sliq 0.2.0

Sliq automatically fixes dataset schema issues, missing values, duplicate rows, and formatting errors so data professionals can get clean data in minutes, not hours or days.

Sliq Python Library

Data cleaning, made effortless.

Sliq is a Python library that provides a simple interface to clean your datasets using AI. With just a few lines of code, you can clean messy data, handle missing values, fix inconsistencies, and prepare your data for analysis or machine learning.

Installation

pip install sliq

Quick Start

Clean a file

import sliq

# Clean a CSV file and save the result
sliq.clean_from_file(
    api_key="your-api-key",
    dataset_path="path/to/your/data.csv",
    dataset_name="My Dataset",
    save_clean_file_path="path/to/output/",
)

Clean a DataFrame

import sliq
import pandas as pd

# Load your data
df = pd.read_csv("path/to/your/data.csv")

# Clean and get the result as a DataFrame
cleaned_df = sliq.clean_from_dataframe(
    api_key="your-api-key",
    dataframe=df,
    dataset_name="My Dataset",
    is_return_dataframe=True,
)

Features

• Simple API: Just two functions to clean any dataset
• Multiple formats: Supports CSV, JSON, JSONL, Excel, and Parquet files
• DataFrame support: Works with both Pandas and Polars DataFrames
• AI-powered: Uses advanced AI to understand and clean your data
• Customizable: Provide context about your data for better cleaning
• Secure: Your data is processed securely and deleted after cleaning

Supported File Formats

Format	Extension
CSV	.csv
JSON	.json
JSONL	.jsonl
Excel	.xlsx, .xls
Parquet	.parquet

API Reference

clean_from_file()

Clean a dataset from a file path.

sliq.clean_from_file(
    api_key: str,                          # Required: Your Sliq API key
    dataset_path: str,                     # Required: Path to your dataset file
    dataset_name: str,                     # Required: Name for your dataset
    save_clean_file_path: str = "",        # Path to save the cleaned file
    save_clean_file_name: str = "",        # Custom name for the saved file
    output_format: str | None = None,      # Output file format (see below)
    dataset_description: str = "",         # Description of your data
    dataset_purpose: str = "",             # What the data will be used for
    column_guide: dict[str, str] = None,   # Column name to description mapping (see below)
    data_source: str = "",                 # Where the data came from
    user_instructions: str = "",           # Special cleaning instructions
    is_feature_engineering: bool = False,  # Enable feature engineering
    is_detailed_report: bool = False,      # Generate detailed report
    recipe_id: str | None = None,          # Cleaning recipe ID (see below)
    verbose: bool = False,                 # Show progress messages
)

clean_from_dataframe()

Clean a Pandas or Polars DataFrame.

cleaned_df = sliq.clean_from_dataframe(
    api_key: str,                          # Required: Your Sliq API key
    dataframe: pd.DataFrame | pl.DataFrame,# Required: Your DataFrame
    dataset_name: str,                     # Required: Name for your dataset
    output_format: str | None = None,      # Output file format (see below)
    dataset_description: str = "",         # Description of your data
    dataset_purpose: str = "",             # What the data will be used for
    column_guide: dict[str, str] = None,   # Column name to description mapping (see below)
    data_source: str = "",                 # Where the data came from
    user_instructions: str = "",           # Special cleaning instructions
    is_feature_engineering: bool = False,  # Enable feature engineering
    is_detailed_report: bool = False,      # Generate detailed report
    recipe_id: str | None = None,          # Cleaning recipe ID (see below)
    is_return_dataframe: bool = True,      # Return cleaned DataFrame
    verbose: bool = False,                 # Show progress messages
)

Output Format

The output_format parameter controls the file format of the cleaned dataset:

• Valid formats: .csv, .json, .jsonl, .xlsx, .parquet
• Default behavior: If output_format is None or an empty string, the cleaned dataset is saved in the same format as the input file
• The format string can be specified with or without the leading dot (e.g., both "csv" and ".csv" are valid)

Example:

# Save cleaned dataset as Parquet regardless of input format
sliq.clean_from_file(
    api_key="your-api-key",
    dataset_path="data/raw.csv",
    dataset_name="My Dataset",
    output_format=".parquet",
    save_clean_file_path="data/cleaned/",
)

Column Guide Validation

When providing a column_guide, the following rules apply:

1. All columns required: You must provide descriptions for all columns in your dataset
2. Column count must match: The number of keys in column_guide must equal the number of columns in your dataset
3. Names must match: Column names are validated against the dataset (case-insensitive, underscore-insensitive)

Example:

# For a dataset with columns: "first_name", "last_name", "Age"
column_guide = {
    "first_name": "Customer's first name",
    "last_name": "Customer's last name", 
    "age": "Customer's age in years",  # Case-insensitive: "age" matches "Age"
}

If column names don't match, you'll get a helpful error message listing the mismatched columns and the actual dataset columns.

Recipe-Driven Cleaning

You can apply a saved cleaning recipe to your dataset by providing a recipe_id. Recipes are pre-configured cleaning instructions created through the Sliq web application that define column-level rules, global cleaning standards, and schema transformations.

Find your Recipe ID at sliqdata.com/dashboard/recipes.

# Apply a saved recipe to a new dataset
sliq.clean_from_file(
    api_key="your-api-key",
    dataset_path="data/raw_sales.csv",
    dataset_name="Sales Data Q4",
    recipe_id="r-6hchd6160aj3c77c4c",
    save_clean_file_path="data/cleaned/",
)

When a recipe_id is provided:

• The cleaning pipeline fetches the recipe from the Sliq platform
• Column descriptions embedded in the recipe are merged into the column guide
• Recipe rules guide the AI cleaning agents for consistent, reproducible results

Note: You can still provide column_guide, user_instructions, and other parameters alongside a recipe. User-provided column descriptions take precedence over recipe-derived descriptions.

Getting an API Key

To use Sliq, you need an API key. Visit sliqdata.com to get your API key.

Documentation

For full documentation, visit sliqdata.com/docs.

License

MIT License - see LICENSE for details.

For developers of the library

---

To install Sliq locally for development

pip install -e "Sliq-python-lib"

Publishing to PyPI

To upload a new version of Sliq-python-lib to PyPI:

1. Check your files:

   • Ensure pyproject.toml and README.md are present and correct.
   • Source code should be in src/sliq/.
   • License and metadata should be set.

2. Build the distribution: In your project root (Sliq-python-lib), run:

   python -m build
   

   (If you don’t have build, install it: pip install build)

3. Check the distribution: Optionally, check for errors:

   twine check dist/*
   

4. Upload to PyPI: Make sure you have a PyPI account and twine installed (pip install twine). Then run:

   twine upload dist/*
   

If you get an error, check the message and resolve it, or ask for help.