AVICA: Automated VLBI pipeline in CASA

Accepted for publication in Astronomy & Astrophysics journal.

Complete demo: avikhagol.github.io/avica-demos

Contents

• Installation
  • Recommended installation
  • Manual installation
  • Full installation script
• Usage
  • Pipeline
  • Manipulating FITS-IDI
  • Configuration
    • Parameter summary
    • Cleaning up intermediate data
• Attribution
• Acknowledgement

Installation

Requirements:

• Ubuntu 18.04+, Debian 10+, or RHEL/CentOS 8+
• Python >= 3.9

The avica package is publicly available on PyPI. Use uv or pipx for an isolated command-line installation.

Recommended installation

Using uv:

uv tool install avica --python 3.10

Using pipx:

pipx install avica

Using pip:

pip install avica

If you install with pip, use a virtual environment unless you already manage Python packages another way.

Manual installation

Clone the repository and install it locally:

git clone https://github.com/avikhagol/avica.git
cd avica/
pip install .

Full installation script

Run install.sh as your normal user. The quickest way is to fetch and run it in one step:

curl -LsSf https://avikhagol.github.io/avica/install.sh | bash
source "$HOME/.local/share/avica-stack/env.sh"

requires rsync, git

The script installs AVICA using uv. If picard is already on PATH, it reuses that installation and skips all CASA/rPICARD downloads and plotting/data setup. Otherwise, it installs rPICARD, jiveplot (with python-pgplot==1.6.1), and monolithic CASA.

For a new rPICARD installation, set CASA_PATH to an existing CASA installation directory (containing bin/casa and bin/mpicasa) or an archive URL:

CASA_PATH=/path/to/casa bash install.sh

Using the one-liner, place the variable after the pipe so it reaches bash, not curl:

curl -LsSf https://avikhagol.github.io/avica/install.sh | CASA_PATH=/path/to/casa bash

If unset, the script prompts for the location. Press Enter to download:

ftp://ftp.mpifr-bonn.mpg.de/outgoing/mjanssen/casa-6.7.5-18-py3.12.el8.tar.xz

Without an interactive terminal, the same default is used. CASA_DIR and CASA_URL are also accepted, after CASA_PATH in that order. Set AVICA_INSTALL_DIR to change the default ~/.local/share/avica-stack directory, or PICARD_REF to select the branch/tag of a new rPICARD clone. Choose a CASA version that matches the selected rPICARD version.

When setting up rPICARD, the downloaded or supplied CASA directory is also saved in AVICA's global configuration using avica pipe config --global casadir=....

The script adds an environment file to ~/.bashrc. It configures AVICA's CASA and input-template paths when it can identify them, preserving other settings and backing up an existing ~/.avica/avica.inp. For an existing picard command, it reads the adjacent rPICARD your_casapath.txt when available; otherwise, existing AVICA settings are left unchanged.

Run bash install.sh --help (or curl -LsSf https://avikhagol.github.io/avica/install.sh | bash -s -- --help for the one-liner) for the environment options. The previous apt and dependency-check options have been removed.

Usage

The pipeline calibration steps rely on rPicard. Follow the rPicard setup instructions first. After rPicard is configured, AVICA only needs a minimal configuration file to get started; see Configuration.

Pipeline

Run the default pipeline:

avica pipe run --fitsfilenames <file1.uvfits,file2.uvfits> --target <source-name>

The default pipeline executes these steps:

• preprocess_fitsidi
• fits_to_ms
• phaseshift
• avica_avg
• avicameta_ms
• avica_snr
• avica_fill_input
• avica_split_ms
• rpicard

You can pass one or more step names to run only part of the pipeline:

avica pipe run preprocess_fitsidi fits_to_ms --fitsfilenames <file.uvfits>

During fits_to_ms, AVICA also discovers station flag files in artifact_dirs, folder_for_fits, the input FITS directories, and <workdir>/raw. AIPS UVFLG files using ANT_NAME (such as EVN .flag files) are converted and added after the existing FITS-IDI/MS flags.

apply_flag_from_artifacts = True
artifact_flag_extensions = .fg;.uvflag;.uvflg;.uvfg;.flag;.flg;.uvflags;.uvflgs;.uvfgs;.flags;.flgs
artifact_flagfiles = []

Extensions are case-insensitive. A nonempty artifact_flagfiles list overrides automatic discovery and accepts arbitrary filenames. Set apply_flag_from_artifacts=False to disable this pass independently of apply_flag_from_idi. Existing measurement sets are flagged only when apply_flag_to_existing_vis=True.

Supported UVFLG fields are ANT_NAME, TIMERANG, OPCODE='FLAG', REASON, TIMEOFF, DTIMRANG, BIF/EIF, and BCHAN/ECHAN. TIMEOFF and DTIMRANG are in seconds and retain their nonzero settings between entries, following AIPS UVFLG INTEXT semantics. Malformed files and records with unsupported selectors, invalid IF/channel ranges, absent antennas, or unrelated times are reported and skipped.

Generated commands are saved as <MS>.artifact_flags.flagcmd; its .json sidecar records input files, row counts, skipped records and application status. Each application saves uniquely named before_artifact_flags_* and after_artifact_flags_* versions using CASA flagmanager. If application fails, the step reports failure and the before-version remains available for restoration. Source flag files are read in place and are not copied into rPICARD directories.

Common options:

Pipeline results

After each step completes, AVICA appends a row to reductions/<target>_result.csv. Use avica pipe result to render that file:

avica pipe result --target <source-name>

This prints a progress ladder — one row per step — with status, success/fail counts, duration, and a condensed note for any failures. Full failure traces appear in panels below the table.

Common options:

The default view collapses to the most recent attempt per step; --history shows all attempts. The footer prints the command to resume from the next incomplete step.

Manipulating FITS-IDI

Check FITS-IDI files for known issues:

avica fitsidi_check <file.uvfits>

Useful options:

Example output:

avica fitsidi_check VLBA_VSN005412_file3.uvfits
+--------------------+---------+-------+-------+----------------+----------+
| hdu                | fixable | total | fixed | problem_code   | affected |
+==========================================================================+
| ARRAY_GEOMETRY     | 0       | 8     | 0     | []             | []       |
| ANTENNA            | 0       | 16    | 0     | []             | []       |
| FREQUENCY          | 0       | 8     | 0     | []             | []       |
| PHASE-CAL          | 0       | 12    | 0     | []             | []       |
| PRIMARY            | 1       | 10    | 0     | ["extra_byte"] | [""]     |
| SOURCE             | 0       | 8     | 0     | []             | []       |
| FLAG               | 0       | 12    | 0     | []             | []       |
| UV_DATA            | 0       | 8     | 0     | []             | []       |
| GAIN_CURVE         | 0       | 8     | 0     | []             | []       |
| SYSTEM_TEMPERATURE | 0       | 8     | 0     | []             | []       |
+--------------------+---------+-------+-------+----------------+----------+

List observation information:

avica listobs <file.uvfits>

Configuration

The pipeline configuration is a key=value file. By default, AVICA looks for avica.inp in the current directory. See the example configuration for a minimal setup.

Pass a custom configuration file with --configfile:

avica pipe run --configfile <path/to/config/file>

If --configfile is not provided, AVICA uses avica.inp by default.

To store defaults persistently, merge an existing file into ~/.avica/avica.inp:

avica pipe config --default --inpfile <path/to/avica.inp>

You can also set default values directly:

avica pipe config --default key=value key2=value2 key3=value3

To set global defaults in AVICA's installed directory, use --global in the same way:

avica pipe config --global --inpfile <path/to/avica.inp>
avica pipe config --global key=value key2=value2 key3=value3

Each of these commands only touches one file: --default writes to ~/.avica/avica.inp, --global writes to the installed avica.inp, and plain avica pipe config writes to the local avica.inp (or --outfile). Only the keys you give it are changed; everything else already in that file — other settings, comments, # str/# int notes — stays as it was. A backup of the old file is saved as avica.inp.bak.

To supersede the rPicard input_template files (array.inp, observation.inp, array_finetune.inp, flagging.inp, constants.inp) used by the rpicard step, point picard_input_template_update at a folder containing the parameters you want to fix:

picard_input_template_update   =   "path/to/folder/with/fixed/inp/files"

Parameter summary

Use --summary to print a report of every pipeline parameter, its resolved value, and where that value came from (global, user, inpfile, cli, built-in default, or runtime context). The /core and /step suffixes distinguish general parameters from step-specific overrides:

avica pipe config --summary --inpfile <path/to/avica.inp>

The summary overlays the installed global avica.inp, then ~/.avica/avica.inp, then the local avica.inp (or an explicit --inpfile), with command-line key=value overrides applied last. Later layers override matching keys and preserve other settings. --no-inpfile skips automatic local-file discovery; an explicit --inpfile is still used. Summaries do not write configuration files.

Cleaning up intermediate data

Each pipeline step can remove its own intermediate files (temporary files, superseded Measurement Sets, etc.) once it finishes. This is controlled by:

Several steps ship with sensible defaults, e.g.:

delete_removables            =   True
preprocess_fitsidi.removables    =   ["raw/*.tmp"]
rpicard.removables               =   ["wd_[SLKQXPD]/VLBI_*.ms"]
fits_to_ms.removables            =   ["*.old"]

Deletion is always confined to the step's working directory; patterns that resolve outside it are ignored.

Attribution

When using AVICA, please add a link to this repository in a footnote.

Acknowledgement

AVICA was developed within the "Search for Milli-Lenses" (SMILE) project. SMILE has received funding from the European Research Council (ERC) under the HORIZON ERC Grants 2021 programme (grant agreement No. 101040021).