sparse vectors name unknown

[EtienneFerrandi]

Checked other resources

• I added a very descriptive title to this issue.
• I searched the LangChain documentation with the integrated search.
• I used the GitHub search to find a similar question and didn't find it.
• I am sure that this is a bug in LangChain rather than my code.
• The bug is not resolved by updating to the latest stable version of LangChain (or the specific integration package).

Example Code

#Step 1

import os
from langchain_huggingface import HuggingFaceEmbeddings
from langchain_qdrant import Qdrant, FastEmbedSparse, RetrievalMode

embeddings = HuggingFaceEmbeddings(model_name='OrdalieTech/Solon-embeddings-large-0.1', model_kwargs={"device": "cuda"})

sparse_embeddings = FastEmbedSparse(model_name="Qdrant/bm25")

vectordb = Qdrant.from_texts(
    texts=texts,
    embedding=embeddings,
    sparse_embedding=sparse_embeddings,
    sparse_vector_name="sparse-vector"
    path=os.path.join(os.getcwd(), 'manuscrits_biblissima_vectordb'),
    collection_name="manuscrits_biblissima",
    retrieval_mode=RetrievalMode.HYBRID,
)

#Step 2

model_kwargs = {"device": "cuda"}
embeddings = HuggingFaceEmbeddings(
    model_name='OrdalieTech/Solon-embeddings-large-0.1',
    model_kwargs=model_kwargs
    )
sparse_embeddings = FastEmbedSparse(
    model_name="Qdrant/bm25",
    model_kwargs=model_kwargs,
    )
        
qdrant = QdrantVectorStore.from_existing_collection(
    collection_name="manuscrits_biblissima",
    path=os.path.join(os.getcwd(), 'manuscrits_biblissima_vectordb'),
    retrieval_mode=RetrievalMode.HYBRID,
    embedding=embeddings, 
    sparse_embedding=sparse_embeddings,
    sparse_vector_name="sparse-vector"
    )

Error Message and Stack Trace (if applicable)

Traceback (most recent call last):
  File "/local/eferra01/data/get_ref_llama3_70B_gguf.py", line 101, in <module>
    qdrant = QdrantVectorStore.from_existing_collection(
  File "/local/eferra01/miniconda3/envs/llama-cpp-env/lib/python3.9/site-packages/langchain_qdrant/qdrant.py", line 286, in from_existing_collection
    return cls(
  File "/local/eferra01/miniconda3/envs/llama-cpp-env/lib/python3.9/site-packages/langchain_qdrant/qdrant.py", line 87, in __init__
    self._validate_collection_config(
  File "/local/eferra01/miniconda3/envs/llama-cpp-env/lib/python3.9/site-packages/langchain_qdrant/qdrant.py", line 937, in _validate_collection_config
    cls._validate_collection_for_sparse(
  File "/local/eferra01/miniconda3/envs/llama-cpp-env/lib/python3.9/site-packages/langchain_qdrant/qdrant.py", line 1022, in _validate_collection_for_sparse
    raise QdrantVectorStoreError(
langchain_qdrant.qdrant.QdrantVectorStoreError: Existing Qdrant collection manuscrits_biblissima does not contain sparse vectors named None. If you want to recreate the collection, set force_recreate parameter to True.

Description

I first create a qdrant database (#Step 1).

Then, in another script, to do RAG, I try to load the database (#Step 2).

However, I have the error above.

I named the sparse vectors when creating the database (Step 1) and took care to mention this name when loading the database for the RAG, (Step 2) but it doesn't seem to have been taken into account...

System Info

langchain-qdrant==0.1.3
OS : Linux
OS Version : Linux dgx 6.1.0-18-amd64 #1 SMP PREEMPT_DYNAMIC Debian 6.1.76-1 (2024-02-01) x86_64 GNU/Linux
Python Version : 3.9.19 | packaged by conda-forge | (main, Mar 20 2024, 12:50:21) \n[GCC 12.3.0]

[wulifu2hao]

I am able to reproduce your issue. However changing "Qdrant.from_texts(" in your step 1 To "QdrantVectorStore.from_texts" fix the issue for me

import os
from langchain_huggingface import HuggingFaceEmbeddings
from langchain_qdrant import FastEmbedSparse, RetrievalMode, QdrantVectorStore

embeddings = HuggingFaceEmbeddings(model_name='OrdalieTech/Solon-embeddings-large-0.1')

sparse_embeddings = FastEmbedSparse(model_name="Qdrant/bm25")

texts = ["the capital of france is paris", "the capital of germany is berlin", "the capital of italy is rome"]
vectordb = QdrantVectorStore.from_texts(
    # vectordb = Qdrant.from_texts(
    texts=texts,
    embedding=embeddings,
    sparse_embedding=sparse_embeddings,
    sparse_vector_name="sparse-vector",
    path=os.path.join(os.getcwd(), 'manuscrits_biblissima_vectordb'),
    collection_name="manuscrits_biblissima",
    retrieval_mode=RetrievalMode.HYBRID,
)
print(vectordb)
        
qdrant = QdrantVectorStore.from_existing_collection(
    collection_name="manuscrits_biblissima",
    path=os.path.join(os.getcwd(), 'manuscrits_biblissima_vectordb'),
    retrieval_mode=RetrievalMode.HYBRID,
    embedding=embeddings, 
    sparse_embedding=sparse_embeddings,
    sparse_vector_name="sparse-vector"
    )
res = qdrant.search("where is the capital of france",search_type="similarity", k=1)
print(res)

[EtienneFerrandi]

yes, it works for me also when I change Qdrant.from_texts by QdrantVectorStore.from_texts.
Thanks !