OMEGA V21 LoRA Adapter (GGUF)

LoRA adapter for autonomous AI agent on Android/Termux (aarch64).

Usage with llama.cpp

Requirements

• Base model: unsloth/gemma-4-E4B-it-GGUF (any quant)
• Vision projector: mmproj-BF16.gguf (for image/audio/video)

Run

./llama-cli \
  -m gemma-4-E4B-it-UD-Q4_K_XL.gguf \
  --mmproj mmproj-BF16.gguf \
  --lora omega-v21-lora-f16.gguf \
  -p "Your prompt"

Features

• Text understanding (custom OMEGA training)
• Vision (from base)
• Audio (from base, E4B native)
• Video (from base)
• Works on Termux aarch64 with 7.6GB RAM
• Optimized for ReAct reasoning + tool-calling

Architecture

OMEGA v22 four-layer architecture:

• Spiders (data collectors)
• Tools (action executors)
• Orchestrator (Gemma 4 E4B + this LoRA)
• SpatialCache (memory system)