fabric_tools

process_channel.py

Given either a user id or channel id for a you tube user, this script downloads the transcript for all the videos and runs then through fabric to extract+wisdom. That wisdom is saved as a markdown file, which is later pushed to another of my github repos https://github.com/osok/arXiv_papers.
Then an embedding is create using a local LM Studio instance with nomic-ai/nomic-embed-text-v1.5-GGUF of the entire markdown file. details of the video including the summary are stored in the metadata. This can then be searched and display using the tool in https://github.com/osok/arXiv_flask

You will need API Keyys for

PineCone
YouTube

usage: process_channel.py [-h] [--channel_id CHANNEL_ID] [--user_id USER_ID] --out_dir OUT_DIR --index_name INDEX_NAME [--count COUNT]

Process YouTube videos.

options:
  -h, --help            show this help message and exit
  --channel_id CHANNEL_ID
                        YouTube channel ID
  --user_id USER_ID     YouTube user ID
  --out_dir OUT_DIR     Output directory
  --index_name INDEX_NAME
                        Pinecone index name
  --count COUNT         Number of videos to process

osok/fabric_tools

fabric_tools