eliza
eliza copied to clipboard
Allow to use local embedding instead of OpenAI when using OpenAI models
Is your feature request related to a problem? Please describe.
When using OpenAI models, I still want to use local embedding instead of OpenAI's one. Since models for OpenAI's embedding are different from those in text generation, there is no reason to force the use of its when using OpenAI models.
Describe the solution you'd like
Change getEmbeddingType
function to only select OpenAI if explicitly set USE_OPENAI_EMBEDDING
env.
Describe alternatives you've considered
Additional context