Skip to content
/ api Public

Automagically generate questions from a passage of text

Notifications You must be signed in to change notification settings

questo-ai/api

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

5 Commits
 
 
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

Questo API

A Note

After more than three years of working on Questo, the three of us have decided it's time to close the chapter on this part of our lives. Questo has been an amazing experience — we're grown together in ways we never would've thought possible, and become better developers for it. We'll be releasing a writeup shortly to accompany this, but for the time being here is all our source code for the backend.

The generators directory contains all the actual question generation technology, including our novel algorithm for generating "template" style (e.g Who, What, Where) questions.

The research folder contains a utility for remotely serving spaCy. This was very useful during development, because otherwise we'd have to keep reloading the model, wasting lots of time in the process.

The scripts folder contains the shell-script used to push our Docker images to the Kubernetes clusters hosted on Google Cloud.

The utilities folder contains the files needed to build the spaCy server's Docker image, as well as the Docker image for a keyword extractor.

Please feel free to reach out if you'd like to discuss the project further!

design principles

  • Only tabs.
  • Readable and self-documented code is a must. Use comments
  • ZMQ microservices are used for internal messaging, speeeeed baby
  • meaningful commit messages when possible 🤘🏼
  • no unilateral changes to API without conferring with iOS

how we split stuff up

Network Diagram

deployment

  1. travis is triggered if TEMPLATE_ONLY or GAPFILL_ONLY are in commit heads, else, if commits are made to stable.
  2. travis internally connects to gcloud, from which the latest versions of a generator are pulled, rebuilt & tested.
  3. pytest is used to test all generators and the API layer.
  4. travis will optionally deploy iff in branch stable.

how to develop

  1. clone this, use conda to create an environment for generators.
  2. run python -m spacy download en_core_web_sm for easy local development.
  3. if doing rigorous spaCy NLP, use the remote spaCy ZMQ service. refer to utilities/spacy-server.
  4. refer to design principles.

About

Automagically generate questions from a passage of text

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published

Contributors 3

  •  
  •  
  •