[Bug]: QdrantVectorStore crashes with latest qdrant-client – search_batch has been removed

[sudeeshnarayanan]

Bug Description

QdrantVectorStore is currently broken with the latest versions of qdrant-client because it still calls search_batch, which no longer exists on QdrantClient. The correct method is now search_batch_points.

This results in a hard runtime error

Version

llama-index-vector-stores-qdrant 0.8.7

Steps to Reproduce

Steps to reproduce

1. Install the Qdrant integration with a recent qdrant-client:

   pip install llama-index-vector-stores-qdrant qdrant-client
   
    2.	Start Qdrant:
   

docker run -p 6333:6333 qdrant/qdrant

3.	Run this minimal script:

from llama_index.vector_stores.qdrant import QdrantVectorStore
from llama_index.core.schema import TextNode
from llama_index.core.vector_stores.types import VectorStoreQuery
import qdrant_client

client = qdrant_client.QdrantClient("http://localhost:6333")

vs = QdrantVectorStore(
collection_name="test",
client=client,
enable_hybrid=True, # triggers the broken path
)

vs.add([TextNode(text="hi", embedding=[0.1,0.2,0.3])])

query = VectorStoreQuery(query_embedding=[0.1,0.2,0.3], similarity_top_k=1)
vs.query(query) # <-- crashes

4.	You immediately get:

"'QdrantClient' object has no attribute 'search_batch'

Relevant Logs/Tracbacks

[dosubot]

Hi @sudeeshnarayanan! I'm Dosu and I’m helping the llama_index team.

You're correct—QdrantVectorStore is broken with the latest qdrant-client because it still calls the deprecated search_batch method, which was removed and replaced by search_batch_points. This affects both the sync and async query paths when enable_hybrid=True. The fix is to update all usages of self._client.search_batch and self._aclient.search_batch to self._client.search_batch_points and self._aclient.search_batch_points, respectively, passing the same arguments. The response handling logic should remain compatible, as the new method returns a list of SearchResult objects just like before. Until this is patched in the integration, you can work around the issue by pinning qdrant-client to a version before the breaking change (pre-1.16.0) source.

No merged PR or official fix is available yet. If you need to patch locally, search for search_batch in llama_index/vector_stores/qdrant/base.py and replace it with search_batch_points in both the query and aquery methods.

To reply, just mention @dosu.

---

^{How did I do? Good | Irrelevant | Incorrect | Verbose | Hallucination | Report 🐛 | Other}