Method BatchEvalHandler

pkg/server/server.go:929–1095

BatchEvalHandler 返回批量评估的HTTP handler

()

// BatchEvalHandler 返回批量评估的HTTP handler
func (s *Server) BatchEvalHandler() http.Handler {
	return http.HandlerFunc(func(w http.ResponseWriter, r *http.Request) {
		start := time.Now()

		if r.Method != http.MethodPost {
			http.Error(w, "method not allowed", http.StatusMethodNotAllowed)
			return
		}

		var req BatchEvalRequest
		if err := json.NewDecoder(r.Body).Decode(&req); err != nil {
			http.Error(w, "invalid JSON body", http.StatusBadRequest)
			return
		}

		if len(req.TestCases) == 0 {
			http.Error(w, "test_cases is required", http.StatusBadRequest)
			return
		}

		// 默认使用基础 scorers
		scorerNames := req.Scorers
		if len(scorerNames) == 0 {
			scorerNames = []string{"keyword_coverage", "lexical_similarity"}
		}

		ctx := r.Context()
		ctx, cancel := context.WithTimeout(ctx, 5*time.Minute)
		defer cancel()

		// 创建 scorers
		scorers := make([]evals.Scorer, 0, len(scorerNames))
		var llmProvider provider.Provider

		for _, name := range scorerNames {
			switch name {
			case "keyword_coverage":
				scorer := evals.NewKeywordCoverageScorer(evals.KeywordCoverageConfig{
					Keywords:        req.Keywords,
					CaseInsensitive: true,
				})
				scorers = append(scorers, scorer)

			case "lexical_similarity":
				scorer := evals.NewLexicalSimilarityScorer(evals.LexicalSimilarityConfig{
					MinTokenLength: 2,
				})
				scorers = append(scorers, scorer)

			case "faithfulness", "hallucination", "answer_relevancy", "context_relevancy",
				"toxicity", "tone_consistency", "coherence", "completeness":
				// LLM-based scorers 需要 Provider
				if llmProvider == nil {
					if req.ProviderConfig == nil {
						http.Error(w, "provider_config is required for LLM-based scorers", http.StatusBadRequest)
						return
					}
					var err error