OpenAI-to-SQLite

Embeddings The embeddings command can be used to calculate and store OpenAI embeddings for strings of text. Each embedding has a cost, so be sure to familiarize yourself with the pricing for the embedding model. The embeddings from the API will then be saved as binary blobs in the embeddings table of the specified SQLite database - or another table, if you pass the -t/-table option. Data from a SQL query The -sql option can be used to read data to be embedded from the attached SQLite database. The query must return an id column and one or more text columns to be embedded. Batching Embeddings will be sent to the OpenAI embeddings API in batches of 100. Working with the stored embeddings The embedding column is a SQLite blob containing 1536 floating point numbers encoded as a sequence of 4 byte values.