I have crated a large VectorStoreIndex which i have persisted - almost 7 gb. I want to transfer this data into a ChromaDB, but do not see a good way to access transfer to Chroma automatically, or access the underlying nodes and embeddings in order to add them manually. Does anyone have a solution to this?
Awesome - it works. Posted solution on github issue as well. Would not be bad idea to have a "retreive nodes" function on any Index to easily reverse and create different database types.
As a note. As a simple persisted database, it was over 9gb of data all together with docstore and vectors saved as JSON. It took almost 1,5 minutes to do similarity search. With Chroma, it is 1,4gb, and is under 1 second. Should have a note in the docs that the simple index is not intended for significant amounts of data.