AptAI

Landing page
Page with apartment query executed

Inspiration

Lots of vibes Money / Cash prizes Bring joy and happiness to the world

What it does

AptAI allows landlords to post apartment photos without having to describe the details of any room. Apartment finders can search for apartments that suit their interests, and lifestyle choices in natural language. This technology goes beyond what's typically found in the market today.

How we built it

We used OpenAI reverse-image-to-text to obtain structured listing photo metadata which we then uploaded to Vectorize to create a searchable database. A UI was built to execute natural language queries in combination with Vectorize vector-search. The query results contained required metadata including the photo image URLs.

Image URLs and photo metadata (JSON formatted) where uploaded to Google Drive as a data source for Vectorize. Vectorize was then used to import the JSON metadata to create vector embeddings for each apartment / room photo.

Several python scripts where vibe-coded to generate the photo JSON metadata files. A UI was mocked in UX-Pilot, and then an implementation was vibe-coded in Replit using React and ExpressJS for backend. The backend code makes Vectorize API requests to execute the user queries. Photo images were uploaded to Google bucket storage.

Challenges we ran into

We had issues importing data from Google data source to Vectorize. We had to iterate on the JSON metadata generated to add required fields to associate the images to their apartment listing info. General teamwork challenges in sharing info when working independently. Everything else worked like a charm with vibe-coding.

Accomplishments that we're proud of

The prototype worked! Built MVP in less than 4 hours. We formed the team quickly and dynamically. The a clean idea came together, and we were able to execute on it.

What we learned

Vectorize
How to integrate external data sources
Debugging vibe coded apps
Using tools we generally had not used before / quickly came up to speed

What's next for AptAI

Management features to uploaded content
Integration with production apartment rental data sources
Dynamic pricing
Integration with DashQ

Try This / Sample Apartment Queries

I want minions
I have young children, find all apartments that already have interesting wall paintings suitable for children.
I want an apartment with beautiful views of downtown
I want apartments that accomodate musicians