resolve conflicts

app.py

@@ -11,12 +11,7 @@ neural_searcher = NeuralSearcher(collection_name=os.getenv('COLLECTION_NAME'))
 
 @app.get("/api/search")
 async def search(q: str):
-    # if not q:
-    #     raise HTTPException(status_code=400, detail="Bad request.")
+    data = await neural_searcher.search(text=q)
+    return data
     
-    # try:
-        data = await neural_searcher.search(text=q)
-        return data
-    # except:
-    #     raise HTTPException(status_code=500, detail="Internal server error.")
 

neural_searcher.py

@@ -1,5 +1,5 @@
 from qdrant_client import QdrantClient
-from fastembed import SparseTextEmbedding
+from fastembed import SparseTextEmbedding, LateInteractionTextEmbedding
 from qdrant_client import QdrantClient, models
 from sentence_transformers import SentenceTransformer
 import os
@@ -10,34 +10,38 @@ class NeuralSearcher:
         self.collection_name = collection_name
         self.dense_model = SentenceTransformer(os.getenv('DENSE_MODEL'),device="cpu")
         self.sparse_model = SparseTextEmbedding(os.getenv('SPARSE_MODEL'))
+        self.late_interaction_model = LateInteractionTextEmbedding(os.getenv('LATE_INTERACTION_MODEL'))
         self.qdrant_client = QdrantClient(os.getenv('QDRANT_URL'),api_key=os.getenv('QDRANT_API_KEY'))
     
     async def search(self, text: str):
   
-        dense_query = self.dense_model.encode(text).tolist()
-        sparse_query = self.sparse_model.query_embed(text)
-        
-        search_result = self.qdrant_client.query_points_groups(
-            collection_name= self.collection_name,
-            group_by="dbid",
-            prefetch=[
+        dense_query = next(self.dense_model.encode(text))
+        sparse_query = next(self.sparse_model.query_embed(text))
+        late_query = next(self.late_interaction_embedding_model.query_embed(text))
+
+        prefetch = [
                 models.Prefetch(
                     query=dense_query,
                     using=os.getenv('DENSE_MODEL'),
                     limit=100
                 ),
                 models.Prefetch(
-                    query=next(sparse_query).as_object(),
+                    query=models.SparseVector(**sparse_query.as_object()),
                     using=os.getenv('SPARSE_MODEL'),
                     limit=100
                 )
-            ],
-            query=models.FusionQuery(
-                fusion=models.Fusion.RRF
-            ),
+        ]
+
+        search_result = self.qdrant_client.query_points_groups(
+            collection_name= self.collection_name,
+            group_by="dbid",
+            prefetch=prefetch,
+            group_size=3,
+            query=late_query,
+            using=os.getenv('LATE_INTERACTION_MODEL'),
+            with_payload=True,
             score_threshold=0.8,
             limit = 10
-        ).points
-
-        payloads = [hit.payload for hit in search_result]
-        return payloads
+        ).groups
+        
+        return search_result