Docs

How this actually works

What runs where, and how to call the API yourself. Everything below was run against the live service before it was written down.

The shape of it

The site is a Next.js app on Vercel. 50 tools sit on top of one FastAPI backend running in a Docker container on a free Hugging Face Space, which currently exposes 116 endpoints.

Ten tools never touch the backend at all — they run entirely in your browser. Which ones, and what happens to anything you upload to the rest, is set out on the Privacy & Terms page. The request path from browser to model, including the security layer in between, is diagrammed on the home page.

Calling the API

There is no API key and no account. The base URL is:

https://wram1708-ml-unified.hf.space

A request that works right now, and what it returns:

curl https://wram1708-ml-unified.hf.space/health

{"status":"ok","models":["80-cereals","diabetes","insurance",
                         "iris","my-automl-model","optuna-run","titanic"]}

Most tool endpoints are POST and take either JSON with a base64-encoded file or a multipart upload. Rather than list all 116 here and let the list rot, use the generated reference below — it is produced from the running code, so it cannot drift.

Full endpoint reference

FastAPI publishes an interactive reference with every route, its request schema and its response shape. You can send test requests straight from the page:

Interactive API reference (Swagger UI) →

OpenAPI schema (JSON) → — for generating a client.

Limits you will hit

  • Rate limits. 60 requests per minute per client, dropping to 10 per minute on endpoints backed by a paid language model. Over the limit returns 429.
  • Request size.Bodies over 10 MB are rejected with 413, and several upload routes cap lower than that.
  • Browser origin. Requests carrying an Origin header from a site that is not allow-listed get 403. Server-side calls with no Origin header — curl, scripts — are unaffected.
  • Cold starts. The Space sleeps when idle, so the first request after a quiet spell takes a few seconds.
  • Daily budget. Endpoints that call a paid model have a per-day cap, after which they return 429 until it resets. This keeps the service free to run.

Source

Backend and frontend are both public on GitHub. Recent changes are listed on the changelog.