A detailed look at the five core components that make up R.E.V.I.S. and how they work.
Core technology is patent pending (JP Application No. 2026-072886).
This unprecedented cognitive system operates through the seamless integration of five independent core components.
The central orchestrator governing all thoughts and modules. It decodes user intent and manages every cognitive process.
Detects conversational drift. Decides whether to recall memories or create new threads to prevent context pollution.
Operates in the background to organize information, vectorizing and anchoring pure knowledge deep within its memory as an advanced RAG feature.
Double-checks answers behind the scenes and triggers correction processes for high reliability.*4
The advanced task queue system powering R.E.V.I.S.'s nervous system. It organizes and executes multiple tasks based on priority.
Prioritizes user actions while intelligently handling background memory and validation tasks.
Enables advanced logical reasoning that transcends the limits of a single model by decomposing complex problems and chaining dedicated inference processes.*5
Link multiple Macs on your local network to pool computational resources, enabling high-speed parallel agent processing.
Proprietary context extraction technology. Rapidly retrieves only the 'pure facts' needed to boost answer accuracy.
Instantly pulls knowledge from Wikipedia for highly reliable and accurate information.
Accesses the entire Web. Covers latest news and specialized niche information.
An advanced compiler layer that dynamically generates and optimizes prompts based on user input.
Builds system prompts in real-time according to the nature of the task to elicit the best possible response from the AI.
Compresses token consumption to the absolute limit while maintaining context, enabling lightning-fast and low-load inference.
A 'virtual user' agent that operates R.E.V.I.S. on your behalf to complete research and tasks independently. It handles any time-consuming processes in the background.
Predicts potentially needed knowledge, preparing the system to generate faster and more accurate responses in future interactions.
Simply say 'Look into this for me.' It conducts thorough research in the background and quietly anchors the findings into the system's memory.
*Future updates will add even more advanced autonomous task execution capabilities, maximising the AI's independence.*3
| Minimum | Recommended | |
|---|---|---|
| Mac | Mac (M4) or later | Mac (M5) or later |
| Memory (RAM) | 16GB or more | 24GB or more |
| Storage (SSD) | 1GB or more*6 | 2GB or more*6 |
| Local Model | Gemma4-12B-QAT(4bit)*2 equivalent or higher | Gemma4-26BA4B-QAT (4bit) equivalent or higher |