Update README.md

README.md

@@ -14,20 +14,21 @@ tags:
   - local-ai
   - multilingual
   - cli-runtime
+  - ai-runtime
 pipeline_tag: text-generation
 library_name: llama.cpp
 ---
 
-# Offline AI 2.1 – EuroLLM-9B-Q8_0 (GGUF)
+# Offline AI 2.2 – EuroLLM-9B-Q8_0 (GGUF)
 
-Offline AI 2.1 is a fully local AI runtime built around digital sovereignty, privacy, and system autonomy.
+Offline AI 2.2 is a fully local AI runtime environment built around digital sovereignty, privacy, and system autonomy.
 
 No cloud.  
 No telemetry.  
 No tracking.  
 No external dependencies.  
 
-Everything runs locally via llama.cpp.
+Everything runs locally via **llama.cpp**.
 
 ---
 
@@ -37,13 +38,46 @@ Below is the Offline AI runtime interface:
 
 ![Offline AI CLI Help Menu](cli_help_menu.png)
 
-Offline AI is not just a model launcher — it is a structured local AI workspace with:
+Offline AI is no longer just a model launcher.
 
-- Profile handling  
-- Runtime status inspection  
-- Controlled execution flow  
-- Modular architecture foundation  
-- Admin mode (locked access for advanced system control)
+It is a **local AI runtime environment** designed to manage and operate language models fully offline with a structured command interface.
+
+Core capabilities include:
+
+- CLI runtime environment  
+- Model lifecycle management  
+- Profile-based workspace system  
+- Snapshot conversation archiving  
+- Runtime diagnostics and monitoring  
+- Administrative control layer  
+
+The architecture is designed as a foundation for **multi-model local AI systems**.
+
+---
+
+## 🧠 RUNTIME ARCHITECTURE
+
+Offline AI uses a layered architecture:
+
+User (CLI)
+↓
+Python Runtime
+↓
+C++ Inference Engine (llama.cpp)
+↓
+GGUF Language Model
+
+
+The Python runtime acts as the **control layer**, responsible for:
+
+- command handling  
+- model orchestration  
+- workspace profiles  
+- snapshots and notes  
+- system diagnostics  
+- administrative operations  
+
+The inference backend is a lightweight C++ wrapper around **llama.cpp** with real-time token streaming.
 
 ---
 
@@ -52,41 +86,46 @@ Offline AI is not just a model launcher — it is a structured local AI workspac
 Base model: EuroLLM-9B  
 Quantization: Q8_0 (GGUF)  
 Format: llama.cpp compatible  
-Runtime: llama.cpp  
-Offline AI Version: 2.1  
+Inference engine: llama.cpp  
+Offline AI Version: 2.2  
 Recommended RAM: 16 GB  
-Platforms: macOS, Windows  
+Platforms: macOS, Windows, Linux  
 
-This repository distributes a quantized GGUF Q8_0 variant of the EuroLLM-9B model for efficient offline inference.  
-The original model weights are unmodified and not fine-tuned as part of this project.
+This repository distributes a **quantized GGUF Q8_0 variant** of the EuroLLM-9B model optimized for efficient local inference.
+
+The original model weights are **not modified and not fine-tuned** as part of this project.
 
 ---
 
-## 🧠 WHAT'S NEW IN 2.1
+## 🚀 WHAT'S NEW IN 2.2
 
-- Refined CLI architecture  
-- Structured command system  
-- Improved response handling  
-- More stable execution  
-- Admin access layer (locked system control mode)  
-- Cleaner internal logic separation  
+- Structured CLI runtime environment  
+- Model lifecycle management system  
+- Model alias system  
+- Workspace profiles and isolation  
+- Snapshot conversation archiving  
+- Runtime diagnostics and monitoring  
+- Administrative control mode  
+- Improved modular runtime architecture  
 
-Offline AI 2.1 transitions from a simple launcher to a structured local runtime environment.
+Offline AI 2.2 evolves the project from a simple model launcher into a **local AI runtime platform** prepared for managing multiple specialized AI models.
 
 ---
 
 ## 🔐 PROJECT PHILOSOPHY
 
-Offline AI demonstrates that:
+Offline AI demonstrates that modern AI systems can operate fully offline.
+
+The project explores the idea that:
 
-- Modern AI can operate without cloud infrastructure  
-- Open models can run independently  
+- AI does not require cloud infrastructure  
+- Open models can run independently on personal hardware  
 - AI tools can respect user privacy  
-- Local-first computing is viable  
+- Local-first computing is a viable architecture  
 
-The project promotes:
+Offline AI promotes:
 
-- Digital independence  
+- Digital sovereignty  
 - Transparent system design  
 - Offline experimentation  
 - User-controlled AI environments  
@@ -101,8 +140,8 @@ Funded by: European Union research initiatives
 Base model license: Apache License 2.0  
 
 Quantized distribution: GGUF Q8_0  
-Runtime: llama.cpp (MIT License)  
-Offline AI interface and wrapper: © David Káninský  
+Runtime engine: llama.cpp (MIT License)  
+Offline AI runtime interface: © David Káninský  
 
 All components are used in compliance with their respective licenses.
 
@@ -123,4 +162,6 @@ Use beyond personal, research, or educational purposes is at your own responsibi
 
 Website: https://OfflineAI.online  
 Domains: .cz / .sk / .de  
-Author: David Káninský
+
+Offline AI Runtime  
+Author: David Káninský