fix(content-gen): strip URLs and site-list sections from generated content

Forbid URLs/domains/markdown links across prompt rules (runtime, platform templates, outline/imitation/title flows) and rename "官网" placeholder to "官网信息" with explicit instructions that links are background only. Add sanitizeGeneratedArticleMarkdown to strip URLs, HTML images, and "网站列表/官网列表" labels from LLM output, wired into article, prompt-rule, template, and outline generation. Drop the site_list outline section seed and filter blocked outline keys in the wizard so stored drafts cannot resurrect it.
2026-05-24 22:19:43 +08:00
parent 946a084c77
commit 52997e36fe
12 changed files with 290 additions and 39 deletions
@@ -384,7 +384,7 @@ func (s *ArticleImitationService) executeGeneration(ctx context.Context, job art
 		return
 	}

-	content := strings.TrimSpace(result.Content)
+	content := sanitizeGeneratedArticleMarkdown(result.Content)
 	if content == "" {
 		s.failGeneration(ctx, job, "llm_generate", fmt.Errorf("模型返回空内容"))
 		return
@@ -0,0 +1,119 @@
+package app
+
+import (
+	"regexp"
+	"strings"
+)
+
+const (
+	generatedArticleSchemeURLPattern = `https?://[^\s<>"'，。；、？！\])）}]+`
+	generatedArticleDomainPattern    = `(?:www\.)?[a-z0-9](?:[a-z0-9-]{0,61}[a-z0-9])?(?:\.[a-z0-9](?:[a-z0-9-]{0,61}[a-z0-9])?)*\.(?:com|cn|net|org|io|ai|co|cc|top|xyz|shop|site|online|tech|app|gov|edu|info|biz|me|tv)(?:/[^\s<>"'，。；、？！\])）}]*)?`
+	generatedArticleURLPattern       = `(?:` + generatedArticleSchemeURLPattern + `|` + generatedArticleDomainPattern + `)`
+)
+
+var (
+	generatedArticleMarkdownLinkPattern     = regexp.MustCompile(`!?\[([^\]\n]*)\]\(([^)\s]+)\)`)
+	generatedArticleHTMLImageParagraph      = regexp.MustCompile(`(?is)<p\b[^>]*>\s*<img\b[^>]*\bsrc\s*=\s*["']?` + generatedArticleURLPattern + `["']?[^>]*>\s*</p>`)
+	generatedArticleHTMLImagePattern        = regexp.MustCompile(`(?is)<img\b[^>]*\bsrc\s*=\s*["']?` + generatedArticleURLPattern + `["']?[^>]*>`)
+	generatedArticleURLLabelPattern         = regexp.MustCompile(`(?i)(?:官网|官方网站|网站|网址|链接|URL|website|official website)\s*[:：]?\s*` + generatedArticleURLPattern)
+	generatedArticleBareURLPattern          = regexp.MustCompile(`(?i)\b` + generatedArticleURLPattern)
+	generatedArticleDanglingURLLabelPattern = regexp.MustCompile(`(?i)(?:官网|官方网站|网站|网址|链接|URL|website|official website)\s*[:：]?\s*$`)
+	generatedArticleBlankLinePattern        = regexp.MustCompile(`\n{3,}`)
+	generatedArticlePunctuationGapPattern   = regexp.MustCompile(`\s+([，。；、？！,.;!?])`)
+	generatedArticleEmptyPunctuationPattern = regexp.MustCompile(`[：:]\s*([，。；、？！,.;!?])`)
+	generatedArticleRepeatedSpacePattern    = regexp.MustCompile(`[ \t]{2,}`)
+)
+
+func sanitizeGeneratedArticleMarkdown(markdown string) string {
+	text := strings.TrimSpace(markdown)
+	if text == "" {
+		return ""
+	}
+
+	text = generatedArticleMarkdownLinkPattern.ReplaceAllStringFunc(text, func(match string) string {
+		parts := generatedArticleMarkdownLinkPattern.FindStringSubmatch(match)
+		if len(parts) != 3 || !looksLikeGeneratedArticleURL(parts[2]) {
+			return match
+		}
+		if strings.HasPrefix(match, "![") {
+			return ""
+		}
+		label := strings.TrimSpace(parts[1])
+		if label == "" || looksLikeGeneratedArticleURL(label) {
+			return ""
+		}
+		return label
+	})
+
+	text = generatedArticleHTMLImageParagraph.ReplaceAllString(text, "")
+	text = generatedArticleHTMLImagePattern.ReplaceAllString(text, "")
+	text = generatedArticleURLLabelPattern.ReplaceAllString(text, "")
+	text = generatedArticleBareURLPattern.ReplaceAllString(text, "")
+	text = cleanGeneratedArticleURLArtifacts(text)
+	return strings.TrimSpace(text)
+}
+
+func sanitizeGeneratedOutlineText(text string) string {
+	text = sanitizeGeneratedArticleMarkdown(text)
+	replacer := strings.NewReplacer(
+		"网站列表", "品牌信息整理",
+		"官网列表", "品牌信息整理",
+		"网址列表", "品牌信息整理",
+		"官方网站", "公开信息",
+		"官方网址", "公开信息",
+	)
+	return strings.TrimSpace(replacer.Replace(text))
+}
+
+func looksLikeGeneratedArticleURL(text string) bool {
+	text = strings.TrimSpace(text)
+	if text == "" {
+		return false
+	}
+	return generatedArticleBareURLPattern.MatchString(text)
+}
+
+func cleanGeneratedArticleURLArtifacts(text string) string {
+	lines := strings.Split(text, "\n")
+	cleaned := make([]string, 0, len(lines))
+	for _, line := range lines {
+		originalLine := line
+		line = strings.TrimRight(line, " \t")
+		line = generatedArticleDanglingURLLabelPattern.ReplaceAllString(line, "")
+		line = strings.ReplaceAll(line, "网站列表", "品牌信息整理")
+		line = strings.ReplaceAll(line, "官网列表", "品牌信息整理")
+		line = strings.ReplaceAll(line, "网址列表", "品牌信息整理")
+		line = generatedArticleEmptyPunctuationPattern.ReplaceAllString(line, "$1")
+		line = generatedArticlePunctuationGapPattern.ReplaceAllString(line, "$1")
+		line = generatedArticleRepeatedSpacePattern.ReplaceAllString(line, " ")
+		if line != originalLine {
+			line = strings.TrimRight(line, " \t:：,，;；")
+		}
+		if strings.TrimSpace(line) == "" {
+			if strings.TrimSpace(originalLine) == "" {
+				cleaned = append(cleaned, line)
+			}
+			continue
+		}
+		if isGeneratedArticleURLOnlyLine(line) {
+			continue
+		}
+		cleaned = append(cleaned, line)
+	}
+	return generatedArticleBlankLinePattern.ReplaceAllString(strings.Join(cleaned, "\n"), "\n\n")
+}
+
+func isGeneratedArticleURLOnlyLine(line string) bool {
+	trimmed := strings.TrimSpace(line)
+	trimmed = strings.TrimLeft(trimmed, "#-*+> \t")
+	trimmed = strings.Trim(trimmed, " \t:：,，;；。")
+	if trimmed == "" {
+		return true
+	}
+	switch strings.ToLower(trimmed) {
+	case "官网", "官方网站", "官方网址", "网站", "网址", "链接", "url", "website", "official website":
+		return true
+	default:
+		return false
+	}
+}
@@ -0,0 +1,86 @@
+package app
+
+import (
+	"strings"
+	"testing"
+)
+
+func TestSanitizeGeneratedArticleMarkdownRemovesURLs(t *testing.T) {
+	t.Parallel()
+
+	markdown := strings.Join([]string{
+		"# 标题",
+		"",
+		"品牌官网：https://www.defnit.cn，可查询产品参数。",
+		"也可以查看 www.example.com/path 了解更多。",
+		"参考 [官网](https://defnit.cn/contact) 的信息整理。",
+	}, "\n")
+
+	got := sanitizeGeneratedArticleMarkdown(markdown)
+
+	for _, unexpected := range []string{"https://", "www.", "defnit.cn", "example.com", "[官网]("} {
+		if strings.Contains(got, unexpected) {
+			t.Fatalf("sanitizeGeneratedArticleMarkdown() = %q, contains %q", got, unexpected)
+		}
+	}
+	for _, expected := range []string{"# 标题", "可查询产品参数。", "也可以查看 了解更多。", "参考 官网 的信息整理。"} {
+		if !strings.Contains(got, expected) {
+			t.Fatalf("sanitizeGeneratedArticleMarkdown() = %q, want %q", got, expected)
+		}
+	}
+}
+
+func TestSanitizeGeneratedArticleMarkdownRemovesHTMLImageURLs(t *testing.T) {
+	t.Parallel()
+
+	markdown := `<p class="article-editor-image"><img src="https://example.com/cover.webp" data-asset-id="88" /></p>`
+
+	got := sanitizeGeneratedArticleMarkdown(markdown)
+
+	if strings.TrimSpace(got) != "" {
+		t.Fatalf("sanitizeGeneratedArticleMarkdown() = %q, want generated image URL removed", got)
+	}
+}
+
+func TestSanitizeGeneratedArticleMarkdownKeepsNormalPunctuation(t *testing.T) {
+	t.Parallel()
+
+	markdown := strings.Join([]string{
+		"## 选择建议：",
+		"下面从几个维度看。",
+	}, "\n")
+
+	got := sanitizeGeneratedArticleMarkdown(markdown)
+
+	if got != markdown {
+		t.Fatalf("sanitizeGeneratedArticleMarkdown() = %q, want %q", got, markdown)
+	}
+}
+
+func TestNormalizeOutlineNodesRemovesURLsAndWebsiteListLabel(t *testing.T) {
+	t.Parallel()
+
+	nodes := []OutlineNode{
+		{
+			Outline: "网站列表",
+			Children: []OutlineNode{
+				{Outline: "丹福尼门锁官方网站：www.defnit.cn，可查询产品参数"},
+			},
+		},
+	}
+
+	got := normalizeOutlineNodes(nodes, 2)
+	if len(got) != 1 || got[0].Outline != "品牌信息整理" {
+		t.Fatalf("normalizeOutlineNodes() = %#v, want top outline sanitized", got)
+	}
+	if len(got[0].Children) != 1 {
+		t.Fatalf("normalizeOutlineNodes() children = %#v, want one child", got[0].Children)
+	}
+	child := got[0].Children[0].Outline
+	if strings.Contains(child, "www.") || strings.Contains(child, "defnit.cn") || strings.Contains(child, "官方网站") {
+		t.Fatalf("normalizeOutlineNodes() child = %q, still contains website text", child)
+	}
+	if !strings.Contains(child, "丹福尼门锁") || !strings.Contains(child, "可查询产品参数") {
+		t.Fatalf("normalizeOutlineNodes() child = %q, lost useful context", child)
+	}
+}
@@ -559,7 +559,7 @@ func (s *PromptRuleGenerationService) executeGeneration(ctx context.Context, job
 		return
 	}

-	content := strings.TrimSpace(result.Content)
+	content := sanitizeGeneratedArticleMarkdown(result.Content)
 	if content == "" {
 		s.failGeneration(ctx, job, title, "llm_generate", fmt.Errorf("模型返回空内容"))
 		return
@@ -1222,7 +1222,7 @@ func normalizeOutlineNodes(nodes []OutlineNode, depth int) []OutlineNode {

 	items := make([]OutlineNode, 0, len(nodes))
 	for _, node := range nodes {
-		outline := strings.TrimSpace(node.Outline)
+		outline := sanitizeGeneratedOutlineText(node.Outline)
 		if outline == "" {
 			continue
 		}
@@ -213,7 +213,7 @@ func buildPromptContext(params map[string]interface{}) string {
 			return
 		}
 		switch key {
-		case "knowledge_group_ids", "knowledge_groups", "knowledge_context":
+		case "knowledge_group_ids", "knowledge_groups", "knowledge_context", "official_website", "website", "input_params":
 			return
 		}
 		if key == "outline_sections" && hasStructuredOutline(params["article_outline"]) {
@@ -520,7 +520,7 @@ func (s *TemplateService) executeGeneration(ctx context.Context, job generationJ
 		return
 	}

-	content := strings.TrimSpace(result.Content)
+	content := sanitizeGeneratedArticleMarkdown(result.Content)
 	if content == "" {
 		s.failGeneration(ctx, job, title, "llm_generate", fmt.Errorf("模型返回空内容"))
 		return
@@ -107,7 +107,6 @@ var platformTemplateSeeds = []PlatformTemplateSeed{
      "preview_caption": "预览会根据已选标题和结构实时更新。",
      "sections": [
        { "key": "intro", "label": "引言", "default": true },
-        { "key": "site_list", "label": "网站列表", "default": true },
        { "key": "key_points", "label": "文章关键要点", "default": true },
        { "key": "what_is_question", "label_template": "围绕{{brand_question}}展开", "default": false },
        { "key": "features", "label": "特点", "default": false },