About ViewKit

ViewKit is a free tool for looking inside scientific and machine-learning data files, right in your browser. Drop in a file and inspect it as a table, a heatmap or a chart, without installing anything and without uploading your data anywhere.

What ViewKit is

One viewer for the file formats researchers and data scientists work with every day: HDF5, Parquet, Zarr, Apache Arrow and Feather, CSV and TSV, Excel, JSON Lines, NumPy, safetensors, TFRecord, WebDataset and audio. Everything runs as a static web app inside your browser tab, so there is nothing to install and no account to create.

Your data never leaves your browser

When you open a file, it is read and parsed entirely on your device using WebAssembly and JavaScript. Its contents are never uploaded, transmitted, logged or stored on any server. You can disconnect from the internet after the page has loaded and keep working. The only information that leaves your device is anonymous usage analytics, which never includes any part of your files. See the Privacy Policy for the details.

Who it is for

ViewKit is built for ML researchers, data scientists and anyone who needs to quickly and privately check what is inside a data file: the shape of an array, the schema of a table, or the contents of a dataset shard.

Why we write about formats

Most of the questions that bring people here are not about the tool. They are about the formats: why a Parquet read is slow, what the difference between float16 and bfloat16 costs, how WebDataset groups files into samples. We publish articles on those questions, written from the specifications and citing them, because the answers are useful whether or not you ever open the viewer.

Get in touch

ViewKit is an independent project, and the list of supported formats grows based on what people ask for. To request a format, report a bug or ask a question, visit the Contact page or email hello@viewkit.app.