Quickstart¶

This path verifies the installation without downloading mission products or overwriting research artifacts.

Want to see the tool first? Open the CLI preview gallery for recorded help output and a baseline dry-run preview with copyable transcripts.

1. Inspect the command surface¶

python -m src.cli --help

SXS exposes the research workflows baseline, scaleup, search, and validate, plus demo, analyze, report, inject, and evaluate in the analysis workbench.

On the main branch, inspect the development version and run the new read-only installation diagnosis with:

sxs --version
sxs doctor
sxs config-check
sxs init runs/research-a
sxs status runs/research-a
sxs support-bundle runs/research-a --output sxs-support.zip

Add --json for a support-friendly record or --network to test the public MAST and NASA Exoplanet Archive endpoints without downloading observations. config-check validates all four workflow YAML files without executing a scientific stage or requiring downloaded data. init creates a marked workspace, copies those validated configurations, and does not run the pipeline or download observations. status then reports configuration health and recorded workflow state without changing the workspace. Use support-bundle only when you need to share bounded diagnostic metadata; inspect diagnostics.json inside the resulting ZIP before uploading it.

2. Preview the baseline workflow¶

python -m src.cli baseline --workspace runs/research-a --dry-run

The JSON plan should list the six baseline stages from environment validation through catalog validation. A dry run does not execute those stages.

3. Run deterministic tests¶

python -m pip install -r requirements-test.txt
python -m pytest -m "not network"

This checks preprocessing, BLS behavior, features, orchestration contracts, candidate selection, and independent-validation rules using deterministic test fixtures.

4. Choose a reproduction depth¶

Goal	Command	Cost profile
Baseline research	`python -m src.cli baseline --config configs/base.yaml`	Full ML environment; public downloads
Scaled qualification	`python -m src.cli scaleup --config configs/scaleup.yaml`	Larger catalog/sample and model training
Bounded candidate screen	`python -m src.cli search --config configs/candidate_search.yaml`	250 targets, four products each
Independent audit	`python -m src.cli validate --config configs/independent_validation.yaml --stage all`	1,000 shuffles per target plus external queries

Warning

These execution commands write research artifacts. Use a separate checkout for a fresh computation. Add --resume only for an interrupted run with compatible artifacts; archived reports alone do not establish compatibility. search expects the accepted RF v2 model and its selection metadata. validate expects the frozen 20-signal shortlist and processed light curves. Run workflows in order unless you are deliberately inspecting an existing evidence record.

5. Inspect results before interpreting them¶

Start with these versioned records:

Then read Interpreting results. A high RF score or catalog absence is not a discovery claim.