Changed embedding model to MiniLM-L6 as faster. Compressed embeddings are now int8. General improvements to API mode

ea0dd40

seanpedrickcase commited on Jul 3