My discord username is .linkoflegends
Inspiration: After witnessing my dad fall victim to identity theft, with the hassle of freezing his social security information and dealing with unfair charges, I realized how sensitive our data truly is and how crucial it is to protect it. Seeing accounts opened under his name without his knowledge exposed just how easily people can impersonate others, leading to serious consequences. I believe it's time to take control of our data and protect ourselves by proactively finding and removing sensitive information from the internet. Automating this process is essential, especially as companies must stay compliant to ensure the safety of every individual.
What it does:
How we built it: Frontend: The frontend of the system is built using HTML/CSS. It likely includes user interfaces for interacting with the data deletion agents and displaying relevant information.
Backend:
Multi-Agent System: The multi-agent system consists of several agents (backed by gpt 3.5 apis ) designed to perform specific tasks related to data deletion. These agents include:
Web-Scraping Agent: Responsible for collecting data from web sources. Verifier Agent: Validates and verifies the data collected by the web-scraping agent. Page Navigating Agent: Navigates through web pages to gather information or perform actions. Auto Email Sender Agent: Automatically sends emails based on predefined criteria or triggers. Docker: The multi-agent system is containerized using Docker, which provides a way to package the agents and their dependencies into isolated containers, ensuring consistency and ease of deployment across different environments.
Python & its frameworks (flask, CORS, pydantic, flask_mail)
RegEx to handle input/errors
Challenges we ran into: The costs of testing/calling multiple LLM Apis
Narrowing down what actions are specific to each agent
Figuring out how to make each agent talk to eachother
Cutting down on time, the process of generating prompts and making the agents run scripts was very time-intensive in the long run
Accomplishments that we're proud of: I was able to get an MVP and have a framework for the MultiAgent LLMs, eventually the process can be optimized so it doesn't take a huge chunk of time to run the models
What we learned:
I learned more about advanced techniques in webscraping, the typing animation feature in html, learning/getting better at implementing Multi-Agent frameworks.
What's next for ControlStopDelete:
Cutting down on latency and improving on the Webscraping and expanding form automation beyond email
Log in or sign up for Devpost to join the conversation.