WealthAgent

• 

  Landing page

Inspiration

We were inspired by past shellhacks projects and google's adk challenge. We wanted to create something that could help many people. Financial literacy, being paramount to modern life, was the first topic that came to mind.

What it does

This website assesses the user on their financial literacy and provides them with a tailored plan and lessons to improve their knowledge.

How we built it

We used an angular front end and a python backend. We run multiple server instances of gemini agents, each performing a specific task. The agents exchange information via a central database file. We created a flask based api to communicate with the agents.

Challenges we ran into

The AI agents often had a mind of its own. We had to be very precise with our prompts to ensure it stayed on track. Connecting the AI agents to the UI and making it cohesive was a monumental task. Often times, we'd find one agent functioning but the next agent was disconnected from the UI. Or we'd find the agents working but not communicating properly. In the end, we found that the authentication data wasn't being passed to each agent, resulting in the model failing. We store the userID in the database so all models can access it. We made sure that each model was sending and retrieving data from the database properly.

Accomplishments that we're proud of

We are very proud to have multiple agents successfully communicating with each other. The combination of a database, a custom api, and the use of google's adk made this a challenging but fun hurdle to overcome.

What we learned

We learned a lot about how to use api calls to connect seperate parts of our codebase. We became much more familiar with building UIs in angular. We also learned how make use of LLM apis.

What's next for AgentWealth

We plan to implement a more cohesive planning system that automatically generates more modules and quizes. There will be periodical assessments to gauge the user's financial literacy progress. The dashboard will pull all of its information directly from the LLM's and databases.

Built With

• adk
• angular.js
• flask
• python
• sqlite