Gogo - The Assistant
Bringing things to life.
The problem Gogo - The Assistant solves
This app aims to bring inanimate things to life. To understand the utility let us look at a use case. Imagine a person bought a new device (like a vacuum cleaner) and now wants to search for a function of a particular button and for which one has to follow a number of steps starting from-
- Searching for the model number on the device or the box
- Trying to google it
- Find the right website, if available and
- Download the manual
- Then go through the points individually to find the right instructions.
With our technology, an organisation can create an assistant just for that item, by providing the data.
We will generate the QR for the item containing the item id and then the organisation can stick it on the device while giving to customers. And when the customers need help, then they can just scan the QR on our app and talk to the assistant. As simple as that.
The assistant would reply with the appropriate answer.
It is really like bringing inanimate objects to life! Similar situations arise when we visit a place, the assistant can work as a guide. It can also help a new developer to study the codebase easily through it's documentation. It is like giving a voice to every text!
Challenges we ran into
We ran into multiple challenges while doing this:
- One of the team members misread the server url, and spent about 2 hours trying to figure out the bug.
- Then choosing an unconventional name like Gogo was also a matter of debate and heated discussions.
- On a serious note, parsing the text and cleaning of the unwanted unicode characters was a nightmare along with finding the right model for judging semantic similarity.
- The storage of the data was also a serious concern.
- Deploying the server on the popular platform heroku presented many problems mainly because of the model loading. The model couldn't directly be installed by npm and running a script on heroku post install for flask was a hassle. But even after that the model size was crossing the free storage limits on the same. To solve this problem we exposed our local host to the web using the local connect library for node.
Technologies used