Nordlys Model Engine

Nordlys is an AI lab building a Mixture of Models. This repository contains the internal engine and tooling that power Nordlys models.

What is Nordlys?

Nordlys is an intelligent model selection system that:

• Selects optimal LLMs based on prompt analysis using ML clustering
• Optimizes costs while maintaining quality through sophisticated selection algorithms
• Supports multiple providers (OpenAI, Anthropic, local models, etc.)
• Provides high-performance inference with GPU acceleration

For Users

Use the public API or SDKs to access Nordlys models. Default model ID:

• nordlys/hypernova

For Contributors

Python Package Development

git clone https://github.com/Nordlys-Labs/nordlys
cd nordlys
uv sync --package nordlys

Requirements: Python 3.11+.

Run tests:

uv run pytest nordlys/tests/

C++ Core Development

For C++ development (including C bindings):

cd nordlys-core
conan install . --build=missing -s compiler.cppstd=20
cmake --preset conan-release -DNORDLYS_BUILD_C=ON
cmake --build . --target nordlys_c

This builds:

• libnordlys_core.a - Core C++ library
• libnordlys_c.so - C FFI bindings

Package Variants

• nordlys: Python package (CPU-only)
• nordlys[cu12]: Python package with CUDA 12.x support (Linux)
• nordlys-core: C++ core library
• nordlys-core-cu12: C++ core with CUDA 12.x (Linux)

Development Commands

# Install all dependencies
uv sync

# Code quality checks
uv run ruff check .
uv run ruff format .
uv run ty check

# Run tests with coverage
uv run pytest --cov

# Build documentation
uv run mkdocs serve

Links

• Documentation: https://docs.nordlyslabs.com
• Issues: https://github.com/Nordlys-Labs/nordlys/issues
• License: MIT (see LICENSE)