This repo creates a streamlit app that lets you extract information about a podcast episode to quickly decide if you are interested in listening.
- Use feedparser library to get all episodes from a podcast given an RSS link.
- Use the Whisper model to convert the podcast audio to text.
- Use the OpenAI API to extract extra information about the podcast episode, such as summary, guests and key points. We use a function call to get a formatted response.
- Display results using a streamlit app
I use streamlit to create the front end of our application. This also allows me to share the app using https://share.streamlit.io/
I use Modal Labs to run the podcast transcription and the OpenAI API call. The code deployed on modal is under scripts/backend.py
This project was part of the Uplimit Course "Building AI Products with OpenAI" given by Ted Sanders, Sidharth Ramachandran.