Find answers from the community

Updated 2 years ago

If I use a pinecone index is there any

If I use a pinecone index, is there any way to make it only use embeddings when query using a GPTSimpleVectorIndex? My issue is the following, I want to embed chunks of text and store those embeddings in a pinecone index, but I can't upload the plaintext into pinecone, so the text will be encrypted when upserted alongside the vector to pinecone. This will break things, especially querying, for all index types right?
j
K
5 comments
@Kaveen hmm. yes it would. this may not be fully supported within gpt index atm
another option @Kaveen is to use our PineconeReader
you could query with embeddings, and you can maintain a separate id_to_text_map
Would the pinecone reader just give me a list of top_k embeddings?
kind of, it also takes in an id_to_text_map. But you can just keep it a blank string in each entry if you'd like https://github.com/jerryjliu/gpt_index/blob/main/gpt_index/readers/pinecone.py
Add a reply
Sign up and join the conversation on Discord