Error:

[dtthanh1971]

System Info

I try this example code

from langchain.retrievers import ParentDocumentRetriever
from langchain.text_splitter import RecursiveCharacterTextSplitter
from langchain.storage import InMemoryStore

# This text splitter is used to create the parent documents
parent_splitter = RecursiveCharacterTextSplitter(chunk_size=2000)
# This text splitter is used to create the child documents
# It should create documents smaller than the parent
child_splitter = RecursiveCharacterTextSplitter(chunk_size=400)
# The vectorstore to use to index the child chunks
vectorstore = Chroma(embedding_function=OpenAIEmbeddings())
# The storage layer for the parent documents
store = InMemoryStore()

vectorstore = Chroma(collection_name="test", embedding_function=OpenAIEmbeddings())

Initialize the retriever

parent_document_retriever = ParentDocumentRetriever(
vectorstore=vectorstore,
docstore=store,
child_splitter=child_splitter,
parent_splitter=parent_splitter,
)

but I encountered an error:

 1 # Initialize the retriever
----> 2 parent_document_retriever = ParentDocumentRetriever(
      3     vectorstore=vectorstore,
      4     docstore=store,
      5     child_splitter=child_splitter,

TypeError: MultiVectorRetriever.__init__() got an unexpected keyword argument 'child_splitter'

Who can help?

No response

Information

• The official example notebooks/scripts
• My own modified scripts

Related Components

• LLMs/Chat Models
• Embedding Models
• Prompts / Prompt Templates / Prompt Selectors
• Output Parsers
• Document Loaders
• Vector Stores / Retrievers
• Memory
• Agents / Agent Executors
• Tools / Toolkits
• Chains
• Callbacks/Tracing
• Async

Reproduction

from langchain.retrievers import ParentDocumentRetriever
from langchain.text_splitter import RecursiveCharacterTextSplitter
from langchain.storage import InMemoryStore

# This text splitter is used to create the parent documents
parent_splitter = RecursiveCharacterTextSplitter(chunk_size=2000)
# This text splitter is used to create the child documents
# It should create documents smaller than the parent
child_splitter = RecursiveCharacterTextSplitter(chunk_size=400)
# The vectorstore to use to index the child chunks
vectorstore = Chroma(embedding_function=OpenAIEmbeddings())
# The storage layer for the parent documents
store = InMemoryStore()

vectorstore = Chroma(collection_name="test", embedding_function=OpenAIEmbeddings())
# Initialize the retriever
parent_document_retriever = ParentDocumentRetriever(
    vectorstore=vectorstore,
    docstore=store,
    child_splitter=child_splitter,
    parent_splitter=parent_splitter,
)

Expected behavior

I can run.

[dosubot]

🤖

Hello @dtthanh1971! I'm Dosu, a bot here to assist you. I can help with solving bugs, answering questions, and guiding you to become a contributor. While we wait for a human maintainer, feel free to ask me anything. I'm here to help you!

Based on the error message you provided, it seems like the ParentDocumentRetriever class does not accept a child_splitter argument in its __init__() method. The child_splitter and parent_splitter are class attributes, but they are not parameters of the class constructor.

In the provided code, the ParentDocumentRetriever class inherits from MultiVectorRetriever, and there's no __init__() method defined in ParentDocumentRetriever itself. This means it uses the __init__() method from MultiVectorRetriever, which does not include child_splitter or parent_splitter as arguments.

To resolve this issue, you need to define an __init__() method in the ParentDocumentRetriever class that accepts child_splitter and parent_splitter as arguments. Here's an example of how you might do this:

class ParentDocumentRetriever(MultiVectorRetriever):
    def __init__(self, vectorstore, docstore, child_splitter, parent_splitter=None):
        super().__init__(vectorstore, docstore)
        self.child_splitter = child_splitter
        self.parent_splitter = parent_splitter

In this code, super().__init__(vectorstore, docstore) calls the __init__() method of the parent class (MultiVectorRetriever), passing the vectorstore and docstore arguments. Then, it sets the child_splitter and parent_splitter attributes to the values passed in.

You can find more details about the ParentDocumentRetriever class in the LangChain codebase here.

I hope this helps! If you have any other questions, feel free to ask.

Sources

• AttributeError: 'RecursiveCharacterTextSplitter' object has no attribute 'split_documents'
• libs/langchain/langchain/retrievers/parent_document_retriever.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.

[jacoblee93]

I suspect this was caused by #14202 - can you try downgrading one version for now? I am working on a fix ASAP.

[dtthanh1971]

I suspect this was caused by #14202 - can you try downgrading one version for now? I am working on a fix ASAP.

Thank for your support. but what version can I downgrade for now (my current version is 0.0.346).

[jacoblee93]

One patch down should do it (0.0.345)

[dtthanh1971]

One patch down should do it (0.0.345)

Thank you. This workaround is helpful.