Skip to content

Releases: balaji-md/bmass

BMASS Seed v0.5 — First Working AI Operating Shell

Choose a tag to compare

@balaji-md balaji-md released this 20 Jul 00:48

BMASS Seed v0.5

This release marks the first complete BMASS operating loop.

The system now boots directly into a headless AI shell on Alpine Linux.

Architecture:

Operator

BMASS

llama-server

Linux

Real command output

BMASS

New

  • Background llama-server supervisor
  • Headless terminal interface
  • Persistent system prompt
  • Natural conversation
  • Local command execution
  • Non-root execution as user bmass
  • Runtime command bridge
  • Real operating-system feedback
  • Startup diagnostics
  • Local API architecture

The prototype has crossed from an LLM running on Linux into an AI-mediated local operating shell.

Experimental release.
aishell

v0.1.0 Seed Prototype

Choose a tag to compare

@balaji-md balaji-md released this 19 Jul 06:25

BMASS v0.1.0 — Seed Prototype

This is the first public release of BMASS (Bootable Model As System).

BMASS explores a simple question:

What if a computer booted directly into a local AI model instead of booting into a desktop and then launching an AI application?

This prototype demonstrates that it is possible using inexpensive, readily available hardware.

Features

  • Bootable AI environment from an 8 GB USB flash drive
  • Alpine Linux base system
  • Local inference with llama.cpp
  • Qwen3 0.6B GGUF model
  • Offline operation after installation
  • Lightweight BMASS launcher
  • Runs on an entry-level laptop with 4 GB RAM
  • Fully documented build process

Purpose

BMASS is not intended to compete with large cloud AI systems.

It is a proof of concept showing that private, portable and offline AI can be built from ordinary hardware using open-source software.

Demonstration

🎥 https://youtube.com/shorts/HRQl7P-wd1w

Repository

💻 https://github.com/balaji-md/bmass