Inspiration
I took an SIA course that made me realise that the risk levels for London are always high and that, in fact, the UK hasn't had a threat level below substantial in recent history.
What it does
This shows on the map the locations of crimes that happen, directly extracted from a database seeded with crime information from the police database. It also allows users to submit crimes that they observe as they happen and uses AI to moderate those submissions
How we built it
Because I joined the hackathon pretty late, I thought of how quickly I could ideate and bring my project to life. As a result, this project was built with v0 on the frontend and copilot on the backend. The technologies used were React, Next.js, Tailwind, Supabase, ChatGPT and Gemini.
Challenges we ran into
The issues were mostly with using the mapbox and LLM APIs. The mapbox API wasn't too hard to get used to after a while, and it was perfectly suited for the police API because I could get data about the coordinates of the map's current view and feed that into the police API's poly parameters. The LLM APIs always work unreliably for me except when run with a local LLM.
The police API also has minor issues in the way it is formatted in my opinion, because the dates of the crimes are only defined by the month and that drastically reduces the quality of the information they provide.
Accomplishments that we're proud of
Building a quality prototype, albeit of a simple idea in a day.
What we learned
How to work with map data, how to conditionally populate my map to improve performance, how to intelligently switch between querying my database and sending requests to the police API directly depending on whether my database already covers that area.
What's next for Community Safety
Polishing. Most especially performance-wise, but also with how the community reports are handled. There should be assigned moderators who can approve reports even when the AIs fail or deny some posts. The posts should also allow for more details like pictures, which will also improve the possibilities for the AI moderators (e.g., marking images as sensitive when the content is triggering).
Log in or sign up for Devpost to join the conversation.