.

Hardware Project and Devices for Everyone:

This might sound strange, but in Kazakhstan, we have several enthusiasts who make their own smart speakers. Aru would look excellent in such a mode. This means that in the future, we could either try to release smart speakers and wearable devices with Aru, or make the blueprints and guides publicly available.

Article Conclusion