The website

Try it now at plainml-tpua.vercel.app, with nothing to install. That's the in-browser version described below: everything runs on your own computer.

To run the website on your own machine:

pip install "plainml[web]"
plainml web

A browser tab opens at http://localhost:8765. Everything runs on your machine, and results are saved in the same runs/ folder the command line uses, so both see the same history.

Using it

1. Upload. Drag a CSV, Excel, Parquet, JSON or TSV file onto the page. You see every column's type (number, category, date, text, or not used) and a preview of the first rows.

   No data to hand? Under the drop zone are six made-up example datasets (churn, house prices, store sales, daily sales, customers and card payments). One click loads one, picks the task that suits it and fills in its settings, so you only have to press Run. 2. Choose a task.

   Task	What you get
   Predict a column	Models compared and the best one explained, as with plainml train
   Forecast over time	A forecast with an 80% range, optionally one per store/product, with inputs and holidays
   Find groups	Segments and what sets each apart
   Find unusual rows	Every row scored, the unusual ones flagged with reasons
   Rank the columns	Which columns matter, by several importance methods and a consensus
   Check for drift	Whether this data differs from a model's training data or another file
   Profile the data	Every column summarised, with warnings
   Clean the data	A tidied copy to download

   Each task has its main options up front and the rest under More options. 3. Watch it run. A progress bar and a live log show each step. Jobs run one at a time, in order. 4. Read the results. Headline numbers, the key findings in plain English, and the full report. 5. Download what you need. Every file the run wrote is listed with what it is and its size, each with its own Download button. Tables can be previewed first.

Runs lists everything you've run (from the website or the command line), with search and filters. Predict uses any saved model on a new upload, or continues a forecast.

Options

plainml web --port 9000                   # another port
plainml web --runs-dir projects/churn     # keep runs somewhere else
plainml web --max-upload-mb 2000          # allow bigger uploads (default 500 MB)
plainml web --no-browser                  # don't open a tab

Sharing it on a network

By default only your own machine can open the site. To let colleagues use it:

plainml web --host 0.0.0.0 --token SOMETHING-LONG-AND-RANDOM

Visitors enter the token once (their browser remembers it for 30 days). Anyone with it can upload data and run jobs, so share it only on a network you trust, or put the site behind HTTPS (for example a reverse proxy). The token can also come from the PLAINML_WEB_TOKEN environment variable, which keeps it out of your shell history.

In the browser, with no server

plainml web --export site

This writes the same website as static files. Opened in a browser, it starts Python (Pyodide) in the background, installs plainml, and runs every task on the visitor's own computer, so their data is never uploaded: - The first visit downloads about 50 MB of Python libraries; after that they're cached. - Runs are saved in that browser. - Downloads are saved straight from the page. - Uploads are limited to 200 MB, and training uses one CPU core, so very large jobs are better with the command line or the server version.

Host the folder anywhere that serves static files. The repository's vercel.json publishes it on Vercel together with these docs. See Hosting for the steps, including a free Hugging Face Static Space.

Putting the server version online

The server version needs a host that runs an always-on container, such as Render, Railway or Fly.io. The repository includes a Dockerfile for it. See Hosting.

Embedding it

The site is a FastAPI app, so you can mount it inside another one:

from plainml.web.server import create_app

app = create_app("runs", token="change-me")