Inspiration

I have been a meditation practitioner and was inspired by a book written by a Stanford neuroscientist on "manifesting" your dreams with less woo woo (although I think there would always be an element of something mysterious that we don't understand yet). I wanted to tune into my dreams like I did when I was a young kid living in the Philippines and dreaming of becoming a scientist in Europe.

Building

I used replit for the initial attempt and then used Cursor and Trae to fix the remaining issues. For the backend, I used postgresql and hosted it in render and with neon. The text-to-speech (tts) API I use is from eleven labs tts. I used groq API to access different LLMs for cheap including deepseek and llama. Internally, my code needed to clean up the LLM outputs for the tts to read it without awkward symbols and phrases.

Challenges

It was my first time doing tts and I also had to find a way to test it without burning through my LLM credits. The database was also a challenge. There were also some bugs with the deployment on render as it was building the site. It couldn't find the index.js and we had encountered some MIME-type error. But we fixed it by adding a content-security protocol and by adding a vite module that looks for the exact index.js that was being generated by the npm build command.

Share this project:

Updates