refactor: Use Qdrant's Query API

pyproject.toml

@@ -37,7 +37,7 @@ torchvision = { version = ">=0.17.0,<0.18.0", optional = true}
 pillow = { version = ">=10.2.0,<11.0.0", optional = true}
 tiktoken = ">=0.6.0,<1.0.0"
 matplotlib = { version = "^3.8.3", optional = true}
-qdrant-client = {version = "^1.8.0", optional = true}
+qdrant-client = {version = "^1.11.1", optional = true}
 google-cloud-aiplatform = {version = "^1.45.0", optional = true}
 requests-mock = "^1.12.1"
 boto3 = { version = "^1.34.98", optional = true }

semantic_router/index/qdrant.py

@@ -5,7 +5,6 @@
 
 from semantic_router.index.base import BaseIndex
 from semantic_router.schema import Metric
-
 from semantic_router.utils.logger import logger
 
 DEFAULT_COLLECTION_NAME = "semantic-router-index"
@@ -97,7 +96,7 @@
 
     def _initialize_clients(self):
         try:
-            from qdrant_client import QdrantClient, AsyncQdrantClient
+            from qdrant_client import AsyncQdrantClient, QdrantClient
 
             sync_client = QdrantClient(
                 location=self.location,
@@ -264,7 +263,7 @@
         top_k: int = 5,
         route_filter: Optional[List[str]] = None,
     ) -> Tuple[np.ndarray, List[str]]:
-        from qdrant_client import models, QdrantClient
+        from qdrant_client import QdrantClient, models
 
         self.client: QdrantClient
         filter = None
@@ -278,15 +277,17 @@
                 ]
             )
 
-        results = self.client.search(
+        results = self.client.query_points(
             self.index_name,
-            query_vector=vector,
+            query=vector,
             limit=top_k,
             with_payload=True,
             query_filter=filter,
         )
-        scores = [result.score for result in results]
-        route_names = [result.payload[SR_ROUTE_PAYLOAD_KEY] for result in results]
+        scores = [result.score for result in results.points]
+        route_names = [
+            result.payload[SR_ROUTE_PAYLOAD_KEY] for result in results.points
+        ]
         return np.array(scores), route_names
 
     async def aquery(
@@ -295,7 +296,7 @@
         top_k: int = 5,
         route_filter: Optional[List[str]] = None,
     ) -> Tuple[np.ndarray, List[str]]:
-        from qdrant_client import models, AsyncQdrantClient
+        from qdrant_client import AsyncQdrantClient, models
 
         self.aclient: Optional[AsyncQdrantClient]
         if self.aclient is None:
@@ -313,15 +314,17 @@
                 ]
             )
 
-        results = await self.aclient.search(
+        results = await self.aclient.query_points(
             self.index_name,
-            query_vector=vector,
+            query=vector,
             limit=top_k,
             with_payload=True,
             query_filter=filter,
         )
-        scores = [result.score for result in results]
-        route_names = [result.payload[SR_ROUTE_PAYLOAD_KEY] for result in results]
+        scores = [result.score for result in results.points]
+        route_names = [
+            result.payload[SR_ROUTE_PAYLOAD_KEY] for result in results.points
+        ]
         return np.array(scores), route_names
 
     def aget_routes(self):