feat(knowledge): multi-query retrieval with LLM query rewrite

Add a shared structured query builder (knowledge_query.go) and switch
all generation paths to it, so retrieval queries carry task intent
(type, brand, region, keywords, key points) instead of a raw prompt.

ResolveContext now rewrites the query into multiple sub-questions via
the URL-markdown model, embeds and searches each, and merges/dedupes
candidates before rerank; falls back to parsed fallback questions when
rewrite is unavailable or returns too few. Resolve logs gain query
list, count, and rewrite stage/model for observability.

- knowledge_query.go/_test.go: BuildKnowledgeQuery + intent normalization
- knowledge_service.go: per-query embed/search loop, query rewrite, logs
- template_prompt/template_service/prompt_generate/article_imitation/
  kol_generation_worker: adopt the shared query builder
- generation_observability: richer task error logging

server/internal/tenant/app/template_prompt.go

@@ -48,32 +48,39 @@ func buildGenerationPrompt(
 	return strings.Join(sections, "\n\n")
 }
 
-func buildGenerationKnowledgeQuery(params map[string]interface{}) string {
+func buildGenerationKnowledgeQuery(templateKey, templateName string, params map[string]interface{}) string {
 	if len(params) == 0 {
 		return ""
 	}
 
-	parts := make([]string, 0, 8)
-	appendValue := func(value string) {
-		value = strings.TrimSpace(value)
-		if value == "" {
-			return
-		}
-		parts = append(parts, value)
+	intent := knowledgeQueryIntent{
+		TaskType:       templateKnowledgeTaskType(templateKey, templateName),
+		TemplateName:   templateName,
+		Title:          firstNonEmptyPromptString(params, "title", "topic", "product_name", "subject"),
+		Topic:          firstNonEmptyPromptString(params, "topic", "subject", "product_name", "title"),
+		BrandName:      stringValue(params["brand_name"]),
+		Region:         firstNonEmptyPromptString(params, "region", "city", "location", "area", "target_region", "service_area"),
+		PrimaryKeyword: firstNonEmptyPromptString(params, "brand_question", "primary_keyword"),
+		Keywords:       mergeKnowledgeQueryKeywords(params["keywords"], params["supplemental_questions"], params["primary_keyword"]),
+		KeyPoints:      firstNonEmptyPromptString(params, "key_points", "review_intro_hook"),
 	}
+	return buildKnowledgeQueryInputText(intent)
+}
 
-	appendValue(stringValue(params["title"]))
-	appendValue(stringValue(params["topic"]))
-	appendValue(stringValue(params["product_name"]))
-	appendValue(stringValue(params["subject"]))
-	appendValue(stringValue(params["brand_name"]))
-	appendValue(stringValue(params["brand_question"]))
-	appendValue(stringValue(params["primary_keyword"]))
-	appendValue(formatPromptValue(params["supplemental_questions"]))
-	appendValue(stringValue(params["key_points"]))
-	appendValue(stringValue(params["review_intro_hook"]))
-
-	return strings.Join(parts, "\n")
+func templateKnowledgeTaskType(templateKey, templateName string) string {
+	switch strings.TrimSpace(templateKey) {
+	case "top_x_article":
+		return "推荐榜文章"
+	case "product_review":
+		return "产品评测文章"
+	case "research_report":
+		return "研究报告"
+	default:
+		if name := strings.TrimSpace(templateName); name != "" {
+			return name
+		}
+		return "文章生成"
+	}
 }
 
 func buildTemplateSpecificWritingRules(templateKey string, params map[string]interface{}) string {