Installation

To install roman_photoz, simply use pip:

pip install roman_photoz

Development

To install the development version of roman_photoz, you can clone the repository and install it in editable mode:

Clone the repository:

git clone https://github.com/spacetelescope/roman_photoz.git

Navigate to the project directory:
```
cd roman_photoz
```
Install package in editable mode. For example, on macOS:
```
pip install -e ".[dev]"
```

Bootstrap data/model setup

Run roman-photoz --setup to download the LePhare auxiliary data required by the Roman config, create a simulated catalog and Roman filter files, build the model (informer stage), trim LEPHAREDIR to the files needed by the estimator, and verify that the required artifacts are present:

Set the environment variables (recommended). roman-photoz uses LEPHAREDIR and LEPHAREWORK to determine where to store the LePhare data and work directories:
```
export LEPHAREDIR=/path/to/lephare_data
export LEPHAREWORK=/path/to/lephare_work
```
If you don’t set them, lephare falls back to its own default cache directories (typically under ~/Library/Caches/lephare on macOS, or the platform equivalent elsewhere) and prints a notice showing the full paths it chose. Setting the variables yourself gives you control over where these (potentially large) files are stored.

roman-photoz also looks for an INFORMER_MODEL_PATH environment variable to determine where the informer model file (roman_model.pkl by default) is read from/written to. It defaults to LEPHAREWORK if not set, so you only need to set it explicitly if you want to read/store the model somewhere else:
```
export INFORMER_MODEL_PATH=/path/to/model_dir
```
Run:
```
roman-photoz --setup
```
Options:
- --nobj N – number of objects in the simulated catalog (default: 1000).
- --simulated-catalog-filename FILE – simulated catalog filename (default: roman_photoz_simulated_catalog.parquet).
This leaves a catalog behind at $LEPHAREWORK/roman_photoz_simulated_catalog.parquet (it already has redshift estimates added by setup’s own model-building pass). This catalog isn’t meant to be your real input data – it’s a fallback you can use in step 3 below if you don’t have your own catalog yet (see “Creating a fresh simulated catalog” below for another option).

Note

roman-photoz --setup only needs to be run once per LEPHAREDIR/LEPHAREWORK location – it downloads the auxiliary data and builds the informer model, which are then reused for all subsequent roman-photoz runs. There’s no need to run it again unless you delete/change those directories.
Run roman-photoz. Assuming there is a folder named OUTPUT in the current working directory where the updated catalog containing the redshift results will be saved to:
```
roman-photoz \
  --input-filename <path/to/your_catalog.parquet> \
  --output-filename ./OUTPUT/<path/to/your_catalog.parquet>
```
--input-filename should point to your own input catalog. If you don’t have one, you can pass in the catalog created in step 2 instead: $LEPHAREWORK/roman_photoz_simulated_catalog.parquet.

Creating a fresh simulated catalog (optional)

If you’d rather not reuse the catalog from step 2 – for example, you want a different number of objects or different physical parameters – generate a fresh one:
```
roman-photoz-create-simulated-catalog \
  --output-filename roman_photoz_simulated_catalog.parquet \
  --refresh-lib-mag
```
To change the physical parameters used to generate the simulated catalog (e.g. redshift range/step), edit roman_photoz/default_config_file.py.