Show OS: I built a search engine that uses vector embeddings
Posted by: altug
Hello sourcerer, here is janNet, my search engine that works like a modern search engine. It uses vector embeddings to compare the search term with a database of vectors. It also has an alternative search function that does not use vectorization, instead it uses the actual keywords and stores them in a reverse-index. This project was purely made to please my curiosity and is open-source:
Source: https://github.com/altugj...
Score: 7
Category: web crawling/mapping
Added: 2025-12-07 14:43:49
gandhiman
This is really cool. What vector database are you using? Been looking at pgvector lately.
Thomas
The reverse-index fallback is smart. Best of both worlds - semantic search when it works, keyword when you need precision.
daniwild
Vector embeddings for search is brilliant! How are you handling the embedding model? Training your own or using pre-trained?