Inspiration

We got our inspiration from the Capital One app. It has a chatbot called Eno, which is helpful but pretty limited. It mostly answers basic questions or links to website pages. We wanted to take that idea further by building something more interactive and useful. Another big reason we chose this project is because all of us have struggled with overspending and budgeting in the past.

What it does

Our chatbot does real-time budgeting right in front of the user. You can give it some account info and ask it to make or analyze transactions. It either processes that and forecasts your budget or runs an analysis based on what you ask for. You can also just ask for general budgeting help or advice.

How we built it

We built the backend in Python using OpenRouter for LLM credits and LiveKit for the speech-to-text part. The frontend was made with React, JavaScript, and CSS. For the database and chat history, we plan to use Cloudflare’s D1 Database and Workers.

Challenges we ran into

Our biggest challenge was connecting the frontend and backend. There were a lot of moving parts and missing pieces that made it hard to get them working together smoothly. We’re still working on that by testing and improving things step by step, often with help from AI tools like ChatGPT or Claude.

Accomplishments that we're proud of

We’re really proud of getting the speech-to-text system running. At first it was tricky since there were many tools to pick from, but once we decided to use LiveKit, it got easier. It was still a tough task, but we made it work and got full conversations running about finances and budgeting.

What we learned

We learned how speech-to-text systems work, how to connect pipelines to language models, and how to tie the frontend and backend together. We also learned a lot about Cloudflare and how to use tools like D1 Databases.

What's next

Next, we want to make it work with Alexa and Google Home so people can use voice commands on those devices too. We also plan to add chat history through Cloudflare so users can look back at past budgeting sessions.

Built With

Share this project:

Updates