User Data Ingest CLI

Command-line client for uploading data into LOFAR 2.0's Long-Term Archive. Built for projects with non-standard pipelines or independent processing resources that need to submit data while adhering to LTA schemas.

See the backend documentation for more context. There's also a web interface if you prefer that.

Getting Started

Python 3.13+ required (download here).

pip install udicli
udicli login

udicli is now on your PATH.

Commands

udicli login          Authenticate with your SRAM application token
udicli whoami         Show your account details
udicli status         Check server connectivity
udicli list-projects  View projects youre a member of
udicli connect        Configure access to your data storage
udicli ingest         Create and submit an ingest request
udicli logout         Remove cached credentials

For help on any command:

udicli <command> --help

Use -v flag with any command to see raw requests and responses:

udicli -v status

Authentication

login looks for your SRAM application token in this order:

1. --token flag: udicli login --token <your-token>

2. .ingestingrc file in your home directory with api_token=<your-token> macOS/Linux: ~/.ingestingrc Windows: C:\Users\<username>\.ingestingrc

3. Interactive prompt: provides a link to SRAM and asks you to paste the token

Once verified, the token is cached locally at ~/.config/user-ingest-cli/credentials-<env>, so you won't need to authenticate again.

Switching environments with --env dev or --env prod keeps sessions separate.

Ingesting Data

An ingest request has this structure: project → data products → files. Each file needs a checksum value and type.

Interactive Mode

Start with no arguments to be guided through the whole process:

udicli ingest

You'll select a project, site, checksum type, and build data products interactively.

Note:

You should provide (absolute or relative) file locations. For all the ways mentioned below you can use glob patterns like "data/*.m5", which selects all m5 files from folder data. For now the tool checks for * or ? in the path. It does not check [..] and does not support recursion.

Command-Line Flags

Provide all details via flags .

udicli ingest --project APPPP_001 --site surf --checksum-type MD5 --file-format MEASUREMENT_SET --product-type VisibilityDataProduct  --files test_files/file1.m5 --files test_files/file2.m5
udicli ingest --project APPPP_001 --site surf --checksum-type MD5 --file-format MEASUREMENT_SET --product-type VisibilityDataProduct  --files test_files/file1.m5 --files test_files/file2.m5
udicli ingest --project APPPP_001 --site surf --checksum-type MD5 --file-format MEASUREMENT_SET --product-type VisibilityDataProduct  --files test_files/file1.m5 --files /home/<user>/Downloads/file2.m5
udicli ingest --project APPPP_001 --site surf --checksum-type MD5 --file-format MEASUREMENT_SET --product-type VisibilityDataProduct  --files "test_files/*.m5"

Repeat --files for multiple files. Don't repeat other flags (they won't stack; only the last one counts).

Piping File Lists

Use --files-from - to read from stdin:

find test_files -name "*.m5" -mtime -1 | udicli ingest --project APPPP_001 --site surf --checksum-type MD5 --product-type VisibilityDataProduct --file-format MEASUREMENT_SET --files-from -
find test_files -name "*.m5"| udicli ingest --project APPPP_001 --site surf --checksum-type MD5 --product-type VisibilityDataProduct --file-format MEASUREMENT_SET --files-from -

CSV File

Pass a file with one path per line:

udicli ingest --project APPPP_001 --site surf --checksum-type MD5 --product-type VisibilityDataProduct --file-format MEASUREMENT_SET --files-from test_config_files/config3.csv

Empty lines in the file are ignored.

JSON Config File

For complex scenarios with multiple data products and filters:

{
  "project": "APPPP_001",
  "site": "surf",
  "checksum_type": "MD5",
  "data_products": [
    {
      "product_type": "VisibilityDataProduct",
      "format": "MEASUREMENT_SET",
      "files": [
        "test_files/file*.m5"
      ]
    },
    {
      "product_type": "VisibilityDataProduct",
      "format": "MS",
      "files": [
        "test_files/*.MS"
      ],
      "filters": {
        "min_size": "1000000",
        "max_size": 100000000,
        "modified_after": "2024-01-01",
        "exclude": [
          "*test*.m5",
          "*backup*.m5"
        ]
      }
    }
  ]
}

Run with:

udicli ingest --inputfile test_config_files/inputFile1.json
udicli ingest --inputfile test_config_files/inputFile2.json

Filters (JSON Config Only)

Filters refine file selection after glob patterns expand. All are optional; a file must pass every filter to be included.

"filters": {
  "min_size": 1000000,
  "max_size": 100000000,
  "modified_after": "2026-07-22",
  "include": ["file1.m5", "file2.m5"],
  "exclude": ["*test*.m5", "*backup*.m5"]
}

min_size and max_size accept numbers or strings (e.g., 1000000 or "1000000").

modified_after takes ISO 8601 format: "2026-07-22" or "2026-07-22T14:00:00".

include and exclude use glob patterns (*.m5, file?.m5). Character classes [..] and recursive globs ** are not supported.

Request Body Format

What gets sent to the server (for reference):

{
  "project_id": "DEV__P_C1_003",
  "site": "surf",
  "checksum_type": "MD5",
  "estimated_total_size_bytes": 1536770867,
  "data_products": [
    {
      "data_product_type": "VisibilityDataProduct",
      "files_format": "MEASUREMENT_SET",
      "files": [
        {"filename": "obs1.ms", "checksum_value": "abc123..."},
        {"filename": "obs2.ms", "checksum_value": "def456..."}
      ]
    }
  ]
}

With --noinput, the tool fails fast if required flags are missing. All validation errors are reported before any files are sent.

Piped input (e.g., find ... | udicli ingest ... --files-from -) automatically enables non-interactive mode.

Testing and Feedback

We're actively developing this tool. As you test, please share your thoughts on these questions:

1. What do your processing resources typically include and how do you connect to them?

2. In a typical cycle, how many files and data products would you upload?

3. Should duplicate files be automatically filtered out?

4. What's the usual folder and file structure of pipeline output? This helps us improve the ingest interface.

5. Would concurrent request management be useful? For example, draft one request, switch to another, then submit both when ready.

6. File format validation isn't implemented yet. We're waiting on the LOFAR Data Working Group to approve the list of valid data products and formats.

7. Should a data product with zero files be rejected with an error or allowed with a warning?

8. What validation rules matter for data you provide via flags or input files? Check numerical values, empty values, consistency between flags?

9. If a file path fails during processing (permissions, disappeared since resolving), should it be skipped with a warning or should the whole request fail?

Development

Clone and set up:

git clone <this-repo>
cd user-data-ingest-cli
python -m venv .venv
source .venv/bin/activate
make bootstrap
make init
pip install -e .

Runtime dependencies live in pyproject.toml's dependencies. requirements/base.txt is just a pinned lock file compiled from it, used to keep dev/CI environments reproducible; requirements/dev.txt adds dev-only tooling (pytest, ruff, black, mypy, pre-commit) on top and is compiled from requirements/dev.in.

After changing dependencies, recompile the lock files and reinstall:

make compile-requirements
pip install -r requirements/dev.txt
pip install -e .

Run linting, formatting, type-checking, and tests:

ruff check src tests
black src tests
mypy src
pytest

Deployment

Every pipeline builds the package (package_files) with an auto-incrementing version (0.1.<commit count> — no manual version action needed). From there, two manual jobs are available in the pipeline's publish stage:

Test deploy to GitLab — click publish_on_gitlab to start the pipeline. Uploads to this project's own Package Registry, authenticated automatically via CI_JOB_TOKEN

pip install udicli --index-url https://__token__:<your-personal-access-token>@git.astron.nl/api/v4/projects/1019/packages/pypi/simple

Release to PyPI — push a tag, then click publish_on_pypi:

git tag v0.2.0
git push origin v0.2.0

That job only appears on tag pipelines and uploads to the real pypi.org using a PYPI_TOKEN CI/CD variable (set under Settings → CI/CD → Variables, masked + protected). Once it's up:

pip install udicli