Readcap | Devpost

Inspiration

We always had a lot of messages in our group chat. We wanted to find a way to summarize our group chats automatically. This led us to an even bigger idea: integrating AI into text messaging.

What it does

Users can tag ReadCap in conversations to fact-check claims, summarize long discussions, retrieve event details based on past messages' context, or just answer questions and respond to any general queries. ReadCap can be added to group chats by just adding its phone number to the chat. ReadCap can also be used in direct messages. Unlike many AI services that require an internet connection, ReadCap works over SMS, making AI accessible even to those without WiFi or a smartphone.

How we built it

We built an API using Next.JS/Node.JS that processes messages when the user sends them. The messages are then stored in a MongoDB Atlas database temporarily. If ReadCap is tagged then an AI assistant responds the query and if any contextual messages are needed it accesses the messages through the MongoDB Atlas database. The AI assistant was called using the Azure OpenAI GPT-4o service through our API. The API sends back a response which is sent back to the user chat. The number operates through Twilio.

Challenges we ran into

Twilio was hard to set up due to the regulatory applications we needed to do. While we managed to code the app using the Twilio API, the regulatory applications would take time to approve and we would not be able to test it with a permanent number at this time. So we just used a temporary number with our API for the demo of the project.

Accomplishments that we're proud of

We are extremely proud of the fact we managed to make the AI work even when the end-user has no WI-FI which solves a real-world problem: the digital divide. We can see in the sample video that when there is no internet connection in the device, we can still use the features of this app via SMS. This would help with the problem of people having low access to the internet and also not having access to AI features. We are also proud of being able to use the Azure Cloud platform to be able to set up an OpenAI GPT-4o instance and code our API that manages all the different platforms we use. The way the all the different platforms we used we are happy that our code was able to integrate everything perfectly.

What we learned

We learned how to use AI programmatically and connect databases to it. We also learned how to code our API which we used Next.js/node.js for. It was also new to us how to set up a Twilio number and learning it from scratch was an interesting learning process.

What's next for Readcap

We want to integrate more custom features like allowing the bot to access calendar apps like Google Calendar. We also want the AI to be able to play small games, access the internet, and respond to current events.