Quickstart
Prerequisites
- Python 3.10, 3.11, 3.12, or 3.13.
- Bash or Zsh for the tested five-command block. Windows users can use the PowerShell equivalent below.
- An installed checkout:
python -m pip install -e .. After the Alpha is published,python -m pip install "climadc==0.1.0a1"is equivalent.
The built-in study is deterministic, project-owned synthetic data. It needs no API key or network access.
Tested five-command path
Run this block from any directory after installation. The marker on the fence is consumed by the offline documentation test.
```bash quickstart-test export CLIMADC_STUDY="$(mktemp -d)/climadc-quickstart" climadc init "$CLIMADC_STUDY" climadc validate "$CLIMADC_STUDY/study.yaml" climadc benchmark "$CLIMADC_STUDY/study.yaml" climadc report "$CLIMADC_STUDY/runs/latest"
The final command prints the absolute path to `report.html`.
## PowerShell equivalent
```powershell
$Study = Join-Path ([System.IO.Path]::GetTempPath()) ("climadc-" + [guid]::NewGuid())
climadc init $Study
climadc validate (Join-Path $Study "study.yaml")
climadc benchmark (Join-Path $Study "study.yaml")
climadc report (Join-Path $Study "runs/latest")
Published run
runs/latest resolves to an immutable run directory containing exactly:
| Artifact | Purpose |
|---|---|
run.yaml |
frozen configuration and run identity |
lineage.json |
input hashes, model IDs, split IDs, and software version |
splits.parquet |
explicit train/calibration/test membership |
predictions.parquet |
typed point and quantile predictions |
metrics.json |
forecast and shadow-decision metrics |
leakage-report.json |
accepted/rejected rows and violations |
dataset-card.md |
source, license, hash, range, and limitations |
report.html |
static human-readable report |
The synthetic Quickstart checks execution, contracts, and artifact integrity. It does not demonstrate real-site accuracy or savings.