This is greengrep - a greentext search engine. Based off of memegrep.com, which I really like.
The plan is simple
- Scrape reddit or some other website
- Generate embeddings (and possibly OCR too?)
- Store in postgres
- Return search based on similarity
Here's a quick guide I can use.