Skip to content

Releases: dddimcha/embodiOS

EMBODIOS v1.0.0 - First Stable Release

29 Jan 11:59
fb7fafe

Choose a tag to compare

EMBODIOS v1.0.0

The world's first bare-metal AI operating system. Run LLMs directly on hardware without any OS overhead.

Highlights

  • 100% bare-metal - No Linux, no userspace, just transformers and hardware
  • GGUF model support - Load any model from Ollama/HuggingFace
  • Production-ready - Bootable ISO with embedded AI model
  • Simple CLI - ./embodi build && ./embodi run

Features

AI Runtime

  • Full GGUF v3 format support (Q4_K, Q5_K, Q6_K, Q8_0, F16, F32)
  • BPE tokenizer with SentencePiece compatibility
  • Streaming inference engine with KV cache
  • Parallel workers for multi-core utilization
  • SIMD acceleration (SSE2, AVX2, NEON)

Verified Models

Model Size Status
SmolLM-135M 469 MB
TinyLlama-1.1B 638 MB
Phi-2-2.7B 1.7 GB
Mistral-7B 4.2 GB

Kernel

  • x86_64 and ARM64 support
  • Memory management (PMM, VMM, slab, heap)
  • PCI, NVMe, VirtIO drivers
  • Interactive console with talk command

Quick Start

# Install dependencies (macOS)
brew install x86_64-elf-gcc x86_64-elf-binutils x86_64-elf-grub xorriso qemu

# Build and run
git clone https://github.com/dddimcha/embodiOS.git
cd embodiOS
./embodi build
./embodi run

Create Bootable ISO

# Download a model
curl -L -o models/smollm.gguf \
  "https://huggingface.co/HuggingFaceTB/SmolLM-135M-Instruct-GGUF/resolve/main/smollm-135m-instruct-q6_k.gguf"

# Build ISO
./embodi iso --model models/smollm.gguf

# Write to USB
sudo dd if=build/embodios.iso of=/dev/sdX bs=4M status=progress

Performance vs llama.cpp

Metric EMBODIOS llama.cpp
Memory 120 MB 160 MB
Latency jitter ±0.5ms ±5-10ms
Boot time <1 sec N/A
First token <20ms ~50ms

Documentation

What's Next

  • Network model loading
  • Web UI
  • Distributed inference (exo integration)

Full Changelog: https://github.com/dddimcha/embodiOS/commits/v1.0.0

v0.3.0

10 Dec 20:42

Choose a tag to compare

EMBODIOS AI-OS v0.3.0

An experimental AI operating system that runs AI models directly on bare metal.

What's New

  • Fixed CI model download (uses Python API)
  • TinyStories-15M (15M params) embedded
  • Text mode ISOs for USB compatibility
  • UEFI + Legacy BIOS support

Downloads

Choose the ISO for your system:

Text Mode ISO (Recommended)

  • File: `embodios-v0.3.0-text-mode.iso`
  • Best for: USB boot, older hardware
  • Fixes "no suitable video mode" errors

Quick Start

  1. Flash ISO to USB with Balena Etcher
  2. Disable Secure Boot in BIOS
  3. Boot from USB (F12/F2/DEL)
  4. Try: `ai Once upon a time`

System Requirements

  • CPU: x86_64 with SSE2
  • RAM: 512MB minimum
  • USB: For physical boot

v0.2.0

30 Jul 20:00

Choose a tag to compare

EMBODIOS v0.2.0

Installation

pip install embodios

Quick Start

embodi init
embodi build -f Modelfile -t my-ai-os
embodi run my-ai-os

See documentation for more details.

What's Changed

  • feat: [DOC] standardize documentation and remove external dependencies by @dddimcha in #2

Full Changelog: https://github.com/dddimcha/embodiOS/commits/v0.2.0

v0.1.0

29 Jul 21:25

Choose a tag to compare

EMBODIOS v0.1.0

Installation

pip install embodios

Quick Start

embodi init
embodi build -f Modelfile -t my-ai-os
embodi run my-ai-os

See documentation for more details.

Full Changelog: https://github.com/dddimcha/embodiOS/commits/v0.1.0