This repository is a scaffold for implementing flexible human evaluations. It uses:
- React.js for implementing the front-end.
- Node.js for implementing the back-end.
- Next.js for coordinating front-end and back-end, running the server, etc.
- MongoDB as a database.
- MaterialUI as a React component library.
First, run a MongoDB instance:
mkdir db
mongod --dbpath=./dbThe application can be executed in development mode with:
npx next devThen, open http://localhost:3000 to see the result.
Each stage of the app is a "page" implemented in the pages directory.
The pattern is that each page interacts with the user, records relevant data,
then moves on to the next.
Back-end API functions are under pages/api. These can choose which content
is given to each user at what stage, and also record interactions to the database.
lib contains library functions. lib/component contains React components that are
only available in the client. Pages dynamically load these components when needed.
(this separation is necessary because of boring reasons related to server-side rendering and Next.js).
lib/data.js contains the database model for a user session. Add stuff you want to record
here, and see how the existing library functions retrieve and edit user sessions.
For deploying your human study, first push the code to the server where you'll host it, and then build it in production mode there:
npx next build
# Start the production server on an open address on the specified port.
npx next start -H 0.0.0.0 -p 8888You can get your data out as a simple JSON file by running mongoexport on the server:
mongoexport --db your-db --collection usersessions --jsonArray --out=my-data.jsonThat's it!