Install¶
pip install aether-quant is the research install: the library and the aq command on numpy and pandas only. It does not pull torch, Lean, Docker, Redis or Postgres. Add an extra for each thing you need.
| Command | Adds | For |
|---|---|---|
pip install aether-quant |
numpy, pandas | the library, aq hello, every facade (about 145 MB in a fresh venv) |
pip install "aether-quant[research]" |
scipy, scikit-learn, statsmodels, pyarrow, httpx | HRP, the gradient-boosted probes, parquet research files, FRED |
pip install "aether-quant[notebooks]" |
research + matplotlib, ipykernel, nbclient | the tutorials |
pip install "aether-quant[dashboard]" |
fastapi, uvicorn | serving the dashboard API |
pip install "aether-quant[live]" |
redis, psycopg, keyring, ibapi, the Lean CLI | paper and live deployments |
pip install "aether-quant[train]" |
torch, scikit-learn, pyarrow | retraining the models |
pip install "aether-quant[docs]" |
mkdocs-material, mkdocstrings, mkdocs-jupyter | building this site |
pip install "aether-quant[full]" |
everything above | the whole system |
The Lean engine image (about 40 GB) is only for real Lean backtests and deployments of the trading system. No research path needs it.
From a clone¶
git clone https://github.com/leon1706-lol/Aether-quant
cd Aether-quant
pip install -e ".[research]"
aq hello --offline
What a light install can do¶
Every facade imports with only numpy and pandas present (tests/test_lite_install.py blocks every optional package and checks). A function that needs an extra imports it when called. For example, compute_hrp_weights needs scipy, so it raises ModuleNotFoundError on a light install and works after pip install "aether-quant[research]".