changes to fast api and neural searcher

app.py

@@ -1,4 +1,4 @@
-from fastapi import FastAPI
+from fastapi import FastAPI, HTTPException
 from neural_searcher import NeuralSearcher
 from huggingface_hub import login
 import os
@@ -10,5 +10,13 @@ app = FastAPI()
 neural_searcher = NeuralSearcher(collection_name=os.getenv('COLLECTION_NAME'))
 
 @app.get("/api/search")
-def search_startup(q: str):
-    return {"result": neural_searcher.search(text=q)}
+async def search(q: str):
+    if not q:
+        raise HTTPException(status_code=400, detail="Bad request.")
+    
+    try:
+        data = await neural_searcher.search(text=q)
+        return data
+    except:
+        raise HTTPException(status_code=500, detail="Internal server error.")
+

neural_searcher.py

@@ -17,24 +17,26 @@ class NeuralSearcher:
         dense_query = self.dense_model.encode(text).tolist()
         sparse_query = self.sparse_model.query_embed(text)
         
-        search_result = self.qdrant_client.query_points(
+        search_result = self.qdrant_client.query_points_groups(
             collection_name= self.collection_name,
+            group_by="dbid",
             prefetch=[
                 models.Prefetch(
                     query=dense_query,
                     using=os.getenv('DENSE_MODEL'),
-                    limit=5
+                    limit=100
                 ),
                 models.Prefetch(
                     query=next(sparse_query).as_object(),
                     using=os.getenv('SPARSE_MODEL'),
-                    limit=5
+                    limit=100
                 )
             ],
             query=models.FusionQuery(
                 fusion=models.Fusion.RRF
             ),
-            limit = 9
+            score_threshold=0.8,
+            limit = 10
         ).points
 
         payloads = [hit.payload for hit in search_result]

requirements.txt

@@ -8,3 +8,4 @@ python-dotenv
 qdrant-client
 qdrant-client[fastembed]>=1.8.2
 sentence-transformers
+firebase