opt: add standard result col to output excel

packages/dbgpt-serve/src/dbgpt_serve/evaluate/service/benchmark/file_parse_service.py

@@ -84,6 +84,7 @@ class FileParseService(ABC):
                 "selfDefineTags",
                 "prompt",
                 "standardAnswerSql",
+                "standardAnswer",
                 "llmCode",
                 "llmOutput",
                 "executeResult",
@@ -121,6 +122,9 @@ class FileParseService(ABC):
                     cm.selfDefineTags,
                     cm.prompt,
                     cm.standardAnswerSql,
+                    json.dumps(cm.standardAnswer, ensure_ascii=False)
+                    if cm.standardAnswer is not None
+                    else "",
                     cm.llmCode,
                     cm.llmOutput,
                     json.dumps(cm.executeResult, ensure_ascii=False)

packages/dbgpt-serve/src/dbgpt_serve/evaluate/service/benchmark/models.py

@@ -117,6 +117,7 @@ class RoundAnswerConfirmModel:
     selfDefineTags: Optional[str]
     prompt: Optional[str]
     standardAnswerSql: Optional[str] = None
+    standardAnswer: Optional[Dict[str, List[str]]] = None
     strategyConfig: Optional[DataCompareStrategyConfig] = None
     llmOutput: Optional[str] = None
     executeResult: Optional[Dict[str, List[str]]] = None

packages/dbgpt-serve/src/dbgpt_serve/evaluate/service/benchmark/user_input_execute_service.py

@@ -128,14 +128,24 @@ class UserInputExecuteService:
 
             strategy_cfg = None
             standard_sql = None
+            standard_answer = None
             if left is not None:
                 standard_sql = left.llmOutput
                 if config.benchmark_mode_type == BenchmarkModeTypeEnum.EXECUTE:
                     strategy_cfg = left.strategyConfig
+                    # 优先使用左侧的执行结果作为标准答案；若无，则尝试从策略配置的 standard_result 取第一项
+                    if left.executeResult is not None:
+                        standard_answer = left.executeResult
+                    elif left.strategyConfig and left.strategyConfig.standard_result:
+                        try:
+                            standard_answer = left.strategyConfig.standard_result[0]
+                        except Exception:
+                            standard_answer = None
                 else:
                     standard_result_list = []
                     if left.executeResult:
                         standard_result_list.append(left.executeResult)
+                        standard_answer = left.executeResult
                     strategy_cfg = DataCompareStrategyConfig(
                         strategy="EXACT_MATCH",
                         order_by=True,
@@ -173,6 +183,7 @@ class UserInputExecuteService:
                     selfDefineTags=inp.self_define_tags,
                     prompt=inp.prompt,
                     standardAnswerSql=standard_sql,
+                    standardAnswer=standard_answer,
                     strategyConfig=strategy_cfg,
                     llmOutput=right.llmOutput if right else None,
                     executeResult=right.executeResult if right else None,