Mock AgentCore GetEvaluator and Evaluate

tools/gen-ai/mock-server/README.md

@@ -20,7 +20,7 @@ and supplies its URL as `${MOCK_SERVER_URL}`.
 | `anthropic_agents`, `mistral_agents` | hosted-agent creation |
 | `assistants` | OpenAI Assistants, threads and runs |
 | `bedrock` | `converse`, `converse-stream`, `invoke` |
-| `bedrock_agent`, `bedrock_agentcore` | agent invocation, knowledge-base retrieval/generation (including streaming), and the memory APIs |
+| `bedrock_agent`, `bedrock_agentcore` | agent invocation, knowledge-base retrieval/generation (including streaming), the memory APIs, and evaluator lookup and evaluation |
 | `cohere` | `/v1` and `/v2` chat and embed |
 | `google_genai` | Gemini and Vertex content generation, code execution, and interaction creation/retrieval with Google Search |
 | `mistral` | chat completions, FIM and embeddings, under `/mistral/v1/…` |

tools/gen-ai/mock-server/src/genai_mock_server/bedrock_agentcore.py

@@ -144,3 +144,55 @@ def delete_memory_record(memory_id, record_id):
         return _json({"message": f"Memory record {record_id} not found"}, status=404)
     _RECORDS[memory_id] = [r for r in store if r["memoryRecordId"] != record_id]
     return _json({"memoryRecordId": record_id})
+
+
+# Evaluations -----------------------------------------------------------------
+
+_EVALUATOR_RATING_SCALE = {
+    "numerical": [
+        {"value": 1.0, "label": "Poor", "definition": "The response does not address the request."},
+        {"value": 3.0, "label": "Fair", "definition": "The response partly addresses the request."},
+        {"value": 5.0, "label": "Good", "definition": "The response fully addresses the request."},
+    ]
+}
+
+
+@bp.route("/evaluators/<evaluator_id>", methods=["GET"])
+def get_evaluator(evaluator_id):
+    return _json(
+        {
+            "evaluatorArn": f"arn:aws:bedrock-agentcore:us-east-1:123456789012:evaluator/{evaluator_id}",
+            "evaluatorId": evaluator_id,
+            "evaluatorName": "response_quality",
+            "evaluatorConfig": {
+                "llmAsAJudge": {
+                    "instructions": "Rate how well the response addresses the request.",
+                    "ratingScale": _EVALUATOR_RATING_SCALE,
+                    "modelConfig": {"bedrockEvaluatorModelConfig": {"modelId": "mock-judge-model"}},
+                }
+            },
+            "level": "TRACE",
+            "status": "ACTIVE",
+            "createdAt": 1735689600.0,
+            "updatedAt": 1735689600.0,
+        }
+    )
+
+
+@bp.route("/evaluations/evaluate/<evaluator_id>", methods=["POST"])
+def evaluate(evaluator_id):
+    return _json(
+        {
+            "evaluationResults": [
+                {
+                    "evaluatorArn": f"arn:aws:bedrock-agentcore:us-east-1:123456789012:evaluator/{evaluator_id}",
+                    "evaluatorId": evaluator_id,
+                    "evaluatorName": "response_quality",
+                    "explanation": "The response fully addresses the request.",
+                    "value": 5.0,
+                    "label": "Good",
+                    "tokenUsage": {"inputTokens": 120, "outputTokens": 30, "totalTokens": 150},
+                }
+            ]
+        }
+    )

tools/gen-ai/mock-server/tests/test_mock_server.py

@@ -198,6 +198,13 @@
         "/api/embed",
         {"model": "nomic-embed-text", "input": ["hi", "there"]},
     ),
+    ("bedrock-agentcore-get-evaluator", "get", "/evaluators/eval-mock-001", None),
+    (
+        "bedrock-agentcore-evaluate",
+        "post",
+        "/evaluations/evaluate/eval-mock-001",
+        {"evaluationInput": {"sessionSpans": []}},
+    ),
 ]
 
 # Resource-creating endpoints mint a fresh id per call, so only the shape is