by eneuman on 10/25/23, 11:53 AM with 4 comments
Why I Built This: Traditional keyword search isn't cutting it. I've used LLM-embeddings to provide more nuanced, relevant results.
How It Works: LLM-embedding similarity on curated datasets for semantically similar results. No need to iterate over keywords any more.
Current Datasets:
- YC Companies - Show HN Posts, - Ask HN Posts - ProductHunt Startups - Github Top 200k Repos
Use Cases:
- Validate a product idea's existence - Check if someone already Asked HN something - Have fun - search random terms and see what pops up
Want to see other datasets? Got one in mind? What would you use it for? (I'm even thinking of adding all PG's tweets - so feel free to be creative.)
Eager for your feedback and ideas. Hope you find this useful and fun!
by simonw on 10/25/23, 1:37 PM
The vector search is using https://lancedb.com/ and OpenAI embeddings.
by omarfarooq on 10/25/23, 12:46 PM