unite

*Unified liNe Integration Turbo Engine

unite is a Python package for Bayesian inference of emission lines from astronomical spectra. It is built on JAX, NumPyro, and Astropy, and supports fitting multiple spectra simultaneously with shared kinematics, calibration tokens, and flexible priors.

Key features:

• Fast pixel integration of line profiles (Gaussian, Voigt, Hermite, and more) with analytic CDF-based integration or numerical LSF convolution for exact absorption modelling

• Simultaneous fitting across multiple spectra and gratings

• Shared kinematic parameters (redshift, FWHM) across lines and components

• Dependent priors with arbitrary-depth parameter chains

• Calibration tokens (flux scale, resolution scale, pixel offset) with free or fixed priors

• YAML serialization for reproducible, human-editable configurations

• Instrument support for JWST/NIRSpec (all gratings + PRISM), SDSS, and generic spectrographs

• User-controlled sampler — ModelBuilder returns (model_fn, model_args) for use with any NumPyro backend (NUTS, SVI, nested sampling, …)

---

Getting Started

• Installation
• Core Concepts
• Contributing

Tutorials

• NIRSpec Fitting
• Full Generic Workflow
• Combining Configurations
• Computation times

Usage

• Priors
  • Supported Priors
    • Uniform
    • TruncatedNormal
    • Fixed
  • Dependent Priors
    • Basic Example
    • Supported Arithmetic
    • Deep Dependency Chains
    • Using with TruncatedNormal
  • Topological Sorting
  • Serialization
• Line Configuration
  • Creating a Line Configuration
  • Adding Lines
    • Basic Usage
    • Adding a Single Line
    • Adding Multiple Lines
  • Parameter Sharing
    • Shared Kinematics
    • Independent Parameters
    • Parameter Names
  • Dependent Priors
  • Line Profiles
    • Gaussian (default)
    • PseudoVoigt
    • Cauchy (Lorentzian)
    • Laplace (Exponential)
    • SEMG — Symmetric Exponentially Modified Gaussian
    • GaussHermite
    • SplitNormal
    • SkewNormal
    • SkewVoigt
    • BoxGauss
    • GaussianSplitLaplace (Asymmetric EMG)
  • Merging Configurations
  • Absorption Lines
    • Adding Tau-Parametrized Lines
    • Tau vs Flux
    • Sharing Tau Parameters
    • Mixing Emission and Absorption
    • Depth Ordering (zorder)
  • Serialization
• Continuum Configuration
  • Creating a Continuum Configuration
    • Automatic from Line Centers
    • Manual Regions
    • Combining Continuum Configurations
    • Depth Ordering (zorder)
  • Continuum Parameters and Priors
    • Custom Priors on Continuum Parameters
    • Parameter Naming
    • Sharing Parameters Across Regions
  • Continuum Forms
    • Linear
    • PowerLaw
    • Polynomial
    • Chebyshev
    • Clamped BSpline
    • Bernstein
    • Blackbody
    • ModifiedBlackbody
    • AttenuatedBlackbody
    • Template
  • Serialization
• Instruments & Spectrum Loading
  • Dispersers
    • Calibration Tokens
    • JWST/NIRSpec
    • SDSS
    • DESI
    • Generic Dispersers
    • InstrumentConfig & Serialization
  • Loading Spectra
    • from_edges
    • from_centers
    • from_arrays
    • from_DJA
    • from_sdss_fits
    • from_desi_fits
  • Mixing Instruments
• Configuration Serialization
  • The Configuration Container
  • Saving and Loading
    • Top-Level Configuration
    • Sub-Configuration Serialization
  • YAML Format
    • Dependent Priors in YAML
    • Dispersers in YAML
  • Combining Configurations
    • Combining Full Configurations
    • Combining Sub-Configurations
    • Handling Missing Sub-Configs
    • Strict Mode — Name Collisions
    • Combined Config Serialization
  • Round-Trip Guarantee
  • Editing Configurations Offline
  • Dict Interface
• Building the Model
  • Spectra Collection
    • Canonical Wavelength Unit
  • Coverage Filtering
  • Flux and Error Scales
    • Continuum Fitting
    • Flux Scale
    • Error Scaling
    • Inspecting the Continuum Fit
    • Manual Scale Override
  • Building the Model
    • Integration Mode
    • Component Depth Ordering (zorder)
• Sampling & Optimization
  • Setup & Performance
    • JAX 64-bit Mode
    • GPU Acceleration
  • Running the Model
  • Inference Methods
    • NUTS (No-U-Turn Sampler)
    • SVI — Stochastic Variational Inference
    • Using SVI to Initialize NUTS
    • Nested Sampling with JAXNS
• Results and Output
  • Parameter Table
    • Full posterior (default)
    • Percentile summaries
    • Column units
    • Rest equivalent widths
    • Table metadata
  • Per-Spectrum Model Tables
    • Columns
    • Array shapes
    • NaN separators between regions
    • Saving spectra tables to FITS
  • FITS Output
    • HDU structure
    • Primary header keywords
    • Reading the FITS file
  • Evaluating the Model at Arbitrary Samples
  • Model Diagnostics
    • Degrees of Freedom
    • Reduced Chi-Square
    • Log-Likelihood and Log-Posterior
    • WAIC
    • Going further with ArviZ

API Reference

• API Reference

Derivations

• Gauss-Hermite
• SEMG
• Polynomial
• Skew Voigt
• Skew Normal

Indices and Tables

• Index

• Module Index