Inspiration
We wanted to help people with ADHD, Alzheimers, and who are hard of hearing/blind.
What it does
We record audio as an input through a computer microphone. Based on what's said, we return various things, such as a transcript of the audio, a new schedule based on new events introduced in the audio ,a summary of the audio, and a recording of the audio. The new schedule helps people with ADHD/ Alzheimers because they can easily forget things that are said a long time ago. A summary is for people with ADHD who have trouble focussing. The recording of the audio is for people who are blind, and the transcript of the audio is for people who have trouble hearing.
How we built it
We built the voice to text software using deepgram and python libraries. We then used Llama and ChromaDB to store the data as well as allow the user to access the data. We used gemini and Langchain to help analyze and parse the text generated by our voice to text software. Finally, we used groq to get api keys for Llama.
Challenges we ran into
The first big challenge we ran into was getting our speech to text script to work. Deepgrams had problems with API keys which prevented us from getting any results and that was the first major roadblock we had in completing our project. The next roadblock came in the form of storing and recording. We had issues getting gemini and Langchain to find and analyze past data that was stored in our data structures which we eventually solved by becoming more specific with our queries for the A.I.
Accomplishments that we're proud of
Our first major accomplishment was getting the speech to text script to work. It is a core part of our project so its completion was a step towards our end goal. The next big accomplishment was when we successfully linked the speech to text script to our data structure script. After they were linked, we were able to save and access data stored in our data structure and use it to do things such as store reminders.
What we learned
We learned how to use different A.I models to streamline our project and allow for easy analysis of data. We also learned how to use frameworks to make efficient data structures and to speed up our completion of our product which gave us more time to refine other parts of our project.
What's next for Smartware
Our final goal is to convert our software into hardware, specifically a pendent. It could be worn by anyone and the above features and more would be available. The final product would have a speaker and a camera so its use cases could expand and it could benefit more people with impairments.
Log in or sign up for Devpost to join the conversation.