Changed embedding model to MiniLM-L6 as faster. Compressed embeddings are now int8. General improvements to API mode ea0dd40 seanpedrickcase commited on Jul 3, 2024