Post AmbeXovsWsAucPSqHY by leafmuncher@mastodon.online
 (DIR) More posts by leafmuncher@mastodon.online
 (DIR) Post #AmbOooKARfg4O8u9wG by alexgarciaxyz@data-folks.masto.host
       2024-10-02T17:06:00Z
       
       0 likes, 0 repeats
       
       New blog post:  How to do "Hybrid search" with SQLite FTS5 and sqlite-vecReciprocal rank fusion, FTS-first, re-order by semantics, and more!Featuring the great snowflake-arctic-embed-m-v1.5 model and sqlite-lembed, using pure SQL examples.https://alexgarcia.xyz/blog/2024/sqlite-vec-hybrid-search/index.html
       
 (DIR) Post #AmbZ0vv71qpuxmbY6C by alexgarciaxyz@data-folks.masto.host
       2024-10-02T19:00:16Z
       
       0 likes, 0 repeats
       
       @markus Tiny models are super useful! I personally can't tell much difference btwn openai and OS models like Snowflake Artic/Nomic/MixedBread.Also, fine-tuning scared me at first, but much easier/cheaper to do that than relying on 3rd parties, if you really wanna squeeze perf. Most of these models claim better benchmarks than OpenAI anyway. I personally never use OpenAI embeddings, but I'm sure they have their place. Ex maybe it's better at multi-lingual/code, but haven't seen that yet
       
 (DIR) Post #Ambdz0HsJvvjhQ9QK8 by alexgarciaxyz@data-folks.masto.host
       2024-10-02T19:55:56Z
       
       0 likes, 0 repeats
       
       @markus Thanks! llamafile does support embeddings models btw, here's a guide: https://github.com/Mozilla-Ocho/llamafile/blob/main/llamafile/server/doc/getting_started.mdThe `llamafiler` binary is available is the .zip release: https://github.com/Mozilla-Ocho/llamafile/releases/tag/0.8.13
       
 (DIR) Post #AmbeXovsWsAucPSqHY by leafmuncher@mastodon.online
       2024-10-02T20:02:13Z
       
       0 likes, 0 repeats
       
       @alexgarciaxyz Great article!
       
 (DIR) Post #Ambf05BjgTrp8wMFEm by alexgarciaxyz@data-folks.masto.host
       2024-10-02T20:07:21Z
       
       0 likes, 0 repeats
       
       @leafmuncher thank you!