Inspiration & Functionality
Named after the Greek goddess of wisdom, Athena takes user queries regarding datasets, and displays the data in an user-friendly way. This will allow our end-users, CBRE employees, to be able to casually use the database without any prior technical knowledge.
How we built it
Initially accompanied by a made-from-scratch database using MySql, most of our back-end was written in Python before we pivoted to Javascript to cater to our then-needs. OpenAI is used in this product as a server taking in user queries as a JSON object, processing it, and outputting a response. Displaying the chatbot's responses is our front-end geared to user-friendliness and designed with React.
Challenges we ran into
Determining which large language model to use was a struggle. Although we knew of more well-known models, we weren't sure if there was a pre-trained model that could cater to our needs more. Time was dedicated to figuring out which large language model was to be used and the relationship between the large language model and the langchain itself. After the initial research into these concepts, challenges primarily appeared in connecting the database to the back-end and connecting the back-end to the front-end. OpenAI's API was also a struggle to wrangle through; with multiple usages of libraries, there were inconsistent cases of the API's performance. Overall, there were steep learning curves and challenges that were overcome.
What we learned
Before, large language models (llms) and langchains were foreign, daunting terms. After the challenge, they are less foreign yet still offer an exciting challenge. With the advent of large language models, the digital landscape is constantly changing---including the documentation for some technologies we were interested in. Creating a database with MySql also proved to be difficult without proper formatting for commands.
What's next for Athena
Coming soon to Athena, users will be able to have their own accounts with user authentication, allowing users to have a history of previous requests and reports generated. Athena will soon support text-to-voice reports, verbally sharing its wisdom whether a user is on the go, visually-impaired, or laying low on screen time. Additionally, in the future horizons for Athena, is a UI/UX design with multiple viewing modes supporting most varieties of color blindness.
Log in or sign up for Devpost to join the conversation.