6 Commits

Author	SHA1	Message	Date
csd4ni3l	1ad2d1b59c	try using jina clip v2(v3 is not multimodal)	2025-08-21 19:25:31 +02:00
csd4ni3l	e9d0bf914a	fix the word embedding cache being included in git and try switching to v3 which might fix jina being so slow	2025-08-21 19:23:13 +02:00
csd4ni3l	a8ee1c90ae	add word vector cache with JSON and add a constant for the word to compare with	2025-08-09 10:30:57 +02:00
csd4ni3l	3f304071e3	fix model again and fix images being inputted as text	2025-08-08 23:14:45 +02:00
csd4ni3l	881ccbda13	fix wrong model	2025-08-08 22:58:19 +02:00
csd4ni3l	8ab56b218f	First commit	2025-08-08 18:08:41 +02:00