Hugging Face has introduced AI Sheets, an open-source, no-code tool designed to help users build, transform, and enrich datasets with AI models. The tool offers a spreadsheet-style interface, allowing users to interact with data and AI models without writing code.

AI Sheets can be run directly in the browser via Hugging Face Spaces or deployed locally from its GitHub repository. It connects seamlessly with the Hugging Face Hub, giving access to thousands of open models through inference providers or locally hosted models, including OpenAI’s gpt-oss amongst many others.

Key Features

1. No-code interface: Work with datasets in a familiar spreadsheet layout.
2. Model flexibility: Compare multiple models side-by-side, refine prompts, and validate results directly in cells.
3. Data transformation: Clean text, classify entries, extract key points, and enrich records with missing details.
4. Synthetic dataset generation: Create test datasets for research or prototyping.
5. Feedback loop: Improve AI outputs by editing or approving generated cells, which are then used as examples for future generations.
6. Import/export support: Load CSV, TSV, XLS, or Parquet files (up to 1,000 rows) and export completed datasets to the Hub for reuse.

Example Use Cases

1. Comparing AI model outputs for the same prompt.
2. Classifying and tagging large text datasets.
3. Generating realistic placeholder data while preserving privacy.
4. Enriching addresses with missing ZIP codes (with web search enabled).
5. Testing prompt variations for customer service automation.

You can try AI Sheets instantly here or explore its source code on GitHub.