NAVI — Local AI Assistant

Cyberpunk-aesthetic AI assistant running entirely on your local machine. No cloud. No data leaks. Just raw power.

Features

• 🧠 Three-tier model router — auto-selects between fast/smart/heavy models based on task
• 💾 Persistent memory — SQLite database remembers context across sessions
• 🎨 Image generation — Stable Diffusion via ComfyUI (JuggernautXL, NatViS, WAI-illustrious)
• 🎬 GIF & Video generation — AnimateDiff + Wan2.1 T2V
• ⚡ Runs 100% locally — RTX 3070, no API costs, no internet required
• 🖥️ CLI interface — cyberpunk green aesthetic, matrix-style

Stack

• Runtime: Node.js (ESM)
• LLM backend: LM Studio (Qwen2.5-3B / Qwen2.5-Coder-7B / Dolphin 24B)
• Image gen: ComfyUI
• Memory: better-sqlite3
• Style: Serial Experiments Lain / cyberpunk

Requirements

• Node.js 18+
• LM Studio running on localhost:1234
• ComfyUI running on localhost:8188 (optional, for image gen)

Quick Start

git clone https://github.com/BlessedsCode/Navi.git
cd Navi
npm install
node NAVI.mjs

Roadmap

• Python rewrite (Navi-Core)
• PostgreSQL RAG memory
• TTS voice output
• Screen vision
• VRAM manager
• Packaged .exe release