adjust vespa fields

backend/danswer/document_index/vespa/app_config/schemas/danswer_chunk.sd

@@ -31,6 +31,8 @@ schema danswer_chunk {
         # https://docs.vespa.ai/en/attributes.html potential enum store for speed, but probably not worth it
         field source_type type string {
             indexing: summary | attribute
+            rank: filter
+            attribute: fast-search
         }
         # Can also index links https://docs.vespa.ai/en/reference/schema-reference.html#attribute
         # URL type matching
@@ -61,6 +63,8 @@ schema danswer_chunk {
         }
         field hidden type bool {
             indexing: summary | attribute
+            rank: filter
+            attribute: fast-search
         }
         field metadata type string {
             indexing: summary | attribute
@@ -82,10 +86,12 @@ schema danswer_chunk {
         }
         field access_control_list type weightedset<string> {
             indexing: summary | attribute
+            rank: filter
             attribute: fast-search
         }
         field document_sets type weightedset<string> {
             indexing: summary | attribute
+            rank: filter
             attribute: fast-search
         }
     }

backend/danswer/document_index/vespa/index.py

@@ -436,10 +436,20 @@ def _query_vespa(query_params: Mapping[str, str | int]) -> list[InferenceChunk]:
         raise ValueError("No/empty query received")
 
     logger.info("Making query with params: %s", query_params)
-    response = requests.get(SEARCH_ENDPOINT, params=query_params)
+    response = requests.get(
+        SEARCH_ENDPOINT,
+        params=dict(
+            **query_params,
+            **{
+                "presentation.timing": True,
+            },
+        ),
+    )
     response.raise_for_status()
 
-    hits = response.json()["root"].get("children", [])
+    response_json = response.json()
+    logger.info("Response: %s", response_json)
+    hits = response_json["root"].get("children", [])
 
     for hit in hits:
         if hit["fields"].get(CONTENT) is None:

backend/scripts/benchmark_search_isolated.py

@@ -133,10 +133,12 @@ def _measure_vespa_latency(filters: dict = {}):
         "hits": num_to_retrieve,
         "offset": 0,
         "ranking.profile": os.environ.get("VESPA_RANKING_PROFILE", "hybrid_search"),
+        "presentation.timing": True,
         "timeout": "10s",
     }
     start = time.monotonic()
-    _query_vespa(params)
+    response = _query_vespa(params)
+    print(response)
     return time.monotonic() - start