Platform support

This package ships one complete headless library with SPICE and all seventeen curated scientific inputs. There is no profile selection or feature extra. The historical extended value returned by capabilities().profile is an internal compatibility label. No GUI is included or planned for GMAT Py.

Qualification on 2026-10-16 for source b52941e9830940c3086839431fd1fb64812076ff built full headless wheels that each passed 194 tests without skips on Linux x86_64 and ARM64 and macOS Intel/ARM64. Linux x86_64 was qualified locally; the other three targets passed native run 38970516533. The builds used standard-ABI CPython 3.14.7 (Linux ARM64: 3.14.8), glibc 2.28 on Linux and macOS 15.7.9. Deployment tags do not establish execution on every older host. These results identify that exact source revision; successor release files require their own source/build receipts and validation.

The earlier minimal/extended wheel variants and Windows results are historical. Windows is unqualified for this candidate. Python 3.10+ metadata permits source build attempts beyond the qualified CPython 3.14 scope; other Python versions, interpreters and free-threaded Python remain unqualified. Wheels are specific to their Python ABI and platform. Runtime needs only the standard library and bundled native binaries/data; source-build requirements are separate. This qualification does not imply public binary availability.

Documentation

Read the hosted documentation for installation, quickstart, TLE Tailor integration, time/frames/units, public Python exports, scripts, native access, examples and development. The Sphinx sources are in docs/. To render locally, install docs/requirements.txt and run python -m sphinx -W --keep-going -n -b html docs docs/_build/html from the repository root. This does not install or compile GMAT.

Propagate an orbit

import gmatpy as gmat

samples = gmat.propagate(
    [7000, 0, 0, 0, 7.5, 1.0],
    epoch="2020-07-20T12:00:00Z",
    times=range(0, 3601, 60),
    forces=gmat.ForceModel(degree=70, order=70, drag=True),
    spacecraft=gmat.Spacecraft(mass_kg=1000, drag_area_m2=5, srp_area_m2=5),
)
print(samples[-1].time_s, samples[-1].state)

Inputs and results are Cartesian Earth-centered mean equator/equinox J2000 (EarthMJ2000Eq), in km and km/s. The epoch must include a timezone; output times are nonnegative SI seconds from the epoch in nondecreasing order. Each call starts a new orbit. Empty times return an empty list.

• Adaptive Prince-Dormand 7/8 integration (accuracy 1e-12, max step 60 s).

• EGM96 gravity through 360x360 or JGM3 through 70x70. Default EGM96 8x8; degree=0 selects point-mass Earth.

• Sun/Moon point-mass gravity and spherical SRP enabled by default, using DE405.

• Optional Jacchia-Roberts drag, with constant F10.7, average F10.7 and Kp (defaults 150, 150, 3); disabled by default.

• Constant spacecraft mass, Cd, Cr and drag/SRP areas.

State, time, frame and ephemeris utilities

These functions call GMAT’s native routines and return plain Python values:

elements = gmat.convert_state(samples[0].state, to_type="Keplerian")
cartesian = gmat.convert_state(elements, from_type="Keplerian", to_type="Cartesian")
fixed = gmat.transform_state(cartesian, "2020-07-20T12:00:00Z",
                             to_frame="EarthFixed")
tt_mjd = gmat.epoch_mjd("2020-07-20T12:00:00Z", scale="TT")
moon = gmat.body_state("Moon", "2020-07-20T12:00:00Z", origin="Earth")

convert_state accepts Cartesian, Keplerian, ModifiedKeplerian, Equinoctial, ModifiedEquinoctial, SphericalRADEC and SphericalAZFPA. Cartesian is km/km/s; Keplerian is [SMA_km, eccentricity, inclination_deg, RAAN_deg, AOP_deg, anomaly_deg]. Other formats use GMAT’s native element definitions and angular units in degrees. The default anomaly is TA (true); MA, EA and HA may be selected. An explicit mu in km^3/s^2 supports conversion around other central bodies; the default is GMAT’s Earth value. These are osculating, not mean/TLE elements.

epoch_mjd accepts a timezone-aware datetime or ISO8601 UTC timestamp and returns conventional MJD (JD - 2400000.5) in UTC, TAI, TT, TDB, UT1 or A1. It does not expose GMAT’s different internal ModJulian origin. Python datetime precision applies, and literal leap-second timestamps (23:59:60) are unsupported.

transform_state handles EarthMJ2000Eq, EarthMJ2000Ec, EarthICRF and EarthFixed, including the rotating-frame velocity contribution. EarthFixed uses GMAT’s FK5/IAU1980 convention; it is not advertised as ITRF or TEME. The native MJ2000Eq/ICRF link applies the same table-derived rotation to position and velocity and uses a zero rotation derivative; it does not differentiate the ICRF table with respect to time.

body_state returns a geometric six-state in MJ2000Eq, in km and km/s, relative to a selected origin. Available bodies are Sun, Moon/Luna, Mercury, Venus, Earth, Mars, Jupiter, Saturn, Uranus, Neptune and Pluto. No light-time or aberration correction is applied. All helpers preserve the caller’s directory.

Advanced native bindings

The bundled core’s existing SWIG API is available through a scoped session:

with gmat.native_session() as api:
    sat = api.Construct("Spacecraft", "Sat")
    sat.SetField("DryMass", 125.0)
    print(sat.GetField("DryMass"))

Setup, serialization and object cleanup are automatic. Native object handles belong to the session: never retain or use them after the block. Convenience functions and nested sessions cannot be called inside an active session; they raise an error instead of clearing its objects or deadlocking. The full raw API retains GMAT’s ownership conventions; use its documented API patterns when attaching owned native objects. GUI and optional-plugin APIs are unavailable. The value-based helpers are the preferred interface for ordinary calculations.

GMAT has process-global state. Calls are serialized; for parallel jobs use multiprocessing with spawn. Calling after fork is rejected. Do not load another GMAT installation/API in the same process. The package uses the same working gmatpy name as GMAT’s separately installed Python API; use an isolated virtual environment when replacing that API. Startup configuration and logs are written to a private temporary directory, never site-packages.

Scientific inputs and limits

The wheel includes gravity coefficients, DE405, orientation, time and the exponential-atmosphere table from this checkout. They are snapshots and do not refresh automatically. AGI/CSSI space-weather files, Schatten forecasts and the unused marini.dat table are deliberately excluded from wheels and source archives while their redistribution terms/provenance remain unresolved. The propagate helper uses constant F10.7/F10.7A/Kp values and does not need these files. Advanced file-based weather models require caller-supplied licensed inputs via CSSISpaceWeatherFile/SchattenFile; absent selected inputs raise an exception, with no automatic download or substitute predictions. Native DE405 covers 1941-01-06 to 2201-02-20; the inherited Earth-orientation snapshot ends at 2026-09-14 and includes predicted rows. GMAT can extrapolate orientation outside the snapshot with reduced fidelity. The restored NASA ICRF_Table.txt retains its irregular grid and internal gaps, with table coverage approximately 1960-01-01 through 2100-01-01. The restored March 2026 Earth high-precision PCK ends on 2026-06-12; that SPICE-frame limit is separate from the default FK5/EOP path. The legacy SVN updater and Xerces are excluded. A new data updater is deferred. Numerical regressions are provided, but this is not validated as an operational mission-analysis tool or an STK HPOP replacement. The convenience API exposes impulsive maneuvers, multi-spacecraft propagation, bounded SMA targeting and mission scripts. Finite burns use scripts. Estimation and STM helpers remain deferred. Other core facilities are accessible experimentally through native_session.

Licensing and distribution status

GMAT and GMAT Py code use Apache-2.0; CSPICE and embedded f2c runtime components have their own notices. gmatpy/NOTICE.txt, gmatpy/LICENSES.md and gmatpy/licenses/ are included in the wheel and source archive. The component manifests record source URLs, content hashes, applicable terms, numerical-data reuse assessments and scientific changes. They are an engineering inventory, not a worldwide legal guarantee. The optional archive-wide SPDX expression is omitted for this mixed distribution; component licenses remain attached. The owner adopted the scoped product-specific determination on 2026-10-16, recorded in gmatpy/licenses/EXPORT_DETERMINATION.json. Its applicability remains limited to the recorded scope; package publication requires separate approval. Check an artifact:

python python/tools/check_distribution.py dist/gmatpy-0.2.0.tar.gz
python python/tools/check_distribution.py /path/to/gmatpy-0.2.0-*.whl --release

The first command verifies contents, notices and exclusions. --release also fails on unresolved manifest review items, instead of presenting an unverified bundle as cleared. Upstream reference data remain in the Git checkout/history; the unreplaced historical msise90_sub.for source is also excluded from the source archive. Excluding these files from wheel/sdist does not clear publication of the entire repository. The retained CubicSpline/MSISE replacements are the project’s equation-derived implementations, with scoped numerical repairs; their independent-origin descriptions remain project assertions. Brent uses NASA’s proposal formulas with focused safeguards. The specific baseline and provenance decisions are recorded in python/ci/CORE_BASELINE.md. They do not establish a blanket proprietary dependency, rights clearance, or an exhaustive audit of every inherited routine. Choose an independent project brand and use NASA/GMAT names to identify origin.

Building and validation

Source builds provide Linux and gated native macOS/Windows x64 qualification paths. Windows source builds remain experimental and unqualified for this candidate. They require a C++17 compiler, CMake, Ninja (or Make), CPython development headers and a PIC CSPICE archive. SWIG and scikit-build-core are Python build requirements. Obtain the official NAIF CSPICE toolkit and compile its C library with position-independent code, or point CSPICE_DIR at an existing include/SpiceUsr.h / lib/cspice.a tree. The source archive does not bundle CSPICE binaries; python/ci/build_cspice.py can rebuild the recorded target-native dependency from its pinned source archive:

python -m pip install build 'scikit-build-core>=1.0' 'swig>=4.2,<5' auditwheel patchelf ninja
CSPICE_DIR=/path/to/cspice bash python/tools/build_wheel.sh

GMAT_PY_PYTHON selects the interpreter and GMAT_PY_WHEEL_DIR the repaired wheel output directory. The helper builds and repairs the wheel. Building in an older manylinux image is necessary for portable Linux wheels; repair cannot lower a compiler’s glibc requirements. For a local host-specific candidate, set GMAT_PY_REPAIR=0 to retain the checked linux_x86_64 wheel without a manylinux claim. Failed builds retain their intermediate directory for diagnosis and recovery. Direct python -m build --wheel builds need -Ccmake.define.CSPICE_DIR=/path/to/cspice. Normal GMAT builds retain the legacy updater by default. Wheels omit duplicate unversioned linker aliases.

python -m build --wheel --no-isolation ^
    -Ccmake.define.GMAT_PY_EXPERIMENTAL_WINDOWS=ON ^
    -Ccmake.define.CSPICE_DIR=C:/path/to/cspice

python -m build --sdist --no-isolation --outdir /path/to/seed
python python/tools/record_source.py /path/to/seed/gmatpy-0.2.0.tar.gz \
    --revision <full-git-commit> --working-tree-sha256 <manifest-sha256> \
    --output python/gmatpy/BUILD_PROVENANCE.json
python -m build --sdist --no-isolation --outdir /path/to/candidate
python python/tools/check_distribution.py /path/to/candidate/gmatpy-0.2.0.tar.gz \
    --source-manifest python/gmatpy/BUILD_PROVENANCE.json

python -m pip install /path/to/gmatpy-0.2.0-*.whl pytest
python -m pytest /path/to/checkout/python/tests

Package contents and mission workflows

See python/PROFILES.rst for package contents, the TLE Tailor contract, supported impulse/multi-spacecraft/targeting/script workflows and remaining coverage. gmatpy.capabilities() reports available features. The internal extended label is retained for compatibility; all current standard wheels contain the full headless feature set and require no profile flags or extras.

Contributor dependencies

Testing and source-archive tools are in the local dev dependency group. With uv, use uv sync --group dev; this installs development tools and also builds the project, so the native source-build prerequisites above still apply. To obtain only contributor tools without building GMAT, use uv sync --group dev --no-install-project. Pip contributors can install build and pytest explicitly in an isolated environment. Documentation uses its separate pinned docs/requirements.txt. Dependency groups are contributor tooling, not alternate published package editions.