add relevance scores and to search response models (#156)

* Update vespa test files

* bump version to 1.9.7

* add hits relevance scoring

* add family relevance

* add rank features to hit response models

* bump version to 1.10.0

---------

Co-authored-by: kdutia <[email protected]>

src/cpr_sdk/models/search.py

@@ -381,6 +381,8 @@ class Hit(BaseModel):
     corpus_import_id: Optional[str] = None
     metadata: Optional[Sequence[dict[str, str]]] = None
     concepts: Optional[Sequence[Concept]] = None
+    relevance: Optional[float] = None
+    rank_features: Optional[dict[str, float]] = None
 
     @classmethod
     def from_vespa_response(cls, response_hit: dict) -> "Hit":
@@ -445,6 +447,8 @@ def from_vespa_response(cls, response_hit: dict) -> "Document":
             corpus_import_id=fields.get("corpus_import_id"),
             metadata=fields.get("metadata"),
             concepts=fields.get("concepts"),
+            relevance=response_hit.get("relevance"),
+            rank_features=fields.get("summaryfeatures"),
         )
 
 
@@ -498,6 +502,8 @@ def from_vespa_response(cls, response_hit: dict) -> "Passage":
             text_block_coords=fields.get("text_block_coords"),
             metadata=fields.get("metadata"),
             concepts=fields.get("concepts"),
+            relevance=response_hit.get("relevance"),
+            rank_features=fields.get("summaryfeatures"),
         )
 
 
@@ -509,6 +515,7 @@ class Family(BaseModel):
     total_passage_hits: int = 0
     continuation_token: Optional[str] = None
     prev_continuation_token: Optional[str] = None
+    relevance: Optional[float] = None
 
 
 class SearchResponse(BaseModel):

src/cpr_sdk/version.py

@@ -1,6 +1,6 @@
 _MAJOR = "1"
-_MINOR = "9"
-_PATCH = "8"
+_MINOR = "10"
+_PATCH = "0"
 _SUFFIX = ""
 
 VERSION_SHORT = "{0}.{1}".format(_MAJOR, _MINOR)

src/cpr_sdk/vespa.py

@@ -131,6 +131,7 @@ def parse_vespa_response(vespa_response: VespaResponse) -> SearchResponse:
         family_hits: List[Hit] = []
         passages_continuation = dig(family, "children", 0, "continuation", "next")
         prev_passages_continuation = dig(family, "children", 0, "continuation", "prev")
+        family_relevance = family.get("relevance")
         for hit in dig(family, "children", 0, "children", default=[]):
             family_hits.append(Hit.from_vespa_response(response_hit=hit))
         families.append(
@@ -140,6 +141,7 @@ def parse_vespa_response(vespa_response: VespaResponse) -> SearchResponse:
                 total_passage_hits=total_passage_hits,
                 continuation_token=passages_continuation,
                 prev_continuation_token=prev_passages_continuation,
+                relevance=family_relevance,
             )
         )
 

tests/test_search_adaptors.py

@@ -65,6 +65,27 @@ def test_vespa_search_adaptor__works(test_vespa):
     assert total_passage_count == response.total_hits
 
 
+@pytest.mark.vespa
+def test_vespa_search_adaptor_relevance_scoring(test_vespa):
+    request = SearchParameters(query_string="the")
+    response = vespa_search(test_vespa, request)
+
+    for family in response.families:
+        assert isinstance(family.relevance, float)
+        for hit in family.hits:
+            assert isinstance(hit.relevance, float)
+
+
+@pytest.mark.vespa
+def test_vespa_search_adaptor_rank_features(test_vespa):
+    request = SearchParameters(query_string="the")
+    response = vespa_search(test_vespa, request)
+
+    for family in response.families:
+        for hit in family.hits:
+            assert isinstance(hit.rank_features, dict)
+
+
 @pytest.mark.vespa
 def test_vespa_search_adaptor__exact_search(test_vespa):
     """Test that exact search works"""