diff --git a/backend/scripts/benchmark_search.py b/backend/scripts/benchmark_search.py
index 88d94a830..e3d80505e 100644
--- a/backend/scripts/benchmark_search.py
+++ b/backend/scripts/benchmark_search.py
@@ -126,7 +126,7 @@ def generate_random_sentence():
     return sentence
 
 
-def _measure_hybrid_search_latency(filters: dict = {}):
+def _measure_hybrid_search_latency(filters: dict | None = None):
     search_type = os.environ.get("VESPA_RANKING_PROFILE", "hybrid_search")
 
     start = time.monotonic()
@@ -135,7 +135,7 @@ def _measure_hybrid_search_latency(filters: dict = {}):
         json={
             "query": generate_random_sentence(),
             "collection": DOCUMENT_INDEX_NAME,
-            "filters": filters,
+            "filters": filters or {},
             "enable_auto_detect_filters": False,
             "search_type": SearchType.HYBRID.value
             if search_type == "hybrid_search"
diff --git a/backend/scripts/benchmark_search_isolated.py b/backend/scripts/benchmark_search_isolated.py
index 21d9d1366..a97ba799e 100644
--- a/backend/scripts/benchmark_search_isolated.py
+++ b/backend/scripts/benchmark_search_isolated.py
@@ -1,12 +1,14 @@
 import os
 import random
 import time
+from collections.abc import Mapping
 
 import nltk
+import requests
 
 from danswer.configs.app_configs import DOC_TIME_DECAY
 from danswer.configs.app_configs import DOCUMENT_INDEX_NAME
-from danswer.document_index.vespa.index import _query_vespa
+from danswer.document_index.vespa.index import SEARCH_ENDPOINT
 from danswer.search.search_runner import embed_query
 
 # Download the wordlist
@@ -127,6 +129,25 @@ def generate_random_sentence():
     return sentence
 
 
+def _query_vespa(query_params: Mapping[str, str | int]) -> list:
+    response = requests.get(
+        SEARCH_ENDPOINT,
+        params=dict(
+            **query_params,
+            **{
+                "presentation.timing": True,
+            },
+        ),
+    )
+    response.raise_for_status()
+
+    response_json = response.json()
+    print("timing info", response_json.get("timing"))
+    hits = response_json["root"].get("children", [])
+
+    return hits
+
+
 def _measure_vespa_latency(filters: dict = {}):
     # yql = (
     #     VespaIndex.yql_base
@@ -135,8 +156,8 @@ def _measure_vespa_latency(filters: dict = {}):
     # )
     yql = (
         f"select "
-        f"documentid, "
-        f"content "
+        f"documentid "
+        # f"content "
         f"from {DOCUMENT_INDEX_NAME} where " + '({grammar: "weakAnd"}userInput(@query))'
     )
     query = generate_random_sentence()