User tiers

The package is built to be useful at three depths. Each tier reuses the same evaluator and provenance-tagged catalogs, so results are comparable across them.

Tier 1 — compare and diagnose

Run a scenario against an Earth baseline and read the binding constraints. No optional extras are required.

uv run orbitdc compare space.yaml earth.yaml --tornado --monte-carlo 500

You get the delivered-compute waterfall, the levelized cost of compute (LCOC), the binding constraint (what fails first), the feasibility thresholds (what would have to change for space to match Earth), and a tornado / Monte-Carlo view of which assumptions move the answer. Start in examples/notebooks/00_quick_start.ipynb and 01_compare.ipynb.

Tier 2 — optimize and explore sensitivity

With the mdao extra, search the design space rather than evaluating one point.

uv run orbitdc optimize space.yaml --objective lcoc
uv run orbitdc optimize space.yaml --pareto lcoc,kg_per_kw
uv run orbitdc sobol space.yaml --objective lcoc
uv run orbitdc doe space.yaml --metrics lcoc,kg_per_kw
uv run orbitdc robust space.yaml earth_*.yaml

Single-objective optimization (gradient-free), NSGA-II Pareto fronts including mixed-integer architecture (satellites, accelerators/sat, altitude), a Latin-hypercube DOE, and SALib Sobol indices. See 03_pareto_exploration.ipynb and 04_monte_carlo_uncertainty.ipynb.

Tier 3 — extend, deepen, and visualize

• Custom catalogs. Add an accelerator, launch vehicle, coating, or coolant as a provenance-tagged YAML entry; it flows through every model and the provenance table.
• Higher fidelity (Phase 4C, opt-in). Formation-keeping Δv and a collision margin, parametric radiator view factors (Level 4), mission-integrated degradation (Level 5), Skyfield ground-station access (orbit extra), and a time-stepped graceful-degradation availability curve.
• Dashboard. With the viz extra, an interactive Panel app:

uv run panel serve examples/dashboard_app.py --show

The model architecture page shows how these couple.