Courtbot

Courtbot is a simple web service for handling court case data. It offers a basic HTTP endpoint for integration with websites, and a set of advanced twilio workflows to handle text-based lookup.

Specifically, the twilio features include:

• Requests. If a case requires a court appearance, the app allows users to sign up for reminders, served 24 hours in advance of the case.
• Unmatched Cases. If a case isn't in the system (usually because it takes two weeks for paper citations to be put into the computer), the app allows users to get information when it becomes available. The app continues checking each day for a number of days (set by config QUEUE_TTL_DAYS) and sends the case information when found (or an apology if not).

Datamodel

The main features of the app use three tables in a PostgreSQL database:

1. hearings | This table has the data about upcoming cases. It is recreated each time runners/load.js is exectued from the csv files found at urls set in config variable DATA_URL. It is ephemeral — it is recreated from scratch every day so the app must be prepared for cases that are there one day and not there the next. It is possible for the CSV to have duplicate rows. The load script enforces unique case_ids.
2. requests | This table stores the requests for notifications. Each row requires a phone number, which is encrypted using config PHONE_ENCRYPTION_KEY, and a case_id. The table also has columns known_case and active. known_case allows the app to distinguish between cases that we have seen at some point in the hearings table. Requests for cases where known_case is false will expire after QUEUE_TTL_DAYS and active will be set to false. If the case appears at anytime before that known_case will be set to true and the request will not expire unless a user manually turns it off by texting DELETE after sending the case. The requests table uses the column updated_at to determine if an unmatched case has expired rather than created_at. These will generally be the same, but it allows for the future possibility of allowing unmatched cases to be extended.
3. notifications | Rows in the notifications table are added whenever the app sends the user a notification. These can include notifications the day before a hearing or notifications that an unmatched case was not found within QUEUE_TTL_DAYS. The table has columns for case_id and phone_number which link the case to the person recieving the notification. It also has the following columns:
   • created_at timestamp, which should correspond to the time the notification is sent
   • event_date the date of the hearing at the time the notification was sent. This may or may not be the date in current versions of the csv as this changes frequently.
   • type enumeration to distinguish between hearing notifications [reminder], matched [matched] cases, and expired cases that were not found within QUEUE_TTL_DAYS [expired]
   • error an error string if sending a notification failed (perhaps due to a twilio error or bad phone number).

See sendReminders.js and sendUnmatched.js for examples of SQL using these tables.

The database also has tables log_hits and log_runners. These log activity of the app.

Running Locally

First, install node (atleast version 7.6), and postgres (at least version 9.5).

Then clone the repository into a folder called courtbot:

git clone git@github.com:codeforanchorage/courtbot.git courtbot
cd courtbot

Since the app uses twilio to send text messages, it requires a bit of configuration. Get a twilio account, create a .env file by running cp .env.sample .env, and add your twilio authentication information. While you're there, add a cookie secret and an encryption key (long random strings).

Install node dependencies

npm install

Define a new PostgreSQL user account, give it a password. You might have to create a postgres account for yourself first with superuser permissions if you don't have one already, or use sudo -u postgres before these commands.

createuser courtbot --pwprompt

Create a new PostgreSQL database and a database to run tests.

createdb courtbotdb -O courtbot
createdb courtbotdb_test -O courtbot

Set up your environment variables. This may require some customization-- especially the DATABASE_TEST_URL.

cp .env.sample .env

Then, to create the tables and load in initial data:

node utils/createTables.js
node runners/load.js

To start the web service:

npm start

Now you can interact with a mock of the service at http://localhost:5000.

Deploying to Heroku

First, get a twilio account and auth token as described above. Then:

heroku create <app name>
heroku addons:add heroku-postgresql
heroku addons:add scheduler
heroku addons:create rollbar:free (only add if you do NOT have another rollbar account you want to use)
heroku config:set COOKIE_SECRET=<random string>
heroku config:set ROLLBAR_ACCESS_TOKEN = <rollbar access token> (only needed if you did NOT use the heroku addon for rollbar)
heroku config:set ROLLBAR_ENDPOINT = <rollbar endpoint> (only needed if you did NOT use the heroku addon for rollbar)
heroku config:set TWILIO_ACCOUNT_SID=<twilio account>
heroku config:set TWILIO_AUTH_TOKEN=<twilio auth token>
heroku config:set TWILIO_PHONE_NUMBER=<twilio phone number>
heroku config:set PHONE_ENCRYPTION_KEY=<random string>
heroku config:set DATA_URL=<court records csv location>
heroku config:set COURT_PUBLIC_URL=<where to send people for more info>
heroku config:set COURT_NAME=<name of court system>
heroku config:set QUEUE_TTL_DAYS=<# days to keep a citation on the search queue>
heroku config:set TZ=<standard timezone ex. America/Anchorage>
heroku config:set TEST_TOMORROW_DATES=<1 if you want all court dates to be tomorrow to test reminders>
heroku config:set ADMIN_LOGIN=<user name for access to admin api>
heroku config:set ADMIN_PASSWORD=<password for access to admin api>
heroku config:set JWT_SECRET=<random string to be used to create json web token when authenticating admin api>
heroku config:set TESTCASE=<case number for testing>
git push heroku master
heroku run node utils/createRequestsTable.js
heroku run node utils/createNotificationsTable.js
heroku run node runners/load.js
heroku open

The dotenv module will try to load the .env file to get the environment variables as an alternative to the above "heroku config" commands. If you don't have this file, dotenv will throw an ENOENT error, but things will still work. To get rid of this error, do this:

heroku run bash --app <APP_NAME>
touch .env
exit

Setting up the Scheduler

Finally, you'll want to set up the scheduler addon to run the various tasks each day. Here's the recommended configuration:

Task	Dyno Size	Frequency	At
node runners/load.js	1X	Daily	8am
node runners/sendReminders.js	1X	Daily	5pm
node runners/sendUnmatched.js	1X	Daily	5:30pm

Running Tests

npm test