fix(audio): remove speed from TTS/cache/MinIO keys; fix presigned URL host rewrite

- Speed is no longer part of Kokoro generation, in-memory cache keys, or MinIO object keys — audio is always generated at 1.0 and playback speed is applied client-side via audioEl.playbackRate - presignAudio() now calls rewriteHost() so chapter audio URLs use the public MinIO endpoint (same as presignVoiceSample already did) - docker-compose.yml: rename MINIO_PUBLIC_ENDPOINT → PUBLIC_MINIO_PUBLIC_URL for the ui service so SvelteKit's $env/dynamic/public picks it up
2026-03-04 19:30:46 +05:00
parent c8e0cf2813
commit acbfafb8cd
9 changed files with 45 additions and 72 deletions
--- a/docker-compose.yml
+++ b/docker-compose.yml
@@ -175,7 +175,7 @@ services:
      POCKETBASE_URL: "http://pocketbase:8090"
      POCKETBASE_ADMIN_EMAIL: "${POCKETBASE_ADMIN_EMAIL:-admin@libnovel.local}"
      POCKETBASE_ADMIN_PASSWORD: "${POCKETBASE_ADMIN_PASSWORD:-changeme123}"
-      MINIO_PUBLIC_ENDPOINT: "${MINIO_PUBLIC_ENDPOINT:-http://localhost:9000}"
+      PUBLIC_MINIO_PUBLIC_URL: "${MINIO_PUBLIC_ENDPOINT:-http://localhost:9000}"
    ports:
      - "${UI_PORT:-5252}:3000"
    healthcheck:
--- a/scraper/internal/server/server.go
+++ b/scraper/internal/server/server.go
@@ -425,9 +425,12 @@ func (s *Server) handleChapterText(w http.ResponseWriter, r *http.Request) {
 // It calls Kokoro's POST /v1/audio/speech with return_download_link=true.
 // Kokoro generates the audio, saves it to its own temp storage, and returns
 // the download filename in the X-Download-Path response header.
-// We cache that filename (in memory, keyed by slug/chapter/voice/speed) and
+// We cache that filename (in memory, keyed by slug/chapter/voice) and
 // return a proxy URL that the browser sets as audio.src.
 //
+// TTS is always generated at speed 1.0; playback speed is controlled
+// client-side via the <audio> element's playbackRate.
+//
 // On a cache hit the proxy URL is returned immediately without re-generating.
 // Concurrent requests for the same key are deduplicated.
 func (s *Server) handleAudioGenerate(w http.ResponseWriter, r *http.Request) {
@@ -438,13 +441,12 @@ func (s *Server) handleAudioGenerate(w http.ResponseWriter, r *http.Request) {
 		return
 	}

-	// Parse optional voice/speed from JSON body.
+	// Parse optional voice from JSON body. Speed is intentionally ignored —
+	// TTS is always generated at 1.0; playback speed is applied client-side.
 	voice := s.kokoroVoice
-	speed := 1.0
 	var body struct {
-		Voice    string  `json:"voice"`
-		Speed    float64 `json:"speed"`
-		MaxChars int     `json:"max_chars"`
+		Voice    string `json:"voice"`
+		MaxChars int    `json:"max_chars"`
 	}
 	if r.Body != nil {
 		_ = json.NewDecoder(r.Body).Decode(&body)
@@ -452,15 +454,12 @@ func (s *Server) handleAudioGenerate(w http.ResponseWriter, r *http.Request) {
 	if body.Voice != "" {
 		voice = body.Voice
 	}
-	if body.Speed > 0 {
-		speed = body.Speed
-	}

-	cacheKey := fmt.Sprintf("%s/%d/%s/%.2f", slug, n, voice, speed)
+	cacheKey := fmt.Sprintf("%s/%d/%s", slug, n, voice)

 	// Fast path: already generated (check persistent store first).
 	if filename, ok := s.store.GetAudioCache(r.Context(), cacheKey); ok {
-		s.writeAudioResponse(w, slug, n, voice, speed, filename)
+		s.writeAudioResponse(w, slug, n, voice, filename)
 		return
 	}

@@ -476,7 +475,7 @@ func (s *Server) handleAudioGenerate(w http.ResponseWriter, r *http.Request) {
 		}
 		// Check store again after waiting.
 		if filename, ok := s.store.GetAudioCache(r.Context(), cacheKey); ok {
-			s.writeAudioResponse(w, slug, n, voice, speed, filename)
+			s.writeAudioResponse(w, slug, n, voice, filename)
 		} else {
 			http.Error(w, `{"error":"audio generation failed"}`, http.StatusInternalServerError)
 		}
@@ -512,10 +511,10 @@ func (s *Server) handleAudioGenerate(w http.ResponseWriter, r *http.Request) {
 		return
 	}

-	// Call Kokoro POST /v1/audio/speech with return_download_link=true.
+	// Call Kokoro POST /v1/audio/speech at speed 1.0.
 	// Kokoro saves the generated audio to its own temp storage and returns the
 	// download path in the X-Download-Path response header.
-	filename, err := s.generateSpeech(r.Context(), text, voice, speed)
+	filename, err := s.generateSpeech(r.Context(), text, voice, 1.0)
 	if err != nil {
 		s.log.Error("kokoro speech generation failed", "slug", slug, "chapter", n, "err", err)
 		http.Error(w, `{"error":"speech generation failed"}`, http.StatusBadGateway)
@@ -528,7 +527,7 @@ func (s *Server) handleAudioGenerate(w http.ResponseWriter, r *http.Request) {

 	// Download generated audio from Kokoro and persist to MinIO synchronously
 	// so that the presigned URL returned to the client is immediately valid.
-	minioKey := s.store.AudioObjectKey(slug, n, voice, speed)
+	minioKey := s.store.AudioObjectKey(slug, n, voice)
 	audioData, dlErr := s.downloadFromKokoro(r.Context(), filename)
 	if dlErr != nil {
 		s.log.Warn("audio MinIO upload skipped: kokoro download failed",
@@ -541,7 +540,7 @@ func (s *Server) handleAudioGenerate(w http.ResponseWriter, r *http.Request) {
 	}

 	s.log.Info("audio generated", "slug", slug, "chapter", n, "filename", filename)
-	s.writeAudioResponse(w, slug, n, voice, speed, filename)
+	s.writeAudioResponse(w, slug, n, voice, filename)
 }

 // generateSpeech calls POST /v1/audio/speech on Kokoro with return_download_link=true
@@ -618,8 +617,8 @@ func (s *Server) downloadFromKokoro(ctx context.Context, filename string) ([]byt

 // writeAudioResponse writes the JSON response for a generated audio chapter.
 // The URL points to our proxy handler GET /api/audio-proxy/{slug}/{n}.
-func (s *Server) writeAudioResponse(w http.ResponseWriter, slug string, n int, voice string, speed float64, filename string) {
-	proxyURL := fmt.Sprintf("/api/audio-proxy/%s/%d?voice=%s&speed=%.1f", slug, n, voice, speed)
+func (s *Server) writeAudioResponse(w http.ResponseWriter, slug string, n int, voice string, filename string) {
+	proxyURL := fmt.Sprintf("/api/audio-proxy/%s/%d?voice=%s", slug, n, voice)
 	w.Header().Set("Content-Type", "application/json")
 	_ = json.NewEncoder(w).Encode(map[string]interface{}{
 		"url":      proxyURL,
@@ -628,7 +627,7 @@ func (s *Server) writeAudioResponse(w http.ResponseWriter, slug string, n int, v
 }

 // handleAudioProxy handles GET /api/audio-proxy/{slug}/{n}.
-// It looks up the Kokoro download filename for this chapter (voice/speed) and
+// It looks up the Kokoro download filename for this chapter (voice) and
 // proxies GET /v1/download/{filename} from the Kokoro server back to the browser.
 func (s *Server) handleAudioProxy(w http.ResponseWriter, r *http.Request) {
 	slug := r.PathValue("slug")
@@ -641,15 +640,8 @@ func (s *Server) handleAudioProxy(w http.ResponseWriter, r *http.Request) {
 	if voice == "" {
 		voice = s.kokoroVoice
 	}
-	speedStr := r.URL.Query().Get("speed")
-	speed := 1.0
-	if speedStr != "" {
-		if v, err := strconv.ParseFloat(speedStr, 64); err == nil && v > 0 {
-			speed = v
-		}
-	}

-	cacheKey := fmt.Sprintf("%s/%d/%s/%.2f", slug, n, voice, speed)
+	cacheKey := fmt.Sprintf("%s/%d/%s", slug, n, voice)
 	filename, ok := s.store.GetAudioCache(r.Context(), cacheKey)
 	if !ok {
 		http.Error(w, "audio not generated yet", http.StatusNotFound)
@@ -709,7 +701,7 @@ func (s *Server) handlePresignChapter(w http.ResponseWriter, r *http.Request) {

 // handlePresignAudio handles GET /api/presign/audio/{slug}/{n}.
 // Returns a presigned MinIO URL for the audio object (if it has been generated).
-// Query params: voice, speed (optional, defaults to server defaults).
+// Query params: voice (optional, defaults to server default).
 func (s *Server) handlePresignAudio(w http.ResponseWriter, r *http.Request) {
 	slug := r.PathValue("slug")
 	n, err := strconv.Atoi(r.PathValue("n"))
@@ -722,14 +714,8 @@ func (s *Server) handlePresignAudio(w http.ResponseWriter, r *http.Request) {
 	if voice == "" {
 		voice = s.kokoroVoice
 	}
-	speed := 1.0
-	if sv := r.URL.Query().Get("speed"); sv != "" {
-		if v, err := strconv.ParseFloat(sv, 64); err == nil && v > 0 {
-			speed = v
-		}
-	}

-	key := s.store.AudioObjectKey(slug, n, voice, speed)
+	key := s.store.AudioObjectKey(slug, n, voice)

 	// Return 404 when the object hasn't been uploaded yet — the client treats
 	// this as "audio not ready" and will either poll or trigger generation.
--- a/scraper/internal/storage/hybrid.go
+++ b/scraper/internal/storage/hybrid.go
@@ -287,8 +287,8 @@ func (h *HybridStore) DeleteProgress(ctx context.Context, sessionID, slug string

 // ─── AudioObjectKey ───────────────────────────────────────────────────────────

-func (h *HybridStore) AudioObjectKey(slug string, n int, voice string, speed float64) string {
-	return AudioObjectKey(slug, n, voice, speed)
+func (h *HybridStore) AudioObjectKey(slug string, n int, voice string) string {
+	return AudioObjectKey(slug, n, voice)
 }

 func (h *HybridStore) AudioExists(ctx context.Context, key string) bool {
--- a/scraper/internal/storage/minio.go
+++ b/scraper/internal/storage/minio.go
@@ -154,10 +154,10 @@ func (m *MinioClient) CountChapters(ctx context.Context, slug string) int {
 // ─── Audio objects ────────────────────────────────────────────────────────────

 // AudioObjectKey returns the MinIO key for a cached audio file.
-// Key: {slug}/ch{n}-{voice}-{speed:.1f}.mp3
-func AudioObjectKey(slug string, n int, voice string, speed float64) string {
+// Key: {slug}/ch{n}-{voice}.mp3
+func AudioObjectKey(slug string, n int, voice string) string {
 	safe := sanitiseVoice(voice)
-	return fmt.Sprintf("%s/ch%d-%s-%.1f.mp3", slug, n, safe, speed)
+	return fmt.Sprintf("%s/ch%d-%s.mp3", slug, n, safe)
 }

 // PutAudio stores an audio file in the audio bucket.
--- a/scraper/internal/storage/store.go
+++ b/scraper/internal/storage/store.go
@@ -129,7 +129,7 @@ type Store interface {
 	// ── Audio object paths (MinIO) ─────────────────────────────────────────

 	// AudioObjectKey returns the MinIO object key for a cached audio file.
-	AudioObjectKey(slug string, n int, voice string, speed float64) string
+	AudioObjectKey(slug string, n int, voice string) string
 	// AudioExists returns true when the audio object is present in the bucket.
 	AudioExists(ctx context.Context, key string) bool

--- a/ui/src/lib/components/AudioPlayer.svelte
+++ b/ui/src/lib/components/AudioPlayer.svelte
@@ -314,14 +314,12 @@
 	async function tryPresign(
 		targetSlug: string,
 		targetChapter: number,
-		targetVoice: string,
-		targetSpeed: number
+		targetVoice: string
 	): Promise<string | null> {
 		const params = new URLSearchParams({
 			slug: targetSlug,
 			n: String(targetChapter),
-			voice: targetVoice,
-			speed: String(targetSpeed)
+			voice: targetVoice
 		});
 		const res = await fetch(`/api/presign/audio?${params}`);
 		if (res.status === 404) return null;
@@ -337,7 +335,6 @@
 		if (audioStore.nextStatus !== 'none') return; // already running or done

 		const voice = audioStore.voice;
-		const speed = audioStore.speed;

 		audioStore.nextStatus = 'prefetching';
 		audioStore.nextChapterPrefetched = nextChapter;
@@ -345,7 +342,7 @@

 		try {
 			// Fast path: already generated
-			const url = await tryPresign(slug, nextChapter, voice, speed);
+			const url = await tryPresign(slug, nextChapter, voice);
 			if (url) {
 				stopNextProgress();
 				audioStore.nextProgress = 100;
@@ -358,14 +355,14 @@
 			const res = await fetch(`/api/audio/${slug}/${nextChapter}`, {
 				method: 'POST',
 				headers: { 'Content-Type': 'application/json' },
-				body: JSON.stringify({ voice, speed })
+				body: JSON.stringify({ voice })
 			});
 			if (!res.ok) throw new Error(`Prefetch generation failed: HTTP ${res.status}`);

 			stopNextProgress();
 			audioStore.nextProgress = 100;

-			const url2 = await tryPresign(slug, nextChapter, voice, speed);
+			const url2 = await tryPresign(slug, nextChapter, voice);
 			if (!url2) throw new Error('Prefetch: audio generated but presign returned 404');

 			audioStore.nextAudioUrl = url2;
@@ -402,7 +399,6 @@

 	async function startPlayback() {
 		const voice = audioStore.voice;
-		const speed = audioStore.speed;

 		// Populate store metadata so layout + mini-bar have track info.
 		audioStore.slug = slug;
@@ -435,7 +431,7 @@
 			}

 			// Fast path B: audio already in MinIO (presign check).
-			const url = await tryPresign(slug, chapter, voice, speed);
+			const url = await tryPresign(slug, chapter, voice);
 			if (url) {
 				audioStore.audioUrl = url;
 				audioStore.status = 'ready';
@@ -453,13 +449,13 @@
 			const res = await fetch(`/api/audio/${slug}/${chapter}`, {
 				method: 'POST',
 				headers: { 'Content-Type': 'application/json' },
-				body: JSON.stringify({ voice, speed })
+				body: JSON.stringify({ voice })
 			});
 			if (!res.ok) throw new Error(`Generation failed: HTTP ${res.status}`);

 			await finishProgress();

-			const url2 = await tryPresign(slug, chapter, voice, speed);
+			const url2 = await tryPresign(slug, chapter, voice);
 			if (!url2) throw new Error('Audio generated but presign returned 404');
 			audioStore.audioUrl = url2;
 			audioStore.status = 'ready';
--- a/ui/src/lib/server/minio.ts
+++ b/ui/src/lib/server/minio.ts
@@ -98,14 +98,12 @@ export async function presignVoiceSample(voice: string): Promise<string> {
 export async function presignAudio(
 	slug: string,
 	n: number,
-	voice?: string,
-	speed?: number
+	voice?: string
 ): Promise<string> {
 	const params = new URLSearchParams();
 	if (voice) params.set('voice', voice);
-	if (speed) params.set('speed', String(speed));
 	const qs = params.toString() ? `?${params.toString()}` : '';
-	log.debug('minio', 'presigning audio', { slug, n, voice, speed });
+	log.debug('minio', 'presigning audio', { slug, n, voice });
 	let res: Response;
 	try {
 		res = await fetch(`${SCRAPER_URL}/api/presign/audio/${slug}/${n}${qs}`);
@@ -126,7 +124,5 @@ export async function presignAudio(
 	}
 	const data = (await res.json()) as { url: string };
 	log.debug('minio', 'presign audio ok', { slug, n });
-	// The scraper now signs audio URLs with the public endpoint directly,
-	// so no host rewrite is needed here.
-	return data.url;
+	return rewriteHost(data.url);
 }
--- a/ui/src/routes/api/audio/[slug]/[n]/+server.ts
+++ b/ui/src/routes/api/audio/[slug]/[n]/+server.ts
@@ -10,9 +10,9 @@ const SCRAPER_URL = env.SCRAPER_API_URL ?? 'http://localhost:8080';
 * Proxies the audio generation request to the scraper's /api/audio endpoint.
 * Keeps the scraper URL server-side — the browser never needs to know it.
 *
- * Body: { voice?: string, speed?: number }
+ * Body: { voice?: string }
 * Response: { url: string, filename: string }
- *   where `url` is a relative path to GET /api/audio/[slug]/[n]?voice=...&speed=...
+ *   where `url` is a relative path to GET /api/audio/[slug]/[n]?voice=...
 */
 export const POST: RequestHandler = async ({ params, request }) => {
 	const { slug, n } = params;
@@ -21,7 +21,7 @@ export const POST: RequestHandler = async ({ params, request }) => {
 		error(400, 'Invalid slug or chapter number');
 	}

-	let body: { voice?: string; speed?: number } = {};
+	let body: { voice?: string } = {};
 	try {
 		body = await request.json();
 	} catch {
@@ -45,10 +45,8 @@ export const POST: RequestHandler = async ({ params, request }) => {
 	// The scraper returns a proxy URL pointing to /api/audio-proxy/... — we rewrite
 	// it to our own /api/audio/[slug]/[n]?... so the browser never calls the scraper directly.
 	const voice = body.voice ?? '';
-	const speed = body.speed ?? 1.0;
 	const qs = new URLSearchParams();
 	if (voice) qs.set('voice', voice);
-	qs.set('speed', String(speed));

 	return new Response(
 		JSON.stringify({
@@ -60,7 +58,7 @@ export const POST: RequestHandler = async ({ params, request }) => {
 };

 /**
- * GET /api/audio/[slug]/[n]?voice=...&speed=...
+ * GET /api/audio/[slug]/[n]?voice=...
 * Proxies the audio stream from the scraper's /api/audio-proxy endpoint.
 * This is the URL the browser's <audio> element uses as its src.
 */
@@ -72,10 +70,8 @@ export const GET: RequestHandler = async ({ params, url }) => {
 	}

 	const voice = url.searchParams.get('voice') ?? '';
-	const speed = url.searchParams.get('speed') ?? '1';
 	const qs = new URLSearchParams();
 	if (voice) qs.set('voice', voice);
-	qs.set('speed', speed);

 	const scraperRes = await fetch(`${SCRAPER_URL}/api/audio-proxy/${slug}/${chapter}?${qs.toString()}`);

--- a/ui/src/routes/api/presign/audio/+server.ts
+++ b/ui/src/routes/api/presign/audio/+server.ts
@@ -4,7 +4,7 @@ import { presignAudio } from '$lib/server/minio';
 import { log } from '$lib/server/logger';

 /**
- * GET /api/presign/audio?slug=...&n=...&voice=...&speed=...
+ * GET /api/presign/audio?slug=...&n=...&voice=...
 * Returns a presigned MinIO URL for the audio file so the browser
 * can stream it directly without going through the server.
 * Returns 404 when the audio has not been generated yet.
@@ -13,14 +13,13 @@ export const GET: RequestHandler = async ({ url }) => {
 	const slug = url.searchParams.get('slug');
 	const n = parseInt(url.searchParams.get('n') ?? '', 10);
 	const voice = url.searchParams.get('voice') ?? undefined;
-	const speed = parseFloat(url.searchParams.get('speed') ?? '1') || 1;

 	if (!slug || !n || n < 1) {
 		error(400, 'Missing slug or n');
 	}

 	try {
-		const presignedUrl = await presignAudio(slug, n, voice, speed);
+		const presignedUrl = await presignAudio(slug, n, voice);
 		return json({ url: presignedUrl });
 	} catch (e) {
 		const status = (e as { status?: number }).status;