PineCone + Max marginal relevance => how to create code?

[aza1200]

Here I create code
for combining pinecone and mmr(max marginal relevance)

I want to set

fetch_k = 100
and k = 10

vectorstore = Pinecone.from_documents(docs, embeddings, index_name=index_name)
retriever = vectorstore.as_retriever(search_type="mmr")
retriever.get_relevant_documents(query )

model = ChatOpenAI(model_name="gpt-3.5-turbo")  # 'ada' 'gpt-3.5-turbo' 'gpt-4',
qa = ConversationalRetrievalChain.from_llm(model, retriever=retriever)

where can i set paramter {"fetch_k":100, "k":10} in this code?

[dosubot[bot]]

Answer generated by a 🤖

Answer

I understand that you're trying to integrate Pinecone and Max Marginal Relevance (MMR) in your code and you're unsure where to set the parameters fetch_k and k.

In your code, you can set these parameters in the get_relevant_documents method of the retriever object. Here's how you can modify your code:

vectorstore = Pinecone.from_documents(docs, embeddings, index_name=index_name)
retriever = vectorstore.as_retriever(search_type="mmr")

# Set fetch_k and k parameters in the get_relevant_documents method
matched_docs = retriever.get_relevant_documents(query, k=10, fetch_k=100)
for i, d in enumerate(matched_docs):
    print(f"\n## Document {i}\n")
    print(d.page_content)

model = ChatOpenAI(model_name="gpt-3.5-turbo")  # 'ada' 'gpt-3.5-turbo' 'gpt-4',
qa = ConversationalRetrievalChain.from_llm(model, retriever=retriever)

In this code, the k parameter determines the number of documents to return, while fetch_k specifies the number of documents to fetch before passing them to the MMR algorithm.

It's generally recommended to set fetch_k to a value significantly larger than k. This is because fetch_k is the number of documents that will be fetched before filtering. If you set fetch_k to a low number, you might not get enough documents to filter from.

Please note that this is one potential solution based on the information provided. If you find this solution helpful and believe it could benefit others, you might consider making a pull request to update the documentation. This way, other users facing the same issue can also benefit from your findings.

Sources

• docs/extras/modules/data_connection/vectorstores/integrations/pinecone.ipynb
• langchain/vectorstores/pinecone.py
• docs/extras/modules/data_connection/vectorstores/integrations/cassandra.ipynb
• docs/extras/modules/data_connection/vectorstores/integrations/faiss.ipynb

---

This response is meant to be useful, save you time, and share context. It is not meant to be a precise solution, but rather a starting point for your own research.

Help me be more useful! Please leave a 👍 if this is helpful and 👎 if it is irrelevant.