MeepMeep

Fast Keplerian orbits for exoplanet modelling.

MeepMeep computes Keplerian orbit quantities (transit geometry, projected separations, radial velocities, and phase curves) using 4th-order Taylor expansions around a set of expansion points distributed along the orbit. This makes it up to two orders of magnitude faster than per-point Newton-Raphson while keeping the approximation error well below the photometric noise of current instruments. Optional analytic gradients with respect to the orbital parameters make it suitable for gradient-based inference (HMC, optimisers).

All hot paths are Numba-jitted and can be called directly from your own @njit kernels with no wrapper overhead.

The method is described in Parviainen & Korth (2020), MNRAS 499, 3356.

Installation

pip install meepmeep

The optional backends have their own extras: pip install "meepmeep[jax]" for the experimental JAX backend and pip install "meepmeep[opencl]" for running the OpenCL device functions. MeepMeep needs Python 3.10 or newer.

For a development checkout:

git clone https://github.com/hpparvi/meepmeep.git
cd meepmeep
pip install -e ".[test]"

Quickstart

import numpy as np
from meepmeep import Orbit

o = Orbit(npt=15, ep_placement="ea")
o.set_pars(tc=0.0, p=3.4, a=8.0, i=1.55, e=0.1, w=0.4)   # times in days, angles in radians
o.set_data(np.linspace(-0.15, 0.15, 500))

x, y, z = o.xyz()                       # sky-frame position (R_star); z > 0 toward observer
rv  = o.radial_velocity(k=120.0)        # radial velocity in the units of k

Bind tp=... instead of tc=... to anchor the orbit at periastron passage.

Analytic gradients

o = Orbit(derivatives=True)
o.set_pars(tc=0.0, p=3.4, a=8.0, i=1.55, e=0.1, w=0.4)
o.set_data(times)
x, y, z, dx, dy, dz = o.xyz()           # gradients w.r.t. (tc, p, a, i, e, w, lan), shape (N, 7)

C library

The evaluators are also available as a plain C99 library built from the same sources as the OpenCL backend, with expansion-point placement and the orbit-wide solvers included so it stands on its own. It is built and installed independently of the Python package:

cmake -S c -B c/build
cmake --build c/build
cmake --install c/build --prefix "$HOME/.local"   # meepmeep.h + libmeepmeep

See docs/source/c_library.rst and c/examples/transit.c.

Citing

If MeepMeep contributes to work that leads to a publication, please cite Parviainen and Korth (2020):

@ARTICLE{2020MNRAS.499.3356P,
       author = {{Parviainen}, H. and {Korth}, J.},
        title = "{Going back to basics: accelerating exoplanet transit modelling using Taylor-series expansion of the orbital motion}",
      journal = {Monthly Notices of the Royal Astronomical Society},
         year = 2020,
        month = dec,
       volume = {499},
       number = {3},
        pages = {3356-3361},
          doi = {10.1093/mnras/staa2953},
       adsurl = {https://ui.adsabs.harvard.edu/abs/2020MNRAS.499.3356P},
}

Generative AI policy

MeepMeep grew out of code written for PyTransit from 2019 onwards and published in Parviainen & Korth (2020). That original code was written without AI or LLM assistance. Development since mid-2026 has been assisted by LLMs, mainly Claude Code with the Opus and Fable models.

Contributions developed with LLMs are welcome, but each one needs a human author who oversees the work and takes responsibility for it. Purely agentic contributions, made without human oversight, are not accepted.

License

MeepMeep is released under the GNU General Public License v3.0. See LICENSE for details.