Combine langchain retrievers

[MS-CICADA]

Checked other resources

• I added a very descriptive title to this question.
• I searched the LangChain documentation with the integrated search.
• I used the GitHub search to find a similar question and didn't find it.

Commit to Help

• I commit to help with one of those options 👆

Example Code

return new ConversationalRetrievalQAChain({
    retriever: vectorstore.asRetriever(!!sourceCount ? sourceCount : 6),
    combineDocumentsChain: docChain,
    questionGeneratorChain: questionGenerator,
    returnSourceDocuments: true
  })

Description

I am using ConversationalRetrievalQAChain in typescript. I have created two vector stores and I want the query from ConversationalRetrievalQAChain to be compared against both vector stores and results from both vector stores to be used to create the final answer.

So I have decided to create two retrievers

const retriever1 = vectorstore1.asRetriever(!!sourceCount ? sourceCount : 6)
const retriever2 = vectorstore2.asRetriever(!!sourceCount ? sourceCount : 10)

How can I now override the Retriever class so that when the query is compared against my custom_retriever, it is compared against documents from both retrievers, and documents from both retrievers are used to create the prompt.

Note: i don't want to merge the vectorstores because that messes up the similarity search.

System Info

"langchain": "^0.1.34"

[KeenanFernandes2000]

I had a similar use case; however, I used a different chain.

Have look at this, it helped me with my project.
https://js.langchain.com/docs/modules/chains/additional/multi_retrieval_qa_router