-
Our logo!
-
Example questions on the form, asking the user about their mouse
-
The start of the troubleshooting form, asking the user to input their image
Inspiration
Many people are discarding their electronic waste, due to limited knowledge of what to do with it. We believe that this can change. On a larger scale, companies have to spend lots of money to buy new electronics, when there there are second-hand devices which are still in good condition.
What it does
We guide users through a troubleshooting process, asking questions about their device. Starting with the age and type of the device, the user is met with further questions depending on their previous responses. Finally, the user is told whether they are able to recycle it for parts, reuse it for a different purpose, or reduce it by selling it off for a discounted price. If the answer was still ambiguous, the user could upload a picture of their device, on which Gemini AI would advise the user.
How we built it
For the front-end, we used streamlit which allowed us to easily place javascript elements onto the screen, such as a multiselect/radio question. The questions were taken from a manually-curated SQLITE database of questions, which would customise the questions asked based on the user's previous answers.
Challenges we ran into
When attempting to integrate the LLM with the troubleshooting questionnaire, we found that the Gemini API key was paid, so we had to look for an alternative. We found that openrouter offers a free-to-try API key for OpenAI chatbots.
Accomplishments that we're proud of
We liked how the questionnaire form turned out; we had a recursive questions function to load in questions, and with the dynamic way streamlit works, any edit to previous question answers would automatically disregard further answers that stemmed from the original answer.
What we learned
Many of us were not previously familiar with streamlit, so this was an opportunity to learn a new library for designing web applications. It also increased our knowledge of normal forms for databases and how to turn that theory into practice.
What's next for Re-E-valuate
Despite problems with AI integration due to the API key issues, we have many more plans for the future. Firstly, today's project focused on a single user, but we hope to find a way for companies with a larger collection of devices to be put to good use, for example selling them off to people who need them. Additionally, with a handmade database, an image recognition model would be able to autofill much of the form (e.g. estimating the device's age and model).
Log in or sign up for Devpost to join the conversation.