Relevance Evaluator: Support Conversational Format in specs (#4571)

Updated the dataMappingSchema to allow 'query' and 'response' to be either a string or an array of objects in Relevance Evaluator specs

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

Author: m7md7sien

assets/evaluators/builtin/relevance/spec.yaml

@@ -1,6 +1,6 @@
 type: "evaluator"
 name: "builtin.relevance"
-version: 3
+version: 4
 displayName: "Relevance-Evaluator"
 description: "Assesses how well the response matches the user’s intent or question. Higher scores mean better alignment with the prompt. It’s best used for generative business writing such as summarizing meeting notes, creating marketing materials, and drafting email."
 evaluatorType: "builtin"
@@ -23,14 +23,22 @@ dataMappingSchema:
   type: "object"
   properties:
     query:
-      type: "string"
+      anyOf:
+        - type: "string"
+        - type: "array"
+          items:
+            type: "object"
     response:
-      type: "string"
+      anyOf:
+        - type: "string"
+        - type: "array"
+          items:
+            type: "object"
   required: ["query", "response"]
 outputSchema:
   relevance:
     type: "ordinal"
     desirable_direction: "increase"
     min_value: 1
     max_value: 5
-path: ./evaluator
+path: ./evaluator