TamedTable is an AI-powered data preparation tool that allows users to transform tabular data using natural language commands. It eliminates the need for complex formulas or coding, making data manipulation accessible to a wider audience. Users can load various data formats including CSV, JSONL, Parquet, and Arrow, either locally or from a URL. The tool understands context, enabling it to perform cleaning, enrichment, classification, and validation tasks that go beyond traditional methods. All transformations are saved as steps, which can be replayed on new data or exported as Python scripts for automation. TamedTable offers a preview of changes with lazy AI execution, ensuring cost-effectiveness. Key features include:
- Natural Language Transformation: Describe desired data changes in plain language (e.g., English, Chinese) and TamedTable executes them.
- Contextual Understanding: The AI interprets the context of data rows to perform intelligent cleaning, normalization, and consistency checks.
- Data Enrichment and Extraction: Automatically fill missing information, extract structured data from free text, and infer relationships (e.g., country from city).
- Classification: Sort and categorize rows based on their meaning, sentiment, or other criteria.
- Validation: Flag rows with potentially incorrect or impossible data, such as fake emails or invalid dates.
- Language Processing: Summarize, translate, and classify text in any language, with voice command support.
- Standard Spreadsheet Operations: Includes deterministic operations like deduplication, joining tables, filtering, pivoting, and sorting.
- Load and Save Functionality: Supports loading from various sources and saving data in multiple formats (CSV, JSONL, Parquet, Arrow). Transformations can be saved as TamedTable scripts or Python scripts.
- Undo/Redo: All changes are tracked, allowing users to easily undo or redo any step without altering the original file.
- Source-Available and API Key Integration: Runs on the user's own API keys for enhanced privacy and control. The source code is available for inspection.
TamedTable provides a guided tour for users without an API key and is designed for trustworthy, repeatable data processing with no vendor lock-in.