GETTING STARTED / INSTALLATION

Install the core.
Add what you need.

The default package is local and offline. Model evaluation, the dashboard, AI planning, and contributor tools are opt-in extras so a data-cleaning script does not install a full platform.

Install now →

1. Install DATADOC

Use Python 3.10, 3.11, or 3.12. Create a virtual environment for each project so the pipeline version is explicit.

Terminalrecommended
python -m venv .venv
python -m pip install --upgrade pip
python -m pip install datadoc-cli

Or with uv:

uvfast environment setup
uv venv
uv pip install datadoc-cli

2. Choose optional extras

ExtraInstallAdds
Coredatadoc-cliProfile, plan, fit, transform, export, CSV, and Parquet.
MLdatadoc-cli[ml]scikit-learn baselines and evaluation reports.
UIdatadoc-cli[ui]FastAPI, Uvicorn, and multipart support for the local dashboard.
AIdatadoc-cli[ai]Optional advisory planning through LiteLLM. It never executes code.
Devdatadoc-cli[dev]Pytest, Ruff, build tooling, and scikit-learn for contributors.
Terminalcommon project setup
python -m pip install "datadoc-cli[ml,ui]"

3. Verify the install

Check that the command is available and run a read-only profile on your own data.

Terminalno data is modified
datadoc --help
datadoc version
datadoc profile data.csv
Expected behavior

The profile command prints JSON-like output containing the schema, null counts, cardinality, roles, findings, duplicate count, and schema fingerprint. It does not write a cleaned dataset.

4. Input and output formats

CSV

Comma-separated files

Read with Polars schema inference. Use a stable header and consistent types between train and inference.

PARQUET

Typed columnar files

Preserves types more reliably and is recommended for transformed feature outputs.

LIMIT

Tabular only

Free text, image, joins, geospatial, and distributed workflows are reported as boundaries rather than guessed.

5. Troubleshooting

“Only CSV and Parquet files are supported.”

Convert spreadsheets or JSON upstream, or export them as CSV/Parquet with an explicit schema.

“Evaluation requires optional ML dependencies.”

Install python -m pip install "datadoc-cli[ml]" in the same environment that owns the datadoc command.

Command shows an old version

Activate the intended virtual environment, reinstall the package, and check where datadoc on Windows or which datadoc on macOS/Linux.

DiagnosticsWindows + Unix
python -c "import datadoc; print(datadoc.__file__)"
where datadoc  # Windows
which datadoc   # macOS / Linux