Reactive AI Ecosystem Documentation

Complete installation guide for the Reactive Neural Networks framework

Complete guide to using and implementing SQA in your models

It includes also many breakthroughs in the performance/computational efficiency with next-gen algorithms like Sequential Three-Way Splice, and next-gen features, especially components pooling and re-mounting, to avoid recreating new DOM nodes.

More details will be provided soon and base features implementation is available on our GitHub

Most Important Features

• x:JSX - Next-Gen Dynamic Template Language
• View Injection design pattern and advanced, 2-level X View Injection System (rx:VI)
• Closure as an instance - all view layer primitives like Components & Directives are functions, that are called only once, on create. Then, initialized closure acts like their "instance", existing as an encapsulated scope and execution context, until Component/Directive is removed (or even longer, when using Pools)
• Rich set of built-in view layer primitives, made for the best cooperation with rx:VI, x:JSX and other rx:WP features/concepts
• Higher-order Factories - as all Component/Directive functions are called only once, they can act as factories for exactly everything - rx:WP introduces special, functional factories (functions creating functions), to create different kinds of view layer primitives, with different rendering behaviors
• Cascade Closures & Cascade Closure Services patterns - sharing service logic with multi-level closures, created on different execution contexts - Higher-order Factories are often expecting Cascade Closures
• rx:Pools - concept of storing and re-using objects/expressions that are expensive in create time - mainly DOM elements - instead of creating new ones - could greatly decrease the number of most expensive DOM operations, but increase the memory usage - depending on the Pool type and the use case
• pools could also act as explicit memory management solution - especially in environments without Garbage Collection
• Basic, Advanced and Alternative Render Modes - different variations of keyed, non-keyed and hybrid modes, based on most advanced and efficient algorithms, providing world's top level performance for every possible use case
• Re-mounting concept - re-mountable components and elements - over the scale performance boost, especially in shared, hybrid modes - re-using old (disposed) Component instances, that would be garbage collected normally, with all their connected DOM trees
• Pre-mounting concept - re-mountable components and elements, can be also pre-created in background (on idle) with some default props and stored in shared pools for re-use - i.e. list (in hybrid mode), that's using data fetched from server, could pre-create expected number of rows, when waiting for first response with data. Then, after receiving data, only re-mount (re-activate reactivity) pre-created elements, instead of creating new ones. That's the world's fastest possible way to mount the async-data based list elements.
• Move everything/everywhere concept - every component and element, could be moved anywhere else in the DOM, without re-creating DOM nodes and component "instances" - not only within the same parent - with different built-in APIs
• rx:Backend Server Components - inspired by upcoming React Server Components concept, but with key difference - in React RFC Server Components are just alternative one-time data fetching method, in rx:WP they are another example of Inversion of Control - they are parts of UI, completely controlled by server.
• rx:Thread off-thread components

World's first experimental (PoC) real-time Reactive Language Model (RxLM) based on revolutionary Reactive Transformer architecture - processing only single interactions/messages, with all the context moved to Short-Term Memory, managed by Attention-Based Memory System.

That's not all from the advantages - event-driven real-time processing with memory is a lot more natural and human-like, than LLMs data-driven approach (processing full conversation history everytime). It's a crucial milestone in development of AGI and awareness models.

MRL Training in progress

"Currently, it's the intermediate version from Memory Reinforcement Learning - training is still in progress, model will be updated soon!"

Advanced library for efficient integration of RxLM/RxAM with cloud/server runtimes. Reactive models require stateful processing, so it needs dedicated solutions for efficiency. Reactive Cloud will provide multi-level memory management patterns, for persistent memory for multiple users

Introduction to multiplatform web/native chatbot client - it will handle stateful event-driven processing and will be recommended for integrations. It will be mostly based on rx:Web Platform. More details will be available soon.