Inspiration
The increasing rental rate coupled with the increasing dense population of San Francisco was the inspiration behind this project. Despite the urgent need for housing solutions, we recognized a lack of optimized platforms for matching compatible house mates. Selecting the right house mate is crucial, and our project aims to address this gap by providing an efficient and effective way to connect individuals seeking shared living arrangements.
What it does
AgreeMate.ai revolutionizes the shared living landscape by addressing the unique challenges faced by urban renters. This platform focuses on facilitating difficult conversations and generating AI-powered housemate agreements. By leveraging OpenAI's APIs, AgreeMate.ai simplifies the process of establishing clear expectations and resolving potential conflicts among housemates. This innovative approach aims to create a more harmonious and stress-free living environment for all individuals sharing a home.
How we built it
We built a web application using React, Firebase, and OpenAI APIs, that enables users to engage in real time communication while being guided by the AgreeMate AI. This AI component acts as a facilitator between potential house mates. The AgreeMate AI can understand the context and nuances of the users' interactions, catching potential conflicts in their schedules, lifestyles, and habits, and simultaneously provides valuable insights and recommendations to ensure smooth and productive discussions. The use of React and Firebase technologies ensures a seamless and responsive UI, allowing for instant messaging and real-time updates.
Challenges we ran into
Working with OpenAI's API and Google Cloud functions can present some unique challenges. Some challenges we ran into were:
- OpenAI's API is triggered after one user input: This behavior is problematic when you want to maintain a continuous conversation flow. Our solution was to trigger the API once it detects all users have sent at least one message stored in a chatroom's Firestore server before AgreeMate continued on.
- ChatGPT's lack of memory of the previous conversation of previous conversations: To overcome this, we implemented a secondary helper AI that kept track of a "risk of conflict" value amongst potential housemates in certain topics. This allowed two AgreeMate AIs to split the load and provide a seamless and continuous context-aware conversation.
- Connectivity with OpenAI & Cloud function: Integrating different services and ensuring smooth connectivity can be challenging. We faced some issues early on and thoroughly tested the integration, configurations, and retrying mechanisms to mitigate these problems.
Accomplishments that we're proud of
Coming up with a creative solutions to overcome all our challenges & creating a working testable POC in under 2 days Some accomplishments were:
- Chaining user inputs to control OpenAI's triggers
- Using prompt engineering & creative solutions (using multiple APIs) to provide our primary AI sufficient context & "memory" of previous conversations
What we learned
We improved on our technical skills, learned best practices (and avoid WORST practices) & most importantly became more driven in our craft throughout this hackathon.
What's next for AgreeMate.ai
Some future plans are using computer vision to fulfill our initial vision of creating similarity scores through user submitted images of their current rooms based on matching aesthetic and cleanliness.
Log in or sign up for Devpost to join the conversation.