Introducing a Modified PineconeHybridSearchRetriever and Custom PineconeHybridVectorCreator for Enhanced Langchain Applications #3379
piedpiper-technologies
started this conversation in
Show and tell
Replies: 0 comments
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
-
Hello Langchain friends and fellow developers:
Over the past month, I've been diving deep into Langchain, exploring its documentation, seeking advice on Reddit, YouTube, and Discord, and watching numerous Pinecone and James Briggs videos. As I started to develop more robust applications with Langchain, I wanted to leverage Pinecone's sparse-dense hybrid indexes, but I encountered some obstacles with the out-of-the-box functionality of the PineconeHybridSearchRetriever.
After initially spending many hours trying to understand the differences between the hybrid retriever and the Pinecone vector store for dense indexes, I decided to modify the PineconeHybridSearchRetriever to better fit my needs. My goal was to create a more flexible and optimized solution for upserting sparse-hybrids into my existing index.
I am excited to share my work at https://github.com/piedpiper-technologies/langchain-modded-pinecone-hybrid-search. This repository contains a modified PineconeHybridSearchRetriever module and a custom PineconeHybridVectorCreator module that I believe can greatly benefit the Langchain community, along with examples and explanations of how it is integrated into some pretty neat solutions I think! 😄
I would be thrilled if these modules could be reviewed, and considered for integration as official Langchain modules. 🤗
Please feel free to check out the repo, provide feedback, and use the modified Pinecone Hybrid Search Retriever in your own projects! If you like what you see a ⭐ would be make me smile!
Happy Hacking!
Beta Was this translation helpful? Give feedback.
All reactions