From 1a855d3e0ebd306206d7ca5db323911aad80a04e Mon Sep 17 00:00:00 2001
From: mark-ly-wang <mark-ly-wang@users.noreply.github.com>
Date: Sat, 25 Apr 2026 19:08:15 +0800
Subject: [PATCH] fix(monitor): extract anthropic text blocks

(cherry picked from commit 76cf1ee834c78e3b15dee6fc23244fbae94a4bd1)
---
 .../service/channel_monitor_checker.go        | 36 ++++++++++++--
 .../channel_monitor_checker_body_test.go      | 47 +++++++++++++++++++
 2 files changed, 79 insertions(+), 4 deletions(-)

diff --git a/backend/internal/service/channel_monitor_checker.go b/backend/internal/service/channel_monitor_checker.go
index ad4058f9e..910a6e80d 100644
--- a/backend/internal/service/channel_monitor_checker.go
+++ b/backend/internal/service/channel_monitor_checker.go
@@ -149,11 +149,11 @@ func pingEndpointOrigin(ctx context.Context, endpoint string) *int {
 	return &ms
 }
 
-// providerAdapter 描述某个 provider 在 challenge 检测中需要的 4 件事：
+// providerAdapter 描述某个 provider 在 challenge 检测中需要的几件事：
 //   - 拼出请求路径（含 model 占位）
 //   - 序列化请求体
 //   - 构造鉴权头
-//   - 从响应 JSON 中按 path 提取文本（gjson path）
+//   - 从响应 JSON 中提取文本（默认按 gjson path；需要时可自定义）
 //
 // 加新 provider 只需要在 providerAdapters 里增加一个条目，无需触碰 callProvider / validateProvider。
 type providerAdapter struct {
@@ -161,6 +161,7 @@ type providerAdapter struct {
 	buildBody    func(model, prompt string) ([]byte, error)
 	buildHeaders func(apiKey string) map[string]string
 	textPath     string // gjson 提取响应文本的 path
+	extractText  func([]byte) string
 }
 
 // providerAdapters 全部已支持的 provider。键值即 MonitorProvider* 字符串。
@@ -184,7 +185,7 @@ var providerAdapters = map[string]providerAdapter{
 				"anthropic-version": monitorAnthropicAPIVersion,
 			}
 		},
-		textPath: "content.0.text",
+		extractText: extractAnthropicMonitorText,
 	},
 	MonitorProviderGemini: {
 		// Gemini 把 model 名写在 URL path 上：/v1beta/models/{model}:generateContent
@@ -293,7 +294,34 @@ func callProvider(ctx context.Context, provider, endpoint, apiKey, model, prompt
 	if provider == MonitorProviderOpenAI && apiMode == MonitorAPIModeResponses {
 		return extractOpenAIResponsesText(respBytes), string(respBytes), status, nil
 	}
-	return gjson.GetBytes(respBytes, adapter.textPath).String(), string(respBytes), status, nil
+	return extractMonitorResponseText(adapter, respBytes), string(respBytes), status, nil
+}
+
+func extractMonitorResponseText(adapter providerAdapter, respBytes []byte) string {
+	if adapter.extractText != nil {
+		return adapter.extractText(respBytes)
+	}
+	return gjson.GetBytes(respBytes, adapter.textPath).String()
+}
+
+func extractAnthropicMonitorText(respBytes []byte) string {
+	content := gjson.GetBytes(respBytes, "content")
+	if !content.IsArray() {
+		return ""
+	}
+
+	parts := make([]string, 0, 1)
+	content.ForEach(func(_, item gjson.Result) bool {
+		if item.Get("type").String() != "text" {
+			return true
+		}
+		text := strings.TrimSpace(item.Get("text").String())
+		if text != "" {
+			parts = append(parts, text)
+		}
+		return true
+	})
+	return strings.Join(parts, "\n")
 }
 
 // extractOpenAIResponsesText 聚合 Responses API 的最终 assistant 文本。
diff --git a/backend/internal/service/channel_monitor_checker_body_test.go b/backend/internal/service/channel_monitor_checker_body_test.go
index bcf7af0b9..3375ee6ac 100644
--- a/backend/internal/service/channel_monitor_checker_body_test.go
+++ b/backend/internal/service/channel_monitor_checker_body_test.go
@@ -451,3 +451,50 @@ func TestRunCheckForModel_ReplaceMode_EmptyResponseIsFailed(t *testing.T) {
 		t.Errorf("failure message should hint replace-mode, got %q", res.Message)
 	}
 }
+
+func TestExtractAnthropicMonitorText(t *testing.T) {
+	tests := []struct {
+		name string
+		body string
+		want string
+	}{
+		{
+			name: "text block after thinking",
+			body: `{"content":[{"type":"thinking","thinking":""},{"type":"text","text":"2"}]}`,
+			want: "2",
+		},
+		{
+			name: "single text block",
+			body: `{"content":[{"type":"text","text":"2"}]}`,
+			want: "2",
+		},
+		{
+			name: "thinking only",
+			body: `{"content":[{"type":"thinking","thinking":""}]}`,
+			want: "",
+		},
+		{
+			name: "multiple text blocks",
+			body: `{"content":[{"type":"text","text":"answer"},{"type":"tool_use","name":"x"},{"type":"text","text":"2"}]}`,
+			want: "answer\n2",
+		},
+	}
+
+	for _, tt := range tests {
+		t.Run(tt.name, func(t *testing.T) {
+			got := extractAnthropicMonitorText([]byte(tt.body))
+			if got != tt.want {
+				t.Fatalf("extractAnthropicMonitorText() = %q, want %q", got, tt.want)
+			}
+		})
+	}
+}
+
+func TestValidateChallenge_AnthropicTextAfterThinking(t *testing.T) {
+	body := []byte(`{"content":[{"type":"thinking","thinking":""},{"type":"text","text":"答案是 2"}]}`)
+	respText := extractAnthropicMonitorText(body)
+
+	if !validateChallenge(respText, "2") {
+		t.Fatalf("validateChallenge(%q, %q) = false, want true", respText, "2")
+	}
+}