Langchain Chroma Warning message

[Shiro-LK]

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

from langchain_community.vectorstores.chroma import Chroma
from langchain.schema.document import Document
from langchain_huggingface.embeddings import HuggingFaceEmbeddings

# embeddings
model_kwargs = {'trust_remote_code': True}
model_name = "Alibaba-NLP/gte-base-en-v1.5"
hf_embeddings = HuggingFaceEmbeddings(
    model_name=model_name, model_kwargs=model_kwargs
)

# Chroma

db = Chroma(persist_directory="chroma_db", embedding_function=hf_embeddings, 
                   collection_name = "full_docs",
                   collection_metadata={"hnsw:space": "cosine"}
                   )
texts = [f'this is document n{i}' for i in range(100)]

documents = [Document(page_content=t) for t in texts]

db.add_documents(documents)
indexes = list(range(30))
db._collection.delete(ids=[db.get()['ids'][i] for i in indexes])

Description

Every time I am deleting documents from my DB with Chroma I had a warning message, and would like to understand if there is a way to remove it (basically not raising this warning). I suppose the way I am deleting is incomplete. It is raised when I am deleting multiples docs.

Here an example of Warning:

WARNING:chromadb.segment.impl.vector.brute_force_index:Delete of nonexisting embedding ID: 0320ada1-8930-4145-9418-fcf3f8f6a3bc

System Info

System Information

OS: Linux
OS Version: #46~22.04.1-Ubuntu SMP PREEMPT_DYNAMIC Wed Jun 7 15:06:04 UTC 20
Python Version: 3.9.16 (main, Dec 7 2022, 01:11:51)
[GCC 9.4.0]

Package Information

langchain_core: 0.2.28
langchain: 0.2.12
langchain_community: 0.2.11
langsmith: 0.1.96
langchain_chroma: 0.1.2
langchain_huggingface: 0.0.3
langchain_text_splitters: 0.2.2

Packages not installed (Not Necessarily a Problem)

The following packages were not found:

langgraph
langserve

[dosubot[bot]]

To suppress the warning message when deleting multiple documents from your Chroma vector store, you can modify the delete method to handle or suppress warnings. Specifically, you can use the suppress_langchain_deprecation_warning context manager provided by LangChain to suppress LangChainDeprecationWarning.

Here is an example of how you might modify the delete method to suppress these warnings:

import warnings
from contextlib import contextmanager

@contextmanager
def suppress_langchain_deprecation_warning():
    """Context manager to suppress LangChainDeprecationWarning."""
    with warnings.catch_warnings():
        warnings.simplefilter("ignore", LangChainDeprecationWarning)
        warnings.simplefilter("ignore", LangChainPendingDeprecationWarning)
        yield

def delete(self, ids: Optional[List[str]] = None, **kwargs: Any) -> None:
    """Delete by vector IDs.

    Args:
        ids: List of ids to delete.
    """
    with suppress_langchain_deprecation_warning():
        self._collection.delete(ids=ids)

This code will suppress all LangChainDeprecationWarning warnings during the execution of the delete method. Adjust the context manager if you need to suppress different types of warnings ^[1][2][3].

To continue talking to Dosu, mention @dosu.

[Shiro-LK]

hello, my goal is not to hide the warning but more understanding why it is happening and how to not make it happen without suppressing warning code

[dosubot[bot]]

The warning message "Delete of nonexisting embedding ID" occurs because the delete method is attempting to delete an ID that does not exist in the collection. This can happen if the IDs you are trying to delete are not present in the collection.

To avoid this warning, you should ensure that the IDs you are trying to delete actually exist in the collection. You can do this by checking the existing IDs in the collection before attempting to delete them.

Here is an updated version of your code snippet that includes a check to ensure the IDs exist before attempting to delete them:

from langchain_community.vectorstores.chroma import Chroma
from langchain.schema.document import Document
from langchain_huggingface.embeddings import HuggingFaceEmbeddings

# embeddings
model_kwargs = {'trust_remote_code': True}
model_name = "Alibaba-NLP/gte-base-en-v1.5"
hf_embeddings = HuggingFaceEmbeddings(
    model_name=model_name, model_kwargs=model_kwargs
)

# Chroma
db = Chroma(persist_directory="chroma_db", embedding_function=hf_embeddings, 
                   collection_name = "full_docs",
                   collection_metadata={"hnsw:space": "cosine"}
                   )
texts = [f'this is document n{i}' for i in range(100)]

documents = [Document(page_content=t) for t in texts]

db.add_documents(documents)
indexes = list(range(30))

# Get existing IDs in the collection
existing_ids = db.get()['ids']

# Filter the IDs to delete to ensure they exist in the collection
ids_to_delete = [db.get()['ids'][i] for i in indexes if db.get()['ids'][i] in existing_ids]

# Delete the filtered IDs
db._collection.delete(ids=ids_to_delete)

This code snippet first retrieves the existing IDs in the collection and then filters the IDs to delete to ensure they exist in the collection before attempting to delete them. This should prevent the warning message from occurring ^[1][2].

To continue talking to Dosu, mention @dosu.

[artur-augustyniak]

If you don't care about trying to delete non-existent ids, you can just silence warnings:

import logging
logging.getLogger("chromadb").setLevel(logging.ERROR)

[XariZaru]

Weirdly, even if I pass every ID in to delete, the collection size never gets smaller in bytes. It's only infinitely growing. It seems like records are deleted but size of db is not shrinking. As if continuing to be preallocated.

[kabir-tesseralabs]

Same issue and the bot answer is useless. The real question is why is this happening - are documents not actually being deleted?