FareHound

FareHound is an SMS application that uses the Wego Flight Api http://www.wego.com/api/flights/docs and the Hoiio Api http://developer.hoiio.com/docs/. The application helps users to find fares which matches their criteria.

For example, they can SMS to a mobile number with the format:

• Singapore to San Francisco On Dec 25 with $1000
• SIN to SFO on Dec 25 with $1000
• SFO on Dec 25 with $1000

Setting Up

The project uses a couple of third-party APIs, and the Typesafe Stack.

Firstly, install the Typesafe Stack (Scala + Akka + Play!).

Then get the API keys for the following services:

1. API Key From Hoiio http://developer.hoiio.com/docs/
   • HOIIO_APP_ID
   • HOIIO_ACCESS_TOKEN
2. API Key From Wego http://www.wego.com/api/flights/docs
   • WEGO_API_KEY
3. API Key From Open Exchange http://openexchangerates.org
   • OPEN_EXCHANGE_KEY
4. REDIS_URL e.g. redis://user:password@localhost:6379/
   • REDIS_URL
5. API Key From Bit.ly http://bitly.com/a/your_api_key
   • BITLY_LOGIN
   • BITLY_API_KEY

You would need to set these environment variables in your .bashrc, .zshrc, .env for heroku, etc.

To start the web server,

play run

How It works

Steps

1. User send in an SMS with a query
   • Singapore to San Francisco On Dec 25 with $1000
2. Confirmation SMS will be sent to confirm successful registration of query
3. An initial search will be started and if the search price is below the user expected price, an SMS is sent to the user with current search price. In addition, the best price for the same search for the past 3 months and a booking url will also be sms back to the user
4. Every 1 hour, A scheduler run all the registered user queries and fire off the searches.

User can at anytime send back the same search with a different price expectation.

The system will only allow 1 price expectation for the same search query for the same user

Developers

1. Choon Kee, Developer @ Wego
2. Junda, Developer @ Hoiio

Technology Stack

All this is made possible by the Typesafe Stack

• Scala 2.9.1 http://scala-lang.org
• Akka 2.0.3 http://akka.io
• Play! 2.0.4 http://www.playframework.org/
• Scala Redis 2.5 https://github.com/debasishg/scala-redis/
• Dispatch 0.9.2 http://dispatch.databinder.net/Dispatch.html
• Kryo 2.19 http://code.google.com/p/kryo/

with data persisted on Redis http://redis.io, a key value store

Overview of the solution

These are the component that made up the solution. AlertService is using Cake pattern for dependency injection

Communication Component and Sms Parsing Component

Sms Controller

Responsible for parsing the SMS request and sending the SMS. Created using Play Framework Controller

Alert Service

Scheduler Component

Responsible for the running hourly update

Search Component

LoadBalancedActor: Run as a Akka FSM. Conduct the Search and the Pull operation and Finding the minimum price of each search. Queued Pending Search Request.

Checker Component

LoadBalancedActor: Receive the minimum price of each search request and check against the user criteria. If it matches the user criteria, send the cheapest price to the user

BestPrice Component

LoadBalancedActor: Call the Wego airfare API to get the cheapest price for the search for the past 3 month

Notifier Component

LoadBalancedActor: Send out the sms to notify the user

Exchange Rate Component

Update the exchange rate using the Open Exchange rate api http://openexchangerates.org

Benefits

• Asynchonous Polling for Flight Prices
• Communication between various components are done through Actor message passing
• Small Memory Footprint. Only User Criteria and the Cheapest Price are stored
• Horizontally scalable. Just add more actors at each layer
• Very Extensible. Just add additional layer to do more transformation before sending the SMS notification

Future Works

1. Support more form of queries
2. Web Interface
3. Email Notification