Skip to content

References and Third-Party Integrations

Last Updated: December 12, 2025
Purpose: Attribution and acknowledgment of upstream projects and libraries


This document lists the third-party libraries, tools, and upstream projects that AI-OS depends on or is inspired by. We are grateful to the open-source community for making these tools available.


Core Dependencies

Machine Learning & Deep Learning

Library License Description Links
PyTorch BSD-3-Clause Deep learning framework Website · GitHub
Transformers Apache-2.0 State-of-the-art NLP models Website · GitHub
Accelerate Apache-2.0 Distributed training utilities GitHub
Datasets Apache-2.0 Dataset loading and processing GitHub
PEFT Apache-2.0 Parameter-efficient fine-tuning GitHub
Safetensors Apache-2.0 Safe tensor serialization GitHub
Hugging Face Hub Apache-2.0 Model hub integration GitHub

Memory Optimization & Performance

Library License Description Links
DeepSpeed Apache-2.0 Deep learning optimization library Website · GitHub
bitsandbytes MIT 8-bit optimizers and quantization GitHub
FlashAttention BSD-3-Clause Fast and memory-efficient attention GitHub

Tokenization

Library License Description Links
SentencePiece Apache-2.0 Unsupervised text tokenizer GitHub
Protobuf BSD-3-Clause Protocol buffers for tokenizers GitHub

Evaluation

Library License Description Links
lm-eval MIT Language model evaluation harness GitHub
math-verify MIT Mathematical verification utilities PyPI

Application Dependencies

Web & Networking

Library License Description Links
aiohttp Apache-2.0 Async HTTP client/server GitHub
httpx BSD-3-Clause Modern HTTP client GitHub
Playwright Apache-2.0 Browser automation GitHub
BeautifulSoup4 MIT HTML/XML parsing Website
lxml BSD-3-Clause XML and HTML processing GitHub
Trafilatura Apache-2.0 Web scraping and text extraction GitHub

Data & Validation

Library License Description Links
Pydantic MIT Data validation using Python types GitHub
orjson Apache-2.0 / MIT Fast JSON library GitHub
PyYAML MIT YAML parser and emitter GitHub
NumPy BSD-3-Clause Numerical computing Website · GitHub

CLI & UI

Library License Description Links
Typer MIT CLI application framework GitHub
Rich MIT Rich text and formatting GitHub
Pillow HPND Python Imaging Library GitHub
pystray LGPL-3.0 System tray icon library GitHub
matplotlib PSF Plotting library GitHub
Markdown BSD-3-Clause Markdown parser GitHub
tkinterweb MIT Web browser widget for Tkinter GitHub
rapidfuzz MIT Fast string matching GitHub

System & Utilities

Library License Description Links
psutil BSD-3-Clause Process and system utilities GitHub
watchdog Apache-2.0 Filesystem event monitoring GitHub
tenacity Apache-2.0 Retry library GitHub
tqdm MIT / MPL-2.0 Progress bars GitHub
dbus-next MIT D-Bus for Python (Linux) GitHub

Development Dependencies

Library License Description Links
pytest MIT Testing framework GitHub
pytest-asyncio Apache-2.0 Async test support GitHub
Ruff MIT Fast Python linter GitHub
Black MIT Code formatter GitHub

Documentation

Tool License Description Links
MkDocs BSD-2-Clause Static site generator GitHub
Material for MkDocs MIT MkDocs theme GitHub

Research References

AI-OS implements concepts from the following research:

Hierarchical Reasoning Models (HRM)

Mixture of Experts (MoE)

  • Switch Transformers: Paper - Scaling to trillion parameter models
  • GShard: Paper - Scaling giant models with conditional computation

Memory Optimization Techniques

  • Gradient Checkpointing: Trade compute for memory during backpropagation
  • Mixed Precision Training: FP16/BF16 training with loss scaling
  • ZeRO Optimization: Memory-efficient distributed training (DeepSpeed)

Acknowledgments

Special thanks to: - The PyTorch team for the foundational deep learning framework - Hugging Face for the transformers ecosystem and model hub - Microsoft DeepSpeed team for training optimizations - The EleutherAI team for the evaluation harness - All contributors to the open-source ML community


License Compliance

All dependencies are used in compliance with their respective licenses. AI-OS is licensed under the AI-OS Non‑Selling Attribution License (ANSL) v1.0 - see LICENSE for details.

For questions about licensing or attribution, please open an issue on GitHub.


Back to Home | Back to Guide