App Safe

Inspiration

Last summer working while one of our group members was working at Scotiabank, he was asked to analyze, and provide recommendations as to how to improve tracking, and efficiency aspects of a pricing request intake platform. As he delved into potential solutions he came to the same two roadblock again and again. Software for enterprise (namely Microsoft) offer very little variability in their functional abilities. When software did have this function it was heavily restricted, and almost impossible to get approval for due to cyber security concerns. Never mind the costs that were majorly inflated. This is why we decided to develop AppSafe. Providing a robust, segregated architecture, with ultimate variability, and zero requirment for technical skills and abilities. Say goodbye to spending hours on a PowerBI platform, or getting those SharePoint permissions just right, and say hello to AppSafe.

What it does

Our project allows users to create applications by describing their various wants/needs in a few text prompts. Our application will then create the described application with an associated secure database to manage access permission and other data.

How we built it

The project utilized natural language processing agents provided by OpenAI to interpret what the user wanted and return the application and database. This allows our application to be able to adapt to whatever the user requests their application to be. To ensure their is consistency between built apps, the AI agent was provided some application templates which had functions that could be applied to various types of apps. Along with this, the user was allowed to correct the AI's interpretation for more accurate results.

Challenges we ran into

A major challenge our group encountered was connecting the MongoDB and Docker Network. This part of the project was the most important as it created our secure backend for our app. However, although team members had experience with MongoDB, it was our first time working with Docker Network making it tough to learn and implement within the Hackathon time limit.

Accomplishments that we're proud of

Our group is proud of implementing the natural language agent in order to produce code that can be immediately applied to run an application. Along with this, our group is proud of successfully implementing a secure background within the Hackathon deadline.

What we learned

Our group learned how to effectively use and teach OpenAI agents to produce accurate results. Along with this, our group learned how to utilize the Streamlit framework to produce simple intractable web applications.

What's next for App Safe

As the application was built within 48 hours, there is much room for improvement. For more important changes, the AI agent will need more fine tuning to generate more accurate results by feeding it more example cases and templates to choose from. The database generation code will also require some adjustment to allow reviewers to generate the database by a click of a button. As for minor tasks, the UI will require a complete overhaul to improve visual aesthetics.