Inspiration
Almost everyone I know has signed a lease without really reading it. It's 20 pages of legal wording, and the stuff that matters (fees, penalties, how you're allowed to leave) is buried somewhere in the middle. You only find out what it says when something goes wrong. We wanted a way to just ask the lease a question and get a straight answer.
What it does
You upload your rental agreement and ask things like "How much notice do I need to give before moving out?" or "Can they keep my deposit?" Leasely answers from the text of your lease only, shows you the passage it got the answer from so you can check it yourself, and flags clauses that look risky, like heavy penalties or one-sided terms.
How we built it
The interface is Streamlit. When you upload a lease, we split it into chunks and turn each chunk into an embedding. When you ask a question, we find the chunks closest to it using plain numpy, then hand only those to the language model to write the answer. That keeps it grounded in your actual document instead of guessing. The model is called through an OpenAI-compatible API, so we could swap providers without rewriting the app.
Challenges we ran into
Honestly, time. We started with only a few hours left, so we had to cut a lot. We also lost a good chunk of that time on API access. One provider wanted card details we didn't want to give, so we switched, and then spent ages fixing endpoints, model names and embedding settings that didn't line up.
What we learned
Making the model point back to the exact lines in the document makes it feel far more trustworthy than a normal chatbot. We also learned to get the API connection working before building anything else.
What's next
We'd like to support more languages and document types, score how risky each clause is, and let you compare two leases side by side. Leasely is meant to help you understand a document, not replace a lawyer, so we'd add clearer pointers to real legal help too.
Log in or sign up for Devpost to join the conversation.