Installation¶
Everything in this project — every script and every notebook — is intended to run
inside the ADRQ Conda environment. Setting that up is the only prerequisite for
reading and running the workflow notebooks.
Prerequisites¶
1 — Clone the repository¶
git clone https://github.com/Philip-Brohan-MO/Auto-Daily-Rainfall-QC-MO.git
cd Auto-Daily-Rainfall-QC-MO
2 — Create the Conda environment¶
The environments/ADRQ.yml file pins all dependencies, including DuckDB,
PyArrow, XGBoost, scikit-learn, Cartopy, and the Milvus vector database:
conda env create -f environments/ADRQ.yml
conda activate ADRQ
Important
Activate ADRQ before running any script or notebook. This is non-negotiable
for reproducibility — the notebooks assume this environment and its ipykernel
kernel.
3 — Configure the project paths¶
The environment file sets two variables that the code relies on:
PYTHONPATH— the repository root, so the project modules import cleanly.PDIR— the data directory that holds the Parquet datasets and all SLURM shard outputs (defaults to/data/scratch/philip.brohan/ADRQ).
Edit these in environments/ADRQ.yml (or export them in your shell) to match
your own installation. The SLURM jobs read the same values from
scripts/slurm/config.sh.
4 — Select the kernel in Jupyter / VS Code¶
Open any notebook under notebooks/ and select the ADRQ kernel. Each
notebook’s header cell repeats this reminder.
Updating¶
git pull
conda env update -f environments/ADRQ.yml --prune