Pinecone text client for node.js

Feature Requests
1

To the best of my knowledge, the pinecone text client is only available for python. Would it be possible to make it available as a node package?

2

hi @lele

Roy from the team behind this package

May I ask what parts you intend to use? does it have to be a package or maybe running those capabilities as-a-service and calling them from the node application is also good?

Best

3

Hi @roy,

Thanks for your quick reply. I would like to use the package to dynamically create sparse vectors in the backend of an application, so I only need the BM25Encoder method. In this case, I would have to use BM25Encoder.default() because it would not be feasible to fit the encoder on my own documents during runtime. So actually, posting the text that needs to be encoded to an endpoint would probably be the better solution, especially if this would allow me to fit the encoder on the data in my index.

All the best

4

Hi @roy, I would like to use your package in node.js for the same purpose, do you know if this will be possible soon ?
Best regards

5

I would like this also

6

I could also really use this feature. With the python-text-client you essentially have everything you need to build the hybrid search functionality with both dense and sparse vector training and generation. In javascript we don’t have the ability to use the BM25 encoder through the pinecone client so we need a 3rd party library but it’s less efficient and has compatibility issues. For example if you save a model with the BM25encoder in the pinecone python client and export to JSON, you can’t load that same JSON model file to the winkNLP model.

7

check out this package: pineconebm25. It does exactly the same thing as the python api.

8

A post was split to a new topic: Pricing details?