Where does my mind wonder?

This is a submission for the Hacktoberfest Open-Source AI Challenge Week 1: Touch Grass

What I Built

Building an open source AI Powered, Social and Outing finder. Works without any net, so no matter where you are - ALWAYS HAS YOUR BACK! The best part about this is that it learns more about you as you interact with it.

It is hyper-personalised so that more you interact with it more it knows about you, better it can recommend what you should do / go out for what mood.

If you are sad -> Give you comforting places
If you are energetic -> Recommends you outdoor activities as per your hobbies
If its end of the month & you have a tight budget -> Recommends you pocket friendly options

Demo

Couldn't get time to make a demo or a deployed link:(

Code

Still in the building process, will update once completed!

How I Built It

The workflow goes like:

1. Spatial Ingestion & Candidate Retrival
   We first fetch the geo-spatial co-ordinates from the users device.
   The client transmits (latitude, longitude) along with device state metadata. The backend executes a geospatial query fetching candidate POIs within a specified radius ($r$).

2. Semantic Context Synthesis
   After we have our candidate, we fetch the users past records. The candidate venues are merged with the users vector embeddings. A lightweight ranker scores candidate POIs based on proximity, ambient context (time of day, day of week), and user activity history.

3. Inferenced Structured Generation
   Find the top-k spots and give our what the candidate actuals has based on the intersection of the user likings, mood, budget & location.

4. User provides Feedback
   Simple thumbs up thumbs down to tell if the LLM has worked properly or not.

Used Gemma 4 for building this interface. Further used the SERP API to get the locations and the surrounds activities. Used Mongo DB for vector embeddings.

Why Does Open Innovation Matter?

People would only go out if there is something in it for them. This can be the activities that they like doing. If we give them something that is way difficult or requires a lot of extra effort without a reward then they won't do it.

And why make this open source?

More users can improve this and we can get to what we want to do for free rather than spending just to figure out what we want to do. Closed APIs would incur cost and that eventually comes down to the user.

I love reading psychology too!

Prize Categories

None, I have not submitted code yet. Will update this when I submit the code.