dosubot[bot]
commented
1 year ago 🤖
Hello,
Thank you for bringing up this issue. It seems like you're trying to limit the retrieval of documents to a specific partition when using the ConversationalRetrievalChain and RetrievalQA with a Milvus vector store.
Based on a similar issue in the LangChain repository, you can extend the VectorStoreRetriever class and use it to filter the documents based on their source path. Here's how you can do it:
class FilteredRetriever(VectorStoreRetriever):
vectorstore: VectorStoreRetriever
search_type: str = "similarity"
search_kwargs: dict = Field(default_factory=dict)
filter_prefix: str
def get_relevant_documents(self, query: str) -> List[Document]:
results = self.vectorstore.get_relevant_documents(query=query)
return [doc for doc in results if doc.metadata['source'].startswith(self.filter_prefix)]And use it like this:
filtered_retriever = FilteredRetriever(vectorstore=store.as_retriever(), filter_prefix=source_filter)
chain = ConversationalRetrievalChain.from_llm(
llm=llm,
retriever=filtered_retriever,
memory=st.session_state.memory,
verbose=True,
return_source_documents=True,
)This solution was suggested in the issue Filtering retrieval with ConversationalRetrievalChain and received positive reactions.
Additionally, you can use the 'filter' option in the 'search_kwargs' argument of the 'as_retriever' method to limit the retrieval of documents to a specific partition. Here is an example:
docsearch.as_retriever(
search_kwargs={'filter': {'partition_name':'Your Partition Name'}}
)Replace 'Your Partition Name' with the name of the partition you want to limit the retrieval to. This is supported by the 'as_retriever' method documentation:
# Use a filter to only retrieve documents from a specific paper
docsearch.as_retriever(
search_kwargs={'filter': {'paper_title':'GPT-4 Technical Report'}}
)In this example, the filter is used to limit the retrieval to documents from a specific paper. You can use the same approach to limit the retrieval to a specific partition.
As for your question about whether LangChain version 0.0.281 supports partitioning in Milvus vector store, I wasn't able to find a definitive answer in the repository. It might be worth checking the documentation or release notes for that specific version.
I hope this helps! If you have any other questions or need further clarification, feel free to ask.
Sources
- Filtering retrieval with ConversationalRetrievalChain
- libs/langchain/langchain/schema/vectorstore.py
This response is meant to be useful and save you time. 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.
If you want to continue the conversation, start your reply with @dosu-bot.
Quicksilver466
Issue you'd like to raise.
I am using langchain version 0.0.281 and using Milvus as the vector store. For now, I have a Milvus-lite instance on my device. I tried using paritions with base vectorstore operations like
add_documentsandsimilarity_searchand its working fine i.e it inserts and searches from the mentioned partition only. But when I try to use partitions with ConversationalRetrievalChain and even RetrievalQA it doesn't work i.e it retrieves documents through all partitions. When usingfrom_llmclassmethod I tried giving the partition names like below:QA = ConversationalRetrievalChain.from_llm( llm=LLM, retriever=VECTOR_DB.vec_db.as_retriever(partition_names=partition_names), memory=MEMORY, return_source_documents=True )But this doesn't ensure that the vector store will only search the mentioned partitions, rather the vectorestore searches through all the documents present regardless of partition. Is there anyway I can limit the retrieval of documents to only those present in mentioned partition.Suggestion:
No response