This PR adds support for using Pinecone as a vector store. This is still a WIP as no tests have been written. From simple usage testing, it appears to all work together. Two things to note:
1) This is using fastembed as there is no way to generate embeddings with pinecone directly.
2) There is also no way to get all vectors in an index / namespace (for get_training_data) so instead, I set the top_k to the max and provide an empty vector.
3) I havent tested using other fastembed models and would also like to add support for using something like text-ada-embedding-2
This PR adds support for using Pinecone as a vector store. This is still a WIP as no tests have been written. From simple usage testing, it appears to all work together. Two things to note:
1) This is using
fastembed
as there is no way to generate embeddings with pinecone directly. 2) There is also no way to get all vectors in an index / namespace (forget_training_data
) so instead, I set the top_k to the max and provide an empty vector. 3) I havent tested using other fastembed models and would also like to add support for using something liketext-ada-embedding-2