fix: search visual cards show localized language first (JA before EN)

- Bilingual format: 【JA】... then 【English】...
- Summary uses localized language (vlm_description_lang) as primary
- Load More function now also supports bilingual format
- Fallback: only vlm_description shows single language without labels

src/views/SearchView.vue

@@ -1178,18 +1178,22 @@ async function sendMessage() {
         const frameNumber = s.frame_number ?? Math.round(ts * fps)
         const windowSec = 5
         
-        // Build bilingual description
+        // Build bilingual description: localized language first, then English
         let vlmDesc = s.vlm_description || ''
         let textContent = vlmDesc
         
         if (bilingual && s.vlm_description_en && s.vlm_description_lang) {
-          // Show both English and preferred language
+          // Show localized language first, then English
-          textContent = `【English】\n${s.vlm_description_en}\n\n【${outputLang.toUpperCase()}】\n${s.vlm_description_lang}`
+          textContent = `【${outputLang.toUpperCase()}】\n${s.vlm_description_lang}\n\n【English】\n${s.vlm_description_en}`
-          vlmDesc = textContent
+          vlmDesc = s.vlm_description_lang
+        } else if (bilingual && s.vlm_description_lang) {
+          // Only localized language available
+          textContent = `【${outputLang.toUpperCase()}】\n${s.vlm_description_lang}`
+          vlmDesc = s.vlm_description_lang
         } else if (bilingual && s.vlm_description_en) {
           // Only English available
           textContent = `【English】\n${s.vlm_description_en}`
-          vlmDesc = textContent
+          vlmDesc = s.vlm_description_en
         }
         
         return {
@@ -1217,16 +1221,22 @@ async function sendMessage() {
         const ts = t.start_time ?? t.timestamp_sec ?? (frame / fps)
         const windowSec = 5
         
-        // Build bilingual description
+        // Build bilingual description: localized language first, then English
         let vlmDesc = t.vlm_description || ''
         let textContent = vlmDesc
         
         if (bilingual && t.vlm_description_en && t.vlm_description_lang) {
-          textContent = `【English】\n${t.vlm_description_en}\n\n【${outputLang.toUpperCase()}】\n${t.vlm_description_lang}`
+          // Show localized language first, then English
-          vlmDesc = textContent
+          textContent = `【${outputLang.toUpperCase()}】\n${t.vlm_description_lang}\n\n【English】\n${t.vlm_description_en}`
+          vlmDesc = t.vlm_description_lang
+        } else if (bilingual && t.vlm_description_lang) {
+          // Only localized language available
+          textContent = `【${outputLang.toUpperCase()}】\n${t.vlm_description_lang}`
+          vlmDesc = t.vlm_description_lang
         } else if (bilingual && t.vlm_description_en) {
+          // Only English available
           textContent = `【English】\n${t.vlm_description_en}`
-          vlmDesc = textContent
+          vlmDesc = t.vlm_description_en
         }
         
         return {
@@ -1323,6 +1333,7 @@ async function loadMore() {
       const base = getApiBase()
       const maxResults = 10
       const outputLang = searchConfig.value.vlm_output_language || 'en'
+      const bilingual = outputLang !== 'en'
       
       // Use offset-based pagination instead of page
       const offset = nextOffset.value
@@ -1339,33 +1350,72 @@ async function loadMore() {
         }).then(r => r.json()).catch(() => ({}))
       ])
       
-      const sceneResults = (scenesRes.scenes || []).map((s: any) => ({
+      const sceneResults = (scenesRes.scenes || []).map((s: any) => {
-        file_uuid: s.file_uuid,
+        let vlmDesc = s.vlm_description || ''
-        file_name: s.file_name || '',
+        let textContent = vlmDesc
-        summary: `[SCENE] ${s.vlm_description || ''}`,
+        
-        similarity: s.score ?? null,
+        if (bilingual && s.vlm_description_en && s.vlm_description_lang) {
-        start_frame: s.frame_number ?? Math.round((s.timestamp_sec || 0) * 30),
+          textContent = `【${outputLang.toUpperCase()}】\n${s.vlm_description_lang}\n\n【English】\n${s.vlm_description_en}`
-        end_frame: (s.frame_number ?? Math.round((s.timestamp_sec || 0) * 30)) + 30,
+          vlmDesc = s.vlm_description_lang
-        start_time: s.timestamp_sec,
+        } else if (bilingual && s.vlm_description_lang) {
-        end_time: (s.timestamp_sec || 0) + 1,
+          textContent = `【${outputLang.toUpperCase()}】\n${s.vlm_description_lang}`
-        source_type: 'scene',
+          vlmDesc = s.vlm_description_lang
-        thumbnail_url: s.thumbnail_url,
+        } else if (bilingual && s.vlm_description_en) {
-        vlm_tags: s.vlm_tags || [],
+          textContent = `【English】\n${s.vlm_description_en}`
-        search_query: q
+          vlmDesc = s.vlm_description_en
-      }))
+        }
+        
+        return {
+          file_uuid: s.file_uuid,
+          file_name: s.file_name || '',
+          summary: vlmDesc ? `[SCENE] ${vlmDesc}` : `[SCENE] ${s.vlm_description || ''}`,
+          text_content: textContent,
+          similarity: s.score ?? null,
+          start_frame: s.frame_number ?? Math.round((s.timestamp_sec || 0) * 30),
+          end_frame: (s.frame_number ?? Math.round((s.timestamp_sec || 0) * 30)) + 30,
+          start_time: s.timestamp_sec,
+          end_time: (s.timestamp_sec || 0) + 1,
+          source_type: 'scene',
+          thumbnail_url: s.thumbnail_url,
+          vlm_tags: s.vlm_tags || [],
+          bilingual,
+          vlm_description_en: s.vlm_description_en,
+          vlm_description_lang: s.vlm_description_lang,
+          search_query: q
+        }
+      })
       
-      const traceResults = (tracesRes.traces || []).map((t: any) => ({
+      const traceResults = (tracesRes.traces || []).map((t: any) => {
-        file_uuid: t.file_uuid,
+        let vlmDesc = t.vlm_description || ''
-        file_name: t.file_name || '',
+        let textContent = vlmDesc
-        summary: `[TRACE] ${t.vlm_description || ''}`,
+        
-        similarity: t.score ?? null,
+        if (bilingual && t.vlm_description_en && t.vlm_description_lang) {
-        start_frame: t.start_frame,
+          textContent = `【${outputLang.toUpperCase()}】\n${t.vlm_description_lang}\n\n【English】\n${t.vlm_description_en}`
-        end_frame: t.end_frame,
+          vlmDesc = t.vlm_description_lang
-        source_type: 'trace',
+        } else if (bilingual && t.vlm_description_lang) {
-        trace_id: t.trace_id,
+          textContent = `【${outputLang.toUpperCase()}】\n${t.vlm_description_lang}`
-        vlm_description: t.vlm_description,
+          vlmDesc = t.vlm_description_lang
-        search_query: q
+        } else if (bilingual && t.vlm_description_en) {
-      }))
+          textContent = `【English】\n${t.vlm_description_en}`
+          vlmDesc = t.vlm_description_en
+        }
+        
+        return {
+          file_uuid: t.file_uuid,
+          file_name: t.file_name || '',
+          summary: vlmDesc || t.text_content || t.vlm_description || 'Person appearance',
+          text_content: textContent,
+          similarity: t.score ?? null,
+          start_frame: t.start_frame,
+          end_frame: t.end_frame,
+          source_type: 'trace',
+          trace_id: t.trace_id,
+          bilingual,
+          vlm_description_en: t.vlm_description_en,
+          vlm_description_lang: t.vlm_description_lang,
+          search_query: q
+        }
+      })
       
       newResults = [...sceneResults, ...traceResults].slice(0, maxResults)
     } else if (currentMode === 'people') {