refactor: [Coda] simplify evaluatorConfiger to single loader with standardized key

(LogID: 202509021957170100911101349068BD9)

Co-Authored-By: Coda <[email protected]>

Author: tpfz

backend/modules/evaluation/pkg/conf/evaluator.go

@@ -32,13 +32,8 @@ func NewEvaluatorConfiger(configFactory conf.IConfigLoaderFactory) IConfiger {
 	if err != nil {
 		return nil
 	}
-	codeLoader, err := configFactory.NewConfigLoader("code_evaluator_config.json")
-	if err != nil {
-		return nil
-	}
 	return &evaluatorConfiger{
-		loader:     loader,
-		codeLoader: codeLoader,
+		loader: loader,
 	}
 }
 func (c *evaluatorConfiger) GetEvaluatorTemplateConf(ctx context.Context) (etf map[string]map[string]*evaluatordto.EvaluatorContent) {
@@ -111,8 +106,8 @@ func DefaultEvaluatorPromptMapping() map[string]string {
 	return make(map[string]string)
 }
 func (c *evaluatorConfiger) GetCodeEvaluatorTemplateConf(ctx context.Context) (etf map[string]map[string]*evaluatordto.EvaluatorContent) {
-	// 直接从根级别读取配置，因为JSON文件的根就是我们需要的结构
-	if c.codeLoader.UnmarshalKey(ctx, "", &etf) == nil && len(etf) > 0 {
+	const key = "code_evaluator_template_conf"
+	if c.loader.UnmarshalKey(ctx, key, &etf) == nil && len(etf) > 0 {
 		return etf
 	}
 	return DefaultCodeEvaluatorTemplateConf()
@@ -123,6 +118,5 @@ func DefaultCodeEvaluatorTemplateConf() map[string]map[string]*evaluatordto.Eval
 }
 
 type evaluatorConfiger struct {
-	loader     conf.IConfigLoader
-	codeLoader conf.IConfigLoader
+	loader conf.IConfigLoader
 }

release/deployment/docker-compose/conf/evaluation.yaml

@@ -2140,4 +2140,56 @@ evaluator_template_conf_en-US:
         prompt_source_type: 1
         prompt_template_key: builtin_template_detail
         prompt_template_name: "Detail"
+       receive_chat_history: false
+
+code_evaluator_template_conf:
+  equals_checker:
+    Python:
+      receive_chat_history: false
+      code_evaluator:
+        language_type: "Python"
+        code_content: "def exec_evaluation(turn_data):\n    try:\n        # 获取实际输出和参考输出\n        actual_text = turn_data[\"turn\"][\"eval_target\"][\"actual_output\"][\"text\"]\n        reference_text = turn_data[\"turn\"][\"eval_set\"][\"reference_output\"][\"text\"]\n        \n        # 比较文本相似性或相等性\n        is_equal = actual_text.strip() == reference_text.strip()\n        score = 1.0 if is_equal else 0.0\n        \n        if is_equal:\n            status = \"匹配\"\n        else:\n            status = \"不匹配\"\n        reason = f\"实际输出与参考输出{status}。实际输出: '{actual_text}', 参考输出: '{reference_text}'\"\n        \n        return EvalOutput(score=score, reason=reason, err_msg=\"\")\n        \n    except KeyError as e:\n        return EvalOutput(score=0.0, reason=f\"字段路径未找到: {e}\", err_msg=str(e))\n    except Exception as e:\n        return EvalOutput(score=0.0, reason=f\"评估失败: {e}\", err_msg=str(e))"
+        code_template_key: "equals_checker"
+        code_template_name: "相等性检查器"
+    Python3:
+      receive_chat_history: false
+      code_evaluator:
+        language_type: "Python3"
+        code_content: "def exec_evaluation(turn_data):\n    try:\n        # 获取实际输出和参考输出\n        actual_text = turn_data[\"turn\"][\"eval_target\"][\"actual_output\"][\"text\"]\n        reference_text = turn_data[\"turn\"][\"eval_set\"][\"reference_output\"][\"text\"]\n        \n        # 比较文本相似性或相等性\n        is_equal = actual_text.strip() == reference_text.strip()\n        score = 1.0 if is_equal else 0.0\n        \n        if is_equal:\n            status = \"匹配\"\n        else:\n            status = \"不匹配\"\n        reason = f\"实际输出与参考输出{status}。实际输出: '{actual_text}', 参考输出: '{reference_text}'\"\n        \n        return EvalOutput(score=score, reason=reason, err_msg=\"\")\n        \n    except KeyError as e:\n        return EvalOutput(score=0.0, reason=f\"字段路径未找到: {e}\", err_msg=str(e))\n    except Exception as e:\n        return EvalOutput(score=0.0, reason=f\"评估失败: {e}\", err_msg=str(e))"
+        code_template_key: "equals_checker"
+        code_template_name: "相等性检查器"
+  contains_checker:
+    JavaScript:
+      receive_chat_history: false
+      input_schemas:
+        - name: "input"
+          type: "string"
+          description: "评估输入内容"
+        - name: "reference_output"
+          type: "string"
+          description: "参考输出内容"
+        - name: "actual_output"
+          type: "string"
+          description: "实际输出内容"
+      code_evaluator:
+        language_type: "JavaScript"
+        code_content: "function execEvaluation(turnData) {\n    try {\n        // 获取实际输出和参考输出\n        const actualText = turnData.turn.eval_target.actual_output.text;\n        const referenceText = turnData.turn.eval_set.reference_output.text;\n        \n        // 检查实际输出是否包含参考输出\n        const contains = actualText.includes(referenceText);\n        const score = contains ? 1.0 : 0.0;\n        \n        const status = contains ? \"包含\" : \"不包含\";\n        const reason = `实际输出${status}参考输出。实际输出: '${actualText}', 参考输出: '${referenceText}'`;\n        \n        return { score: score, reason: reason, err_msg: \"\" };\n        \n    } catch (e) {\n        return { score: 0.0, reason: `评估失败: ${e.message}`, err_msg: e.message };\n    }\n}"
+        code_template_key: "contains_checker"
+        code_template_name: "包含性检查器"
+    JS:
       receive_chat_history: false
+      input_schemas:
+        - name: "input"
+          type: "string"
+          description: "评估输入内容"
+        - name: "reference_output"
+          type: "string"
+          description: "参考输出内容"
+        - name: "actual_output"
+          type: "string"
+          description: "实际输出内容"
+      code_evaluator:
+        language_type: "JS"
+        code_content: "function execEvaluation(turnData) {\n    try {\n        // 获取实际输出和参考输出\n        const actualText = turnData.turn.eval_target.actual_output.text;\n        const referenceText = turnData.turn.eval_set.reference_output.text;\n        \n        // 检查实际输出是否包含参考输出\n        const contains = actualText.includes(referenceText);\n        const score = contains ? 1.0 : 0.0;\n        \n        const status = contains ? \"包含\" : \"不包含\";\n        const reason = `实际输出${status}参考输出。实际输出: '${actualText}', 参考输出: '${referenceText}'`;\n        \n        return { score: score, reason: reason, err_msg: \"\" };\n        \n    } catch (e) {\n        return { score: 0.0, reason: `评估失败: ${e.message}`, err_msg: e.message };\n    }\n}"
+        code_template_key: "contains_checker"
+        code_template_name: "包含性检查器"e