While started the index as a query engine loaded from the vector store, I could see it computing embeddings for all the documents via Llama debugger trace. Is it that while it runs the query it computes the embeddings for the documents rather than at index creation time ?
I wanted to know that when I create an index with documents , does it creates embedding at the time or when I initiate the query engine, it does at runtime. This is for the text that is present in the nodes/chunks
I'm not sure where those logs are coming from, BUT in the first screenshot, no embeddings are called or created, except for one embedding for the query text.
Yes, this is exactly what I guessed as well since it is shown in the second screenshot that the node embeddings are created while creating the index. Thanks for the clarification, I will play around with chroma API to find where these embeddings are stored !