Free, open tools for AI safety
Sandbox AI agents, watch what they do with a hidden Observer, and measure it with evals. Open source, local, with evidence anyone can check.
Stage geospatial datasets for an event (an area and a time window) into one Zarr store.
Developer toolsPython · ShellGitHub @gaia-hazlabOnlineOwner unconfirmed
Have you tried it? One tap is enough. No sign-in needed.
prototype for a data staging CLI
Translation pending
Loading the full README.
Read the full README ↗Stage geospatial datasets for an event (an area of interest and a time window) into one analysis-ready Zarr store: weather stations, gridded weather (PRISM, HRRR, ERA5, Stage IV), land cover, elevation, soil, burn severity, satellite imagery (Sentinel-1, Sentinel-2, OPERA RTC) and seismic and infrasound waveforms.
Documentation: https://gaia-hazlab.github.io/gaia-cli/
git clone https://github.com/gaia-hazlab/gaia-cli
cd gaia-cli
pixi run gaia stage --config conus-precip \
-i https://raw.githubusercontent.com/DSHydro/skagit-met/refs/heads/main/data/GIS/SkagitBoundary.json \
-s 2025-12-01 -e 2025-12-20 -o skagit
See examples/skagit-precip.ipynb for an analysis of the result, and examples/pioneer-landlab.ipynb for staging Landlab inputs.
This is a convenience for creating datasets without requiring any local setup, but you must have write permissions to this repository to run GitHub actions.
The Stage Event workflow collates and stages data required for landlab landslide modeling in Zarr format. It optionally aligns the datasets and uploads the results to a public object store. Run it from the Actions tab or with the GitHub CLI:
gh workflow run stage-event.yml -R gaia-hazlab/gaia-cli \
-f config=examples/pioneer.yaml -f name=pioneer \
-f align="--crs utm --res 10 --aoi examples/pioneer-aoi.geojson" -f landlab_inputs=true
See the workflow file for all inputs, and
stage files for writing
your own config.
Staged stores shared publicly can be browsed in STAC Browser or on a map in stac-map.
pixi run test # tests
pixi run lint # ruff
pixi run typecheck # ty
pixi run docs # preview the user docs (docs/) at http://localhost:3000, running the notebooks
pixi run strip-notebooks # clear example notebook outputs before committing
User documentation lives in docs/ and is published to GitHub Pages from
main. Developer notes (design, roadmap, per-dataset details) live in dev/.