Improve translated search flow and modal layout

2026-03-13 11:57:58 +09:00
parent 1fc06fb785
commit 6852e07607
4 changed files with 186 additions and 70 deletions
@@ -348,7 +348,9 @@ func rankSearchResults(query string, results []services.SearchResult) []services
 	}
 	negativeTerms := []string{
 		"shocking", "amazing", "crazy", "must watch", "reaction", "gossip", "celebrity",
-		"thumbnail", "meme", "prank", "drama", "breaking", "viral",
+		"thumbnail", "meme", "prank", "drama", "breaking", "viral", "tutorial",
+		"how to", "review", "walkthrough", "course", "lesson", "podcast", "interview",
+		"premiere pro", "after effects", "explained", "breakdown", "vlog",
 	}
 	type scoredResult struct {
 		item  services.SearchResult
@@ -379,11 +381,11 @@ func rankSearchResults(query string, results []services.SearchResult) []services
 		}
 		switch result.Source {
 		case "Google Video":
-			score += 3
+			score += 2
 		case "Envato":
-			score += 4
+			score += 5
 		case "Artgrid":
-			score += 4
+			score += 5
 		}
 		scored = append(scored, scoredResult{item: result, score: score})
 	}
@@ -52,30 +52,30 @@ func (s *SearchService) SearchMedia(queries []string) ([]SearchResult, error) {
 		categories   string
 		engine       string
 		queryBuilder func(string) string
-		match        func(string) bool
+		match        func(SearchResult) bool
 	}{
 		{
 			name:       "Google Video",
 			categories: "videos",
 			engine:     s.GoogleVideoEngine,
 			queryBuilder: func(query string) string {
-				return query
+				return buildGoogleVideoQuery(query)
 			},
-			match: func(string) bool { return true },
+			match: isUsefulGoogleVideoResult,
 		},
 		{
 			name:         "Envato",
 			categories:   "general",
 			engine:       s.WebEngine,
 			queryBuilder: buildEnvatoQuery,
-			match:        isEnvatoURL,
+			match:        isRenderableEnvatoResult,
 		},
 		{
 			name:         "Artgrid",
 			categories:   "general",
 			engine:       s.WebEngine,
 			queryBuilder: buildArtgridQuery,
-			match:        func(link string) bool { return strings.Contains(strings.ToLower(link), "artgrid.io") },
+			match:        isRenderableArtgridResult,
 		},
 	}

@@ -88,10 +88,7 @@ func (s *SearchService) SearchMedia(queries []string) ([]SearchResult, error) {
 			continue
 		}
 		for _, source := range sources {
-			searchQuery := query
-			if source.queryBuilder != nil {
-				searchQuery = source.queryBuilder(query)
-			}
+			searchQuery := source.queryBuilder(query)

 			items, err := s.search(searchQuery, source.categories, source.engine, source.name)
 			if err != nil {
@@ -112,10 +109,7 @@ func (s *SearchService) SearchMedia(queries []string) ([]SearchResult, error) {
 				if item.Link == "" || seen[item.Link] {
 					continue
 				}
-				if source.match != nil && !source.match(item.Link) {
-					continue
-				}
-				if !isRenderableLink(item.Link, item.Source) {
+				if !source.match(item) {
 					continue
 				}
 				seen[item.Link] = true
@@ -138,7 +132,7 @@ func (s *SearchService) search(query, categories, engine, source string) ([]Sear
 	values.Set("q", query)
 	values.Set("format", "json")
 	values.Set("safesearch", "0")
-	values.Set("language", "ko-KR")
+	values.Set("language", "en-US")
 	if categories != "" {
 		values.Set("categories", categories)
 	}
@@ -189,6 +183,71 @@ func (s *SearchService) search(query, categories, engine, source string) ([]Sear
 	return results, nil
 }

+func buildGoogleVideoQuery(query string) string {
+	return fmt.Sprintf(`"%s" ("stock footage" OR "b-roll" OR cinematic OR "establishing shot" OR editorial) -tutorial -"how to" -review -reaction -course -podcast -vlog -interview -breakdown -edit -editing`, query)
+}
+
+func buildEnvatoQuery(query string) string {
+	return fmt.Sprintf(`"%s" ("stock footage" OR "stock video" OR "motion graphics" OR cinematic OR "b-roll") (site:elements.envato.com OR site:videohive.net/item) -site:elements.envato.com/stock-video -site:elements.envato.com/video-templates -site:elements.envato.com/stock-video/stock-footage`, query)
+}
+
+func buildArtgridQuery(query string) string {
+	return fmt.Sprintf(`"%s" ("stock footage" OR "b-roll" OR cinematic OR editorial) site:artgrid.io/clip/`, query)
+}
+
+func isUsefulGoogleVideoResult(result SearchResult) bool {
+	text := strings.ToLower(result.Title + " " + result.Snippet)
+	for _, banned := range []string{
+		"tutorial", "how to", "review", "reaction", "podcast", "interview", "walkthrough",
+		"course", "lesson", "edit tutorial", "editing tutorial", "premiere pro", "after effects",
+		"breakdown", "explained", "vlog",
+	} {
+		if strings.Contains(text, banned) {
+			return false
+		}
+	}
+	for _, desired := range []string{
+		"b-roll", "stock footage", "cinematic", "footage", "establishing shot", "4k",
+	} {
+		if strings.Contains(text, desired) {
+			return true
+		}
+	}
+	lowerLink := strings.ToLower(result.Link)
+	return strings.Contains(lowerLink, "youtube.com/watch") || strings.Contains(lowerLink, "youtu.be/")
+}
+
+func isRenderableEnvatoResult(result SearchResult) bool {
+	parsed, err := url.Parse(result.Link)
+	if err != nil {
+		return false
+	}
+	host := strings.ToLower(parsed.Host)
+	path := strings.Trim(parsed.Path, "/")
+	if strings.Contains(host, "videohive.net") {
+		return strings.HasPrefix(path, "item/")
+	}
+	if strings.Contains(host, "elements.envato.com") {
+		if path == "" || strings.Contains(path, "/") {
+			return false
+		}
+		return regexp.MustCompile(`-[A-Z0-9]{6,}$`).MatchString(path)
+	}
+	return false
+}
+
+func isRenderableArtgridResult(result SearchResult) bool {
+	parsed, err := url.Parse(result.Link)
+	if err != nil {
+		return false
+	}
+	if !strings.Contains(strings.ToLower(parsed.Host), "artgrid.io") {
+		return false
+	}
+	path := strings.Trim(parsed.Path, "/")
+	return regexp.MustCompile(`^clip/[0-9]+/`).MatchString(path)
+}
+
 func firstNonEmpty(values ...string) string {
 	for _, value := range values {
 		if strings.TrimSpace(value) != "" {
@@ -225,39 +284,6 @@ func inferDisplayLink(link string, parsed []any) string {
 	return ""
 }

-func isEnvatoURL(link string) bool {
-	lower := strings.ToLower(link)
-	return strings.Contains(lower, "envato") || strings.Contains(lower, "videohive.net")
-}
-
-func isRenderableLink(link, source string) bool {
-	parsed, err := url.Parse(link)
-	if err != nil {
-		return false
-	}
-	path := strings.Trim(parsed.Path, "/")
-	if path == "" {
-		return false
-	}
-	lower := strings.ToLower(link)
-	switch source {
-	case "Envato":
-		return strings.Contains(lower, "/item/") || strings.Contains(lower, "/stock-video/") || strings.Contains(lower, "/video-templates/")
-	case "Artgrid":
-		return strings.Contains(lower, "artgrid.io") && len(strings.Split(path, "/")) >= 2
-	default:
-		return true
-	}
-}
-
-func buildEnvatoQuery(query string) string {
-	return fmt.Sprintf(`%s ("stock video" OR footage OR "video template" OR cinematic) (site:elements.envato.com/stock-video OR site:elements.envato.com/video-templates OR site:videohive.net/item)`, query)
-}
-
-func buildArtgridQuery(query string) string {
-	return fmt.Sprintf(`%s ("stock footage" OR "b-roll" OR cinematic OR editorial) (site:artgrid.io)`, query)
-}
-
 func deriveThumbnail(link string) string {
 	if videoID := extractYouTubeID(link); videoID != "" {
 		return "https://i.ytimg.com/vi/" + videoID + "/hqdefault.jpg"
@@ -39,9 +39,11 @@ func NewGeminiService(apiKey string) *GeminiService {

 func (g *GeminiService) ExpandQuery(query string) ([]string, error) {
 	if g.APIKey == "" {
-		return fallbackQueryExpansion(query), nil
+		return fallbackQueryExpansion(query, query), nil
 	}

+	englishBase := g.TranslateQuery(query)
+
 	body := map[string]any{
 		"systemInstruction": map[string]any{
 			"parts": []map[string]string{
@@ -55,12 +57,13 @@ func (g *GeminiService) ExpandQuery(query string) ([]string, error) {
 				"parts": []map[string]string{
 					{
 						"text": `Return JSON only in this shape: {"querywords":["..."]}.
-Generate at most 10 concise search variations for media discovery across Google Video, Envato, and Artgrid.
-If the user query is in Korean, include strong English search variants that a stock footage editor would use.
+Generate at most 10 concise English search variations for media discovery across Google Video, Envato, and Artgrid.
+The queries must be usable directly in English search engines for stock footage discovery.
 Prioritize media, video footage, stock footage, cinematic b-roll, editorial footage, and scene-based search terms.
 Avoid celebrity gossip, reaction-style phrasing, clickbait phrasing, and generic web search wording.
-Mix Korean and English when useful, but make sure several queries are clean English production keywords.
-User query: ` + query,
+Do not output Korean unless it is part of a proper noun.
+Original user query: ` + query + `
+English base translation: ` + englishBase,
 					},
 				},
 			},
@@ -86,7 +89,7 @@ User query: ` + query,

 	rawText, err := g.generateText(body)
 	if err != nil {
-		return fallbackQueryExpansion(query), nil
+		return fallbackQueryExpansion(query, englishBase), nil
 	}

 	jsonText, err := extractJSONObject(rawText)
@@ -108,8 +111,9 @@ Output must start with { and end with }.
 Do not add prose, explanations, markdown, code fences, or labels.
 Return exactly this shape: {"querywords":["..."]}.
 Generate up to 10 search queries for media discovery across Google Video, Envato, and Artgrid.
-If the original query is Korean, include strong English stock-footage search phrases.
-User query: ` + query,
+Every query must be in natural English and suitable for stock-footage search.
+Original user query: ` + query + `
+English base translation: ` + englishBase,
 						},
 					},
 				},
@@ -134,20 +138,20 @@ User query: ` + query,
 		}
 		rawText, retryErr := g.generateText(strictBody)
 		if retryErr != nil {
-			return fallbackQueryExpansion(query), nil
+			return fallbackQueryExpansion(query, englishBase), nil
 		}
 		jsonText, err = extractJSONObject(rawText)
 		if err != nil {
-			return fallbackQueryExpansion(query), nil
+			return fallbackQueryExpansion(query, englishBase), nil
 		}
 	}

 	var parsed QueryExpansion
 	if err := json.Unmarshal([]byte(jsonText), &parsed); err != nil {
-		return fallbackQueryExpansion(query), nil
+		return fallbackQueryExpansion(query, englishBase), nil
 	}

-	queries := fallbackQueryExpansion(query)
+	queries := fallbackQueryExpansion(query, englishBase)
 	seen := map[string]bool{}
 	for _, existing := range queries {
 		seen[strings.ToLower(strings.TrimSpace(existing))] = true
@@ -167,6 +171,47 @@ User query: ` + query,
 	return queries, nil
 }

+func (g *GeminiService) TranslateQuery(query string) string {
+	if strings.TrimSpace(query) == "" || looksMostlyASCII(query) || g.APIKey == "" {
+		return strings.TrimSpace(query)
+	}
+
+	body := map[string]any{
+		"systemInstruction": map[string]any{
+			"parts": []map[string]string{
+				{
+					"text": "You translate media search intents into natural English. Output one plain English search phrase only. No labels, no quotes, no explanations.",
+				},
+			},
+		},
+		"contents": []map[string]any{
+			{
+				"parts": []map[string]string{
+					{
+						"text": "Translate this user query into concise English suitable for stock-footage search: " + query,
+					},
+				},
+			},
+		},
+		"generationConfig": map[string]any{
+			"responseMimeType": "text/plain",
+			"temperature":      0.1,
+			"maxOutputTokens":  40,
+		},
+	}
+
+	rawText, err := g.generateText(body)
+	if err != nil {
+		return strings.TrimSpace(query)
+	}
+
+	translated := sanitizePlainEnglishLine(rawText)
+	if translated == "" {
+		return strings.TrimSpace(query)
+	}
+	return translated
+}
+
 func (g *GeminiService) generateText(body map[string]any) (string, error) {
 	rawBody, _ := json.Marshal(body)
 	endpoint := "https://generativelanguage.googleapis.com/v1beta/models/gemini-2.5-flash:generateContent?key=" + g.APIKey
@@ -402,8 +447,11 @@ func truncateForError(text string, limit int) string {
 	return trimmed[:limit] + "..."
 }

-func fallbackQueryExpansion(query string) []string {
-	base := strings.TrimSpace(query)
+func fallbackQueryExpansion(originalQuery, englishQuery string) []string {
+	base := strings.TrimSpace(englishQuery)
+	if base == "" {
+		base = strings.TrimSpace(originalQuery)
+	}
 	candidates := []string{
 		base,
 		base + " b-roll",
@@ -416,6 +464,9 @@ func fallbackQueryExpansion(query string) []string {
 		base + " 4k footage",
 		base + " cinematic b-roll",
 	}
+	if strings.TrimSpace(originalQuery) != "" && !strings.EqualFold(strings.TrimSpace(originalQuery), strings.TrimSpace(englishQuery)) {
+		candidates = append(candidates, strings.TrimSpace(originalQuery))
+	}

 	seen := map[string]bool{}
 	queries := make([]string, 0, len(candidates))
@@ -433,3 +484,38 @@ func fallbackQueryExpansion(query string) []string {
 	}
 	return queries
 }
+
+func sanitizePlainEnglishLine(text string) string {
+	lines := strings.Split(text, "\n")
+	for _, line := range lines {
+		line = strings.TrimSpace(strings.Trim(line, "\"'`"))
+		if line == "" {
+			continue
+		}
+		lower := strings.ToLower(line)
+		for _, prefix := range []string{"translation:", "english:", "translated query:"} {
+			if strings.HasPrefix(lower, prefix) {
+				line = strings.TrimSpace(line[len(prefix):])
+				lower = strings.ToLower(line)
+			}
+		}
+		if strings.HasPrefix(lower, "here is") || strings.HasPrefix(lower, "the translation") {
+			continue
+		}
+		if line != "" {
+			return line
+		}
+	}
+	return ""
+}
+
+func looksMostlyASCII(text string) bool {
+	ascii := 0
+	runes := []rune(text)
+	for _, r := range runes {
+		if r <= 127 {
+			ascii++
+		}
+	}
+	return ascii >= len(runes)*8/10
+}
@@ -70,7 +70,7 @@
    </main>

    <div id="previewModal" class="fixed inset-0 z-50 hidden items-center justify-center bg-black/80 px-4">
-      <div class="w-full max-w-4xl rounded-3xl border border-white/10 bg-zinc-950 p-5 shadow-2xl">
+      <div class="w-full max-w-5xl rounded-3xl border border-white/10 bg-zinc-950 p-5 shadow-2xl">
        <div class="flex items-start justify-between gap-4">
          <div>
            <p class="text-xs uppercase tracking-[0.3em] text-zinc-500">Download Preview</p>
@@ -78,12 +78,12 @@
          </div>
          <button id="closePreviewModal" class="rounded-full border border-white/10 px-3 py-2 text-xs uppercase tracking-[0.2em] text-zinc-300">Close</button>
        </div>
-        <div class="mt-5 grid gap-5 xl:grid-cols-[minmax(0,1.15fr)_360px]">
+        <div class="mt-5 flex flex-col gap-5">
          <div id="previewMediaFrame" class="min-w-0 flex min-h-[320px] items-center justify-center overflow-hidden rounded-3xl border border-white/10 bg-black/30 p-2">
            <video id="previewVideo" class="hidden max-h-[60vh] w-full bg-black object-contain" controls playsinline></video>
            <img id="previewThumbnail" class="max-h-[60vh] w-full object-contain" alt="" />
          </div>
-          <div class="min-w-0 space-y-4">
+          <div class="min-w-0 grid gap-4 lg:grid-cols-2">
            <div class="rounded-2xl border border-white/10 bg-white/[0.03] p-4">
              <div class="flex items-center justify-between text-sm text-zinc-400">
                <span>Detected duration</span>
@@ -116,7 +116,9 @@
              <span class="text-sm text-zinc-400">Quality</span>
              <select id="qualitySelect" class="w-full rounded-2xl border border-white/10 bg-black/40 px-4 py-3 text-sm text-white"></select>
            </label>
-            <button id="confirmDownload" class="w-full rounded-2xl border border-white bg-white px-5 py-3 text-sm font-medium text-black transition hover:bg-zinc-200">Confirm Download</button>
+            <div class="lg:col-span-2">
+              <button id="confirmDownload" class="w-full rounded-2xl border border-white bg-white px-5 py-3 text-sm font-medium text-black transition hover:bg-zinc-200">Confirm Download</button>
+            </div>
          </div>
        </div>
      </div>
@@ -136,6 +138,6 @@
      </a>
    </template>

-    <script src="/app.js?v=20260313c" defer></script>
+    <script src="/app.js?v=20260313d" defer></script>
  </body>
 </html>