pyampp 1.0.1

Python Automatic Model Production Pipeline

pyAMPP is a Python implementation of the Automatic Model Production Pipeline (AMPP) for solar coronal modeling. It streamlines the process of generating realistic 3D solar atmosphere models with minimal user input.

Documentation

Full documentation is available at: https://pyampp.readthedocs.io/en/latest/

Format and viewer references:

• Canonical model I/O and contract enforcement: docs/model_io.rst

• Upgraded HDF5 stage format (NONE/BND/POT/NAS/NAS.GEN/NAS.CHR): docs/model_hdf5_format.rst

• Public geometry API and migration status: docs/geometry_api.rst

• Practical gx-fov2box CLI guide: docs/gx_fov2box_usage.rst

• Main GUI workflow and command mapping: docs/gui_workflow.rst

• Viewer guide (2D selector, 3D viewer, and refmap browser): docs/viewers.rst

• Release notes: CHANGELOG.rst

Overview

AMPP automates the production of 3D solar models by:

• Downloading vector magnetic field data from the Helioseismic and Magnetic Imager (HMI) onboard the Solar Dynamics Observatory (SDO)

• Optionally downloading contextual Atmospheric Imaging Assembly (AIA) data

• Performing magnetic field extrapolations (Potential and/or Nonlinear Force-Free Field)

• Generating synthetic plasma emission models assuming either steady-state or impulsive heating

• Producing non-LTE chromospheric models constrained by photospheric measurements

• Enabling interactive inspection and customization through focused GUIs

Installation

pyAMPP can be installed directly from PyPI.

0. Setting up a Python 3.10 Environment (Recommended)

We strongly recommend running pyAMPP in a dedicated Python 3.10 environment. You may use any of the following tools to install Python and create an isolated environment:

• Miniforge: https://github.com/conda-forge/miniforge

• Miniconda: https://www.anaconda.com/docs/getting-started/miniconda/install

• Anaconda: https://www.anaconda.com/docs/getting-started/anaconda/install

• Conda: https://docs.conda.io/projects/conda/en/latest/user-guide/install/index.html

• pyenv: https://github.com/pyenv/pyenv?tab=readme-ov-file#installation

Please refer to the official installation instructions for your chosen tool. Their documentation is comprehensive and up to date, so we will not repeat it here.

You can skip this environment setup section if you already have a suitable Python environment.

The instructions below use Conda as an example:

1. Install Conda If you don’t have Conda yet, follow the instructions for your platform here: https://docs.conda.io/projects/conda/en/latest/user-guide/install/index.html

2. Create and activate your environment Open an Anaconda Prompt or Command Prompt and run:

   conda create -n suncast python=3.10
   conda activate suncast

3. Upgrade pip and install pyampp

   pip install -U pip setuptools wheel
   pip install -U pyampp

4. (Optional) Install additional dependencies

   For a full scientific Python stack (e.g., SunPy and related tools):

   pip install -U sunpy[all]

Development And Test Setup

If you are developing pyAMPP from a local checkout, install it in editable mode from the repository root:

pip install -e .

This installs the package and its core runtime dependencies, but it does not install optional dependency groups such as the test tools.

The optional dependency groups are declared in pyproject.toml under [project.optional-dependencies]. Use the extras that match your task:

pip install -e ".[tests]"
pip install -e ".[docs]"
pip install -e ".[tests,docs]"

For local test runs, the recommended setup is:

pip install -U pip setuptools wheel
pip install -e ".[tests]"
python -m pytest -q

Using python -m pytest is recommended because it guarantees that pytest runs from the same environment as the interpreter where pyAMPP and its test extras were installed. If pytest reports an unknown option such as --doctest-rst, the active shell is usually picking up a different environment that does not have pytest-doctestplus installed.

We currently keep test and documentation dependencies in pyproject.toml extras rather than separate requirements files. That keeps the install metadata in one authoritative place and lets users opt into tests and docs explicitly.

Real-Data Sanity Checks

The repository also includes a small set of manual real-data sanity checks under real_data_checks/.

These scripts are intentionally separate from both the normal pytest suite and the usage examples because they:

• depend on local real-data fixtures that are not expected on every machine,

• can take minutes to run,

• are meant for targeted validation rather than routine development.

See real_data_checks/README.md for their purpose, runtime expectations, required fixture paths, and command-line usage.

Main Applications

pyAMPP exposes one main GUI and three focused command-line tools:

1. pyampp – The main desktop application. It is the front-end to gx-fov2box: - captures model parameters, - builds the exact reproducible CLI command, - runs the command, - opens the follow-up viewers when needed.

2. gx-fov2box – Headless model generator (IDL gx_fov2box equivalent).

3. gxbox-view2d – 2D FOV / box selector for existing .h5 or .sav models.

4. gxbox-view3d – 3D viewer for existing model HDF5 files.

5. gxrefmap-view – 2D browser for base maps and reference maps stored in model HDF5 files.

Canonical Model I/O And Geometry Contract

pyAMPP now provides a centralized model I/O surface via pyampp.io. This is the recommended path for application-level model loading and saving.

Key behavior:

• all model restores (H5 and SAV) are normalized through one loader path,

• missing Tier 1 + Tier 2 geometry contract fields in legacy models are inferred at load time,

• completed contract metadata is persisted on save,

• geometry consumers can assume one normalized metadata state.

Use these APIs:

from pyampp import io, geometry

model = io.load_model("/path/to/model.h5")
contract = model["metadata"]["geometry_contract"]
corners = geometry.world_corners_from_geometry_contract(contract)

Low-level readers under pyampp.gxbox.boxutils remain available for internal and compatibility use, but they are not the canonical application-level load entrypoint for contract-enforced workflows.

For portability and collaborator handoff, use pyampp.io.export_thin_model to produce a lightweight metadata-only HDF5 artifact containing full metadata plus optional observer sections.

Usage Examples

1. Launch the time/coord selector (pyampp)

pyampp

2. Generate a model via CLI (gx-fov2box)

gx-fov2box \
  --time "2022-03-30T17:22:37" \
  --coords 34.44988566346035 14.26110705696788 \
  --hgs \
  --box-dims 360 180 200 \
  --dx-km 1400 \
  --pad-frac 0.25 \
  --data-dir /path/to/download_dir \
  --gxmodel-dir /path/to/gx_models_dir \
  --save-potential \
  --save-bounds

3. Open the 2D FOV / box selector for an existing model

gxbox-view2d /path/to/model.h5

4. Open the 3D viewer for an existing model

gxbox-view3d /path/to/model.h5

Notes:

• –coords takes two floats, separated by space (no brackets or commas).

• One of –hpc, –hgc, or –hgs must be specified to define the coordinate system.

• Remaining parameters are optional and have default values.

• The default downloader backend is DRMS.

• Use --use-fido to switch to the legacy SunPy/Fido downloader.

• Use --force-download to bypass cache reuse and benchmark the downloader path directly.

• The GUI mirrors these controls with a Downloader selector and a Use cache checkbox.

• The GUI bottom command toolbar includes command-export actions: - copy the current gx-fov2box command - save the current command as a shell script

• Set PYAMPP_JSOC_NOTIFY_EMAIL to override the JSOC export notification email (default: [e-mail]).

Entry-Box Resume / Jump Rules

gx-fov2box supports --entry-box with .h5 or .sav input for stage-aware resume/recompute.

• --rebuild: ignore stage payload and recompute from NONE using resolved entry parameters.

• pyampp-export-model: canonicalize a legacy .sav or older .h5 into normalized HDF5 without recomputation.

• Without --jump2*, the pipeline starts from the detected entry stage.

• When --entry-box contains metadata/execute, --data-dir and --gxmodel-dir default from that execute string.

• Explicit CLI values for --data-dir / --gxmodel-dir always override execute-derived defaults.

• Jump validation rules are enforced: - backward jumps are allowed, - forward jumps are allowed by one stage, - POT -> NAS is explicitly allowed (implicit POT -> BND -> NAS), - POT -> GEN is explicitly allowed (skip both BND and NAS/NLFFF; keep true POT vectors).

• Save requests for stages before the selected start stage are ignored with a warning.

Command-Line Tools

After installation, the following commands become available:

• pyampp: Launch the command-builder GUI.

• gx-fov2box: Run the model-generation pipeline headlessly.

• gxbox-view2d: Open the 2D FOV / box selector on an existing model or entry box.

• gxbox-view3d: Open an existing HDF5 model in the 3D viewer.

• gxrefmap-view: Open base/refmaps from an HDF5 model in a 2D map browser.

• h5tree: Print an HDF5 tree (metadata shown by default, plus observer/name, optional observer/label / observer/source, and observer/pb0r/* when present; --no-metadata hides these value lines, --meta prints them only).

• gx-idl2fov2box: Translate IDL gx_fov2box execute strings (or SAV EXECUTE) into Python gx-fov2box commands.

• gx-fov2box2idl: Translate Python gx-fov2box commands (or HDF5 metadata/execute) into simple IDL gx_fov2box calls.

• show-model-metadata: Inspect canonical metadata and optional observer metadata from any supported model file.

• export-model-metadata: Generate a metadata-only thin HDF5 (full metadata + optional observer) from any supported model file.

License

Copyright (c) 2024, SUNCAST team. Released under the 3-clause BSD license.