Lucky you was just reading this https

Right, but that's more if you want to generate your own embeddings during index construction, rather than using OpenAI's embeddings

In my case, I already have OpenAI generated embeddings, that I want to feed into the index

So for example, if I knew the format of the index.json file that GPT-Index creates, I could export my embeddings in that format

And have GPT-Index reuse them

You can set the embedding field manually on a Document

if set, then when you construct an index from Documents, then the embeddings you provide will be used instead of us embedding it (with openai by dfeault)

Thanks, that's exactly what I was looking for