app in action
We like to travel a lot, so building tools that make this easier and takes out a lot of the leg work is a natural choice. No one does their planning on one website. We set out to develop an application that solves the problem of travel planning by bringing all content to one place for your once in a life time inter continental trip.
What it does
The first thing the user does is enter a budget. The site gives you a random itinerary based on your budget and country of choice(can go around the world as well). Each stop has it's own card, which can be extended and added to later. You can then manually add notes or use the chrome/firefox extension to add smart, context aware, notes. You can select text on a given website and we'll use alchemy and wit.ai to find out where to place the note and if we can find out any key information.
How I built it
Last year I had built a bare bones back end web app called RouteRush which created a random trip based on budget. UniTrip has restrictions on randomizer to make for a country specific experience. We added a lot of things over the randomizer. Redis-queue was implemented to keep track of the currently processed city and left balance. Shifted the backend away from Djang and into Flask and the front end to Angular. The most interesting new feature for the user experience is the concept of enriching cards with contextualized snippets which the user adds via extensions. This intelligent text parsing is achieved with via alchemyapi and wit.ai. The extensions were built using Firefox addon api and Chrome extension api. A further feature is the ability to edit the travel history and itinerary.
Challenges I ran into
Getting wit.ai to work was a pain since the model we wanted is depreciated (but still fine!) and the machine learning will continue to sometimes act unreliably until more user data is submitted. None of us are designers, so designing and picking up angular was also time consuming.
Accomplishments that I'm proud of
- Created an app I see myself using when improved further
- Created a decent front end
- Fleshed out a usable product from a bare bones previous project.
- Extensions were made for the two most popular 3rd party browsers.
What I learned
wit.ai, angularjs, alchemyapi, firefox extensions
What's next for UniTrip
- Better models for Wit.ai. Currently the text processing is very poor.
- Community Intelligence - If someone adds something important, it can be shown to others as well
- Proper Authentication
- Better Design
- More Cities
- Ability to add new, user defined, cities to the trip list.