Deploying Potato
Choose how to get a Potato study to your annotators, from a no-terminal guide for researchers to one-command cloud deploys and your own Docker server.
To deploy a Potato study, you run the task on a server that your annotators can reach through a web address, decide who can sign in, and download the annotations when the study ends. Researchers who have never used a terminal should start with Put Your Potato Study Online, which walks through each route step by step.
A server here is a computer that stays switched on and connected to the internet, so annotators can open the task in a browser from anywhere. The Study Designer builds a task and downloads the project folder every route starts from. The Designer's preview runs only in your own browser, and nobody else can reach that copy.
Choosing how to deploy
The right page depends on who does the technical work and how long the study runs:
| Situation | Start here |
|---|---|
| You have never used a terminal, or you are not sure where to start | Put Your Potato Study Online |
| Your institution's IT or research computing team will host the study | Potato Server Requirements for Your IT Team, to send to the team |
| A pilot from your own laptop for an afternoon | Running Potato Locally or in the Cloud, the potato share section |
| A study on a rented cloud server, set up with one command | Running Potato Locally or in the Cloud, then the guide for AWS Lightsail, Jetstream2, Hetzner, Vultr and Linode, or Hugging Face Spaces |
| A server you already run, with Docker | Run Potato in Docker, then Production Setup and Reverse Proxy |
| Comparing hosts for a study that runs for weeks | Deploying an Annotation Server |
Sign-in
Who can sign in decides who can see and change the study. A list of usernames suits a known group of annotators, as Password Management describes. Institutional and Google, GitHub, or Hugging Face accounts go through SSO and OAuth, and crowdworkers arrive through the links described in Crowdsourcing Integration. Passwordless Login covers username-only access for demos and classroom exercises.
Who sees which items
Task Assignment decides which items each annotator sees and how many annotators label each item. Heterogeneous Annotator Coverage covers studies in which different annotators label different amounts, and Data Directory covers adding items while the server runs.
Results and shutdown
The admin dashboard at /admin shows progress while the study runs, and the dashboard's Publish Dataset page downloads the annotations as a CSV or JSONL archive. Download the results and check the download before you delete a server, because a deleted server takes the annotations with the server unless a backup holds a copy.