search

FlowiseAI / Flowise

Drag & drop UI to build your customized LLM flow
https://flowiseai.com
Apache License 2.0
29.39k stars 15.19k forks source link

[BUG] The OpenAI Embedding Node for LamaIndex does not work for 3072 vector dimension #1864

Closed toi500 closed 4 months ago

toi500 commented 6 months ago

The OpenAI Embedding Node for LamaIndex is fixed at vector dimension 1536, independently if you select the text-embedding-3-large model which is 3072.

This produces a error message and it does not index any data at Pinecone.

The node works for text-embedding-ada-002 / 1536 as expected.

Version 1.6.0

0001 0002 0003
HenryHengZJ commented 6 months ago

I dont think we hard-coded 1536, its undefined and we let the OpenAI SDK to figure out what dimension to use depending on the model.

Codechilla commented 6 months ago

Same problem, but with Qdrant (local) instead of Pincone.