System Design for Project Sophia
One of the issues I've come up against while building the Sophia app has been getting stuck in specific parts.
In the past, it was easy because it was small and I would keep on moving.
One of the biggest problems was the dependence on 11labs, and the poor quality.
I will need to address this in a timely fashion when I get a chance to audit every audio file.
This might be something I do a little bit of every day, that's just the only way to make progress.
But today I want to lay out the overall system in a cleaner way and make sure that I can break things down so that trouble shooting one point doesn't regress the entire system.
This will ideally be in the form of a visual diagram, but first I'm going to jot down the different components:
- Surface YouTube Lists
- Enable the Selection of a Video
- Add the Video to the Compaction pipeline
The Compaction Pipeline
- Create a record, a Compaction
- Extract the Audio as an Audio File
- Create the Audio File as a Text Transcript
- Take the Transcript through an LLM to get a summary
- Take the summary through Voice Generator to get Audio
- Make the Audio available via the Firebase database
- Push the Audio out onto a Podcast
- Push the Audio out onto YouTube
What tool will I use for the visual?
Whimsical