NPU 🧠 AI LLM Keyboard: Joint Venture Core

Architects: [JUSTIN BOOTH] & Gemini AI (The "Mastermind" Partnership) Status: Android Studio Nuked 🛡️ | GitHub-First Protocol Active 🚀 Hardware Target: Qualcomm Hexagon NPU / DSP (2026)

---

🤝 The Joint Venture Manifesto

This project is a collaborative effort between human architectural intuition and AI logic.

• The Human: Provides the system requirements Logical build protocol architexture and design.
• The AI: Provides assistance, automation logic, and "Forensic" error-checking.
• The Goal: To prove that high-level mobile engineering belongs on GitHub, not in a bloated IDE.

---

🛠 Project Pillars

1. The "No-Trash" UI (/ui)

• Number Row: Permanent, high-efficiency access.
• Español Support: Native mapping for ñ and accents without "spazzing."
• Data-First: Layouts are stored as JSON maps to keep the logic portable.

2. The Hardware Brain (/core)

• NPU Integration: Direct C++ bindings for the Qualcomm AI Stack.
• Low Energy: 4-bit quantization for localized, private LLM inference.

---

🛠 System Architecture

This project is decoupled from OS-specific "trash." The logic is stored in pure data structures to ensure low-latency performance and battery efficiency.

1. The UI Layer (/ui)

• Number Row: Permanent dedicated row (no long-press required).
• Spanish Support: Native ñ and accented vowel mapping (á, é, í, ó, ú) defined in master_layout.json.
• Engine: Designed for Jetpack Compose (Android) or SwiftUI (iOS) consumption.

2. The Core Engine (/core)

• Logic: C++ Bridge for direct NPU/DSP interaction.
• Memory: Optimized for low-footprint LLM inference (4-bit quantization).

3. The Forensic Log

• Status: Android Studio Nuked.
• Platform: GitHub-First Development.
• Protocol: "Data Jar" sync via Google Keep.

---

🚀 Build Instructions

1. Edit ui/master_layout.json to modify key positions.
2. Use GitHub Actions (CI/CD) for compilation—Avoid local IDE bloat.
3. Commit with precision.