This repo contains the very simple backend used to power Rambla, a walking meditation companion app I created in the fall of 2023. You can read more about Rambla here.

In a nutshell, Rambla is an iOS app that helps the user explore complex topics by breaking them down into simpler questions, which are represented as nearby locations in a map. As the user walks to these locations and answers the questions, more questions appear, based on the overall meditation.

The backend works as follows:

• It receives requests from the iOS app containing the user's input (initial topic and subsequent replies), and responds with simple, relevant questions.
• backend.py initializes a Flask server (good enough for personal/testing use) and a custom LLM chain.
• chain.py contains the implementation of the LLM chain. Specifically, Rambla uses theory-of-mind reasoning to produce an answer in two steps: it first generates a hypothesis about the user's mental state based on their input, and then produces a list of helpful questions based on both the user's input and the hypothesis.
• The code for the iOS app is here. I've never coded for iOS and wrote the entire frontend with the help of ChatGPT, so apologies in advance for the mess.