mitodl/lmod_proxy

Flask application for proxying requests to the MIT Learning Modules API from edx-platform

MIT is deprecating the Learning Modules application and we have stopped using this app. This repository is archived.

lmod_proxy

Flask application for proxying requests to the MIT Learning Modules API from edx-platform.

Quick Start

• Download the latest tar ball or clone from github.
• Run pip install -r requirements.txt
• Make sure you have htpasswd available/installed
• Run htpasswd -c ~/.htpasswd <your username>
• Create environment variable with path to password file: export LMODP_HTPASSWD_PATH=~/.htpasswd
• Run lmod_proxy
• Go to https://localhost:5000/ and use the form there to try things out

You will also likely need to customize your settings to be able to do authentication to LMod via a proper certificate. To do that, determine the path to your certificate and export that path for the configuration with a configuration environment variable: export LMODP_CERT=/path/to/mycert.pem. The certificate needs to be a plain (no passphrase) base64 encoded file with both your certificate and private key from MIT.

macOS tip (likely relevant for other OSes, too): You can export your MIT certificate from Keychain Access to a PKCS file (with the .p12 extension). Then you can convert it to PEM with the following command: openssl pkcs12 -in your_certificate.p12 -out your_certificate.pem -nodes

Running on Heroku

There is an included Procfile to run the application in heroku, to do so just create a heroku app, configure the appropriate environment variables in heroku via heroku config:set LMODP_...=.... The one catch being that your certificate file and HTPASSWD files probably shouldn't be checked in with the repository on heroku, so you can actually copy your entire HTPASSWD file into the LMODP_HTPASSWD variable and your certificate into LMODP_CERT_STRING variable.

Configuring edX Platform

To configure your edX platform to use this as your remote gradebook endpoint, you need to configure a few things, first you need to enable the feature and point it at the URL of your lmod_proxy server. Adding something like: FEATURES['REMOTE_GRADEBOOK_URL'] = 'https://<htpasswd_user>:<htpasswd_password>@myapp.herokuapp.com/edx_grades' to your config will enable the feature, and then in your course, you will just need to specify the GradeBook UUID to point at in your advanced settings with something like:

"remote_gradebook": {
  "name": "STELLAR:/project/mitxdemosite",
  "section": "devops01"
}

Development Notes

See the Development Notes

lmod_proxy's Issues

Normalize grades before submitting to Stellar

As an instructor, I would like to have grades submitted to the lmod gradebook be "out of" the max score for the assignment. Grades coming from edX are between 0 and 1.0 and it would be clearer for instructors and students to have them be between 0 and the max score.

• add a configuration setting to control whether grades should be normalized at all. In our current use of lmod-proxy this would be an application setting, similar to LMODP_APPROVE_GRADES
• fetch the max scores for all problems in the gradebook (possibly using get_assignments()
• normalize the grades coming in from edX before submitting them in L32-L35

This may be very slow and we'll need to consider the possibility of timeouts.

Update build for new library requirements

lmod_proxy needs maintenance to build properly due to new features in its imported libraries.

Setup initial project structure

Add nice index page

Add page that is descriptive of what the application does as well as offering a user interface to explore the features (i.e. calls to edx_grades "API")

Transmit raw score and max score to remote gradebook

Currently lmod_proxy sends calculated scores to the remote gradebook. The requested change is to send the raw score and the max score.

To send the max score, the assignment must be created by lmod_proxy. (In the existing implementation, assignments are created in the body of the PyLmod call that sends scores to the remote gradebook.)

Upgrade to Python 3

With the official end of life for Python 2 at the end of this year we should upgrade this app to run on Python 3

grade export shouldn't overwrite excused absences

a side issue, mentioned in ZenDesk #2001:

a warning: some instructors may have set excused absences, which the MITx upload presently overwrites

I'm a little surprised this happens, so this may require some investigation.

Flesh out README

Add quickstart, heroku setup, and configuration for edx-platform to README. Also badges.

Update versions in requirements.txt

Majority of pinned versions in requirements need to be updated.

Merging enrollment on non-existent section raises a 500

@itsbenweeks reports that merging or overloading enrollment from a course from a non-existent section causes a non-graceful 500 error.

Use non-root user for Dockerfile

conditionally set grades, to avoid overriding manually set grades

I'm not sure if this is an issue for lmod_proxy, or for pylmod.

I'd like to an option for setting grades to be conditional -- only set the grade if one hasn't been set already.

This would avoid overriding grades that have been set manually in the gradebook, and may obviate the immediate need for #5 (even though it's still a good idea).

Move deployment off of dot cloud

Dot cloud is going out of business. We need to find another host that won't timeout on a 60 second request.

Will require a deploy to staging and rp as well.

explicity set isGradeApproved when setting a grade

In the current version of Grades2Stellar, there is no value set for isGradeApproved when setting a student's grade.

This should be explicitly set to false or, even better, the API should pass a value through.

This will help us address issues like Zendesk #1435.

Hmmm. Maybe this needs to be an issue on https://github.com/mitodl/PyLmod

Basic Authentication

Support basic authentication (and the configuration of that), to allow for some security.

Add heroku/foreman support to project

Add relevant config to run as a foreman/heroku project

Use tox for tests

Migrate Travis -> Github Actions

Travis will be limited open source projects build time soon, so we need to transition to github actions.

Take a look at how we did it open discussions or bootcamps or micromasters

Move pylmod to setup.py when available

release 0.2.0

• Use release script.
• push to lmod_proxy_qa
• integration testing
• deploy to lmod_proxy

Test cmd.run_server

Currently this is untested, either mark it as uncovered or add a test

Add logging setup ala gitreload

API compatibility with grades2stellar

Add docker-compose volume to link /src and /tmp

Add configuration style

Use a "first found" yaml configuration file loader from one of three locations: $(pwd)/lmod_proxy.yml, ~/.lmod_proxy.yml, and then /etc/lmod_proxy.yml. Also allow environment variable overrides prefixed with LMODP_.

Create Continuous Integration Deployment

It is convenient for integration and system testing to have an installation of the application that is always running the latest master version of the repo that works. We can add this easily with Travis as a step to do after a successful test has occurred. It does this by packaging our app up into a heroku deployable unit, and pushing it.

Make certificate usable from environment variable

Currently LMODP_CERT takes the path to a cert, but for heroku, it really needs to take the whole environment variable. So similar to http://mit-lti-flask-example.readthedocs.org/en/latest/ lmod_proxy should also support this. This is already done for HTPASSWD in PR #15