fix(ui): resolve avatar URL from MinIO; fall back to OAuth provider URL

Add resolveAvatarUrl(userId, storedValue) helper that tries MinIO first, then falls back to the stored HTTP URL for OAuth users (Google/GitHub) who have never uploaded a custom avatar. Add getUserById() to pocketbase helpers for batch avatar resolution in comments. Update all 6 call sites to use the new helper.
ci: remove Docker builds from CI; keep vet/build/test/type-check only
2026-03-28 19:23:30 +05:00 · 2026-03-28 19:08:17 +05:00 · 2026-03-28 19:01:58 +05:00 · 2026-03-28 16:28:36 +05:00 · 2026-03-28 16:15:04 +05:00 · 2026-03-28 16:02:46 +05:00
46 changed files with 2474 additions and 398 deletions
--- a/.gitea/workflows/ci.yaml
+++ b/.gitea/workflows/ci.yaml
@@ -2,20 +2,14 @@ name: CI

 on:
  push:
-    branches: ["main", "master"]
    paths:
      - "backend/**"
      - "ui/**"
-      - "caddy/**"
-      - "docker-compose.yml"
      - ".gitea/workflows/ci.yaml"
  pull_request:
-    branches: ["main", "master"]
    paths:
      - "backend/**"
      - "ui/**"
-      - "caddy/**"
-      - "docker-compose.yml"
      - ".gitea/workflows/ci.yaml"

 concurrency:
@@ -23,10 +17,13 @@ concurrency:
  cancel-in-progress: true

 jobs:
-  # ── backend: vet & test ───────────────────────────────────────────────────────
-  test-backend:
-    name: Test backend
+  # ── Go: vet + build + test ────────────────────────────────────────────────
+  backend:
+    name: Backend
    runs-on: ubuntu-latest
+    defaults:
+      run:
+        working-directory: backend
    steps:
      - uses: actions/checkout@v4

@@ -36,16 +33,23 @@ jobs:
          cache-dependency-path: backend/go.sum

      - name: go vet
-        working-directory: backend
        run: go vet ./...

+      - name: Build backend
+        run: go build -o /dev/null ./cmd/backend
+
+      - name: Build runner
+        run: go build -o /dev/null ./cmd/runner
+
+      - name: Build healthcheck
+        run: go build -o /dev/null ./cmd/healthcheck
+
      - name: Run tests
-        working-directory: backend
        run: go test -short -race -count=1 -timeout=60s ./...

-  # ── ui: type-check & build ────────────────────────────────────────────────────
-  check-ui:
-    name: Check ui
+  # ── UI: type-check + build ────────────────────────────────────────────────
+  ui:
+    name: UI
    runs-on: ubuntu-latest
    defaults:
      run:
@@ -67,57 +71,3 @@ jobs:

      - name: Build
        run: npm run build
-
-  # ── docker: validate Dockerfiles build (no push) ──────────────────────────────
-  docker-backend:
-    name: Docker / backend
-    runs-on: ubuntu-latest
-    needs: [test-backend]
-    steps:
-      - uses: actions/checkout@v4
-      - uses: docker/setup-buildx-action@v3
-      - name: Build
-        uses: docker/build-push-action@v6
-        with:
-          context: backend
-          target: backend
-          push: false
-
-  docker-runner:
-    name: Docker / runner
-    runs-on: ubuntu-latest
-    needs: [test-backend]
-    steps:
-      - uses: actions/checkout@v4
-      - uses: docker/setup-buildx-action@v3
-      - name: Build
-        uses: docker/build-push-action@v6
-        with:
-          context: backend
-          target: runner
-          push: false
-
-  docker-ui:
-    name: Docker / ui
-    runs-on: ubuntu-latest
-    needs: [check-ui]
-    steps:
-      - uses: actions/checkout@v4
-      - uses: docker/setup-buildx-action@v3
-      - name: Build
-        uses: docker/build-push-action@v6
-        with:
-          context: ui
-          push: false
-
-  docker-caddy:
-    name: Docker / caddy
-    runs-on: ubuntu-latest
-    steps:
-      - uses: actions/checkout@v4
-      - uses: docker/setup-buildx-action@v3
-      - name: Build
-        uses: docker/build-push-action@v6
-        with:
-          context: caddy
-          push: false
--- a/.gitea/workflows/release.yaml
+++ b/.gitea/workflows/release.yaml
@@ -136,52 +136,51 @@ jobs:
          cache-to: type=inline

  # ── ui: source map upload ─────────────────────────────────────────────────────
-  # Builds the UI with source maps and uploads them to GlitchTip so that error
-  # stack traces resolve to original .svelte/.ts file names and line numbers.
-  # Runs in parallel with docker-ui (both need check-ui to pass first).
-  upload-sourcemaps:
-    name: Upload source maps
-    runs-on: ubuntu-latest
-    needs: [check-ui]
-    defaults:
-      run:
-        working-directory: ui
-    steps:
-      - uses: actions/checkout@v4
-
-      - uses: actions/setup-node@v4
-        with:
-          node-version: "22"
-          cache: npm
-          cache-dependency-path: ui/package-lock.json
-
-      - name: Install dependencies
-        run: npm ci
-
-      - name: Build with source maps
-        run: npm run build
-
-      - name: Download glitchtip-cli
-        run: |
-          curl -L "https://gitlab.com/glitchtip/glitchtip-cli/-/jobs/artifacts/v0.1.0/raw/artifacts/glitchtip-cli-linux-x86_64?job=build-linux-x86_64" \
-            -o /usr/local/bin/glitchtip-cli
-          chmod +x /usr/local/bin/glitchtip-cli
-
-      - name: Inject debug IDs into build artifacts
-        run: glitchtip-cli sourcemaps inject ./build
-        env:
-          SENTRY_URL: https://errors.libnovel.cc/
-          SENTRY_AUTH_TOKEN: ${{ secrets.GLITCHTIP_AUTH_TOKEN }}
-          SENTRY_ORG: libnovel
-          SENTRY_PROJECT: libnovel-ui
-
-      - name: Upload source maps to GlitchTip
-        run: glitchtip-cli sourcemaps upload ./build --release ${{ gitea.ref_name }}
-        env:
-          SENTRY_URL: https://errors.libnovel.cc/
-          SENTRY_AUTH_TOKEN: ${{ secrets.GLITCHTIP_AUTH_TOKEN }}
-          SENTRY_ORG: libnovel
-          SENTRY_PROJECT: libnovel-ui
+  # Commented out: GlitchTip project/auth token needs to be recreated after
+  # the GlitchTip DB wipe. Re-enable once GLITCHTIP_AUTH_TOKEN is updated.
+  # upload-sourcemaps:
+  #   name: Upload source maps
+  #   runs-on: ubuntu-latest
+  #   needs: [check-ui]
+  #   defaults:
+  #     run:
+  #       working-directory: ui
+  #   steps:
+  #     - uses: actions/checkout@v4
+  #
+  #     - uses: actions/setup-node@v4
+  #       with:
+  #         node-version: "22"
+  #         cache: npm
+  #         cache-dependency-path: ui/package-lock.json
+  #
+  #     - name: Install dependencies
+  #       run: npm ci
+  #
+  #     - name: Build with source maps
+  #       run: npm run build
+  #
+  #     - name: Download glitchtip-cli
+  #       run: |
+  #         curl -L "https://gitlab.com/glitchtip/glitchtip-cli/-/jobs/artifacts/v0.1.0/raw/artifacts/glitchtip-cli-linux-x86_64?job=build-linux-x86_64" \
+  #           -o /usr/local/bin/glitchtip-cli
+  #         chmod +x /usr/local/bin/glitchtip-cli
+  #
+  #     - name: Inject debug IDs into build artifacts
+  #       run: glitchtip-cli sourcemaps inject ./build
+  #       env:
+  #         SENTRY_URL: https://errors.libnovel.cc/
+  #         SENTRY_AUTH_TOKEN: ${{ secrets.GLITCHTIP_AUTH_TOKEN }}
+  #         SENTRY_ORG: libnovel
+  #         SENTRY_PROJECT: libnovel-ui
+  #
+  #     - name: Upload source maps to GlitchTip
+  #       run: glitchtip-cli sourcemaps upload ./build --release ${{ gitea.ref_name }}
+  #       env:
+  #         SENTRY_URL: https://errors.libnovel.cc/
+  #         SENTRY_AUTH_TOKEN: ${{ secrets.GLITCHTIP_AUTH_TOKEN }}
+  #         SENTRY_ORG: libnovel
+  #         SENTRY_PROJECT: libnovel-ui

  # ── docker: ui ────────────────────────────────────────────────────────────────
  docker-ui:
@@ -261,7 +260,7 @@ jobs:
  release:
    name: Gitea Release
    runs-on: ubuntu-latest
-    needs: [docker-backend, docker-runner, docker-ui, docker-caddy, upload-sourcemaps]
+    needs: [docker-backend, docker-runner, docker-ui, docker-caddy]
    steps:
      - uses: actions/checkout@v4
        with:
--- a/24
+++ b/24
@@ -234,3 +234,27 @@ search.libnovel.cc {
 	reverse_proxy meilisearch:7700
 }

+# ── Redis TCP proxy: exposes homelab Redis over TLS for Asynq ─────────────────
+# The backend (prod) connects to rediss://redis.libnovel.cc:6380 to enqueue
+# Asynq jobs. Caddy terminates TLS (Let's Encrypt cert for redis.libnovel.cc)
+# and proxies the raw TCP stream to the homelab Redis via this reverse proxy.
+#
+# NOTE: Redis is NOT running on the prod server — it runs on the homelab
+# (192.168.0.109:6379) and is exposed to the internet via this Caddy proxy.
+# The homelab Redis is protected by REDIS_PASSWORD (requirepass).
+#
+# Caddy layer4 app handles this; requires the caddy-l4 module in the build.
+{
+	layer4 {
+		redis.libnovel.cc:6380 {
+			route {
+				tls
+				proxy {
+					# Homelab Redis — replace with actual homelab IP or FQDN
+					upstream {$HOMELAB_REDIS_ADDR:192.168.0.109:6379}
+				}
+			}
+		}
+	}
+}
+}
--- a/backend/cmd/backend/main.go
+++ b/backend/cmd/backend/main.go
@@ -22,12 +22,16 @@ import (
 	"time"

 	"github.com/getsentry/sentry-go"
+	"github.com/hibiken/asynq"
+	"github.com/libnovel/backend/internal/asynqqueue"
 	"github.com/libnovel/backend/internal/backend"
 	"github.com/libnovel/backend/internal/config"
 	"github.com/libnovel/backend/internal/kokoro"
 	"github.com/libnovel/backend/internal/meili"
 	"github.com/libnovel/backend/internal/otelsetup"
+	"github.com/libnovel/backend/internal/pockettts"
 	"github.com/libnovel/backend/internal/storage"
+	"github.com/libnovel/backend/internal/taskqueue"
 )

 // version and commit are set at build time via -ldflags.
@@ -100,6 +104,15 @@ func run() error {
 		kokoroClient = &noopKokoro{}
 	}

+	// ── Pocket-TTS (voice list + sample generation; audio generation is the runner's job) ──
+	var pocketTTSClient pockettts.Client
+	if cfg.PocketTTS.URL != "" {
+		pocketTTSClient = pockettts.New(cfg.PocketTTS.URL)
+		log.Info("pocket-tts voices enabled", "url", cfg.PocketTTS.URL)
+	} else {
+		log.Info("POCKET_TTS_URL not set — pocket-tts voices unavailable in backend")
+	}
+
 	// ── Meilisearch (search reads only; indexing is the runner's job) ────────
 	var searchIndex meili.Client
 	if cfg.Meilisearch.URL != "" {
@@ -110,6 +123,24 @@ func run() error {
 		searchIndex = meili.NoopClient{}
 	}

+	// ── Task Producer ────────────────────────────────────────────────────────
+	// When REDIS_ADDR is set the backend dual-writes: PocketBase record (audit)
+	// + Asynq job (immediate delivery). Otherwise it writes to PocketBase only
+	// and the runner picks up on the next poll tick.
+	var producer taskqueue.Producer = store
+	if cfg.Redis.Addr != "" {
+		redisOpt, parseErr := parseRedisOpt(cfg.Redis)
+		if parseErr != nil {
+			return fmt.Errorf("parse REDIS_ADDR: %w", parseErr)
+		}
+		asynqProducer := asynqqueue.NewProducer(store, redisOpt)
+		defer asynqProducer.Close() //nolint:errcheck
+		producer = asynqProducer
+		log.Info("backend: asynq task dispatch enabled", "addr", cfg.Redis.Addr)
+	} else {
+		log.Info("backend: poll-mode task dispatch (REDIS_ADDR not set)")
+	}
+
 	// ── Backend server ───────────────────────────────────────────────────────
 	srv := backend.New(
 		backend.Config{
@@ -125,10 +156,11 @@ func run() error {
 			PresignStore:  store,
 			ProgressStore: store,
 			CoverStore:    store,
-			Producer:      store,
+			Producer:      producer,
 			TaskReader:    store,
 			SearchIndex:   searchIndex,
 			Kokoro:        kokoroClient,
+			PocketTTS:     pocketTTSClient,
 			Log:           log,
 		},
 	)
@@ -165,3 +197,16 @@ func (n *noopKokoro) GenerateAudio(_ context.Context, _, _ string) ([]byte, erro
 func (n *noopKokoro) ListVoices(_ context.Context) ([]string, error) {
 	return nil, nil
 }
+
+// parseRedisOpt converts a config.Redis into an asynq.RedisConnOpt.
+// Handles full "redis://" / "rediss://" URLs and plain "host:port".
+func parseRedisOpt(cfg config.Redis) (asynq.RedisConnOpt, error) {
+	addr := cfg.Addr
+	if len(addr) > 7 && (addr[:8] == "redis://" || (len(addr) > 8 && addr[:9] == "rediss://")) {
+		return asynq.ParseRedisURI(addr)
+	}
+	return asynq.RedisClientOpt{
+		Addr:     addr,
+		Password: cfg.Password,
+	}, nil
+}
--- a/backend/cmd/runner/main.go
+++ b/backend/cmd/runner/main.go
@@ -20,6 +20,7 @@ import (
 	"time"

 	"github.com/getsentry/sentry-go"
+	"github.com/libnovel/backend/internal/asynqqueue"
 	"github.com/libnovel/backend/internal/browser"
 	"github.com/libnovel/backend/internal/config"
 	"github.com/libnovel/backend/internal/kokoro"
@@ -29,6 +30,7 @@ import (
 	"github.com/libnovel/backend/internal/pockettts"
 	"github.com/libnovel/backend/internal/runner"
 	"github.com/libnovel/backend/internal/storage"
+	"github.com/libnovel/backend/internal/taskqueue"
 )

 // version and commit are set at build time via -ldflags.
@@ -150,10 +152,25 @@ func run() error {
 		OrchestratorWorkers:         workers,
 		MetricsAddr:                 cfg.Runner.MetricsAddr,
 		CatalogueRefreshInterval:    cfg.Runner.CatalogueRefreshInterval,
+		CatalogueRequestDelay:       cfg.Runner.CatalogueRequestDelay,
 		SkipInitialCatalogueRefresh: cfg.Runner.SkipInitialCatalogueRefresh,
+		RedisAddr:                   cfg.Redis.Addr,
+		RedisPassword:               cfg.Redis.Password,
 	}
+
+	// In Asynq mode the Consumer is a thin wrapper: claim/heartbeat/reap are
+	// no-ops, but FinishAudioTask / FinishScrapeTask / FailTask write back to
+	// PocketBase as before.
+	var consumer taskqueue.Consumer = store
+	if cfg.Redis.Addr != "" {
+		log.Info("runner: asynq mode — using Redis for task dispatch", "addr", cfg.Redis.Addr)
+		consumer = asynqqueue.NewConsumer(store)
+	} else {
+		log.Info("runner: poll mode — using PocketBase for task dispatch")
+	}
+
 	deps := runner.Dependencies{
-		Consumer:    store,
+		Consumer:    consumer,
 		BookWriter:  store,
 		BookReader:  store,
 		AudioStore:  store,
--- a/backend/go.mod
+++ b/backend/go.mod
@@ -9,6 +9,7 @@ require (

 require (
 	github.com/andybalholm/brotli v1.1.1 // indirect
+	github.com/beorn7/perks v1.0.1 // indirect
 	github.com/cenkalti/backoff/v5 v5.0.3 // indirect
 	github.com/cespare/xxhash/v2 v2.3.0 // indirect
 	github.com/davecgh/go-spew v1.1.1 // indirect
@@ -22,16 +23,25 @@ require (
 	github.com/golang-jwt/jwt/v5 v5.3.1 // indirect
 	github.com/google/uuid v1.6.0 // indirect
 	github.com/grpc-ecosystem/grpc-gateway/v2 v2.28.0 // indirect
+	github.com/hibiken/asynq v0.26.0 // indirect
+	github.com/hibiken/asynq/x v0.0.0-20260203063626-d704b68a426d // indirect
 	github.com/klauspost/compress v1.18.2 // indirect
 	github.com/klauspost/cpuid/v2 v2.2.11 // indirect
 	github.com/klauspost/crc32 v1.3.0 // indirect
 	github.com/meilisearch/meilisearch-go v0.36.1 // indirect
 	github.com/minio/crc64nvme v1.1.1 // indirect
 	github.com/minio/md5-simd v1.1.2 // indirect
+	github.com/munnerz/goautoneg v0.0.0-20191010083416-a7dc8b61c822 // indirect
 	github.com/philhofer/fwd v1.2.0 // indirect
 	github.com/pmezard/go-difflib v1.0.0 // indirect
+	github.com/prometheus/client_golang v1.23.2 // indirect
+	github.com/prometheus/client_model v0.6.2 // indirect
+	github.com/prometheus/common v0.66.1 // indirect
+	github.com/prometheus/procfs v0.16.1 // indirect
 	github.com/redis/go-redis/v9 v9.18.0 // indirect
+	github.com/robfig/cron/v3 v3.0.1 // indirect
 	github.com/rs/xid v1.6.0 // indirect
+	github.com/spf13/cast v1.10.0 // indirect
 	github.com/tinylib/msgp v1.6.1 // indirect
 	go.opentelemetry.io/auto/sdk v1.2.1 // indirect
 	go.opentelemetry.io/contrib/bridges/otelslog v0.17.0 // indirect
@@ -47,10 +57,12 @@ require (
 	go.opentelemetry.io/otel/trace v1.42.0 // indirect
 	go.opentelemetry.io/proto/otlp v1.9.0 // indirect
 	go.uber.org/atomic v1.11.0 // indirect
+	go.yaml.in/yaml/v2 v2.4.2 // indirect
 	go.yaml.in/yaml/v3 v3.0.4 // indirect
 	golang.org/x/crypto v0.48.0 // indirect
 	golang.org/x/sys v0.41.0 // indirect
 	golang.org/x/text v0.34.0 // indirect
+	golang.org/x/time v0.14.0 // indirect
 	google.golang.org/genproto/googleapis/api v0.0.0-20260209200024-4cfbd4190f57 // indirect
 	google.golang.org/genproto/googleapis/rpc v0.0.0-20260209200024-4cfbd4190f57 // indirect
 	google.golang.org/grpc v1.79.2 // indirect
--- a/backend/go.sum
+++ b/backend/go.sum
@@ -1,5 +1,7 @@
 github.com/andybalholm/brotli v1.1.1 h1:PR2pgnyFznKEugtsUo0xLdDop5SKXd5Qf5ysW+7XdTA=
 github.com/andybalholm/brotli v1.1.1/go.mod h1:05ib4cKhjx3OQYUY22hTVd34Bc8upXjOLL2rKwwZBoA=
+github.com/beorn7/perks v1.0.1 h1:VlbKKnNfV8bJzeqoa4cOKqO6bYr3WgKZxO8Z16+hsOM=
+github.com/beorn7/perks v1.0.1/go.mod h1:G2ZrVWU2WbWT9wwq4/hrbKbnv/1ERSJQ0ibhJ6rlkpw=
 github.com/cenkalti/backoff/v5 v5.0.3 h1:ZN+IMa753KfX5hd8vVaMixjnqRZ3y8CuJKRKj1xcsSM=
 github.com/cenkalti/backoff/v5 v5.0.3/go.mod h1:rkhZdG3JZukswDf7f0cwqPNk4K0sa+F97BxZthm/crw=
 github.com/cespare/xxhash/v2 v2.3.0 h1:UL815xU9SqsFlibzuggzjXhog7bL6oX9BbNZnL2UFvs=
@@ -27,6 +29,10 @@ github.com/google/uuid v1.6.0 h1:NIvaJDMOsjHA8n1jAhLSgzrAzy1Hgr+hNrb57e+94F0=
 github.com/google/uuid v1.6.0/go.mod h1:TIyPZe4MgqvfeYDBFedMoGGpEw/LqOeaOT+nhxU+yHo=
 github.com/grpc-ecosystem/grpc-gateway/v2 v2.28.0 h1:HWRh5R2+9EifMyIHV7ZV+MIZqgz+PMpZ14Jynv3O2Zs=
 github.com/grpc-ecosystem/grpc-gateway/v2 v2.28.0/go.mod h1:JfhWUomR1baixubs02l85lZYYOm7LV6om4ceouMv45c=
+github.com/hibiken/asynq v0.26.0 h1:1Zxr92MlDnb1Zt/QR5g2vSCqUS03i95lUfqx5X7/wrw=
+github.com/hibiken/asynq v0.26.0/go.mod h1:Qk4e57bTnWDoyJ67VkchuV6VzSM9IQW2nPvAGuDyw58=
+github.com/hibiken/asynq/x v0.0.0-20260203063626-d704b68a426d h1:Ld5m8EIK5QVOq/owOexKIbETij3skACg4eU1pArHsrw=
+github.com/hibiken/asynq/x v0.0.0-20260203063626-d704b68a426d/go.mod h1:hhpStehaxSGg3ib9wJXzw5AXY1YS6lQ9BNavAgPbIhE=
 github.com/klauspost/compress v1.18.2 h1:iiPHWW0YrcFgpBYhsA6D1+fqHssJscY/Tm/y2Uqnapk=
 github.com/klauspost/compress v1.18.2/go.mod h1:R0h/fSBs8DE4ENlcrlib3PsXS61voFxhIs2DeRhCvJ4=
 github.com/klauspost/cpuid/v2 v2.0.1/go.mod h1:FInQzS24/EEf25PyTYn52gqo7WaD8xa0213Md/qVLRg=
@@ -42,14 +48,36 @@ github.com/minio/md5-simd v1.1.2 h1:Gdi1DZK69+ZVMoNHRXJyNcxrMA4dSxoYHZSQbirFg34=
 github.com/minio/md5-simd v1.1.2/go.mod h1:MzdKDxYpY2BT9XQFocsiZf/NKVtR7nkE4RoEpN+20RM=
 github.com/minio/minio-go/v7 v7.0.98 h1:MeAVKjLVz+XJ28zFcuYyImNSAh8Mq725uNW4beRisi0=
 github.com/minio/minio-go/v7 v7.0.98/go.mod h1:cY0Y+W7yozf0mdIclrttzo1Iiu7mEf9y7nk2uXqMOvM=
+github.com/munnerz/goautoneg v0.0.0-20191010083416-a7dc8b61c822 h1:C3w9PqII01/Oq1c1nUAm88MOHcQC9l5mIlSMApZMrHA=
+github.com/munnerz/goautoneg v0.0.0-20191010083416-a7dc8b61c822/go.mod h1:+n7T8mK8HuQTcFwEeznm/DIxMOiR9yIdICNftLE1DvQ=
 github.com/philhofer/fwd v1.2.0 h1:e6DnBTl7vGY+Gz322/ASL4Gyp1FspeMvx1RNDoToZuM=
 github.com/philhofer/fwd v1.2.0/go.mod h1:RqIHx9QI14HlwKwm98g9Re5prTQ6LdeRQn+gXJFxsJM=
 github.com/pmezard/go-difflib v1.0.0 h1:4DBwDE0NGyQoBHbLQYPwSUPoCMWR5BEzIk/f1lZbAQM=
 github.com/pmezard/go-difflib v1.0.0/go.mod h1:iKH77koFhYxTK1pcRnkKkqfTogsbg7gZNVY4sRDYZ/4=
+github.com/prometheus/client_golang v1.20.5 h1:cxppBPuYhUnsO6yo/aoRol4L7q7UFfdm+bR9r+8l63Y=
+github.com/prometheus/client_golang v1.20.5/go.mod h1:PIEt8X02hGcP8JWbeHyeZ53Y/jReSnHgO035n//V5WE=
+github.com/prometheus/client_golang v1.23.2 h1:Je96obch5RDVy3FDMndoUsjAhG5Edi49h0RJWRi/o0o=
+github.com/prometheus/client_golang v1.23.2/go.mod h1:Tb1a6LWHB3/SPIzCoaDXI4I8UHKeFTEQ1YCr+0Gyqmg=
+github.com/prometheus/client_model v0.6.1 h1:ZKSh/rekM+n3CeS952MLRAdFwIKqeY8b62p8ais2e9E=
+github.com/prometheus/client_model v0.6.1/go.mod h1:OrxVMOVHjw3lKMa8+x6HeMGkHMQyHDk9E3jmP2AmGiY=
+github.com/prometheus/client_model v0.6.2 h1:oBsgwpGs7iVziMvrGhE53c/GrLUsZdHnqNwqPLxwZyk=
+github.com/prometheus/client_model v0.6.2/go.mod h1:y3m2F6Gdpfy6Ut/GBsUqTWZqCUvMVzSfMLjcu6wAwpE=
+github.com/prometheus/common v0.55.0 h1:KEi6DK7lXW/m7Ig5i47x0vRzuBsHuvJdi5ee6Y3G1dc=
+github.com/prometheus/common v0.55.0/go.mod h1:2SECS4xJG1kd8XF9IcM1gMX6510RAEL65zxzNImwdc8=
+github.com/prometheus/common v0.66.1 h1:h5E0h5/Y8niHc5DlaLlWLArTQI7tMrsfQjHV+d9ZoGs=
+github.com/prometheus/common v0.66.1/go.mod h1:gcaUsgf3KfRSwHY4dIMXLPV0K/Wg1oZ8+SbZk/HH/dA=
+github.com/prometheus/procfs v0.15.1 h1:YagwOFzUgYfKKHX6Dr+sHT7km/hxC76UB0learggepc=
+github.com/prometheus/procfs v0.15.1/go.mod h1:fB45yRUv8NstnjriLhBQLuOUt+WW4BsoGhij/e3PBqk=
+github.com/prometheus/procfs v0.16.1 h1:hZ15bTNuirocR6u0JZ6BAHHmwS1p8B4P6MRqxtzMyRg=
+github.com/prometheus/procfs v0.16.1/go.mod h1:teAbpZRB1iIAJYREa1LsoWUXykVXA1KlTmWl8x/U+Is=
 github.com/redis/go-redis/v9 v9.18.0 h1:pMkxYPkEbMPwRdenAzUNyFNrDgHx9U+DrBabWNfSRQs=
 github.com/redis/go-redis/v9 v9.18.0/go.mod h1:k3ufPphLU5YXwNTUcCRXGxUoF1fqxnhFQmscfkCoDA0=
+github.com/robfig/cron/v3 v3.0.1 h1:WdRxkvbJztn8LMz/QEvLN5sBU+xKpSqwwUO1Pjr4qDs=
+github.com/robfig/cron/v3 v3.0.1/go.mod h1:eQICP3HwyT7UooqI/z+Ov+PtYAWygg1TEWWzGIFLtro=
 github.com/rs/xid v1.6.0 h1:fV591PaemRlL6JfRxGDEPl69wICngIQ3shQtzfy2gxU=
 github.com/rs/xid v1.6.0/go.mod h1:7XoLgs4eV+QndskICGsho+ADou8ySMSjJKDIan90Nz0=
+github.com/spf13/cast v1.10.0 h1:h2x0u2shc1QuLHfxi+cTJvs30+ZAHOGRic8uyGTDWxY=
+github.com/spf13/cast v1.10.0/go.mod h1:jNfB8QC9IA6ZuY2ZjDp0KtFO2LZZlg4S/7bzP6qqeHo=
 github.com/stretchr/testify v1.9.0 h1:HtqpIVDClZ4nwg75+f6Lvsy/wHu+3BoSGCbBAcpTsTg=
 github.com/stretchr/testify v1.9.0/go.mod h1:r2ic/lqez/lEtzL7wO/rwa5dbSLXVDPFyf8C91i36aY=
 github.com/stretchr/testify v1.11.1 h1:7s2iGBzp5EwR7/aIZr8ao5+dra3wiQyKjjFuvgVKu7U=
@@ -84,6 +112,8 @@ go.opentelemetry.io/proto/otlp v1.9.0 h1:l706jCMITVouPOqEnii2fIAuO3IVGBRPV5ICjce
 go.opentelemetry.io/proto/otlp v1.9.0/go.mod h1:xE+Cx5E/eEHw+ISFkwPLwCZefwVjY+pqKg1qcK03+/4=
 go.uber.org/atomic v1.11.0 h1:ZvwS0R+56ePWxUNi+Atn9dWONBPp/AUETXlHW0DxSjE=
 go.uber.org/atomic v1.11.0/go.mod h1:LUxbIzbOniOlMKjJjyPfpl4v+PKK2cNJn91OQbhoJI0=
+go.yaml.in/yaml/v2 v2.4.2 h1:DzmwEr2rDGHl7lsFgAHxmNz/1NlQ7xLIrlN2h5d1eGI=
+go.yaml.in/yaml/v2 v2.4.2/go.mod h1:081UH+NErpNdqlCXm3TtEran0rJZGxAYx9hb/ELlsPU=
 go.yaml.in/yaml/v3 v3.0.4 h1:tfq32ie2Jv2UxXFdLJdh3jXuOzWiL1fo0bu/FbuKpbc=
 go.yaml.in/yaml/v3 v3.0.4/go.mod h1:DhzuOOF2ATzADvBadXxruRBLzYTpT36CKvDb3+aBEFg=
 golang.org/x/crypto v0.48.0 h1:/VRzVqiRSggnhY7gNRxPauEQ5Drw9haKdM0jqfcCFts=
@@ -94,6 +124,8 @@ golang.org/x/sys v0.41.0 h1:Ivj+2Cp/ylzLiEU89QhWblYnOE9zerudt9Ftecq2C6k=
 golang.org/x/sys v0.41.0/go.mod h1:OgkHotnGiDImocRcuBABYBEXf8A9a87e/uXjp9XT3ks=
 golang.org/x/text v0.34.0 h1:oL/Qq0Kdaqxa1KbNeMKwQq0reLCCaFtqu2eNuSeNHbk=
 golang.org/x/text v0.34.0/go.mod h1:homfLqTYRFyVYemLBFl5GgL/DWEiH5wcsQ5gSh1yziA=
+golang.org/x/time v0.14.0 h1:MRx4UaLrDotUKUdCIqzPC48t1Y9hANFKIRpNx+Te8PI=
+golang.org/x/time v0.14.0/go.mod h1:eL/Oa2bBBK0TkX57Fyni+NgnyQQN4LitPmob2Hjnqw4=
 google.golang.org/genproto/googleapis/api v0.0.0-20260209200024-4cfbd4190f57 h1:JLQynH/LBHfCTSbDWl+py8C+Rg/k1OVH3xfcaiANuF0=
 google.golang.org/genproto/googleapis/api v0.0.0-20260209200024-4cfbd4190f57/go.mod h1:kSJwQxqmFXeo79zOmbrALdflXQeAYcUbgS7PbpMknCY=
 google.golang.org/genproto/googleapis/rpc v0.0.0-20260209200024-4cfbd4190f57 h1:mWPCjDEyshlQYzBpMNHaEof6UX1PmHcaUODUywQ0uac=
--- a/backend/internal/asynqqueue/consumer.go
+++ b/backend/internal/asynqqueue/consumer.go
@@ -0,0 +1,56 @@
+package asynqqueue
+
+import (
+	"context"
+	"time"
+
+	"github.com/libnovel/backend/internal/domain"
+	"github.com/libnovel/backend/internal/taskqueue"
+)
+
+// Consumer wraps the PocketBase-backed Consumer for result write-back only.
+//
+// When using Asynq, the runner no longer polls for work — Asynq delivers
+// tasks via the ServeMux handlers. The only Consumer operations the handlers
+// need are:
+//   - FinishAudioTask / FinishScrapeTask — write result back to PocketBase
+//   - FailTask — mark PocketBase record as failed
+//
+// ClaimNextAudioTask, ClaimNextScrapeTask, HeartbeatTask, and ReapStaleTasks
+// are all no-ops here because Asynq owns those responsibilities.
+type Consumer struct {
+	pb taskqueue.Consumer // underlying PocketBase consumer (for write-back)
+}
+
+// NewConsumer wraps an existing PocketBase Consumer.
+func NewConsumer(pb taskqueue.Consumer) *Consumer {
+	return &Consumer{pb: pb}
+}
+
+// ── Write-back (delegated to PocketBase) ──────────────────────────────────────
+
+func (c *Consumer) FinishScrapeTask(ctx context.Context, id string, result domain.ScrapeResult) error {
+	return c.pb.FinishScrapeTask(ctx, id, result)
+}
+
+func (c *Consumer) FinishAudioTask(ctx context.Context, id string, result domain.AudioResult) error {
+	return c.pb.FinishAudioTask(ctx, id, result)
+}
+
+func (c *Consumer) FailTask(ctx context.Context, id, errMsg string) error {
+	return c.pb.FailTask(ctx, id, errMsg)
+}
+
+// ── No-ops (Asynq owns claiming / heartbeating / reaping) ───────────────────
+
+func (c *Consumer) ClaimNextScrapeTask(_ context.Context, _ string) (domain.ScrapeTask, bool, error) {
+	return domain.ScrapeTask{}, false, nil
+}
+
+func (c *Consumer) ClaimNextAudioTask(_ context.Context, _ string) (domain.AudioTask, bool, error) {
+	return domain.AudioTask{}, false, nil
+}
+
+func (c *Consumer) HeartbeatTask(_ context.Context, _ string) error { return nil }
+
+func (c *Consumer) ReapStaleTasks(_ context.Context, _ time.Duration) (int, error) { return 0, nil }
--- a/backend/internal/asynqqueue/producer.go
+++ b/backend/internal/asynqqueue/producer.go
@@ -0,0 +1,90 @@
+package asynqqueue
+
+import (
+	"context"
+	"encoding/json"
+	"fmt"
+
+	"github.com/hibiken/asynq"
+	"github.com/libnovel/backend/internal/taskqueue"
+)
+
+// Producer dual-writes every task: first to PocketBase (via pb, for audit /
+// UI status), then to Redis via Asynq so the runner picks it up immediately.
+type Producer struct {
+	pb     taskqueue.Producer // underlying PocketBase producer
+	client *asynq.Client
+}
+
+// NewProducer wraps an existing PocketBase Producer with Asynq dispatch.
+func NewProducer(pb taskqueue.Producer, redisOpt asynq.RedisConnOpt) *Producer {
+	return &Producer{
+		pb:     pb,
+		client: asynq.NewClient(redisOpt),
+	}
+}
+
+// Close shuts down the underlying Asynq client connection.
+func (p *Producer) Close() error {
+	return p.client.Close()
+}
+
+// CreateScrapeTask creates a PocketBase record then enqueues an Asynq job.
+func (p *Producer) CreateScrapeTask(ctx context.Context, kind, targetURL string, fromChapter, toChapter int) (string, error) {
+	id, err := p.pb.CreateScrapeTask(ctx, kind, targetURL, fromChapter, toChapter)
+	if err != nil {
+		return "", err
+	}
+
+	payload := ScrapePayload{
+		PBTaskID:    id,
+		Kind:        kind,
+		TargetURL:   targetURL,
+		FromChapter: fromChapter,
+		ToChapter:   toChapter,
+	}
+	taskType := TypeScrapeBook
+	if kind == "catalogue" {
+		taskType = TypeScrapeCatalogue
+	}
+	if err := p.enqueue(ctx, taskType, payload); err != nil {
+		// Non-fatal: PB record exists; runner will pick it up on next poll.
+		return id, fmt.Errorf("asynq enqueue scrape (task still in PB): %w", err)
+	}
+	return id, nil
+}
+
+// CreateAudioTask creates a PocketBase record then enqueues an Asynq job.
+func (p *Producer) CreateAudioTask(ctx context.Context, slug string, chapter int, voice string) (string, error) {
+	id, err := p.pb.CreateAudioTask(ctx, slug, chapter, voice)
+	if err != nil {
+		return "", err
+	}
+
+	payload := AudioPayload{
+		PBTaskID: id,
+		Slug:     slug,
+		Chapter:  chapter,
+		Voice:    voice,
+	}
+	if err := p.enqueue(ctx, TypeAudioGenerate, payload); err != nil {
+		return id, fmt.Errorf("asynq enqueue audio (task still in PB): %w", err)
+	}
+	return id, nil
+}
+
+// CancelTask delegates to PocketBase; Asynq jobs may already be running and
+// cannot be reliably cancelled, so we only update the audit record.
+func (p *Producer) CancelTask(ctx context.Context, id string) error {
+	return p.pb.CancelTask(ctx, id)
+}
+
+// enqueue serialises payload and dispatches it to Asynq.
+func (p *Producer) enqueue(_ context.Context, taskType string, payload any) error {
+	b, err := json.Marshal(payload)
+	if err != nil {
+		return fmt.Errorf("marshal payload: %w", err)
+	}
+	_, err = p.client.Enqueue(asynq.NewTask(taskType, b))
+	return err
+}
--- a/backend/internal/asynqqueue/tasks.go
+++ b/backend/internal/asynqqueue/tasks.go
@@ -0,0 +1,46 @@
+// Package asynqqueue provides Asynq-backed implementations of the
+// taskqueue.Producer and taskqueue.Consumer interfaces.
+//
+// Architecture:
+//   - Producer: dual-writes — creates a PocketBase record for audit/UI, then
+//     enqueues an Asynq job so the runner picks it up immediately (sub-ms).
+//   - Consumer: thin wrapper used only for result write-back (FinishAudioTask,
+//     FinishScrapeTask, FailTask). ClaimNext*/Heartbeat/Reap are no-ops because
+//     Asynq owns those responsibilities.
+//   - Handlers: asynq.HandlerFunc wrappers that decode job payloads and invoke
+//     the existing runner logic (runScrapeTask / runAudioTask).
+//
+// Fallback: when REDIS_ADDR is empty the caller should use the plain
+// storage.Store (PocketBase-polling) implementation unchanged.
+package asynqqueue
+
+// Queue names — keep all jobs on the default queue for now.
+// Add separate queues (e.g. "audio", "scrape") later if you need priority.
+const QueueDefault = "default"
+
+// Task type constants used for Asynq routing.
+const (
+	TypeAudioGenerate   = "audio:generate"
+	TypeScrapeBook      = "scrape:book"
+	TypeScrapeCatalogue = "scrape:catalogue"
+)
+
+// AudioPayload is the Asynq job payload for audio generation tasks.
+type AudioPayload struct {
+	// PBTaskID is the PocketBase record ID created before enqueueing.
+	// The handler uses it to write results back via Consumer.FinishAudioTask.
+	PBTaskID string `json:"pb_task_id"`
+	Slug     string `json:"slug"`
+	Chapter  int    `json:"chapter"`
+	Voice    string `json:"voice"`
+}
+
+// ScrapePayload is the Asynq job payload for scrape tasks.
+type ScrapePayload struct {
+	// PBTaskID is the PocketBase record ID created before enqueueing.
+	PBTaskID    string `json:"pb_task_id"`
+	Kind        string `json:"kind"`         // "catalogue", "book", or "book_range"
+	TargetURL   string `json:"target_url"`   // empty for catalogue tasks
+	FromChapter int    `json:"from_chapter"` // 0 unless Kind=="book_range"
+	ToChapter   int    `json:"to_chapter"`   // 0 unless Kind=="book_range"
+}
--- a/backend/internal/backend/handlers.go
+++ b/backend/internal/backend/handlers.go
@@ -47,6 +47,7 @@ import (
 	"github.com/libnovel/backend/internal/kokoro"
 	"github.com/libnovel/backend/internal/meili"
 	"github.com/libnovel/backend/internal/novelfire/htmlutil"
+	"github.com/libnovel/backend/internal/pockettts"
 	"github.com/libnovel/backend/internal/scraper"
 )

@@ -703,7 +704,7 @@ func (s *Server) handleAudioProxy(w http.ResponseWriter, r *http.Request) {
 // ── Voices ─────────────────────────────────────────────────────────────────────

 // handleVoices handles GET /api/voices.
-// Returns {"voices": [...]} — fetched from Kokoro with built-in fallback.
+// Returns {"voices": [...]} — merged list from Kokoro and pocket-tts.
 func (s *Server) handleVoices(w http.ResponseWriter, r *http.Request) {
 	writeJSON(w, 0, map[string]any{"voices": s.voices(r.Context())})
 }
@@ -763,8 +764,8 @@ const voiceSampleText = "Hello! This is a preview of what I sound like. I hope y

 // handlePresignVoiceSample handles GET /api/presign/voice-sample/{voice}.
 // If the sample has not been generated yet it synthesises it on the fly via
-// Kokoro, stores the result in MinIO, and returns the presigned URL — so the
-// caller always gets a playable URL in a single request.
+// the appropriate TTS engine (Kokoro for kokoro voices, pocket-tts for
+// pocket-tts voices), stores the result in MinIO, and returns the presigned URL.
 func (s *Server) handlePresignVoiceSample(w http.ResponseWriter, r *http.Request) {
 	voice := r.PathValue("voice")
 	if voice == "" {
@@ -777,7 +778,20 @@ func (s *Server) handlePresignVoiceSample(w http.ResponseWriter, r *http.Request
 	// Generate sample on demand when it is not in MinIO yet.
 	if !s.deps.AudioStore.AudioExists(r.Context(), key) {
 		s.deps.Log.Info("generating voice sample on demand", "voice", voice)
-		mp3, err := s.deps.Kokoro.GenerateAudio(r.Context(), voiceSampleText, voice)
+
+		var (
+			mp3 []byte
+			err error
+		)
+		if pockettts.IsPocketTTSVoice(voice) {
+			if s.deps.PocketTTS == nil {
+				jsonError(w, http.StatusServiceUnavailable, "pocket-tts not configured")
+				return
+			}
+			mp3, err = s.deps.PocketTTS.GenerateAudio(r.Context(), voiceSampleText, voice)
+		} else {
+			mp3, err = s.deps.Kokoro.GenerateAudio(r.Context(), voiceSampleText, voice)
+		}
 		if err != nil {
 			s.deps.Log.Error("voice sample generation failed", "voice", voice, "err", err)
 			jsonError(w, http.StatusInternalServerError, "voice sample generation failed")
@@ -1148,9 +1162,9 @@ func stripMarkdown(src string) string {

 // ── Hardcoded Kokoro voice fallback ───────────────────────────────────────────

-// kokoroVoices is the built-in fallback list used when the Kokoro service is
-// unavailable. Matches the list in the old scraper helpers.go.
-var kokoroVoices = []string{
+// kokoroVoiceIDs is the built-in fallback list of Kokoro voice IDs used when
+// the Kokoro service is unavailable.
+var kokoroVoiceIDs = []string{
 	// American English
 	"af_alloy", "af_aoede", "af_bella", "af_heart", "af_jadzia",
 	"af_jessica", "af_kore", "af_nicole", "af_nova", "af_river",
--- a/backend/internal/backend/server.go
+++ b/backend/internal/backend/server.go
@@ -30,8 +30,10 @@ import (

 	sentryhttp "github.com/getsentry/sentry-go/http"
 	"github.com/libnovel/backend/internal/bookstore"
+	"github.com/libnovel/backend/internal/domain"
 	"github.com/libnovel/backend/internal/kokoro"
 	"github.com/libnovel/backend/internal/meili"
+	"github.com/libnovel/backend/internal/pockettts"
 	"github.com/libnovel/backend/internal/taskqueue"
 	"go.opentelemetry.io/contrib/instrumentation/net/http/otelhttp"
 )
@@ -59,9 +61,12 @@ type Dependencies struct {
 	// SearchIndex provides full-text book search via Meilisearch.
 	// If nil, the local-only fallback search is used.
 	SearchIndex meili.Client
-	// Kokoro is the TTS client (used for voice list only in the backend;
+	// Kokoro is the Kokoro TTS client (used for voice list only in the backend;
 	// audio generation is done by the runner).
 	Kokoro kokoro.Client
+	// PocketTTS is the pocket-tts client (used for voice list only in the backend;
+	// audio generation is done by the runner).
+	PocketTTS pockettts.Client
 	// Log is the structured logger.
 	Log *slog.Logger
 }
@@ -84,7 +89,7 @@ type Server struct {

 	// voiceMu guards cachedVoices. Populated lazily on first GET /api/voices.
 	voiceMu      sync.RWMutex
-	cachedVoices []string
+	cachedVoices []domain.Voice
 }

 // New creates a Server from cfg and deps.
@@ -264,10 +269,10 @@ func jsonError(w http.ResponseWriter, status int, msg string) {
 	_ = json.NewEncoder(w).Encode(map[string]string{"error": msg})
 }

-// voices returns the list of available Kokoro voices. On the first call it
-// fetches from the Kokoro service and caches the result. Falls back to the
-// hardcoded list on error.
-func (s *Server) voices(ctx context.Context) []string {
+// voices returns the merged list of available voices from Kokoro and pocket-tts.
+// On the first call it fetches from both services and caches the result.
+// Falls back to the hardcoded Kokoro list on error.
+func (s *Server) voices(ctx context.Context) []domain.Voice {
 	s.voiceMu.RLock()
 	cached := s.cachedVoices
 	s.voiceMu.RUnlock()
@@ -275,23 +280,89 @@ func (s *Server) voices(ctx context.Context) []string {
 		return cached
 	}

-	if s.deps.Kokoro == nil {
-		return kokoroVoices
-	}
-
 	fetchCtx, cancel := context.WithTimeout(ctx, 5*time.Second)
 	defer cancel()
-	list, err := s.deps.Kokoro.ListVoices(fetchCtx)
-	if err != nil || len(list) == 0 {
-		s.deps.Log.Warn("backend: could not fetch kokoro voices, using built-in list", "err", err)
-		return kokoroVoices
+
+	var result []domain.Voice
+
+	// ── Kokoro voices ─────────────────────────────────────────────────────────
+	var kokoroIDs []string
+	if s.deps.Kokoro != nil {
+		ids, err := s.deps.Kokoro.ListVoices(fetchCtx)
+		if err != nil || len(ids) == 0 {
+			s.deps.Log.Warn("backend: could not fetch kokoro voices, using built-in list", "err", err)
+			ids = kokoroVoiceIDs
+		} else {
+			s.deps.Log.Info("backend: fetched kokoro voices", "count", len(ids))
+		}
+		kokoroIDs = ids
+	} else {
+		kokoroIDs = kokoroVoiceIDs
+	}
+	for _, id := range kokoroIDs {
+		result = append(result, kokoroVoice(id))
+	}
+
+	// ── Pocket-TTS voices ─────────────────────────────────────────────────────
+	if s.deps.PocketTTS != nil {
+		ids, err := s.deps.PocketTTS.ListVoices(fetchCtx)
+		if err != nil {
+			s.deps.Log.Warn("backend: could not fetch pocket-tts voices", "err", err)
+		} else {
+			for _, id := range ids {
+				result = append(result, pocketTTSVoice(id))
+			}
+			s.deps.Log.Info("backend: fetched pocket-tts voices", "count", len(ids))
+		}
 	}

 	s.voiceMu.Lock()
-	s.cachedVoices = list
+	s.cachedVoices = result
 	s.voiceMu.Unlock()
-	s.deps.Log.Info("backend: fetched kokoro voices", "count", len(list))
-	return list
+	return result
+}
+
+// kokoroVoice builds a domain.Voice for a Kokoro voice ID.
+// The two-character prefix encodes language and gender:
+//
+//	af/am → en-us f/m  |  bf/bm → en-gb f/m
+//	ef/em → es    f/m  |  ff    → fr    f
+//	hf/hm → hi    f/m  |  if/im → it    f/m
+//	jf/jm → ja    f/m  |  pf/pm → pt    f/m
+//	zf/zm → zh    f/m
+func kokoroVoice(id string) domain.Voice {
+	type meta struct{ lang, gender string }
+	prefixMap := map[string]meta{
+		"af": {"en-us", "f"}, "am": {"en-us", "m"},
+		"bf": {"en-gb", "f"}, "bm": {"en-gb", "m"},
+		"ef": {"es", "f"}, "em": {"es", "m"},
+		"ff": {"fr", "f"},
+		"hf": {"hi", "f"}, "hm": {"hi", "m"},
+		"if": {"it", "f"}, "im": {"it", "m"},
+		"jf": {"ja", "f"}, "jm": {"ja", "m"},
+		"pf": {"pt", "f"}, "pm": {"pt", "m"},
+		"zf": {"zh", "f"}, "zm": {"zh", "m"},
+	}
+	if len(id) >= 2 {
+		if m, ok := prefixMap[id[:2]]; ok {
+			return domain.Voice{ID: id, Engine: "kokoro", Lang: m.lang, Gender: m.gender}
+		}
+	}
+	return domain.Voice{ID: id, Engine: "kokoro", Lang: "en", Gender: ""}
+}
+
+// pocketTTSVoice builds a domain.Voice for a pocket-tts voice ID.
+// All pocket-tts voices are English audiobook narrators.
+func pocketTTSVoice(id string) domain.Voice {
+	femaleVoices := map[string]struct{}{
+		"alba": {}, "fantine": {}, "cosette": {}, "eponine": {},
+		"azelma": {}, "anna": {}, "vera": {}, "mary": {}, "jane": {}, "eve": {},
+	}
+	gender := "m"
+	if _, ok := femaleVoices[id]; ok {
+		gender = "f"
+	}
+	return domain.Voice{ID: id, Engine: "pocket-tts", Lang: "en", Gender: gender}
 }

 // handleHealth handles GET /health.
--- a/backend/internal/config/config.go
+++ b/backend/internal/config/config.go
@@ -86,6 +86,19 @@ type Valkey struct {
 	Addr string
 }

+// Redis holds connection settings for the Asynq task queue Redis instance.
+// This is separate from Valkey (presign cache) — it may point to the same
+// Redis or a dedicated one. An empty Addr falls back to PocketBase polling.
+type Redis struct {
+	// Addr is the host:port (or rediss://... URL) of the Redis instance.
+	// Use rediss:// scheme for TLS (e.g. rediss://:password@redis.libnovel.cc:6380).
+	// An empty string disables Asynq and falls back to PocketBase polling.
+	Addr string
+	// Password is the Redis AUTH password.
+	// Not needed when Addr is a full rediss:// URL that includes the password.
+	Password string
+}
+
 // Runner holds settings specific to the runner/worker binary.
 type Runner struct {
 	// PollInterval is how often the runner checks PocketBase for pending tasks.
@@ -113,6 +126,11 @@ type Runner struct {
 	// is already indexed and a 24h walk would be wasteful.
 	// Controlled by RUNNER_SKIP_INITIAL_CATALOGUE_REFRESH=true.
 	SkipInitialCatalogueRefresh bool
+	// CatalogueRequestDelay is the base delay inserted between per-book metadata
+	// requests during a catalogue refresh. A random jitter of up to 50% is added
+	// on top. Defaults to 2s. Increase to reduce 429 pressure on novelfire.net.
+	// Controlled by RUNNER_CATALOGUE_REQUEST_DELAY (e.g. "3s", "500ms").
+	CatalogueRequestDelay time.Duration
 }

 // Config is the top-level configuration struct consumed by both binaries.
@@ -125,6 +143,7 @@ type Config struct {
 	Runner      Runner
 	Meilisearch Meilisearch
 	Valkey      Valkey
+	Redis       Redis
 	// LogLevel is one of "debug", "info", "warn", "error".
 	LogLevel string
 }
@@ -182,6 +201,7 @@ func Load() Config {
 			MetricsAddr:                 envOr("RUNNER_METRICS_ADDR", ":9091"),
 			CatalogueRefreshInterval:    envDuration("RUNNER_CATALOGUE_REFRESH_INTERVAL", 0),
 			SkipInitialCatalogueRefresh: envBool("RUNNER_SKIP_INITIAL_CATALOGUE_REFRESH", false),
+			CatalogueRequestDelay:       envDuration("RUNNER_CATALOGUE_REQUEST_DELAY", 2*time.Second),
 		},

 		Meilisearch: Meilisearch{
@@ -192,6 +212,11 @@ func Load() Config {
 		Valkey: Valkey{
 			Addr: envOr("VALKEY_ADDR", ""),
 		},
+
+		Redis: Redis{
+			Addr:     envOr("REDIS_ADDR", ""),
+			Password: envOr("REDIS_PASSWORD", ""),
+		},
 	}
 }

--- a/backend/internal/domain/domain.go
+++ b/backend/internal/domain/domain.go
@@ -60,6 +60,20 @@ type RankingItem struct {
 	Updated   time.Time `json:"updated,omitempty"`
 }

+// ── Voice types ───────────────────────────────────────────────────────────────
+
+// Voice describes a single text-to-speech voice available in the system.
+type Voice struct {
+	// ID is the voice identifier passed to TTS clients (e.g. "af_bella", "alba").
+	ID string `json:"id"`
+	// Engine is "kokoro" or "pocket-tts".
+	Engine string `json:"engine"`
+	// Lang is the primary language tag (e.g. "en-us", "en-gb", "en", "es", "fr").
+	Lang string `json:"lang"`
+	// Gender is "f" or "m".
+	Gender string `json:"gender"`
+}
+
 // ── Storage record types ──────────────────────────────────────────────────────

 // ChapterInfo is a lightweight chapter descriptor stored in the index.
--- a/backend/internal/novelfire/scraper.go
+++ b/backend/internal/novelfire/scraper.go
@@ -13,6 +13,7 @@ import (
 	"errors"
 	"fmt"
 	"log/slog"
+	"math/rand"
 	"net/url"
 	"path"
 	"strconv"
@@ -55,6 +56,9 @@ func (s *Scraper) SourceName() string { return "novelfire.net" }
 // ── CatalogueProvider ─────────────────────────────────────────────────────────

 // ScrapeCatalogue streams all CatalogueEntry values across all catalogue pages.
+// Each page fetch uses retryGet with 429-aware exponential backoff.
+// A small inter-page delay (cataloguePageDelay) is inserted between requests to
+// avoid hammering the server when paging through hundreds of catalogue pages.
 func (s *Scraper) ScrapeCatalogue(ctx context.Context) (<-chan domain.CatalogueEntry, <-chan error) {
 	entries := make(chan domain.CatalogueEntry, 64)
 	errs := make(chan error, 16)
@@ -73,8 +77,18 @@ func (s *Scraper) ScrapeCatalogue(ctx context.Context) (<-chan domain.CatalogueE
 			default:
 			}

+			// Polite inter-page delay — skipped on the very first page.
+			if page > 1 {
+				jitter := time.Duration(500+rand.Intn(1000)) * time.Millisecond
+				select {
+				case <-ctx.Done():
+					return
+				case <-time.After(jitter):
+				}
+			}
+
 			s.log.Info("scraping catalogue page", "page", page, "url", pageURL)
-			raw, err := s.client.GetContent(ctx, pageURL)
+			raw, err := retryGet(ctx, s.log, s.client, pageURL, 9, 10*time.Second)
 			if err != nil {
 				errs <- fmt.Errorf("catalogue page %d: %w", page, err)
 				return
@@ -139,10 +153,11 @@ func (s *Scraper) ScrapeCatalogue(ctx context.Context) (<-chan domain.CatalogueE
 // ── MetadataProvider ──────────────────────────────────────────────────────────

 // ScrapeMetadata fetches and parses book metadata from the book's landing page.
+// Uses retryGet with 429-aware exponential backoff (up to 9 attempts).
 func (s *Scraper) ScrapeMetadata(ctx context.Context, bookURL string) (domain.BookMeta, error) {
 	s.log.Debug("metadata fetch starting", "url", bookURL)

-	raw, err := s.client.GetContent(ctx, bookURL)
+	raw, err := retryGet(ctx, s.log, s.client, bookURL, 9, 10*time.Second)
 	if err != nil {
 		return domain.BookMeta{}, fmt.Errorf("metadata fetch %s: %w", bookURL, err)
 	}
--- a/backend/internal/runner/asynq_runner.go
+++ b/backend/internal/runner/asynq_runner.go
@@ -0,0 +1,149 @@
+package runner
+
+// asynq_runner.go — Asynq-based task dispatch for the runner.
+//
+// When cfg.RedisAddr is set, Run() calls runAsynq() instead of runPoll().
+// The Asynq server replaces the polling loop: it listens on Redis for tasks
+// enqueued by the backend Producer and delivers them immediately.
+//
+// Handlers in this file decode Asynq job payloads and call the existing
+// runScrapeTask / runAudioTask methods, keeping all execution logic in one place.
+
+import (
+	"context"
+	"encoding/json"
+	"fmt"
+	"time"
+
+	"github.com/hibiken/asynq"
+	asynqmetrics "github.com/hibiken/asynq/x/metrics"
+	"github.com/libnovel/backend/internal/asynqqueue"
+	"github.com/libnovel/backend/internal/domain"
+)
+
+// runAsynq starts an Asynq server that replaces the PocketBase poll loop.
+// It also starts the periodic catalogue refresh ticker.
+// Blocks until ctx is cancelled.
+func (r *Runner) runAsynq(ctx context.Context) error {
+	redisOpt, err := r.redisConnOpt()
+	if err != nil {
+		return fmt.Errorf("runner: parse redis addr: %w", err)
+	}
+
+	srv := asynq.NewServer(redisOpt, asynq.Config{
+		// Allocate concurrency slots for each task type.
+		// Total concurrency = scrape + audio slots.
+		Concurrency: r.cfg.MaxConcurrentScrape + r.cfg.MaxConcurrentAudio,
+		Queues: map[string]int{
+			asynqqueue.QueueDefault: 1,
+		},
+		// Let Asynq handle retries with exponential back-off.
+		RetryDelayFunc: asynq.DefaultRetryDelayFunc,
+		// Log errors from handlers via the existing structured logger.
+		ErrorHandler: asynq.ErrorHandlerFunc(func(_ context.Context, task *asynq.Task, err error) {
+			r.deps.Log.Error("runner: asynq task failed",
+				"type", task.Type(),
+				"err", err,
+			)
+		}),
+	})
+
+	mux := asynq.NewServeMux()
+	mux.HandleFunc(asynqqueue.TypeAudioGenerate, r.handleAudioTask)
+	mux.HandleFunc(asynqqueue.TypeScrapeBook, r.handleScrapeTask)
+	mux.HandleFunc(asynqqueue.TypeScrapeCatalogue, r.handleScrapeTask)
+
+	// Register Asynq queue metrics with the default Prometheus registry so
+	// the /metrics endpoint (metrics.go) can expose them.
+	inspector := asynq.NewInspector(redisOpt)
+	collector := asynqmetrics.NewQueueMetricsCollector(inspector)
+	if err := r.metricsRegistry.Register(collector); err != nil {
+		r.deps.Log.Warn("runner: could not register asynq prometheus collector", "err", err)
+	}
+
+	// Start the periodic catalogue refresh.
+	catalogueTick := time.NewTicker(r.cfg.CatalogueRefreshInterval)
+	defer catalogueTick.Stop()
+	if !r.cfg.SkipInitialCatalogueRefresh {
+		go r.runCatalogueRefresh(ctx)
+	} else {
+		r.deps.Log.Info("runner: skipping initial catalogue refresh (RUNNER_SKIP_INITIAL_CATALOGUE_REFRESH=true)")
+	}
+
+	r.deps.Log.Info("runner: asynq mode active", "redis_addr", r.cfg.RedisAddr)
+
+	// Run catalogue refresh ticker in the background.
+	go func() {
+		for {
+			select {
+			case <-ctx.Done():
+				return
+			case <-catalogueTick.C:
+				go r.runCatalogueRefresh(ctx)
+			}
+		}
+	}()
+
+	// Start Asynq server (non-blocking).
+	if err := srv.Start(mux); err != nil {
+		return fmt.Errorf("runner: asynq server start: %w", err)
+	}
+
+	// Block until context is cancelled, then gracefully stop.
+	<-ctx.Done()
+	r.deps.Log.Info("runner: context cancelled, shutting down asynq server")
+	srv.Shutdown()
+	return nil
+}
+
+// redisConnOpt parses cfg.RedisAddr into an asynq.RedisConnOpt.
+// Supports full "redis://" / "rediss://" URLs and plain "host:port".
+func (r *Runner) redisConnOpt() (asynq.RedisConnOpt, error) {
+	addr := r.cfg.RedisAddr
+	// ParseRedisURI handles redis:// and rediss:// schemes.
+	if len(addr) > 7 && (addr[:8] == "redis://" || addr[:9] == "rediss://") {
+		return asynq.ParseRedisURI(addr)
+	}
+	// Plain "host:port" — use RedisClientOpt directly.
+	return asynq.RedisClientOpt{
+		Addr:     addr,
+		Password: r.cfg.RedisPassword,
+	}, nil
+}
+
+// handleScrapeTask is the Asynq handler for TypeScrapeBook and TypeScrapeCatalogue.
+func (r *Runner) handleScrapeTask(ctx context.Context, t *asynq.Task) error {
+	var p asynqqueue.ScrapePayload
+	if err := json.Unmarshal(t.Payload(), &p); err != nil {
+		return fmt.Errorf("unmarshal scrape payload: %w", err)
+	}
+	task := domain.ScrapeTask{
+		ID:          p.PBTaskID,
+		Kind:        p.Kind,
+		TargetURL:   p.TargetURL,
+		FromChapter: p.FromChapter,
+		ToChapter:   p.ToChapter,
+	}
+	r.tasksRunning.Add(1)
+	defer r.tasksRunning.Add(-1)
+	r.runScrapeTask(ctx, task)
+	return nil
+}
+
+// handleAudioTask is the Asynq handler for TypeAudioGenerate.
+func (r *Runner) handleAudioTask(ctx context.Context, t *asynq.Task) error {
+	var p asynqqueue.AudioPayload
+	if err := json.Unmarshal(t.Payload(), &p); err != nil {
+		return fmt.Errorf("unmarshal audio payload: %w", err)
+	}
+	task := domain.AudioTask{
+		ID:      p.PBTaskID,
+		Slug:    p.Slug,
+		Chapter: p.Chapter,
+		Voice:   p.Voice,
+	}
+	r.tasksRunning.Add(1)
+	defer r.tasksRunning.Add(-1)
+	r.runAudioTask(ctx, task)
+	return nil
+}
--- a/backend/internal/runner/catalogue_refresh.go
+++ b/backend/internal/runner/catalogue_refresh.go
@@ -6,17 +6,20 @@ package runner
 //
 // Design:
 //   - Runs on its own ticker (CatalogueRefreshInterval, default 24h) inside Run().
-//   - Also fires once on startup.
-//   - ScrapeCatalogue streams CatalogueEntry values over a channel — we iterate
-//     and call ScrapeMetadata for each entry.
-//   - Per-request random jitter (1–3s) prevents hammering novelfire.net.
-//   - Cover images are fetched from the URL embedded in BookMeta.Cover and
-//     stored in MinIO (browse bucket, key: covers/{slug}.jpg).
-//   - WriteMetadata + UpsertBook are called for every successfully scraped book.
-//   - Errors for individual books are logged and skipped; the loop continues.
-//   - The cover URL stored in BookMeta.Cover is rewritten to the internal proxy
-//     path (/api/cover/novelfire.net/{slug}) so the UI always fetches via the
-//     backend, which will serve from MinIO.
+//   - Also fires once on startup (unless SkipInitialCatalogueRefresh is set).
+//   - ScrapeCatalogue streams CatalogueEntry values over a channel — already has
+//     its own inter-page jitter + retryGet (see scraper.go).
+//   - Per-book: only metadata is scraped here (not chapters). Chapters are scraped
+//     on-demand when a user opens a book or via an explicit scrape task.
+//   - Between each metadata request a configurable base delay plus up to 50%
+//     random jitter is applied (CatalogueRequestDelay, default 2s). This keeps
+//     the request rate well below novelfire.net's rate limit even for ~15k books.
+//   - ScrapeMetadata itself uses retryGet with 429-aware exponential backoff
+//     (up to 9 attempts), so transient rate limits are handled gracefully.
+//   - Cover images are fetched and stored in MinIO on first sight; subsequent
+//     refreshes skip covers that already exist (CoverExists check).
+//   - Books already present in Meilisearch are skipped entirely (fast path).
+//   - Errors for individual books are logged and skipped; the loop never aborts.

 import (
 	"context"
@@ -29,7 +32,7 @@ import (

 // runCatalogueRefresh performs one full catalogue walk: scrapes metadata for
 // every book on novelfire.net, downloads covers to MinIO, and upserts to
-// Meilisearch. Errors for individual books are logged and skipped.
+// Meilisearch. Individual book failures are logged and skipped.
 func (r *Runner) runCatalogueRefresh(ctx context.Context) {
 	if r.deps.Novel == nil {
 		r.deps.Log.Warn("runner: catalogue refresh skipped — Novel scraper not configured")
@@ -40,8 +43,9 @@ func (r *Runner) runCatalogueRefresh(ctx context.Context) {
 		return
 	}

+	delay := r.cfg.CatalogueRequestDelay
 	log := r.deps.Log.With("op", "catalogue_refresh")
-	log.Info("runner: catalogue refresh starting")
+	log.Info("runner: catalogue refresh starting", "request_delay", delay)

 	entries, errCh := r.deps.Novel.ScrapeCatalogue(ctx)

@@ -51,26 +55,26 @@ func (r *Runner) runCatalogueRefresh(ctx context.Context) {
 			break
 		}

-		// Skip books already present in Meilisearch — they were indexed on a
-		// previous run. Re-indexing only happens when a scrape task is
-		// explicitly enqueued (e.g. via the admin UI or API).
+		// Fast path: skip books already indexed in Meilisearch.
 		if r.deps.SearchIndex.BookExists(ctx, entry.Slug) {
 			skipped++
 			continue
 		}

-		// Random jitter between books to avoid rate-limiting.
-		jitter := time.Duration(1000+rand.Intn(2000)) * time.Millisecond
+		// Polite delay between metadata requests: base + up to 50% jitter.
+		// This applies before every fetch so we never fire bursts.
+		jitter := time.Duration(rand.Int63n(int64(delay / 2)))
 		select {
 		case <-ctx.Done():
 			break
-		case <-time.After(jitter):
+		case <-time.After(delay + jitter):
 		}

+		// ScrapeMetadata internally retries on 429 with exponential back-off.
 		meta, err := r.deps.Novel.ScrapeMetadata(ctx, entry.URL)
 		if err != nil {
-			log.Warn("runner: catalogue refresh: metadata scrape failed",
-				"url", entry.URL, "err", err)
+			log.Warn("runner: catalogue refresh: metadata scrape failed — skipping book",
+				"slug", entry.Slug, "url", entry.URL, "err", err)
 			errCount++
 			continue
 		}
@@ -81,35 +85,32 @@ func (r *Runner) runCatalogueRefresh(ctx context.Context) {

 		// Persist to PocketBase.
 		if err := r.deps.BookWriter.WriteMetadata(ctx, meta); err != nil {
-			log.Warn("runner: catalogue refresh: WriteMetadata failed",
+			log.Warn("runner: catalogue refresh: WriteMetadata failed — skipping book",
 				"slug", meta.Slug, "err", err)
 			errCount++
 			continue
 		}

-		// Index in Meilisearch.
+		// Index in Meilisearch (non-fatal).
 		if err := r.deps.SearchIndex.UpsertBook(ctx, meta); err != nil {
 			log.Warn("runner: catalogue refresh: UpsertBook failed",
 				"slug", meta.Slug, "err", err)
-			// non-fatal — continue
 		}

-		// Download and store cover image in MinIO if we have a cover URL
-		// and a CoverStore is wired in.
+		// Download cover to MinIO if not already cached (non-fatal).
 		if r.deps.CoverStore != nil && originalCover != "" {
 			if !r.deps.CoverStore.CoverExists(ctx, meta.Slug) {
 				if err := r.downloadCover(ctx, meta.Slug, originalCover); err != nil {
 					log.Warn("runner: catalogue refresh: cover download failed",
 						"slug", meta.Slug, "url", originalCover, "err", err)
-					// non-fatal
 				}
 			}
 		}

 		ok++
-		if ok%100 == 0 {
+		if ok%50 == 0 {
 			log.Info("runner: catalogue refresh progress",
-				"scraped", ok, "errors", errCount)
+				"scraped", ok, "skipped", skipped, "errors", errCount)
 		}
 	}

--- a/backend/internal/runner/metrics.go
+++ b/backend/internal/runner/metrics.go
@@ -1,21 +1,28 @@
 package runner

-// metrics.go — lightweight HTTP metrics endpoint for the runner.
+// metrics.go — Prometheus metrics HTTP endpoint for the runner.
 //
-// GET /metrics returns a JSON document with live task counters and uptime.
-// No external dependency (no Prometheus); plain net/http only.
+// GET /metrics returns a Prometheus text/plain scrape response.
+// Exposes:
+//   - Standard Go runtime metrics (via promhttp)
+//   - Runner task counters (tasks_running, tasks_completed, tasks_failed)
+//   - Asynq queue metrics (registered in asynq_runner.go when Redis is enabled)
+//
+// GET /health — simple liveness probe.

 import (
 	"context"
-	"encoding/json"
 	"fmt"
 	"log/slog"
 	"net"
 	"net/http"
 	"time"
+
+	"github.com/prometheus/client_golang/prometheus"
+	"github.com/prometheus/client_golang/prometheus/promhttp"
 )

-// metricsServer serves GET /metrics for the runner process.
+// metricsServer serves GET /metrics and GET /health for the runner process.
 type metricsServer struct {
 	addr string
 	r    *Runner
@@ -23,21 +30,62 @@ type metricsServer struct {
 }

 func newMetricsServer(addr string, r *Runner, log *slog.Logger) *metricsServer {
-	return &metricsServer{addr: addr, r: r, log: log}
+	ms := &metricsServer{addr: addr, r: r, log: log}
+	ms.registerCollectors()
+	return ms
+}
+
+// registerCollectors registers runner-specific Prometheus collectors.
+// Called once at construction; Asynq queue collector is registered separately
+// in asynq_runner.go after the Redis connection is established.
+func (ms *metricsServer) registerCollectors() {
+	// Runner task gauges / counters backed by the atomic fields on Runner.
+	ms.r.metricsRegistry.MustRegister(prometheus.NewGaugeFunc(
+		prometheus.GaugeOpts{
+			Namespace: "runner",
+			Name:      "tasks_running",
+			Help:      "Number of tasks currently being processed.",
+		},
+		func() float64 { return float64(ms.r.tasksRunning.Load()) },
+	))
+	ms.r.metricsRegistry.MustRegister(prometheus.NewCounterFunc(
+		prometheus.CounterOpts{
+			Namespace: "runner",
+			Name:      "tasks_completed_total",
+			Help:      "Total number of tasks completed successfully since startup.",
+		},
+		func() float64 { return float64(ms.r.tasksCompleted.Load()) },
+	))
+	ms.r.metricsRegistry.MustRegister(prometheus.NewCounterFunc(
+		prometheus.CounterOpts{
+			Namespace: "runner",
+			Name:      "tasks_failed_total",
+			Help:      "Total number of tasks that ended in failure since startup.",
+		},
+		func() float64 { return float64(ms.r.tasksFailed.Load()) },
+	))
+	ms.r.metricsRegistry.MustRegister(prometheus.NewGaugeFunc(
+		prometheus.GaugeOpts{
+			Namespace: "runner",
+			Name:      "uptime_seconds",
+			Help:      "Seconds since the runner process started.",
+		},
+		func() float64 { return time.Since(ms.r.startedAt).Seconds() },
+	))
 }

 // ListenAndServe starts the HTTP server and blocks until ctx is cancelled or
 // a fatal listen error occurs.
 func (ms *metricsServer) ListenAndServe(ctx context.Context) error {
 	mux := http.NewServeMux()
-	mux.HandleFunc("GET /metrics", ms.handleMetrics)
+	mux.Handle("GET /metrics", promhttp.HandlerFor(ms.r.metricsRegistry, promhttp.HandlerOpts{}))
 	mux.HandleFunc("GET /health", ms.handleHealth)

 	srv := &http.Server{
 		Addr:         ms.addr,
 		Handler:      mux,
 		ReadTimeout:  5 * time.Second,
-		WriteTimeout: 5 * time.Second,
+		WriteTimeout: 10 * time.Second,
 		BaseContext:  func(_ net.Listener) context.Context { return ctx },
 	}

@@ -58,35 +106,8 @@ func (ms *metricsServer) ListenAndServe(ctx context.Context) error {
 	}
 }

-// handleMetrics handles GET /metrics.
-// Response shape (JSON):
-//
-//	{
-//	  "tasks_running":   N,
-//	  "tasks_completed": N,
-//	  "tasks_failed":    N,
-//	  "uptime_seconds":  N
-//	}
-func (ms *metricsServer) handleMetrics(w http.ResponseWriter, _ *http.Request) {
-	uptimeSec := int64(time.Since(ms.r.startedAt).Seconds())
-	metricsWriteJSON(w, 0, map[string]int64{
-		"tasks_running":   ms.r.tasksRunning.Load(),
-		"tasks_completed": ms.r.tasksCompleted.Load(),
-		"tasks_failed":    ms.r.tasksFailed.Load(),
-		"uptime_seconds":  uptimeSec,
-	})
-}
-
-// handleHealth handles GET /health — simple liveness probe for the metrics server.
+// handleHealth handles GET /health — simple liveness probe.
 func (ms *metricsServer) handleHealth(w http.ResponseWriter, _ *http.Request) {
-	metricsWriteJSON(w, 0, map[string]string{"status": "ok"})
-}
-
-// metricsWriteJSON writes v as a JSON response with the given status code.
-func metricsWriteJSON(w http.ResponseWriter, status int, v any) {
 	w.Header().Set("Content-Type", "application/json")
-	if status != 0 {
-		w.WriteHeader(status)
-	}
-	_ = json.NewEncoder(w).Encode(v)
+	_, _ = w.Write([]byte(`{"status":"ok"}`))
 }
--- a/backend/internal/runner/runner.go
+++ b/backend/internal/runner/runner.go
@@ -34,6 +34,7 @@ import (
 	"github.com/libnovel/backend/internal/pockettts"
 	"github.com/libnovel/backend/internal/scraper"
 	"github.com/libnovel/backend/internal/taskqueue"
+	"github.com/prometheus/client_golang/prometheus"
 )

 // Config tunes the runner behaviour.
@@ -41,6 +42,7 @@ type Config struct {
 	// WorkerID uniquely identifies this runner instance in PocketBase records.
 	WorkerID string
 	// PollInterval is how often the runner checks for new tasks.
+	// Only used in PocketBase-polling mode (RedisAddr == "").
 	PollInterval time.Duration
 	// MaxConcurrentScrape limits simultaneous book-scrape goroutines.
 	MaxConcurrentScrape int
@@ -50,14 +52,20 @@ type Config struct {
 	OrchestratorWorkers int
 	// HeartbeatInterval is how often active tasks PATCH their heartbeat_at
 	// timestamp to signal they are still alive. Defaults to 30s when 0.
+	// Only used in PocketBase-polling mode.
 	HeartbeatInterval time.Duration
 	// StaleTaskThreshold is how old a heartbeat must be (or absent) before the
 	// task is considered orphaned and reset to pending. Defaults to 2m when 0.
+	// Only used in PocketBase-polling mode.
 	StaleTaskThreshold time.Duration
 	// CatalogueRefreshInterval is how often the runner walks the full catalogue,
 	// scrapes per-book metadata, downloads covers, and re-indexes everything in
 	// Meilisearch. Defaults to 24h (expensive — full catalogue walk).
 	CatalogueRefreshInterval time.Duration
+	// CatalogueRequestDelay is the base inter-request pause during a catalogue
+	// refresh metadata walk. Jitter of up to 50% is added on top.
+	// Defaults to 2s. Set via RUNNER_CATALOGUE_REQUEST_DELAY.
+	CatalogueRequestDelay time.Duration
 	// SkipInitialCatalogueRefresh suppresses the immediate catalogue walk that
 	// otherwise fires at startup. The periodic ticker (CatalogueRefreshInterval)
 	// still fires normally. Set RUNNER_SKIP_INITIAL_CATALOGUE_REFRESH=true for
@@ -66,6 +74,15 @@ type Config struct {
 	// MetricsAddr is the HTTP listen address for the /metrics endpoint.
 	// Defaults to ":9091". Set to "" to disable.
 	MetricsAddr string
+	// RedisAddr is the address of the Redis instance used for Asynq task
+	// dispatch. When set the runner switches from PocketBase-polling mode to
+	// Asynq ServeMux mode (immediate task delivery, no polling).
+	// Supports plain "host:port" or a full "rediss://..." URL.
+	// When empty the runner falls back to PocketBase polling.
+	RedisAddr string
+	// RedisPassword is the Redis AUTH password.
+	// Not required when RedisAddr is a full URL that includes credentials.
+	RedisPassword string
 }

 // Dependencies are the external services the runner depends on.
@@ -99,6 +116,8 @@ type Runner struct {
 	cfg  Config
 	deps Dependencies

+	metricsRegistry *prometheus.Registry
+
 	// Atomic task counters — read by /metrics without locking.
 	tasksRunning   atomic.Int64
 	tasksCompleted atomic.Int64
@@ -130,6 +149,9 @@ func New(cfg Config, deps Dependencies) *Runner {
 	if cfg.CatalogueRefreshInterval <= 0 {
 		cfg.CatalogueRefreshInterval = 24 * time.Hour
 	}
+	if cfg.CatalogueRequestDelay <= 0 {
+		cfg.CatalogueRequestDelay = 2 * time.Second
+	}
 	if cfg.MetricsAddr == "" {
 		cfg.MetricsAddr = ":9091"
 	}
@@ -139,15 +161,18 @@ func New(cfg Config, deps Dependencies) *Runner {
 	if deps.SearchIndex == nil {
 		deps.SearchIndex = meili.NoopClient{}
 	}
-	return &Runner{cfg: cfg, deps: deps, startedAt: time.Now()}
+	return &Runner{cfg: cfg, deps: deps, startedAt: time.Now(), metricsRegistry: prometheus.NewRegistry()}
 }

-// Run starts the poll loop and the metrics HTTP server, blocking until ctx is
-// cancelled.
+// Run starts the worker loop and the metrics HTTP server, blocking until ctx
+// is cancelled.
+//
+// When cfg.RedisAddr is set the runner uses Asynq (immediate task delivery).
+// Otherwise it falls back to PocketBase polling (legacy mode).
 func (r *Runner) Run(ctx context.Context) error {
 	r.deps.Log.Info("runner: starting",
 		"worker_id", r.cfg.WorkerID,
-		"poll_interval", r.cfg.PollInterval,
+		"mode", r.mode(),
 		"max_scrape", r.cfg.MaxConcurrentScrape,
 		"max_audio", r.cfg.MaxConcurrentAudio,
 		"catalogue_refresh_interval", r.cfg.CatalogueRefreshInterval,
@@ -164,6 +189,23 @@ func (r *Runner) Run(ctx context.Context) error {
 		}()
 	}

+	if r.cfg.RedisAddr != "" {
+		return r.runAsynq(ctx)
+	}
+	return r.runPoll(ctx)
+}
+
+// mode returns a short string describing the active dispatch mode.
+func (r *Runner) mode() string {
+	if r.cfg.RedisAddr != "" {
+		return "asynq"
+	}
+	return "poll"
+}
+
+// runPoll is the legacy PocketBase-polling dispatch loop.
+// Used when cfg.RedisAddr is empty.
+func (r *Runner) runPoll(ctx context.Context) error {
 	scrapeSem := make(chan struct{}, r.cfg.MaxConcurrentScrape)
 	audioSem := make(chan struct{}, r.cfg.MaxConcurrentAudio)
 	var wg sync.WaitGroup
@@ -181,6 +223,8 @@ func (r *Runner) Run(ctx context.Context) error {
 		r.deps.Log.Info("runner: skipping initial catalogue refresh (RUNNER_SKIP_INITIAL_CATALOGUE_REFRESH=true)")
 	}

+	r.deps.Log.Info("runner: poll mode active", "poll_interval", r.cfg.PollInterval)
+
 	// Run one poll immediately on startup, then on each tick.
 	for {
 		r.poll(ctx, scrapeSem, audioSem, &wg)
--- a/backend/internal/storage/store.go
+++ b/backend/internal/storage/store.go
@@ -706,7 +706,7 @@ func (s *Store) ListAudioTasks(ctx context.Context) ([]domain.AudioTask, error)
 }

 func (s *Store) GetAudioTask(ctx context.Context, cacheKey string) (domain.AudioTask, bool, error) {
-	filter := fmt.Sprintf(`cache_key=%q`, cacheKey)
+	filter := fmt.Sprintf(`cache_key='%s'`, cacheKey)
 	items, err := s.pb.listAll(ctx, "audio_jobs", filter, "-started")
 	if err != nil || len(items) == 0 {
 		return domain.AudioTask{}, false, err
--- a/caddy/Dockerfile
+++ b/caddy/Dockerfile
@@ -2,7 +2,8 @@ FROM caddy:2-builder AS builder

 RUN xcaddy build \
    --with github.com/mholt/caddy-ratelimit \
-    --with github.com/hslatman/caddy-crowdsec-bouncer/http
+    --with github.com/hslatman/caddy-crowdsec-bouncer/http \
+    --with github.com/mholt/caddy-l4

 FROM caddy:2-alpine
 COPY --from=builder /usr/bin/caddy /usr/bin/caddy
--- a/docker-compose.yml
+++ b/docker-compose.yml
@@ -15,7 +15,7 @@ x-infra-env: &infra-env
  POCKETBASE_ADMIN_EMAIL: "${POCKETBASE_ADMIN_EMAIL}"
  POCKETBASE_ADMIN_PASSWORD: "${POCKETBASE_ADMIN_PASSWORD}"
  # Meilisearch
-  MEILI_URL: "http://meilisearch:7700"
+  MEILI_URL: "${MEILI_URL:-http://meilisearch:7700}"
  MEILI_API_KEY: "${MEILI_MASTER_KEY}"
  # Valkey
  VALKEY_ADDR: "valkey:6379"
@@ -154,15 +154,21 @@ services:
    # No public port — all traffic is routed via Caddy.
    expose:
      - "8080"
-    environment:
+     environment:
      <<: *infra-env
      BACKEND_HTTP_ADDR: ":8080"
      LOG_LEVEL: "${LOG_LEVEL}"
      KOKORO_URL: "${KOKORO_URL}"
      KOKORO_VOICE: "${KOKORO_VOICE}"
+      POCKET_TTS_URL: "${POCKET_TTS_URL}"
      GLITCHTIP_DSN: "${GLITCHTIP_DSN}"
      OTEL_EXPORTER_OTLP_ENDPOINT: "${OTEL_EXPORTER_OTLP_ENDPOINT}"
      OTEL_SERVICE_NAME: "backend"
+      # Asynq task queue — backend enqueues jobs to homelab Redis via Caddy TLS proxy.
+      # Set to "rediss://:password@redis.libnovel.cc:6380" in Doppler prd config.
+      # Leave empty to fall back to PocketBase polling.
+      REDIS_ADDR: "${REDIS_ADDR}"
+      REDIS_PASSWORD: "${REDIS_PASSWORD}"
    healthcheck:
      test: ["CMD", "/healthcheck", "http://localhost:8080/health"]
      interval: 15s
@@ -353,13 +359,16 @@ services:


  # ─── Caddy (reverse proxy + automatic HTTPS) ──────────────────────────────────
-  # Custom build includes github.com/mholt/caddy-ratelimit and
-  # github.com/hslatman/caddy-crowdsec-bouncer/http.
+  # Custom build includes github.com/mholt/caddy-ratelimit,
+  # github.com/hslatman/caddy-crowdsec-bouncer/http, and
+  # github.com/mholt/caddy-l4 (TCP layer4 proxy for Redis).
  caddy:
    image: kalekber/libnovel-caddy:${GIT_TAG:-latest}
    build:
      context: ./caddy
      dockerfile: Dockerfile
+    labels:
+      com.centurylinklabs.watchtower.enable: "true"
    restart: unless-stopped
    depends_on:
      backend:
@@ -372,9 +381,12 @@ services:
      - "80:80"
      - "443:443"
      - "443:443/udp" # HTTP/3 (QUIC)
+      - "6380:6380"   # Redis TCP proxy (TLS) for homelab → Asynq
    environment:
      DOMAIN: "${DOMAIN}"
      CADDY_ACME_EMAIL: "${CADDY_ACME_EMAIL}"
+      # Homelab Redis address — Caddy TCP-proxies inbound :6380 to this.
+      HOMELAB_REDIS_ADDR: "${HOMELAB_REDIS_ADDR:?HOMELAB_REDIS_ADDR required for Redis TCP proxy}"
    env_file:
      - path: ./crowdsec/.crowdsec.env
        required: false
--- a/homelab/docker-compose.yml
+++ b/homelab/docker-compose.yml
@@ -394,7 +394,7 @@ services:
  # Voices match existing IDs: af_bella, af_sky, af_heart, etc.
  # The runner reaches it at http://kokoro-fastapi:8880 via the Docker network.
  kokoro-fastapi:
-    image: ghcr.io/remsky/kokoro-fastapi-gpu:latest
+    image: kokoro-fastapi:latest
    restart: unless-stopped
    deploy:
      resources:
--- a/homelab/otel/grafana/provisioning/alerting/contact-points.yaml
+++ b/homelab/otel/grafana/provisioning/alerting/contact-points.yaml
@@ -0,0 +1,16 @@
+# Grafana alerting provisioning — contact points
+# Sends all alerts to Gotify (self-hosted push notifications).
+apiVersion: 1
+
+contactPoints:
+  - orgId: 1
+    name: Gotify
+    receivers:
+      - uid: gotify-webhook
+        type: webhook
+        settings:
+          url: "http://gotify/message?token=ABZrZgCY-4ivcmt"
+          httpMethod: POST
+          title: "{{ .CommonLabels.alertname }}"
+          message: "{{ range .Alerts }}{{ .Annotations.summary }}\n{{ .Annotations.description }}{{ end }}"
+        disableResolveMessage: false
--- a/homelab/otel/grafana/provisioning/alerting/notification-policy.yaml
+++ b/homelab/otel/grafana/provisioning/alerting/notification-policy.yaml
@@ -0,0 +1,15 @@
+# Grafana alerting provisioning — notification policies
+# Routes all alerts to Gotify by default.
+apiVersion: 1
+
+policies:
+  - orgId: 1
+    receiver: Gotify
+    group_by: ["alertname", "service"]
+    group_wait: 30s
+    group_interval: 5m
+    repeat_interval: 4h
+    routes:
+      - receiver: Gotify
+        matchers:
+          - severity =~ "critical|warning"
--- a/homelab/otel/grafana/provisioning/alerting/rules.yaml
+++ b/homelab/otel/grafana/provisioning/alerting/rules.yaml
@@ -0,0 +1,214 @@
+# Grafana alerting provisioning — alert rules
+# Covers: runner down, high task failure rate, audio error spike, backend error spike.
+apiVersion: 1
+
+groups:
+  - orgId: 1
+    name: LibNovel Runner
+    folder: LibNovel
+    interval: 1m
+    rules:
+
+      - uid: runner-down
+        title: Runner Down
+        condition: C
+        for: 2m
+        annotations:
+          summary: "LibNovel runner is not reachable"
+          description: "The Prometheus scrape of runner:9091 has been failing for >2 minutes. Tasks are not being processed."
+        labels:
+          severity: critical
+          service: runner
+        data:
+          - refId: A
+            datasourceUid: prometheus
+            relativeTimeRange: { from: 300, to: 0 }
+            model:
+              expr: "up{job=\"libnovel-runner\"}"
+              instant: true
+              intervalMs: 1000
+              maxDataPoints: 43200
+          - refId: C
+            datasourceUid: __expr__
+            relativeTimeRange: { from: 300, to: 0 }
+            model:
+              type: classic_conditions
+              conditions:
+                - evaluator: { params: [1], type: lt }
+                  operator: { type: and }
+                  query: { params: [A] }
+                  reducer: { params: [], type: last }
+
+      - uid: runner-high-failure-rate
+        title: Runner High Task Failure Rate
+        condition: C
+        for: 5m
+        annotations:
+          summary: "Runner task failure rate is above 20%"
+          description: "More than 20% of runner tasks have been failing for the last 5 minutes. Check runner logs."
+        labels:
+          severity: warning
+          service: runner
+        data:
+          - refId: A
+            datasourceUid: prometheus
+            relativeTimeRange: { from: 600, to: 0 }
+            model:
+              expr: "rate(libnovel_runner_tasks_failed_total[5m]) / clamp_min(rate(libnovel_runner_tasks_completed_total[5m]) + rate(libnovel_runner_tasks_failed_total[5m]), 0.001)"
+              instant: true
+              intervalMs: 1000
+              maxDataPoints: 43200
+          - refId: C
+            datasourceUid: __expr__
+            relativeTimeRange: { from: 600, to: 0 }
+            model:
+              type: classic_conditions
+              conditions:
+                - evaluator: { params: [0.2], type: gt }
+                  operator: { type: and }
+                  query: { params: [A] }
+                  reducer: { params: [], type: last }
+
+      - uid: runner-tasks-stalled
+        title: Runner Tasks Stalled
+        condition: C
+        for: 10m
+        annotations:
+          summary: "Runner has tasks running for >10 minutes with no completions"
+          description: "tasks_running > 0 but rate(tasks_completed) is 0. Tasks may be stuck or the runner is in a crash loop."
+        labels:
+          severity: warning
+          service: runner
+        data:
+          - refId: Running
+            datasourceUid: prometheus
+            relativeTimeRange: { from: 900, to: 0 }
+            model:
+              expr: "libnovel_runner_tasks_running"
+              instant: true
+              intervalMs: 1000
+              maxDataPoints: 43200
+          - refId: Rate
+            datasourceUid: prometheus
+            relativeTimeRange: { from: 900, to: 0 }
+            model:
+              expr: "rate(libnovel_runner_tasks_completed_total[10m])"
+              instant: true
+              intervalMs: 1000
+              maxDataPoints: 43200
+          - refId: C
+            datasourceUid: __expr__
+            relativeTimeRange: { from: 900, to: 0 }
+            model:
+              type: classic_conditions
+              conditions:
+                - evaluator: { params: [0], type: gt }
+                  operator: { type: and }
+                  query: { params: [Running] }
+                  reducer: { params: [], type: last }
+                - evaluator: { params: [0.001], type: lt }
+                  operator: { type: and }
+                  query: { params: [Rate] }
+                  reducer: { params: [], type: last }
+
+  - orgId: 1
+    name: LibNovel Backend
+    folder: LibNovel
+    interval: 1m
+    rules:
+
+      - uid: backend-high-error-rate
+        title: Backend High Error Rate
+        condition: C
+        for: 5m
+        annotations:
+          summary: "Backend API error rate above 5%"
+          description: "More than 5% of backend HTTP requests are returning 5xx status codes (as seen from UI OTel instrumentation)."
+        labels:
+          severity: warning
+          service: backend
+        data:
+          - refId: A
+            datasourceUid: prometheus
+            relativeTimeRange: { from: 600, to: 0 }
+            model:
+              expr: "sum(rate(http_client_request_duration_seconds_count{job=\"ui\", server_address=\"backend\", http_response_status_code=~\"5..\"}[5m])) / clamp_min(sum(rate(http_client_request_duration_seconds_count{job=\"ui\", server_address=\"backend\"}[5m])), 0.001)"
+              instant: true
+              intervalMs: 1000
+              maxDataPoints: 43200
+          - refId: C
+            datasourceUid: __expr__
+            relativeTimeRange: { from: 600, to: 0 }
+            model:
+              type: classic_conditions
+              conditions:
+                - evaluator: { params: [0.05], type: gt }
+                  operator: { type: and }
+                  query: { params: [A] }
+                  reducer: { params: [], type: last }
+
+      - uid: backend-high-p95-latency
+        title: Backend High p95 Latency
+        condition: C
+        for: 5m
+        annotations:
+          summary: "Backend p95 latency above 2s"
+          description: "95th percentile latency of backend spans has exceeded 2 seconds for >5 minutes."
+        labels:
+          severity: warning
+          service: backend
+        data:
+          - refId: A
+            datasourceUid: prometheus
+            relativeTimeRange: { from: 600, to: 0 }
+            model:
+              expr: "histogram_quantile(0.95, sum(rate(traces_spanmetrics_latency_bucket{service=\"backend\"}[5m])) by (le))"
+              instant: true
+              intervalMs: 1000
+              maxDataPoints: 43200
+          - refId: C
+            datasourceUid: __expr__
+            relativeTimeRange: { from: 600, to: 0 }
+            model:
+              type: classic_conditions
+              conditions:
+                - evaluator: { params: [2], type: gt }
+                  operator: { type: and }
+                  query: { params: [A] }
+                  reducer: { params: [], type: last }
+
+  - orgId: 1
+    name: LibNovel OTel Pipeline
+    folder: LibNovel
+    interval: 2m
+    rules:
+
+      - uid: otel-collector-down
+        title: OTel Collector Down
+        condition: C
+        for: 3m
+        annotations:
+          summary: "OTel collector is not reachable"
+          description: "Prometheus cannot scrape otel-collector:8888. Traces and logs may be dropping."
+        labels:
+          severity: warning
+          service: otel-collector
+        data:
+          - refId: A
+            datasourceUid: prometheus
+            relativeTimeRange: { from: 600, to: 0 }
+            model:
+              expr: "up{job=\"otel-collector\"}"
+              instant: true
+              intervalMs: 1000
+              maxDataPoints: 43200
+          - refId: C
+            datasourceUid: __expr__
+            relativeTimeRange: { from: 600, to: 0 }
+            model:
+              type: classic_conditions
+              conditions:
+                - evaluator: { params: [1], type: lt }
+                  operator: { type: and }
+                  query: { params: [A] }
+                  reducer: { params: [], type: last }
--- a/homelab/otel/grafana/provisioning/dashboards/backend.json
+++ b/homelab/otel/grafana/provisioning/dashboards/backend.json
@@ -0,0 +1,338 @@
+{
+  "uid": "libnovel-backend",
+  "title": "Backend API",
+  "description": "Request rate, error rate, and latency for the LibNovel backend. Powered by Tempo span metrics and UI OTel instrumentation.",
+  "tags": ["libnovel", "backend", "api"],
+  "timezone": "browser",
+  "refresh": "30s",
+  "time": { "from": "now-3h", "to": "now" },
+  "schemaVersion": 39,
+  "panels": [
+    {
+      "id": 1,
+      "type": "stat",
+      "title": "Request Rate (RPS)",
+      "gridPos": { "x": 0, "y": 0, "w": 4, "h": 4 },
+      "options": {
+        "reduceOptions": { "calcs": ["lastNotNull"] },
+        "colorMode": "value",
+        "graphMode": "area",
+        "textMode": "auto"
+      },
+      "fieldConfig": {
+        "defaults": {
+          "unit": "reqps",
+          "color": { "mode": "thresholds" },
+          "thresholds": { "mode": "absolute", "steps": [{ "color": "green", "value": null }] }
+        }
+      },
+      "targets": [
+        {
+          "datasource": { "type": "prometheus", "uid": "prometheus" },
+          "expr": "sum(rate(traces_spanmetrics_calls_total{service=\"backend\"}[5m]))",
+          "legendFormat": "rps",
+          "instant": true
+        }
+      ]
+    },
+    {
+      "id": 2,
+      "type": "stat",
+      "title": "Error Rate",
+      "gridPos": { "x": 4, "y": 0, "w": 4, "h": 4 },
+      "options": {
+        "reduceOptions": { "calcs": ["lastNotNull"] },
+        "colorMode": "background",
+        "graphMode": "none"
+      },
+      "fieldConfig": {
+        "defaults": {
+          "unit": "percentunit",
+          "thresholds": {
+            "mode": "absolute",
+            "steps": [
+              { "color": "green", "value": null },
+              { "color": "yellow", "value": 0.01 },
+              { "color": "red", "value": 0.05 }
+            ]
+          }
+        }
+      },
+      "targets": [
+        {
+          "datasource": { "type": "prometheus", "uid": "prometheus" },
+          "expr": "sum(rate(traces_spanmetrics_calls_total{service=\"backend\", status_code=\"STATUS_CODE_ERROR\"}[5m])) / clamp_min(sum(rate(traces_spanmetrics_calls_total{service=\"backend\"}[5m])), 0.001)",
+          "legendFormat": "error rate",
+          "instant": true
+        }
+      ]
+    },
+    {
+      "id": 3,
+      "type": "stat",
+      "title": "p50 Latency",
+      "gridPos": { "x": 8, "y": 0, "w": 4, "h": 4 },
+      "options": { "reduceOptions": { "calcs": ["lastNotNull"] }, "colorMode": "value", "graphMode": "area" },
+      "fieldConfig": {
+        "defaults": {
+          "unit": "s",
+          "thresholds": {
+            "mode": "absolute",
+            "steps": [
+              { "color": "green", "value": null },
+              { "color": "yellow", "value": 0.2 },
+              { "color": "red", "value": 1 }
+            ]
+          }
+        }
+      },
+      "targets": [
+        {
+          "datasource": { "type": "prometheus", "uid": "prometheus" },
+          "expr": "histogram_quantile(0.50, sum(rate(traces_spanmetrics_latency_bucket{service=\"backend\"}[5m])) by (le))",
+          "legendFormat": "p50",
+          "instant": true
+        }
+      ]
+    },
+    {
+      "id": 4,
+      "type": "stat",
+      "title": "p95 Latency",
+      "gridPos": { "x": 12, "y": 0, "w": 4, "h": 4 },
+      "options": { "reduceOptions": { "calcs": ["lastNotNull"] }, "colorMode": "value", "graphMode": "area" },
+      "fieldConfig": {
+        "defaults": {
+          "unit": "s",
+          "thresholds": {
+            "mode": "absolute",
+            "steps": [
+              { "color": "green", "value": null },
+              { "color": "yellow", "value": 0.5 },
+              { "color": "red", "value": 2 }
+            ]
+          }
+        }
+      },
+      "targets": [
+        {
+          "datasource": { "type": "prometheus", "uid": "prometheus" },
+          "expr": "histogram_quantile(0.95, sum(rate(traces_spanmetrics_latency_bucket{service=\"backend\"}[5m])) by (le))",
+          "legendFormat": "p95",
+          "instant": true
+        }
+      ]
+    },
+    {
+      "id": 5,
+      "type": "stat",
+      "title": "p99 Latency",
+      "gridPos": { "x": 16, "y": 0, "w": 4, "h": 4 },
+      "options": { "reduceOptions": { "calcs": ["lastNotNull"] }, "colorMode": "value", "graphMode": "area" },
+      "fieldConfig": {
+        "defaults": {
+          "unit": "s",
+          "thresholds": {
+            "mode": "absolute",
+            "steps": [
+              { "color": "green", "value": null },
+              { "color": "yellow", "value": 1 },
+              { "color": "red", "value": 5 }
+            ]
+          }
+        }
+      },
+      "targets": [
+        {
+          "datasource": { "type": "prometheus", "uid": "prometheus" },
+          "expr": "histogram_quantile(0.99, sum(rate(traces_spanmetrics_latency_bucket{service=\"backend\"}[5m])) by (le))",
+          "legendFormat": "p99",
+          "instant": true
+        }
+      ]
+    },
+    {
+      "id": 6,
+      "type": "stat",
+      "title": "5xx Errors / min",
+      "gridPos": { "x": 20, "y": 0, "w": 4, "h": 4 },
+      "options": { "reduceOptions": { "calcs": ["lastNotNull"] }, "colorMode": "background", "graphMode": "none" },
+      "fieldConfig": {
+        "defaults": {
+          "unit": "short",
+          "thresholds": {
+            "mode": "absolute",
+            "steps": [
+              { "color": "green", "value": null },
+              { "color": "yellow", "value": 1 },
+              { "color": "red", "value": 5 }
+            ]
+          }
+        }
+      },
+      "targets": [
+        {
+          "datasource": { "type": "prometheus", "uid": "prometheus" },
+          "expr": "sum(rate(http_client_request_duration_seconds_count{job=\"ui\", server_address=\"backend\", http_response_status_code=~\"5..\"}[5m])) * 60",
+          "legendFormat": "5xx/min",
+          "instant": true
+        }
+      ]
+    },
+    {
+      "id": 10,
+      "type": "timeseries",
+      "title": "Request Rate by Status",
+      "gridPos": { "x": 0, "y": 4, "w": 12, "h": 8 },
+      "options": {
+        "tooltip": { "mode": "multi" },
+        "legend": { "displayMode": "list", "placement": "bottom" }
+      },
+      "fieldConfig": {
+        "defaults": { "unit": "reqps", "custom": { "lineWidth": 2, "fillOpacity": 10 } },
+        "overrides": [
+          { "matcher": { "id": "byFrameRefID", "options": "errors" }, "properties": [{ "id": "color", "value": { "fixedColor": "red", "mode": "fixed" } }] }
+        ]
+      },
+      "targets": [
+        {
+          "refId": "success",
+          "datasource": { "type": "prometheus", "uid": "prometheus" },
+          "expr": "sum(rate(http_client_request_duration_seconds_count{job=\"ui\", server_address=\"backend\", http_response_status_code=~\"2..\"}[5m]))",
+          "legendFormat": "2xx"
+        },
+        {
+          "refId": "notfound",
+          "datasource": { "type": "prometheus", "uid": "prometheus" },
+          "expr": "sum(rate(http_client_request_duration_seconds_count{job=\"ui\", server_address=\"backend\", http_response_status_code=~\"4..\"}[5m]))",
+          "legendFormat": "4xx"
+        },
+        {
+          "refId": "errors",
+          "datasource": { "type": "prometheus", "uid": "prometheus" },
+          "expr": "sum(rate(http_client_request_duration_seconds_count{job=\"ui\", server_address=\"backend\", http_response_status_code=~\"5..\"}[5m]))",
+          "legendFormat": "5xx"
+        }
+      ]
+    },
+    {
+      "id": 11,
+      "type": "timeseries",
+      "title": "Latency Percentiles (backend spans)",
+      "gridPos": { "x": 12, "y": 4, "w": 12, "h": 8 },
+      "options": {
+        "tooltip": { "mode": "multi" },
+        "legend": { "displayMode": "list", "placement": "bottom" }
+      },
+      "fieldConfig": {
+        "defaults": { "unit": "s", "custom": { "lineWidth": 2, "fillOpacity": 10 } }
+      },
+      "targets": [
+        {
+          "datasource": { "type": "prometheus", "uid": "prometheus" },
+          "expr": "histogram_quantile(0.50, sum(rate(traces_spanmetrics_latency_bucket{service=\"backend\"}[5m])) by (le))",
+          "legendFormat": "p50"
+        },
+        {
+          "datasource": { "type": "prometheus", "uid": "prometheus" },
+          "expr": "histogram_quantile(0.95, sum(rate(traces_spanmetrics_latency_bucket{service=\"backend\"}[5m])) by (le))",
+          "legendFormat": "p95"
+        },
+        {
+          "datasource": { "type": "prometheus", "uid": "prometheus" },
+          "expr": "histogram_quantile(0.99, sum(rate(traces_spanmetrics_latency_bucket{service=\"backend\"}[5m])) by (le))",
+          "legendFormat": "p99"
+        }
+      ]
+    },
+    {
+      "id": 12,
+      "type": "timeseries",
+      "title": "Requests / min by HTTP method (UI → Backend)",
+      "gridPos": { "x": 0, "y": 12, "w": 12, "h": 8 },
+      "options": {
+        "tooltip": { "mode": "multi" },
+        "legend": { "displayMode": "list", "placement": "bottom" }
+      },
+      "fieldConfig": {
+        "defaults": { "unit": "short", "custom": { "lineWidth": 2, "fillOpacity": 5 } }
+      },
+      "targets": [
+        {
+          "datasource": { "type": "prometheus", "uid": "prometheus" },
+          "expr": "sum(rate(http_client_request_duration_seconds_count{job=\"ui\", server_address=\"backend\"}[5m])) by (http_request_method) * 60",
+          "legendFormat": "{{http_request_method}}"
+        }
+      ]
+    },
+    {
+      "id": 13,
+      "type": "timeseries",
+      "title": "Requests / min — UI → PocketBase",
+      "gridPos": { "x": 12, "y": 12, "w": 12, "h": 8 },
+      "description": "Traffic from SvelteKit server to PocketBase (auth, collections, etc.).",
+      "options": {
+        "tooltip": { "mode": "multi" },
+        "legend": { "displayMode": "list", "placement": "bottom" }
+      },
+      "fieldConfig": {
+        "defaults": { "unit": "short", "custom": { "lineWidth": 2, "fillOpacity": 5 } }
+      },
+      "targets": [
+        {
+          "datasource": { "type": "prometheus", "uid": "prometheus" },
+          "expr": "sum(rate(http_client_request_duration_seconds_count{job=\"ui\", server_address=\"pocketbase\"}[5m])) by (http_request_method, http_response_status_code) * 60",
+          "legendFormat": "{{http_request_method}} {{http_response_status_code}}"
+        }
+      ]
+    },
+    {
+      "id": 14,
+      "type": "timeseries",
+      "title": "UI → Backend Latency (p50 / p95)",
+      "gridPos": { "x": 0, "y": 20, "w": 12, "h": 8 },
+      "description": "HTTP client latency as seen from the SvelteKit SSR layer calling backend.",
+      "options": {
+        "tooltip": { "mode": "multi" },
+        "legend": { "displayMode": "list", "placement": "bottom" }
+      },
+      "fieldConfig": {
+        "defaults": { "unit": "s", "custom": { "lineWidth": 2, "fillOpacity": 5 } }
+      },
+      "targets": [
+        {
+          "datasource": { "type": "prometheus", "uid": "prometheus" },
+          "expr": "histogram_quantile(0.50, sum(rate(http_client_request_duration_seconds_bucket{job=\"ui\", server_address=\"backend\"}[5m])) by (le))",
+          "legendFormat": "p50"
+        },
+        {
+          "datasource": { "type": "prometheus", "uid": "prometheus" },
+          "expr": "histogram_quantile(0.95, sum(rate(http_client_request_duration_seconds_bucket{job=\"ui\", server_address=\"backend\"}[5m])) by (le))",
+          "legendFormat": "p95"
+        }
+      ]
+    },
+    {
+      "id": 20,
+      "type": "logs",
+      "title": "Backend Errors",
+      "gridPos": { "x": 0, "y": 28, "w": 24, "h": 10 },
+      "options": {
+        "showTime": true,
+        "showLabels": false,
+        "wrapLogMessage": true,
+        "prettifyLogMessage": true,
+        "enableLogDetails": true,
+        "sortOrder": "Descending",
+        "dedupStrategy": "none"
+      },
+      "targets": [
+        {
+          "datasource": { "type": "loki", "uid": "loki" },
+          "expr": "{service_name=\"backend\"} | json | level =~ `(WARN|ERROR|error|warn)`",
+          "legendFormat": ""
+        }
+      ]
+    }
+  ]
+}
--- a/homelab/otel/grafana/provisioning/dashboards/catalogue.json
+++ b/homelab/otel/grafana/provisioning/dashboards/catalogue.json
@@ -0,0 +1,275 @@
+{
+  "uid": "libnovel-catalogue",
+  "title": "Catalogue & Content Progress",
+  "description": "Scraping progress, audio generation coverage, and catalogue health derived from runner structured logs.",
+  "tags": ["libnovel", "catalogue", "content"],
+  "timezone": "browser",
+  "refresh": "1m",
+  "time": { "from": "now-24h", "to": "now" },
+  "schemaVersion": 39,
+  "panels": [
+    {
+      "id": 1,
+      "type": "stat",
+      "title": "Books Scraped (last 24h)",
+      "description": "Count of unique book slugs appearing in successful scrape task completions.",
+      "gridPos": { "x": 0, "y": 0, "w": 4, "h": 4 },
+      "options": { "reduceOptions": { "calcs": ["sum"] }, "colorMode": "value", "graphMode": "none" },
+      "fieldConfig": {
+        "defaults": {
+          "color": { "fixedColor": "blue", "mode": "fixed" },
+          "thresholds": { "mode": "absolute", "steps": [] }
+        }
+      },
+      "targets": [
+        {
+          "datasource": { "type": "loki", "uid": "loki" },
+          "expr": "sum_over_time({service_name=\"runner\"} | json | msg=`scrape task done` [24h])",
+          "legendFormat": "books scraped"
+        }
+      ]
+    },
+    {
+      "id": 2,
+      "type": "stat",
+      "title": "Chapters Scraped (last 24h)",
+      "gridPos": { "x": 4, "y": 0, "w": 4, "h": 4 },
+      "options": { "reduceOptions": { "calcs": ["sum"] }, "colorMode": "value", "graphMode": "none" },
+      "fieldConfig": {
+        "defaults": {
+          "color": { "fixedColor": "blue", "mode": "fixed" },
+          "thresholds": { "mode": "absolute", "steps": [] }
+        }
+      },
+      "targets": [
+        {
+          "datasource": { "type": "loki", "uid": "loki" },
+          "expr": "sum_over_time({service_name=\"runner\"} | json | unwrap scraped [24h])",
+          "legendFormat": "chapters scraped"
+        }
+      ]
+    },
+    {
+      "id": 3,
+      "type": "stat",
+      "title": "Audio Jobs Completed (last 24h)",
+      "gridPos": { "x": 8, "y": 0, "w": 4, "h": 4 },
+      "options": { "reduceOptions": { "calcs": ["sum"] }, "colorMode": "value", "graphMode": "none" },
+      "fieldConfig": {
+        "defaults": {
+          "color": { "fixedColor": "green", "mode": "fixed" },
+          "thresholds": { "mode": "absolute", "steps": [] }
+        }
+      },
+      "targets": [
+        {
+          "datasource": { "type": "loki", "uid": "loki" },
+          "expr": "sum_over_time({service_name=\"runner\"} | json | msg=`audio task done` [24h])",
+          "legendFormat": "audio done"
+        }
+      ]
+    },
+    {
+      "id": 4,
+      "type": "stat",
+      "title": "Audio Jobs Failed (last 24h)",
+      "gridPos": { "x": 12, "y": 0, "w": 4, "h": 4 },
+      "options": { "reduceOptions": { "calcs": ["sum"] }, "colorMode": "background", "graphMode": "none" },
+      "fieldConfig": {
+        "defaults": {
+          "thresholds": {
+            "mode": "absolute",
+            "steps": [
+              { "color": "green", "value": null },
+              { "color": "yellow", "value": 1 },
+              { "color": "red", "value": 5 }
+            ]
+          }
+        }
+      },
+      "targets": [
+        {
+          "datasource": { "type": "loki", "uid": "loki" },
+          "expr": "sum_over_time({service_name=\"runner\"} | json | msg=`audio task failed` [24h])",
+          "legendFormat": "audio failed"
+        }
+      ]
+    },
+    {
+      "id": 5,
+      "type": "stat",
+      "title": "Scrape Errors (last 24h)",
+      "gridPos": { "x": 16, "y": 0, "w": 4, "h": 4 },
+      "options": { "reduceOptions": { "calcs": ["sum"] }, "colorMode": "background", "graphMode": "none" },
+      "fieldConfig": {
+        "defaults": {
+          "thresholds": {
+            "mode": "absolute",
+            "steps": [
+              { "color": "green", "value": null },
+              { "color": "yellow", "value": 1 },
+              { "color": "red", "value": 10 }
+            ]
+          }
+        }
+      },
+      "targets": [
+        {
+          "datasource": { "type": "loki", "uid": "loki" },
+          "expr": "sum_over_time({service_name=\"runner\"} | json | msg=`scrape task failed` [24h])",
+          "legendFormat": "scrape errors"
+        }
+      ]
+    },
+    {
+      "id": 6,
+      "type": "stat",
+      "title": "Catalogue Refresh — Books Indexed",
+      "description": "Total books indexed in the last catalogue refresh cycle (from the ok field in the summary log).",
+      "gridPos": { "x": 20, "y": 0, "w": 4, "h": 4 },
+      "options": { "reduceOptions": { "calcs": ["lastNotNull"] }, "colorMode": "value", "graphMode": "none" },
+      "fieldConfig": {
+        "defaults": {
+          "color": { "fixedColor": "purple", "mode": "fixed" },
+          "thresholds": { "mode": "absolute", "steps": [] }
+        }
+      },
+      "targets": [
+        {
+          "datasource": { "type": "loki", "uid": "loki" },
+          "expr": "last_over_time({service_name=\"runner\"} | json | op=`catalogue_refresh` | msg=`catalogue refresh done` | unwrap ok [7d])",
+          "legendFormat": "indexed"
+        }
+      ]
+    },
+    {
+      "id": 10,
+      "type": "timeseries",
+      "title": "Audio Generation Rate (tasks/min)",
+      "gridPos": { "x": 0, "y": 4, "w": 12, "h": 8 },
+      "description": "Rate of audio task completions and failures over time.",
+      "options": {
+        "tooltip": { "mode": "multi" },
+        "legend": { "displayMode": "list", "placement": "bottom" }
+      },
+      "fieldConfig": {
+        "defaults": { "unit": "short", "custom": { "lineWidth": 2, "fillOpacity": 10 } },
+        "overrides": [
+          { "matcher": { "id": "byName", "options": "failed" }, "properties": [{ "id": "color", "value": { "fixedColor": "red", "mode": "fixed" } }] },
+          { "matcher": { "id": "byName", "options": "completed" }, "properties": [{ "id": "color", "value": { "fixedColor": "green", "mode": "fixed" } }] }
+        ]
+      },
+      "targets": [
+        {
+          "datasource": { "type": "prometheus", "uid": "prometheus" },
+          "expr": "sum(rate(traces_spanmetrics_calls_total{service=\"runner\", span_name=\"runner.audio_task\", status_code!=\"STATUS_CODE_ERROR\"}[5m])) * 60",
+          "legendFormat": "completed"
+        },
+        {
+          "datasource": { "type": "prometheus", "uid": "prometheus" },
+          "expr": "sum(rate(traces_spanmetrics_calls_total{service=\"runner\", span_name=\"runner.audio_task\", status_code=\"STATUS_CODE_ERROR\"}[5m])) * 60",
+          "legendFormat": "failed"
+        }
+      ]
+    },
+    {
+      "id": 11,
+      "type": "timeseries",
+      "title": "Scraping Rate (tasks/min)",
+      "gridPos": { "x": 12, "y": 4, "w": 12, "h": 8 },
+      "description": "Rate of scrape task completions and failures over time.",
+      "options": {
+        "tooltip": { "mode": "multi" },
+        "legend": { "displayMode": "list", "placement": "bottom" }
+      },
+      "fieldConfig": {
+        "defaults": { "unit": "short", "custom": { "lineWidth": 2, "fillOpacity": 10 } },
+        "overrides": [
+          { "matcher": { "id": "byName", "options": "failed" }, "properties": [{ "id": "color", "value": { "fixedColor": "red", "mode": "fixed" } }] },
+          { "matcher": { "id": "byName", "options": "completed" }, "properties": [{ "id": "color", "value": { "fixedColor": "blue", "mode": "fixed" } }] }
+        ]
+      },
+      "targets": [
+        {
+          "datasource": { "type": "prometheus", "uid": "prometheus" },
+          "expr": "sum(rate(traces_spanmetrics_calls_total{service=\"runner\", span_name=\"runner.scrape_task\", status_code!=\"STATUS_CODE_ERROR\"}[5m])) * 60",
+          "legendFormat": "completed"
+        },
+        {
+          "datasource": { "type": "prometheus", "uid": "prometheus" },
+          "expr": "sum(rate(traces_spanmetrics_calls_total{service=\"runner\", span_name=\"runner.scrape_task\", status_code=\"STATUS_CODE_ERROR\"}[5m])) * 60",
+          "legendFormat": "failed"
+        }
+      ]
+    },
+    {
+      "id": 20,
+      "type": "logs",
+      "title": "Scrape Task Events",
+      "description": "One log line per completed or failed scrape task. Fields: task_id, kind, url, scraped, skipped, errors.",
+      "gridPos": { "x": 0, "y": 12, "w": 24, "h": 10 },
+      "options": {
+        "showTime": true,
+        "showLabels": false,
+        "wrapLogMessage": false,
+        "prettifyLogMessage": true,
+        "enableLogDetails": true,
+        "sortOrder": "Descending",
+        "dedupStrategy": "none"
+      },
+      "targets": [
+        {
+          "datasource": { "type": "loki", "uid": "loki" },
+          "expr": "{service_name=\"runner\"} | json | msg =~ `scrape task (done|failed|starting)`",
+          "legendFormat": ""
+        }
+      ]
+    },
+    {
+      "id": 21,
+      "type": "logs",
+      "title": "Audio Task Events",
+      "description": "One log line per completed or failed audio task. Fields: task_id, slug, chapter, voice, key (on success), reason (on failure).",
+      "gridPos": { "x": 0, "y": 22, "w": 24, "h": 10 },
+      "options": {
+        "showTime": true,
+        "showLabels": false,
+        "wrapLogMessage": false,
+        "prettifyLogMessage": true,
+        "enableLogDetails": true,
+        "sortOrder": "Descending",
+        "dedupStrategy": "none"
+      },
+      "targets": [
+        {
+          "datasource": { "type": "loki", "uid": "loki" },
+          "expr": "{service_name=\"runner\"} | json | msg =~ `audio task (done|failed|starting)`",
+          "legendFormat": ""
+        }
+      ]
+    },
+    {
+      "id": 22,
+      "type": "logs",
+      "title": "Catalogue Refresh Progress",
+      "description": "Progress logs from the background catalogue refresh (every 24h). Fields: op=catalogue_refresh, scraped, ok, skipped, errors.",
+      "gridPos": { "x": 0, "y": 32, "w": 24, "h": 8 },
+      "options": {
+        "showTime": true,
+        "showLabels": false,
+        "wrapLogMessage": false,
+        "prettifyLogMessage": true,
+        "enableLogDetails": true,
+        "sortOrder": "Descending",
+        "dedupStrategy": "none"
+      },
+      "targets": [
+        {
+          "datasource": { "type": "loki", "uid": "loki" },
+          "expr": "{service_name=\"runner\"} | json | op=`catalogue_refresh`",
+          "legendFormat": ""
+        }
+      ]
+    }
+  ]
+}
--- a/homelab/otel/grafana/provisioning/dashboards/runner.json
+++ b/homelab/otel/grafana/provisioning/dashboards/runner.json
@@ -0,0 +1,377 @@
+{
+  "uid": "libnovel-runner",
+  "title": "Runner Operations",
+  "description": "Task queue health, throughput, TTS routing, and live logs for the homelab runner.",
+  "tags": ["libnovel", "runner"],
+  "timezone": "browser",
+  "refresh": "30s",
+  "time": { "from": "now-3h", "to": "now" },
+  "schemaVersion": 39,
+  "panels": [
+    {
+      "id": 1,
+      "type": "stat",
+      "title": "Tasks Running",
+      "gridPos": { "x": 0, "y": 0, "w": 4, "h": 4 },
+      "options": {
+        "reduceOptions": { "calcs": ["lastNotNull"] },
+        "colorMode": "background",
+        "graphMode": "none",
+        "textMode": "auto"
+      },
+      "fieldConfig": {
+        "defaults": {
+          "thresholds": {
+            "mode": "absolute",
+            "steps": [
+              { "color": "green", "value": null },
+              { "color": "yellow", "value": 1 },
+              { "color": "red", "value": 3 }
+            ]
+          },
+          "mappings": []
+        }
+      },
+      "targets": [
+        {
+          "datasource": { "type": "prometheus", "uid": "prometheus" },
+          "expr": "libnovel_runner_tasks_running",
+          "legendFormat": "running",
+          "instant": true
+        }
+      ]
+    },
+    {
+      "id": 2,
+      "type": "stat",
+      "title": "Tasks Completed (total)",
+      "gridPos": { "x": 4, "y": 0, "w": 4, "h": 4 },
+      "options": {
+        "reduceOptions": { "calcs": ["lastNotNull"] },
+        "colorMode": "background",
+        "graphMode": "area",
+        "textMode": "auto"
+      },
+      "fieldConfig": {
+        "defaults": {
+          "color": { "fixedColor": "green", "mode": "fixed" },
+          "thresholds": { "mode": "absolute", "steps": [] }
+        }
+      },
+      "targets": [
+        {
+          "datasource": { "type": "prometheus", "uid": "prometheus" },
+          "expr": "libnovel_runner_tasks_completed_total",
+          "legendFormat": "completed",
+          "instant": true
+        }
+      ]
+    },
+    {
+      "id": 3,
+      "type": "stat",
+      "title": "Tasks Failed (total)",
+      "gridPos": { "x": 8, "y": 0, "w": 4, "h": 4 },
+      "options": {
+        "reduceOptions": { "calcs": ["lastNotNull"] },
+        "colorMode": "background",
+        "graphMode": "none",
+        "textMode": "auto"
+      },
+      "fieldConfig": {
+        "defaults": {
+          "thresholds": {
+            "mode": "absolute",
+            "steps": [
+              { "color": "green", "value": null },
+              { "color": "yellow", "value": 1 },
+              { "color": "red", "value": 5 }
+            ]
+          }
+        }
+      },
+      "targets": [
+        {
+          "datasource": { "type": "prometheus", "uid": "prometheus" },
+          "expr": "libnovel_runner_tasks_failed_total",
+          "legendFormat": "failed",
+          "instant": true
+        }
+      ]
+    },
+    {
+      "id": 4,
+      "type": "stat",
+      "title": "Runner Uptime",
+      "gridPos": { "x": 12, "y": 0, "w": 4, "h": 4 },
+      "options": {
+        "reduceOptions": { "calcs": ["lastNotNull"] },
+        "colorMode": "value",
+        "graphMode": "none",
+        "textMode": "auto"
+      },
+      "fieldConfig": {
+        "defaults": {
+          "unit": "s",
+          "thresholds": {
+            "mode": "absolute",
+            "steps": [
+              { "color": "red", "value": null },
+              { "color": "yellow", "value": 60 },
+              { "color": "green", "value": 300 }
+            ]
+          }
+        }
+      },
+      "targets": [
+        {
+          "datasource": { "type": "prometheus", "uid": "prometheus" },
+          "expr": "libnovel_runner_uptime_seconds",
+          "legendFormat": "uptime",
+          "instant": true
+        }
+      ]
+    },
+    {
+      "id": 5,
+      "type": "stat",
+      "title": "Task Failure Rate",
+      "gridPos": { "x": 16, "y": 0, "w": 4, "h": 4 },
+      "options": {
+        "reduceOptions": { "calcs": ["lastNotNull"] },
+        "colorMode": "background",
+        "graphMode": "none",
+        "textMode": "auto"
+      },
+      "fieldConfig": {
+        "defaults": {
+          "unit": "percentunit",
+          "thresholds": {
+            "mode": "absolute",
+            "steps": [
+              { "color": "green", "value": null },
+              { "color": "yellow", "value": 0.05 },
+              { "color": "red", "value": 0.2 }
+            ]
+          }
+        }
+      },
+      "targets": [
+        {
+          "datasource": { "type": "prometheus", "uid": "prometheus" },
+          "expr": "libnovel_runner_tasks_failed_total / clamp_min(libnovel_runner_tasks_completed_total + libnovel_runner_tasks_failed_total, 1)",
+          "legendFormat": "failure rate",
+          "instant": true
+        }
+      ]
+    },
+    {
+      "id": 6,
+      "type": "stat",
+      "title": "Runner Alive",
+      "gridPos": { "x": 20, "y": 0, "w": 4, "h": 4 },
+      "options": {
+        "reduceOptions": { "calcs": ["lastNotNull"] },
+        "colorMode": "background",
+        "graphMode": "none",
+        "textMode": "auto"
+      },
+      "fieldConfig": {
+        "defaults": {
+          "mappings": [
+            { "type": "value", "options": { "1": { "text": "UP", "color": "green" }, "0": { "text": "DOWN", "color": "red" } } }
+          ],
+          "thresholds": { "mode": "absolute", "steps": [] }
+        }
+      },
+      "targets": [
+        {
+          "datasource": { "type": "prometheus", "uid": "prometheus" },
+          "expr": "up{job=\"libnovel-runner\"}",
+          "legendFormat": "runner",
+          "instant": true
+        }
+      ]
+    },
+    {
+      "id": 10,
+      "type": "timeseries",
+      "title": "Task Throughput (per minute)",
+      "gridPos": { "x": 0, "y": 4, "w": 12, "h": 8 },
+      "options": {
+        "tooltip": { "mode": "multi" },
+        "legend": { "displayMode": "list", "placement": "bottom" }
+      },
+      "fieldConfig": {
+        "defaults": {
+          "unit": "ops",
+          "custom": { "lineWidth": 2, "fillOpacity": 10 }
+        },
+        "overrides": [
+          { "matcher": { "id": "byName", "options": "failed" }, "properties": [{ "id": "color", "value": { "fixedColor": "red", "mode": "fixed" } }] },
+          { "matcher": { "id": "byName", "options": "completed" }, "properties": [{ "id": "color", "value": { "fixedColor": "green", "mode": "fixed" } }] }
+        ]
+      },
+      "targets": [
+        {
+          "datasource": { "type": "prometheus", "uid": "prometheus" },
+          "expr": "rate(libnovel_runner_tasks_completed_total[5m]) * 60",
+          "legendFormat": "completed"
+        },
+        {
+          "datasource": { "type": "prometheus", "uid": "prometheus" },
+          "expr": "rate(libnovel_runner_tasks_failed_total[5m]) * 60",
+          "legendFormat": "failed"
+        },
+        {
+          "datasource": { "type": "prometheus", "uid": "prometheus" },
+          "expr": "libnovel_runner_tasks_running",
+          "legendFormat": "running"
+        }
+      ]
+    },
+    {
+      "id": 11,
+      "type": "timeseries",
+      "title": "Audio Task Span Latency (p50 / p95 / p99)",
+      "gridPos": { "x": 12, "y": 4, "w": 12, "h": 8 },
+      "description": "End-to-end latency of runner.audio_task spans from Tempo span metrics.",
+      "options": {
+        "tooltip": { "mode": "multi" },
+        "legend": { "displayMode": "list", "placement": "bottom" }
+      },
+      "fieldConfig": {
+        "defaults": {
+          "unit": "s",
+          "custom": { "lineWidth": 2, "fillOpacity": 10 }
+        }
+      },
+      "targets": [
+        {
+          "datasource": { "type": "prometheus", "uid": "prometheus" },
+          "expr": "histogram_quantile(0.50, sum(rate(traces_spanmetrics_latency_bucket{service=\"runner\", span_name=\"runner.audio_task\"}[5m])) by (le))",
+          "legendFormat": "p50"
+        },
+        {
+          "datasource": { "type": "prometheus", "uid": "prometheus" },
+          "expr": "histogram_quantile(0.95, sum(rate(traces_spanmetrics_latency_bucket{service=\"runner\", span_name=\"runner.audio_task\"}[5m])) by (le))",
+          "legendFormat": "p95"
+        },
+        {
+          "datasource": { "type": "prometheus", "uid": "prometheus" },
+          "expr": "histogram_quantile(0.99, sum(rate(traces_spanmetrics_latency_bucket{service=\"runner\", span_name=\"runner.audio_task\"}[5m])) by (le))",
+          "legendFormat": "p99"
+        }
+      ]
+    },
+    {
+      "id": 20,
+      "type": "timeseries",
+      "title": "Scrape Task Span Latency (p50 / p95 / p99)",
+      "gridPos": { "x": 0, "y": 12, "w": 12, "h": 8 },
+      "description": "End-to-end latency of runner.scrape_task spans from Tempo span metrics.",
+      "options": {
+        "tooltip": { "mode": "multi" },
+        "legend": { "displayMode": "list", "placement": "bottom" }
+      },
+      "fieldConfig": {
+        "defaults": {
+          "unit": "s",
+          "custom": { "lineWidth": 2, "fillOpacity": 10 }
+        }
+      },
+      "targets": [
+        {
+          "datasource": { "type": "prometheus", "uid": "prometheus" },
+          "expr": "histogram_quantile(0.50, sum(rate(traces_spanmetrics_latency_bucket{service=\"runner\", span_name=\"runner.scrape_task\"}[5m])) by (le))",
+          "legendFormat": "p50"
+        },
+        {
+          "datasource": { "type": "prometheus", "uid": "prometheus" },
+          "expr": "histogram_quantile(0.95, sum(rate(traces_spanmetrics_latency_bucket{service=\"runner\", span_name=\"runner.scrape_task\"}[5m])) by (le))",
+          "legendFormat": "p95"
+        },
+        {
+          "datasource": { "type": "prometheus", "uid": "prometheus" },
+          "expr": "histogram_quantile(0.99, sum(rate(traces_spanmetrics_latency_bucket{service=\"runner\", span_name=\"runner.scrape_task\"}[5m])) by (le))",
+          "legendFormat": "p99"
+        }
+      ]
+    },
+    {
+      "id": 21,
+      "type": "timeseries",
+      "title": "Audio vs Scrape Task Rate",
+      "gridPos": { "x": 12, "y": 12, "w": 12, "h": 8 },
+      "description": "Relative throughput of audio generation vs book scraping.",
+      "options": {
+        "tooltip": { "mode": "multi" },
+        "legend": { "displayMode": "list", "placement": "bottom" }
+      },
+      "fieldConfig": {
+        "defaults": {
+          "unit": "ops",
+          "custom": { "lineWidth": 2, "fillOpacity": 10 }
+        }
+      },
+      "targets": [
+        {
+          "datasource": { "type": "prometheus", "uid": "prometheus" },
+          "expr": "sum(rate(traces_spanmetrics_calls_total{service=\"runner\", span_name=\"runner.audio_task\"}[5m]))",
+          "legendFormat": "audio tasks/s"
+        },
+        {
+          "datasource": { "type": "prometheus", "uid": "prometheus" },
+          "expr": "sum(rate(traces_spanmetrics_calls_total{service=\"runner\", span_name=\"runner.scrape_task\"}[5m]))",
+          "legendFormat": "scrape tasks/s"
+        }
+      ]
+    },
+    {
+      "id": 30,
+      "type": "logs",
+      "title": "Runner Logs (errors & warnings)",
+      "gridPos": { "x": 0, "y": 20, "w": 24, "h": 10 },
+      "options": {
+        "showTime": true,
+        "showLabels": false,
+        "showCommonLabels": false,
+        "wrapLogMessage": true,
+        "prettifyLogMessage": true,
+        "enableLogDetails": true,
+        "sortOrder": "Descending",
+        "dedupStrategy": "none"
+      },
+      "targets": [
+        {
+          "datasource": { "type": "loki", "uid": "loki" },
+          "expr": "{service_name=\"runner\"} | json | level =~ `(WARN|ERROR|error|warn)`",
+          "legendFormat": ""
+        }
+      ]
+    },
+    {
+      "id": 31,
+      "type": "logs",
+      "title": "Runner Logs (all)",
+      "gridPos": { "x": 0, "y": 30, "w": 24, "h": 10 },
+      "options": {
+        "showTime": true,
+        "showLabels": false,
+        "showCommonLabels": false,
+        "wrapLogMessage": true,
+        "prettifyLogMessage": true,
+        "enableLogDetails": true,
+        "sortOrder": "Descending",
+        "dedupStrategy": "none"
+      },
+      "targets": [
+        {
+          "datasource": { "type": "loki", "uid": "loki" },
+          "expr": "{service_name=\"runner\"} | json",
+          "legendFormat": ""
+        }
+      ]
+    }
+  ]
+}
--- a/homelab/runner/docker-compose.yml
+++ b/homelab/runner/docker-compose.yml
@@ -1,7 +1,7 @@
 # LibNovel homelab runner
 #
 # Connects to production PocketBase and MinIO via public subdomains.
-# All secrets come from Doppler (project=libnovel, config=prd).
+# All secrets come from Doppler (project=libnovel, config=prd_homelab).
 # Run with: doppler run -- docker compose up -d
 #
 # Differs from prod runner:
@@ -11,12 +11,31 @@
 #   - MEILI_URL               → https://search.libnovel.cc (Caddy-proxied)
 #   - VALKEY_ADDR             → unset (not exposed publicly)
 #   - RUNNER_SKIP_INITIAL_CATALOGUE_REFRESH=true
+#   - Redis service for Asynq task queue (local to homelab, exposed to prod via Caddy TCP proxy)

 services:
+  redis:
+    image: redis:7-alpine
+    restart: unless-stopped
+    volumes:
+      - redis_data:/data
+    command: >
+      redis-server
+      --appendonly yes
+      --requirepass "${REDIS_PASSWORD}"
+    healthcheck:
+      test: ["CMD", "redis-cli", "-a", "${REDIS_PASSWORD}", "ping"]
+      interval: 10s
+      timeout: 5s
+      retries: 5
+
  runner:
    image: kalekber/libnovel-runner:latest
    restart: unless-stopped
    stop_grace_period: 135s
+    depends_on:
+      redis:
+        condition: service_healthy
    environment:
      # ── PocketBase ──────────────────────────────────────────────────────────
      POCKETBASE_URL: "https://pb.libnovel.cc"
@@ -42,6 +61,14 @@ services:
      KOKORO_URL: "${KOKORO_URL}"
      KOKORO_VOICE: "${KOKORO_VOICE}"

+      # ── Pocket TTS ──────────────────────────────────────────────────────────
+      POCKET_TTS_URL: "${POCKET_TTS_URL}"
+
+      # ── Asynq / Redis (local service) ───────────────────────────────────────
+      # The runner connects to the local Redis sidecar.
+      REDIS_ADDR: "redis:6379"
+      REDIS_PASSWORD: "${REDIS_PASSWORD}"
+
      # ── Runner tuning ───────────────────────────────────────────────────────
      RUNNER_WORKER_ID: "${RUNNER_WORKER_ID}"
      RUNNER_POLL_INTERVAL: "${RUNNER_POLL_INTERVAL}"
@@ -60,3 +87,6 @@ services:
      interval: 60s
      timeout: 5s
      retries: 3
+
+volumes:
+  redis_data:
--- a/ui/src/lib/components/AudioPlayer.svelte
+++ b/ui/src/lib/components/AudioPlayer.svelte
@@ -51,6 +51,7 @@
 	import { audioStore } from '$lib/audio.svelte';
 	import { Button } from '$lib/components/ui/button';
 	import { cn } from '$lib/utils';
+	import type { Voice } from '$lib/types';

 	interface Props {
 		slug: string;
@@ -63,8 +64,8 @@
 		nextChapter?: number | null;
 		/** Full chapter list for the book (number + title). Written into the store. */
 		chapters?: { number: number; title: string }[];
-		/** List of available voices from the Kokoro API. */
-		voices?: string[];
+		/** List of available voices from the backend. */
+		voices?: Voice[];
 	}

 	let {
@@ -78,6 +79,10 @@
 		voices = []
 	}: Props = $props();

+	// ── Derived: voices grouped by engine ──────────────────────────────────
+	const kokoroVoices = $derived(voices.filter((v) => v.engine === 'kokoro'));
+	const pocketVoices = $derived(voices.filter((v) => v.engine === 'pocket-tts'));
+
 	// ── Voice selector state ────────────────────────────────────────────────
 	let showVoicePanel = $state(false);
 	/** Voice whose sample is currently being fetched or playing. */
@@ -86,10 +91,33 @@
 	let sampleAudio = $state<HTMLAudioElement | null>(null);

 	/**
-	 * Human-readable label for a voice ID.
-	 * e.g. "af_bella" → "Bella (US F)" | "bm_george" → "George (UK M)"
+	 * Human-readable label for a voice.
+	 * Kokoro: "af_bella" → "Bella (US F)"
+	 * Pocket-TTS: "alba" → "Alba (EN F)"
+	 * Falls back gracefully if called with a bare string (e.g. from the store default).
 	 */
-	function voiceLabel(v: string): string {
+	function voiceLabel(v: Voice | string): string {
+		// Handle plain string IDs stored in audioStore.voice
+		if (typeof v === 'string') {
+			// Try to match against the voices list
+			const found = voices.find((x) => x.id === v);
+			if (found) return voiceLabel(found);
+			// Bare kokoro ID fallback (legacy / default "af_bella")
+			return kokoroLabelFromId(v);
+		}
+
+		if (v.engine === 'pocket-tts') {
+			const langLabel = v.lang.toUpperCase().replace('-', '');
+			const genderLabel = v.gender.toUpperCase();
+			const name = v.id.replace(/_/g, ' ').replace(/\b\w/g, (c) => c.toUpperCase());
+			return `${name} (${langLabel} ${genderLabel})`;
+		}
+
+		// Kokoro
+		return kokoroLabelFromId(v.id);
+	}
+
+	function kokoroLabelFromId(id: string): string {
 		const langMap: Record<string, string> = {
 			af: 'US', am: 'US',
 			bf: 'UK', bm: 'UK',
@@ -112,9 +140,8 @@
 			pf: 'F', pm: 'M',
 			zf: 'F', zm: 'M',
 		};
-		const prefix = v.slice(0, 2);
-		const name = v.slice(3);
-		// Capitalise and strip legacy v0 prefix.
+		const prefix = id.slice(0, 2);
+		const name = id.slice(3);
 		const displayName = name
 			.replace(/^v0/, '')
 			.replace(/^([a-z])/, (c: string) => c.toUpperCase());
@@ -316,23 +343,28 @@

 	// ── API helpers ────────────────────────────────────────────────────────────

+	type PresignResult =
+		| { ready: true; url: string }
+		| { ready: false; enqueued: boolean }; // enqueued=true → presign already POSTed
+
 	async function tryPresign(
 		targetSlug: string,
 		targetChapter: number,
 		targetVoice: string
-	): Promise<string | null> {
+	): Promise<PresignResult> {
 		const params = new URLSearchParams({
 			slug: targetSlug,
 			n: String(targetChapter),
 			voice: targetVoice
 		});
 		const res = await fetch(`/api/presign/audio?${params}`);
-		// 202: TTS was just enqueued by the presign endpoint — audio not ready yet.
+		// 202: presign endpoint already triggered TTS — skip the POST, go straight to polling.
 		// 404: legacy fallback (should no longer occur after endpoint change).
-		if (res.status === 202 || res.status === 404) return null;
+		if (res.status === 202) return { ready: false, enqueued: true };
+		if (res.status === 404) return { ready: false, enqueued: false };
 		if (!res.ok) throw new Error(`presign HTTP ${res.status}`);
 		const data = (await res.json()) as { url: string };
-		return data.url;
+		return { ready: true, url: data.url };
 	}

 	type AudioStatusResponse =
@@ -394,50 +426,52 @@

 		try {
 			// Fast path: already generated
-			const url = await tryPresign(slug, nextChapter, voice);
-			if (url) {
+			const presignResult = await tryPresign(slug, nextChapter, voice);
+			if (presignResult.ready) {
 				stopNextProgress();
 				audioStore.nextProgress = 100;
-				audioStore.nextAudioUrl = url;
+				audioStore.nextAudioUrl = presignResult.url;
 				audioStore.nextStatus = 'prefetched';
 				return;
 			}

-			// Slow path: trigger Kokoro generation (non-blocking POST), then poll.
-			const res = await fetch(`/api/audio/${slug}/${nextChapter}`, {
-				method: 'POST',
-				headers: { 'Content-Type': 'application/json' },
-				body: JSON.stringify({ voice })
-			});
-			if (!res.ok) throw new Error(`Prefetch generation failed: HTTP ${res.status}`);
+			// Slow path: trigger generation (or skip POST if presign already enqueued).
+			if (!presignResult.enqueued) {
+				const res = await fetch(`/api/audio/${slug}/${nextChapter}`, {
+					method: 'POST',
+					headers: { 'Content-Type': 'application/json' },
+					body: JSON.stringify({ voice })
+				});
+				if (!res.ok) throw new Error(`Prefetch generation failed: HTTP ${res.status}`);

-			// Whether the server returned 200 (already cached) or 202 (enqueued),
-			// always presign — the status endpoint no longer returns a proxy URL.
-			if (res.status === 200) {
-				// Body is { status: 'done' } — audio confirmed in MinIO. Presign it.
-				await res.body?.cancel();
-			}
-			// else 202: generation enqueued — fall through to poll.
-
-			if (res.status !== 200) {
-				// 202: poll until done.
-				const final = await pollAudioStatus(slug, nextChapter, voice);
-				stopNextProgress();
-				audioStore.nextProgress = 100;
-
-				if (final.status === 'failed') {
-					throw new Error(`Prefetch failed: ${(final as { error?: string }).error ?? 'unknown'}`);
+				if (res.status === 200) {
+					// Body is { status: 'done' } — audio confirmed in MinIO. Presign it.
+					await res.body?.cancel();
+					stopNextProgress();
+					audioStore.nextProgress = 100;
+					const doneUrl = await tryPresign(slug, nextChapter, voice);
+					if (!doneUrl.ready) throw new Error('Prefetch: audio done but presign returned 404');
+					audioStore.nextAudioUrl = doneUrl.url;
+					audioStore.nextStatus = 'prefetched';
+					return;
 				}
-			} else {
-				stopNextProgress();
-				audioStore.nextProgress = 100;
+				// 202: generation enqueued — fall through to poll.
+			}
+
+			// Poll until done (covers both: presign-enqueued and POST-enqueued paths).
+			const final = await pollAudioStatus(slug, nextChapter, voice);
+			stopNextProgress();
+			audioStore.nextProgress = 100;
+
+			if (final.status === 'failed') {
+				throw new Error(`Prefetch failed: ${(final as { error?: string }).error ?? 'unknown'}`);
 			}

 			// Audio is ready in MinIO — get a direct presigned URL.
 			const doneUrl = await tryPresign(slug, nextChapter, voice);
-			if (!doneUrl) throw new Error('Prefetch: audio done but presign returned 404');
+			if (!doneUrl.ready) throw new Error('Prefetch: audio done but presign returned 404');

-			audioStore.nextAudioUrl = doneUrl;
+			audioStore.nextAudioUrl = doneUrl.url;
 			audioStore.nextStatus = 'prefetched';
 		} catch {
 			stopNextProgress();
@@ -505,9 +539,9 @@
 			}

 			// Fast path B: audio already in MinIO (presign check).
-			const url = await tryPresign(slug, chapter, voice);
-			if (url) {
-				audioStore.audioUrl = url;
+			const presignResult = await tryPresign(slug, chapter, voice);
+			if (presignResult.ready) {
+				audioStore.audioUrl = presignResult.url;
 				audioStore.status = 'ready';
 				// Restore last saved position after the audio element loads
 				restoreSavedAudioTime();
@@ -520,33 +554,44 @@
 			audioStore.status = 'generating';
 			startProgress();

-			const res = await fetch(`/api/audio/${slug}/${chapter}`, {
-				method: 'POST',
-				headers: { 'Content-Type': 'application/json' },
-				body: JSON.stringify({ voice })
-			});
-			if (!res.ok) throw new Error(`Generation failed: HTTP ${res.status}`);
+			// presignResult.enqueued=true means /api/presign/audio already POSTed on our
+			// behalf — skip the duplicate POST and go straight to polling.
+			if (!presignResult.enqueued) {
+				const res = await fetch(`/api/audio/${slug}/${chapter}`, {
+					method: 'POST',
+					headers: { 'Content-Type': 'application/json' },
+					body: JSON.stringify({ voice })
+				});
+				if (!res.ok) throw new Error(`Generation failed: HTTP ${res.status}`);

-			if (res.status !== 200) {
-				// 202: generation enqueued — poll until done.
-				const final = await pollAudioStatus(slug, chapter, voice);
-
-				if (final.status === 'failed') {
-					throw new Error(
-						`Generation failed: ${(final as { error?: string }).error ?? 'unknown error'}`
-					);
+				if (res.status === 200) {
+					// Already cached — body is { status: 'done' }, no url needed.
+					await res.body?.cancel();
+					await finishProgress();
+					const doneUrl = await tryPresign(slug, chapter, voice);
+					if (!doneUrl.ready) throw new Error('Audio generated but presign returned 404');
+					audioStore.audioUrl = doneUrl.url;
+					audioStore.status = 'ready';
+					maybeStartPrefetch();
+					return;
 				}
-			} else {
-				// 200: already cached — body is { status: 'done' }, no url needed.
-				await res.body?.cancel();
+				// 202: fall through to polling below.
+			}
+
+			// Poll until the runner finishes generating.
+			const final = await pollAudioStatus(slug, chapter, voice);
+			if (final.status === 'failed') {
+				throw new Error(
+					`Generation failed: ${(final as { error?: string }).error ?? 'unknown error'}`
+				);
 			}

 			await finishProgress();

 			// Audio is ready in MinIO — always use a presigned URL for direct playback.
 			const doneUrl = await tryPresign(slug, chapter, voice);
-			if (!doneUrl) throw new Error('Audio generated but presign returned 404');
-			audioStore.audioUrl = doneUrl;
+			if (!doneUrl.ready) throw new Error('Audio generated but presign returned 404');
+			audioStore.audioUrl = doneUrl.url;
 			audioStore.status = 'ready';
 			// Don't restore time for freshly generated audio — position is 0
 			// Immediately start pre-generating the next chapter in background.
@@ -627,6 +672,52 @@

 <svelte:window onkeydown={handleKeyDown} />

+<!-- ── Voice row snippet (reused in both engine sections) ──────────────── -->
+{#snippet voiceRow(v: import('$lib/types').Voice)}
+	<div
+		class={cn('flex items-center gap-2 px-3 py-2 hover:bg-zinc-800 transition-colors cursor-pointer', audioStore.voice === v.id && 'bg-amber-400/10')}
+		role="button"
+		tabindex="0"
+		onclick={() => selectVoice(v.id)}
+		onkeydown={(e) => e.key === 'Enter' && selectVoice(v.id)}
+	>
+		<!-- Selected indicator -->
+		<div class="w-4 flex-shrink-0">
+			{#if audioStore.voice === v.id}
+				<svg class="w-3.5 h-3.5 text-amber-400" fill="currentColor" viewBox="0 0 24 24">
+					<path d="M9 16.17L4.83 12l-1.42 1.41L9 19 21 7l-1.41-1.41L9 16.17z"/>
+				</svg>
+			{/if}
+		</div>
+
+		<!-- Voice name -->
+		<span class={cn('flex-1 text-xs', audioStore.voice === v.id ? 'text-amber-400 font-medium' : 'text-zinc-300')}>
+			{voiceLabel(v)}
+		</span>
+		<span class="text-zinc-600 text-xs font-mono">{v.id}</span>
+
+		<!-- Sample play button -->
+		<Button
+			variant="ghost"
+			size="icon"
+			class={cn('h-6 w-6 flex-shrink-0', samplePlayingVoice === v.id ? 'text-amber-400 bg-amber-400/15 hover:bg-amber-400/25' : 'text-zinc-500 hover:text-zinc-200')}
+			onclick={(e) => { e.stopPropagation(); playSample(v.id); }}
+			title={samplePlayingVoice === v.id ? 'Stop sample' : 'Play sample'}
+			aria-label={samplePlayingVoice === v.id ? `Stop ${v.id} sample` : `Play ${v.id} sample`}
+		>
+			{#if samplePlayingVoice === v.id}
+				<svg class="w-3.5 h-3.5" fill="currentColor" viewBox="0 0 24 24">
+					<path d="M6 6h12v12H6z"/>
+				</svg>
+			{:else}
+				<svg class="w-3.5 h-3.5" fill="currentColor" viewBox="0 0 24 24">
+					<path d="M8 5v14l11-7z"/>
+				</svg>
+			{/if}
+		</Button>
+	</div>
+{/snippet}
+
 <div class="mt-6 p-4 rounded-lg bg-zinc-800 border border-zinc-700">
 	<div class="flex items-center justify-between gap-2 mb-3">
 		<div class="flex items-center gap-2">
@@ -674,50 +765,25 @@
 				</Button>
 			</div>
 			<div class="max-h-64 overflow-y-auto">
-				{#each voices as v (v)}
-					<div
-						class={cn('flex items-center gap-2 px-3 py-2 hover:bg-zinc-800 transition-colors cursor-pointer', audioStore.voice === v && 'bg-amber-400/10')}
-						role="button"
-						tabindex="0"
-						onclick={() => selectVoice(v)}
-						onkeydown={(e) => e.key === 'Enter' && selectVoice(v)}
-					>
-						<!-- Selected indicator -->
-						<div class="w-4 flex-shrink-0">
-							{#if audioStore.voice === v}
-								<svg class="w-3.5 h-3.5 text-amber-400" fill="currentColor" viewBox="0 0 24 24">
-									<path d="M9 16.17L4.83 12l-1.42 1.41L9 19 21 7l-1.41-1.41L9 16.17z"/>
-								</svg>
-							{/if}
-						</div>
-
-						<!-- Voice name -->
-						<span class={cn('flex-1 text-xs', audioStore.voice === v ? 'text-amber-400 font-medium' : 'text-zinc-300')}>
-							{voiceLabel(v)}
-						</span>
-						<span class="text-zinc-600 text-xs font-mono">{v}</span>
-
-						<!-- Sample play button (stop propagation so click doesn't select) -->
-						<Button
-							variant="ghost"
-							size="icon"
-							class={cn('h-6 w-6 flex-shrink-0', samplePlayingVoice === v ? 'text-amber-400 bg-amber-400/15 hover:bg-amber-400/25' : 'text-zinc-500 hover:text-zinc-200')}
-							onclick={(e) => { e.stopPropagation(); playSample(v); }}
-							title={samplePlayingVoice === v ? 'Stop sample' : 'Play sample'}
-							aria-label={samplePlayingVoice === v ? `Stop ${v} sample` : `Play ${v} sample`}
-						>
-							{#if samplePlayingVoice === v}
-								<svg class="w-3.5 h-3.5" fill="currentColor" viewBox="0 0 24 24">
-									<path d="M6 6h12v12H6z"/>
-								</svg>
-							{:else}
-								<svg class="w-3.5 h-3.5" fill="currentColor" viewBox="0 0 24 24">
-									<path d="M8 5v14l11-7z"/>
-								</svg>
-							{/if}
-						</Button>
+				<!-- Kokoro (GPU) section -->
+				{#if kokoroVoices.length > 0}
+					<div class="px-3 py-1.5 bg-zinc-800/70 border-b border-zinc-700/50">
+						<span class="text-[10px] font-semibold text-zinc-500 uppercase tracking-widest">Kokoro (GPU)</span>
 					</div>
-				{/each}
+					{#each kokoroVoices as v (v.id)}
+						{@render voiceRow(v)}
+					{/each}
+				{/if}
+
+				<!-- Pocket TTS (CPU) section -->
+				{#if pocketVoices.length > 0}
+					<div class="px-3 py-1.5 bg-zinc-800/70 border-b border-zinc-700/50 {kokoroVoices.length > 0 ? 'border-t border-zinc-700' : ''}">
+						<span class="text-[10px] font-semibold text-zinc-500 uppercase tracking-widest">Pocket TTS (CPU)</span>
+					</div>
+					{#each pocketVoices as v (v.id)}
+						{@render voiceRow(v)}
+					{/each}
+				{/if}
 			</div>
 			<div class="px-3 py-2 border-t border-zinc-700 bg-zinc-800/50">
 				<p class="text-xs text-zinc-500">
--- a/ui/src/lib/server/minio.ts
+++ b/ui/src/lib/server/minio.ts
@@ -40,8 +40,8 @@ export async function presignAvatarUploadUrl(userId: string, mimeType: string):
 }

 /**
- * Returns a presigned GET URL for a user's avatar, rewritten to the public URL.
- * Returns null if no avatar exists.
+ * Returns a presigned GET URL for a user's avatar from MinIO.
+ * Returns null if no avatar object exists in MinIO for this user.
 */
 export async function presignAvatarUrl(userId: string): Promise<string | null> {
 	const res = await backendFetch(`/api/presign/avatar/${encodeURIComponent(userId)}`);
@@ -54,6 +54,42 @@ export async function presignAvatarUrl(userId: string): Promise<string | null> {
 	return data.url ? rewriteHost(data.url) : null;
 }

+/**
+ * Resolves the best available avatar URL for a user.
+ *
+ * Priority:
+ *   1. MinIO — if the user has uploaded a custom avatar it will be found here
+ *      (presigned, short-lived GET URL).
+ *   2. OAuth provider URL — stored in avatar_url when the account was created
+ *      via Google / GitHub OAuth (e.g. https://lh3.googleusercontent.com/...).
+ *      Returned as-is; the browser fetches it directly.
+ *
+ * Pass the raw `avatar_url` field from the PocketBase record as `storedValue`
+ * so this function can distinguish between a MinIO key and a remote URL without
+ * an extra DB round-trip.
+ *
+ * Returns null when neither source yields an avatar.
+ */
+export async function resolveAvatarUrl(
+	userId: string,
+	storedValue: string | null | undefined
+): Promise<string | null> {
+	// 1. Try MinIO first (custom upload takes priority over OAuth picture).
+	try {
+		const minioUrl = await presignAvatarUrl(userId);
+		if (minioUrl) return minioUrl;
+	} catch {
+		// MinIO unavailable — fall through to OAuth fallback.
+	}
+
+	// 2. Fall back to OAuth-provided picture URL if it looks like a remote URL.
+	if (storedValue && storedValue.startsWith('http')) {
+		return storedValue;
+	}
+
+	return null;
+}
+
 /**
 * Rewrites the MinIO host in a presigned URL to the public-facing URL.
 *
--- a/ui/src/lib/server/pocketbase.ts
+++ b/ui/src/lib/server/pocketbase.ts
@@ -541,6 +541,19 @@ export async function getUserByUsername(username: string): Promise<User | null>
 	return listOne<User>('app_users', `username="${username.replace(/"/g, '\\"')}"`);
 }

+/**
+ * Look up a user by their PocketBase record ID. Returns null if not found.
+ */
+export async function getUserById(id: string): Promise<User | null> {
+	const token = await getToken();
+	const res = await fetch(`${PB_URL}/api/collections/app_users/records/${encodeURIComponent(id)}`, {
+		headers: { Authorization: `Bearer ${token}` }
+	});
+	if (res.status === 404) return null;
+	if (!res.ok) return null;
+	return res.json() as Promise<User>;
+}
+
 /**
 * Look up a user by email. Returns null if not found.
 */
--- a/ui/src/lib/types.ts
+++ b/ui/src/lib/types.ts
@@ -6,6 +6,20 @@
 * safe to import in both server and client code.
 */

+// ── Voice ─────────────────────────────────────────────────────────────────────
+
+/** A single TTS voice returned by GET /api/voices. */
+export interface Voice {
+	/** Voice identifier passed to TTS clients (e.g. "af_bella", "alba"). */
+	id: string;
+	/** TTS engine: "kokoro" | "pocket-tts". */
+	engine: string;
+	/** Primary language tag (e.g. "en-us", "en-gb", "en", "es", "fr"). */
+	lang: string;
+	/** Gender: "f" | "m". */
+	gender: string;
+}
+
 // ── Comments ─────────────────────────────────────────────────────────────────

 export interface BookComment {
--- a/ui/src/routes/+layout.svelte
+++ b/ui/src/routes/+layout.svelte
@@ -244,7 +244,7 @@
 				href="/catalogue"
 				class="hidden sm:block text-sm transition-colors {page.url.pathname.startsWith('/catalogue') ? 'text-zinc-100 font-medium' : 'text-zinc-400 hover:text-zinc-100'}"
 			>
-				Discover
+				Catalogue
 			</a>
 			<a
 				href="https://feedback.libnovel.cc"
@@ -332,7 +332,7 @@
 				onclick={() => (menuOpen = false)}
 				class="px-3 py-2.5 rounded-lg text-sm font-medium transition-colors {page.url.pathname.startsWith('/catalogue') ? 'bg-zinc-800 text-zinc-100' : 'text-zinc-400 hover:bg-zinc-800 hover:text-zinc-100'}"
 			>
-				Discover
+				Catalogue
 			</a>
 			<a
 				href="https://feedback.libnovel.cc"
@@ -400,7 +400,7 @@
 			<!-- Top row: site links -->
 			<nav class="flex flex-wrap items-center justify-center gap-x-5 gap-y-2">
 			<a href="/books" class="hover:text-zinc-400 transition-colors">Library</a>
-			<a href="/catalogue" class="hover:text-zinc-400 transition-colors">Discover</a>
+			<a href="/catalogue" class="hover:text-zinc-400 transition-colors">Catalogue</a>
 			<a
 				href="https://feedback.libnovel.cc"
 				target="_blank"
@@ -426,16 +426,26 @@
 					</svg>
 				</a>
 			</nav>
-			<!-- Bottom row: legal links + copyright -->
-			<div class="flex flex-wrap items-center justify-center gap-x-5 gap-y-2 text-zinc-700">
-				<a href="/disclaimer" class="hover:text-zinc-500 transition-colors">Disclaimer</a>
-				<a href="/privacy" class="hover:text-zinc-500 transition-colors">Privacy</a>
-				<a href="/dmca" class="hover:text-zinc-500 transition-colors">DMCA</a>
-				<span>&copy; {new Date().getFullYear()} libnovel</span>
-				{#if env.PUBLIC_BUILD_VERSION && env.PUBLIC_BUILD_VERSION !== 'dev'}
-					<span class="text-zinc-800">{env.PUBLIC_BUILD_VERSION}+{env.PUBLIC_BUILD_COMMIT?.slice(0, 7)}</span>
+		<!-- Bottom row: legal links + copyright -->
+		<div class="flex flex-wrap items-center justify-center gap-x-5 gap-y-2 text-zinc-700">
+			<a href="/disclaimer" class="hover:text-zinc-500 transition-colors">Disclaimer</a>
+			<a href="/privacy" class="hover:text-zinc-500 transition-colors">Privacy</a>
+			<a href="/dmca" class="hover:text-zinc-500 transition-colors">DMCA</a>
+			<span>&copy; {new Date().getFullYear()} libnovel</span>
+		</div>
+		<!-- Build version / commit SHA -->
+		<div class="text-xs tabular-nums font-mono px-2 py-0.5 rounded bg-zinc-800 border border-zinc-700">
+			{#if env.PUBLIC_BUILD_VERSION && env.PUBLIC_BUILD_VERSION !== 'dev'}
+				<span class="text-zinc-300" title="Build version">{env.PUBLIC_BUILD_VERSION}</span>
+				{#if env.PUBLIC_BUILD_COMMIT && env.PUBLIC_BUILD_COMMIT !== 'unknown'}
+					<span class="text-zinc-500 select-all" title="Commit SHA"
+						>+{env.PUBLIC_BUILD_COMMIT.slice(0, 7)}</span
+					>
 				{/if}
-			</div>
+			{:else}
+				<span class="text-zinc-400">dev</span>
+			{/if}
+		</div>
 		</div>
 	</footer>
 </div>
--- a/ui/src/routes/api/auth/me/+server.ts
+++ b/ui/src/routes/api/auth/me/+server.ts
@@ -1,6 +1,7 @@
 import { json, error } from '@sveltejs/kit';
 import type { RequestHandler } from './$types';
 import { getUserByUsername } from '$lib/server/pocketbase';
+import { resolveAvatarUrl } from '$lib/server/minio';

 /**
 * GET /api/auth/me
@@ -13,10 +14,11 @@ export const GET: RequestHandler = async ({ locals }) => {
 	}
 	// Fetch full record from PocketBase to get avatar_url
 	const record = await getUserByUsername(locals.user.username).catch(() => null);
+	const avatarUrl = await resolveAvatarUrl(locals.user.id, record?.avatar_url).catch(() => null);
 	return json({
 		id: locals.user.id,
 		username: locals.user.username,
 		role: locals.user.role,
-		avatar_url: record?.avatar_url ?? null
+		avatar_url: avatarUrl
 	});
 };
--- a/ui/src/routes/api/chapter/[slug]/[n]/+server.ts
+++ b/ui/src/routes/api/chapter/[slug]/[n]/+server.ts
@@ -4,6 +4,7 @@ import type { RequestHandler } from './$types';
 import { getBook, listChapterIdx } from '$lib/server/pocketbase';
 import { log } from '$lib/server/logger';
 import { backendFetch } from '$lib/server/scraper';
+import type { Voice } from '$lib/types';

 /**
 * GET /api/chapter/[slug]/[n]
@@ -48,11 +49,11 @@ export const GET: RequestHandler = async ({ params, url, locals }) => {
 			? '<p>' + chapterData.text.replace(/\n{2,}/g, '</p><p>').replace(/\n/g, '<br>') + '</p>'
 			: '';

-		let voices: string[] = [];
+		let voices: Voice[] = [];
 		try {
 			const vRes = await backendFetch('/api/voices');
 			if (vRes.ok) {
-				const d = (await vRes.json()) as { voices: string[] };
+				const d = (await vRes.json()) as { voices: Voice[] };
 				voices = d.voices ?? [];
 			}
 		} catch {
@@ -85,10 +86,10 @@ export const GET: RequestHandler = async ({ params, url, locals }) => {
 	const chapterIdx = chapters.find((c) => c.number === n);
 	if (!chapterIdx) error(404, `Chapter ${n} not found`);

-	let voices: string[] = [];
+	let voices: Voice[] = [];
 	try {
 		if (voicesRes?.ok) {
-			const data = (await voicesRes.json()) as { voices: string[] };
+			const data = (await voicesRes.json()) as { voices: Voice[] };
 			voices = data.voices ?? [];
 		}
 	} catch {
--- a/ui/src/routes/api/comments/[slug]/+server.ts
+++ b/ui/src/routes/api/comments/[slug]/+server.ts
@@ -5,9 +5,10 @@ import {
 	listReplies,
 	createComment,
 	getMyVotes,
+	getUserById,
 	type CommentSort
 } from '$lib/server/pocketbase';
-import { presignAvatarUrl } from '$lib/server/minio';
+import { resolveAvatarUrl } from '$lib/server/minio';
 import { log } from '$lib/server/logger';

 /**
@@ -38,13 +39,15 @@ export const GET: RequestHandler = async ({ params, url, locals }) => {
 			replies: repliesPerComment[i]
 		}));

-		// Batch-resolve avatar presign URLs for all unique user_ids
+		// Batch-resolve avatar URLs for all unique user_ids
+		// MinIO first (custom upload), fall back to OAuth provider picture.
 		const allComments = [...topLevel, ...allReplies];
 		const uniqueUserIds = [...new Set(allComments.map((c) => c.user_id).filter(Boolean))];
 		const avatarEntries = await Promise.all(
 			uniqueUserIds.map(async (userId) => {
 				try {
-					const url = await presignAvatarUrl(userId);
+					const user = await getUserById(userId);
+					const url = await resolveAvatarUrl(userId, user?.avatar_url);
 					return [userId, url] as [string, string | null];
 				} catch {
 					return [userId, null] as [string, null];
--- a/ui/src/routes/api/profile/avatar/+server.ts
+++ b/ui/src/routes/api/profile/avatar/+server.ts
@@ -1,6 +1,6 @@
 import { json, error } from '@sveltejs/kit';
 import type { RequestHandler } from './$types';
-import { presignAvatarUrl } from '$lib/server/minio';
+import { presignAvatarUrl, resolveAvatarUrl } from '$lib/server/minio';
 import { updateUserAvatarUrl, getUserByUsername } from '$lib/server/pocketbase';
 import { backendFetch } from '$lib/server/scraper';

@@ -63,10 +63,6 @@ export const GET: RequestHandler = async ({ locals }) => {
 	if (!locals.user) error(401, 'Not authenticated');

 	const record = await getUserByUsername(locals.user.username).catch(() => null);
-	if (!record?.avatar_url) {
-		return json({ avatar_url: null });
-	}
-
-	const avatarUrl = await presignAvatarUrl(locals.user.id);
+	const avatarUrl = await resolveAvatarUrl(locals.user.id, record?.avatar_url).catch(() => null);
 	return json({ avatar_url: avatarUrl });
 };
--- a/ui/src/routes/api/users/[username]/+server.ts
+++ b/ui/src/routes/api/users/[username]/+server.ts
@@ -1,7 +1,7 @@
 import { json, error } from '@sveltejs/kit';
 import type { RequestHandler } from './$types';
 import { getPublicProfile, getSubscription } from '$lib/server/pocketbase';
-import { presignAvatarUrl } from '$lib/server/minio';
+import { resolveAvatarUrl } from '$lib/server/minio';
 import { log } from '$lib/server/logger';

 /**
@@ -15,11 +15,9 @@ export const GET: RequestHandler = async ({ params, locals }) => {
 		const profile = await getPublicProfile(username);
 		if (!profile) error(404, `User "${username}" not found`);

-		// Resolve avatar presigned URL if set
+		// Resolve avatar — MinIO first, fall back to OAuth provider picture
 		let avatarUrl: string | null = null;
-		if (profile.avatar_url) {
-			avatarUrl = await presignAvatarUrl(profile.id).catch(() => null);
-		}
+		avatarUrl = await resolveAvatarUrl(profile.id, profile.avatar_url).catch(() => null);

 		// Is the current logged-in user subscribed?
 		let isSubscribed = false;
--- a/ui/src/routes/api/voices/+server.ts
+++ b/ui/src/routes/api/voices/+server.ts
@@ -1,11 +1,12 @@
 import { json } from '@sveltejs/kit';
 import type { RequestHandler } from './$types';
 import { backendFetch } from '$lib/server/scraper';
+import type { Voice } from '$lib/types';

 /**
 * GET /api/voices
- * Proxies the voice list from the backend → Kokoro.
- * Returns { voices: string[] }
+ * Proxies the voice list from the backend (Kokoro + pocket-tts).
+ * Returns { voices: Voice[] }
 */
 export const GET: RequestHandler = async () => {
 	try {
@@ -13,7 +14,7 @@ export const GET: RequestHandler = async () => {
 		if (!res.ok) {
 			return json({ voices: [] });
 		}
-		const data = (await res.json()) as { voices: string[] };
+		const data = (await res.json()) as { voices: Voice[] };
 		return json({ voices: data.voices ?? [] });
 	} catch {
 		return json({ voices: [] });
--- a/ui/src/routes/books/[slug]/chapters/[n]/+page.server.ts
+++ b/ui/src/routes/books/[slug]/chapters/[n]/+page.server.ts
@@ -4,6 +4,7 @@ import type { PageServerLoad } from './$types';
 import { getBook, listChapterIdx } from '$lib/server/pocketbase';
 import { log } from '$lib/server/logger';
 import { backendFetch } from '$lib/server/scraper';
+import type { Voice } from '$lib/types';

 export const load: PageServerLoad = async ({ params, url, locals }) => {
 	const { slug } = params;
@@ -43,11 +44,11 @@ export const load: PageServerLoad = async ({ params, url, locals }) => {
 			: '';

 		// Fetch voices (non-critical for preview)
-		let voices: string[] = [];
+		let voices: Voice[] = [];
 		try {
 			const vRes = await backendFetch('/api/voices');
 			if (vRes.ok) {
-				const d = (await vRes.json()) as { voices: string[] };
+				const d = (await vRes.json()) as { voices: Voice[] };
 				voices = d.voices ?? [];
 			}
 		} catch {
@@ -93,11 +94,11 @@ export const load: PageServerLoad = async ({ params, url, locals }) => {
 	const chapterIdx = chapters.find((c) => c.number === n);
 	if (!chapterIdx) error(404, `Chapter ${n} not found`);

-	// Parse voices — fall back to a minimal default list on error
-	let voices: string[] = [];
+	// Parse voices — fall back to empty list on error
+	let voices: Voice[] = [];
 	try {
 		if (voicesRes?.ok) {
-			const data = (await voicesRes.json()) as { voices: string[] };
+			const data = (await voicesRes.json()) as { voices: Voice[] };
 			voices = data.voices ?? [];
 		}
 	} catch {
--- a/ui/src/routes/catalogue/+page.svelte
+++ b/ui/src/routes/catalogue/+page.svelte
@@ -1,5 +1,6 @@
 <script lang="ts">
 	import { enhance } from '$app/forms';
+	import { goto } from '$app/navigation';
 	import { navigating } from '$app/state';
 	import { untrack } from 'svelte';
 	import type { PageData, ActionData } from './$types';
@@ -7,6 +8,29 @@

 	let { data, form }: { data: PageData; form: ActionData } = $props();

+	// ── Local filter state (mirrors URL params) ──────────────────────────────
+	// These are separate from data.* so we can bind them to selects and keep
+	// the DOM in sync. They sync back from data whenever a navigation completes.
+	let filterSort = $state(untrack(() => data.sort));
+	let filterGenre = $state(untrack(() => data.genre));
+	let filterStatus = $state(untrack(() => data.status));
+
+	// Keep local state in sync whenever SvelteKit re-runs the load (URL changed).
+	$effect(() => {
+		filterSort = data.sort;
+		filterGenre = data.genre;
+		filterStatus = data.status;
+	});
+
+	function navigateWithFilters(overrides: { sort?: string; genre?: string; status?: string }) {
+		const params = new URLSearchParams();
+		params.set('sort', overrides.sort ?? filterSort);
+		params.set('genre', overrides.genre ?? filterGenre);
+		params.set('status', overrides.status ?? filterStatus);
+		params.set('page', '1');
+		goto(`/catalogue?${params.toString()}`);
+	}
+
 	// Track which novel card is currently being navigated to
 	let loadingSlug = $state<string | null>(null);

@@ -389,11 +413,12 @@
 				<select
 					id="filter-sort"
 					name="sort"
-					value={data.sort}
+					bind:value={filterSort}
+					onchange={() => navigateWithFilters({ sort: filterSort })}
 					class="bg-zinc-900 border border-zinc-700 text-zinc-200 text-sm rounded px-3 py-2 focus:outline-none focus:border-amber-400 w-full"
 				>
 					{#each sorts as s}
-						<option value={s.value}>{s.label}</option>
+						<option value={s.value} selected={s.value === filterSort}>{s.label}</option>
 					{/each}
 				</select>
 			</div>
@@ -403,12 +428,13 @@
 				<select
 					id="filter-genre"
 					name="genre"
-					value={data.genre}
+					bind:value={filterGenre}
+					onchange={() => navigateWithFilters({ genre: filterGenre })}
 					disabled={isRankView}
 					class="bg-zinc-900 border border-zinc-700 text-zinc-200 text-sm rounded px-3 py-2 focus:outline-none focus:border-amber-400 disabled:opacity-40 disabled:cursor-not-allowed w-full"
 				>
 					{#each genres as g}
-						<option value={g.value}>{g.label}</option>
+						<option value={g.value} selected={g.value === filterGenre}>{g.label}</option>
 					{/each}
 				</select>
 			</div>
@@ -418,12 +444,13 @@
 				<select
 					id="filter-status"
 					name="status"
-					value={data.status}
+					bind:value={filterStatus}
+					onchange={() => navigateWithFilters({ status: filterStatus })}
 					disabled={isRankView}
 					class="bg-zinc-900 border border-zinc-700 text-zinc-200 text-sm rounded px-3 py-2 focus:outline-none focus:border-amber-400 disabled:opacity-40 disabled:cursor-not-allowed w-full"
 				>
 					{#each statuses as st}
-						<option value={st.value}>{st.label}</option>
+						<option value={st.value} selected={st.value === filterStatus}>{st.label}</option>
 					{/each}
 				</select>
 			</div>
@@ -437,13 +464,6 @@
 			<a href="/catalogue" class="px-4 py-2 rounded bg-zinc-700 text-zinc-300 text-sm hover:bg-zinc-600 transition-colors">
 				Reset
 			</a>
-			<button
-				type="submit"
-				onclick={() => (filtersOpen = false)}
-				class="px-4 py-2 rounded bg-amber-400 text-zinc-900 text-sm font-semibold hover:bg-amber-300 transition-colors"
-			>
-				Apply
-			</button>
 		</div>
 	</form>
 {/if}
--- a/ui/src/routes/profile/+page.server.ts
+++ b/ui/src/routes/profile/+page.server.ts
@@ -1,7 +1,7 @@
 import { fail, redirect } from '@sveltejs/kit';
 import type { Actions, PageServerLoad } from './$types';
 import { changePassword, listUserSessions, getUserByUsername } from '$lib/server/pocketbase';
-import { presignAvatarUrl } from '$lib/server/minio';
+import { resolveAvatarUrl } from '$lib/server/minio';
 import { log } from '$lib/server/logger';

 export const load: PageServerLoad = async ({ locals }) => {
@@ -16,13 +16,11 @@ export const load: PageServerLoad = async ({ locals }) => {
 		log.warn('profile', 'listUserSessions failed (non-fatal)', { err: String(e) });
 	}

-	// Fetch avatar presigned URL if user has one
+	// Fetch avatar — MinIO first, fall back to OAuth provider picture
 	let avatarUrl: string | null = null;
 	try {
 		const record = await getUserByUsername(locals.user.username);
-		if (record?.avatar_url) {
-			avatarUrl = await presignAvatarUrl(locals.user.id);
-		}
+		avatarUrl = await resolveAvatarUrl(locals.user.id, record?.avatar_url);
 	} catch (e) {
 		log.warn('profile', 'avatar fetch failed (non-fatal)', { err: String(e) });
 	}
--- a/ui/src/routes/profile/+page.svelte
+++ b/ui/src/routes/profile/+page.svelte
@@ -5,6 +5,7 @@
 	import type { PageData, ActionData } from './$types';
 	import { audioStore } from '$lib/audio.svelte';
 	import { browser } from '$app/environment';
+	import type { Voice } from '$lib/types';

 	let { data, form }: { data: PageData; form: ActionData } = $props();

@@ -56,14 +57,18 @@
 	}

 	// ── Settings ────────────────────────────────────────────────────────────────
-	let voices = $state<string[]>([]);
+	let voices = $state<Voice[]>([]);
 	let voicesLoaded = $state(false);

+	// Derived: voices grouped by engine
+	const kokoroVoices = $derived(voices.filter((v) => v.engine === 'kokoro'));
+	const pocketVoices = $derived(voices.filter((v) => v.engine === 'pocket-tts'));
+
 	// Load voices on mount
 	$effect(() => {
 		fetch('/api/voices')
 			.then((r) => r.json())
-			.then((d: { voices: string[] }) => {
+			.then((d: { voices: Voice[] }) => {
 				voices = d.voices ?? [];
 				voicesLoaded = true;
 			})
@@ -276,9 +281,20 @@
 					bind:value={voice}
 					class="w-full bg-zinc-700 border border-zinc-600 rounded-lg px-3 py-2 text-zinc-100 text-sm focus:outline-none focus:ring-2 focus:ring-amber-400"
 				>
-					{#each voices as v}
-						<option value={v}>{v}</option>
-					{/each}
+					{#if kokoroVoices.length > 0}
+						<optgroup label="Kokoro (GPU)">
+							{#each kokoroVoices as v}
+								<option value={v.id}>{v.id}</option>
+							{/each}
+						</optgroup>
+					{/if}
+					{#if pocketVoices.length > 0}
+						<optgroup label="Pocket TTS (CPU)">
+							{#each pocketVoices as v}
+								<option value={v.id}>{v.id}</option>
+							{/each}
+						</optgroup>
+					{/if}
 				</select>
 			{/if}
 		</div>
--- a/ui/src/routes/users/[username]/+page.server.ts
+++ b/ui/src/routes/users/[username]/+page.server.ts
@@ -6,7 +6,7 @@ import {
 	getUserPublicLibrary,
 	getUserCurrentlyReading
 } from '$lib/server/pocketbase';
-import { presignAvatarUrl } from '$lib/server/minio';
+import { resolveAvatarUrl } from '$lib/server/minio';
 import { log } from '$lib/server/logger';

 export const load: PageServerLoad = async ({ params, locals }) => {
@@ -15,11 +15,9 @@ export const load: PageServerLoad = async ({ params, locals }) => {
 	const profile = await getPublicProfile(username).catch(() => null);
 	if (!profile) error(404, `User "${username}" not found`);

-	// Resolve avatar
+	// Resolve avatar — MinIO first, fall back to OAuth provider picture
 	let avatarUrl: string | null = null;
-	if (profile.avatar_url) {
-		avatarUrl = await presignAvatarUrl(profile.id).catch(() => null);
-	}
+	avatarUrl = await resolveAvatarUrl(profile.id, profile.avatar_url).catch(() => null);

 	// Subscription state for the logged-in visitor
 	let isSubscribed = false;
Author	SHA1	Message	Date
Admin	428b57732e	fix(ui): resolve avatar URL from MinIO; fall back to OAuth provider URL Some checks failed CI / Backend (push) Successful in 50s Details CI / UI (push) Successful in 34s Details Release / Test backend (push) Successful in 37s Details Release / Check ui (push) Successful in 45s Details Release / Docker / caddy (push) Successful in 35s Details CI / Backend (pull_request) Successful in 49s Details CI / UI (pull_request) Successful in 40s Details Release / Docker / backend (push) Failing after 52s Details Release / Docker / runner (push) Failing after 58s Details Release / Docker / ui (push) Failing after 55s Details Release / Gitea Release (push) Has been skipped Details Add resolveAvatarUrl(userId, storedValue) helper that tries MinIO first, then falls back to the stored HTTP URL for OAuth users (Google/GitHub) who have never uploaded a custom avatar. Add getUserById() to pocketbase helpers for batch avatar resolution in comments. Update all 6 call sites to use the new helper.	2026-03-28 19:23:30 +05:00
Admin	61e77e3e28	ci: remove Docker builds from CI; keep vet/build/test/type-check only All checks were successful CI / Backend (push) Successful in 43s Details CI / UI (push) Successful in 37s Details CI / UI (pull_request) Successful in 35s Details CI / Backend (pull_request) Successful in 1m9s Details Docker image builds belong to the release workflow (tag-triggered). CI now runs: go vet, go build (backend/runner/healthcheck), go test, svelte-check, and UI vite build — fast feedback without Docker overhead. Also triggers on all branches, not just main/master.	2026-03-28 19:08:17 +05:00
Admin	b363c151a5	fix(ui): fix catalogue filters in Svelte 5; improve build badge visibility Some checks failed Release / Test backend (push) Successful in 40s Details Release / Check ui (push) Successful in 49s Details Release / Docker / caddy (push) Failing after 1m0s Details CI / Test backend (pull_request) Successful in 40s Details CI / Docker / caddy (pull_request) Failing after 21s Details CI / Check ui (pull_request) Successful in 48s Details Release / Docker / ui (push) Successful in 2m49s Details Release / Docker / runner (push) Successful in 3m10s Details Release / Docker / backend (push) Successful in 3m47s Details Release / Gitea Release (push) Has been skipped Details CI / Docker / ui (pull_request) Successful in 1m31s Details CI / Docker / backend (pull_request) Successful in 3m13s Details CI / Docker / runner (pull_request) Successful in 3m59s Details - Catalogue filter selects: replace value= initializer with bind:value + onchange goto() so filters navigate immediately on change (no Apply button) - Add selected= on each <option> for correct DOM initialisation in Svelte 5 - Build badge: give distinct bg-zinc-800 pill, visible zinc-300/400/500 text instead of zinc-700/zinc-800 which blended into the footer background	2026-03-28 19:01:58 +05:00
Admin	aef9e04419	fix(runner): harden catalogue scrape against 429s; disable sourcemap upload Some checks failed Release / Test backend (push) Successful in 29s Details Release / Check ui (push) Successful in 43s Details CI / Test backend (pull_request) Successful in 28s Details Release / Docker / caddy (push) Successful in 1m8s Details CI / Docker / caddy (pull_request) Failing after 39s Details CI / Check ui (pull_request) Successful in 55s Details Release / Docker / runner (push) Successful in 2m22s Details Release / Docker / ui (push) Successful in 2m20s Details CI / Docker / backend (pull_request) Successful in 2m32s Details CI / Docker / ui (pull_request) Successful in 1m28s Details CI / Docker / runner (pull_request) Successful in 2m25s Details Release / Docker / backend (push) Successful in 2m33s Details Release / Gitea Release (push) Failing after 2s Details - scraper.go: ScrapeCatalogue now uses retryGet (9 attempts, 10s base) + 500–1500ms inter-page jitter instead of bare GetContent. ScrapeMetadata also switched to retryGet so a single 429 on a book page is retried rather than aborting the whole refresh. - catalogue_refresh.go: per-book delay is now configurable (RUNNER_CATALOGUE_REQUEST_DELAY, default 2s) + up to 50% random jitter applied before every metadata fetch. Only metadata is scraped here — chapters are fetched on-demand, not during catalogue refresh. Progress logged every 50 books instead of 100. - config.go / runner.go / main.go: add CatalogueRequestDelay field wired from RUNNER_CATALOGUE_REQUEST_DELAY env var. - release.yaml: comment out upload-sourcemaps job and remove it from the release needs; GlitchTip auth token needs refreshing after DB wipe.	2026-03-28 16:28:36 +05:00
Admin	58e78cd34d	fix(infra): make MEILI_URL configurable; rename Discover → Catalogue in nav Some checks failed Release / Test backend (push) Failing after 11s Details Release / Docker / caddy (push) Failing after 11s Details Release / Check ui (push) Failing after 11s Details Release / Docker / backend (push) Has been skipped Details Release / Docker / runner (push) Has been skipped Details Release / Upload source maps (push) Has been skipped Details Release / Docker / ui (push) Has been skipped Details Release / Gitea Release (push) Has been skipped Details CI / Check ui (pull_request) Successful in 49s Details CI / Docker / ui (pull_request) Successful in 1m37s Details CI / Docker / caddy (pull_request) Successful in 4m4s Details CI / Test backend (pull_request) Successful in 4m29s Details CI / Docker / runner (pull_request) Failing after 1m27s Details CI / Docker / backend (pull_request) Successful in 1m41s Details - docker-compose.yml: MEILI_URL in x-infra-env was hardcoded to the local meilisearch container, ignoring the Doppler MEILI_URL secret entirely. Changed to "${MEILI_URL:-http://meilisearch:7700}" so prod reads from https://search.libnovel.cc while local dev keeps the container default. - ui/+layout.svelte: rename nav + footer label from "Discover" to "Catalogue" (desktop nav, mobile menu, footer — all three occurrences).	2026-03-28 16:15:04 +05:00
Admin	c5c167035d	fix(tts): fix pocket-tts voices missing in UI and 500 on first TTS enqueue Some checks failed CI / Test backend (pull_request) Successful in 40s Details CI / Check ui (pull_request) Successful in 45s Details Release / Test backend (push) Successful in 40s Details Release / Check ui (push) Successful in 41s Details Release / Docker / caddy (push) Successful in 1m8s Details CI / Docker / runner (pull_request) Failing after 35s Details CI / Docker / caddy (pull_request) Successful in 3m39s Details CI / Docker / ui (pull_request) Successful in 1m12s Details CI / Docker / backend (pull_request) Successful in 3m24s Details Release / Upload source maps (push) Failing after 43s Details Release / Docker / runner (push) Successful in 2m49s Details Release / Docker / ui (push) Successful in 2m51s Details Release / Docker / backend (push) Failing after 7m14s Details Release / Gitea Release (push) Has been skipped Details - Add POCKET_TTS_URL env to backend service in docker-compose.yml so pocket-tts voices appear in the voice selector (Doppler secret existed but the env var was never passed to the container) - Fix GetAudioTask PocketBase filter using %q (double-quotes) instead of single-quoted string, causing the duplicate-task guard to always miss - Fix AudioPlayer double-POST: GET /api/presign/audio already enqueues TTS internally on 404; AudioPlayer now skips the redundant POST and polls directly, eliminating the 500 from the PB unique-key conflict	2026-03-28 16:02:46 +05:00
Admin	4a00d953bb	feat(ui): show build version in footer; enable watchtower for caddy Some checks failed CI / Check ui (pull_request) Successful in 50s Details CI / Docker / ui (pull_request) Failing after 38s Details CI / Test backend (pull_request) Successful in 3m31s Details CI / Docker / backend (pull_request) Failing after 11s Details CI / Docker / runner (pull_request) Failing after 11s Details CI / Docker / caddy (pull_request) Successful in 11m0s Details - Footer now displays build version tag + short commit SHA (from PUBLIC_BUILD_VERSION / PUBLIC_BUILD_COMMIT env vars baked in at image build time); falls back to 'dev' in local builds - docker-compose.yml: add watchtower label to caddy service so it auto-updates alongside backend/ui/runner on new image pushes - homelab/docker-compose.yml: use locally-built kokoro-fastapi:latest image (consistent with actual homelab setup)	2026-03-28 15:42:22 +05:00
Admin	fe1a933fd0	feat(queue): replace PocketBase polling with Asynq + Redis Some checks failed CI / Docker / caddy (pull_request) Failing after 50s Details CI / Check ui (pull_request) Successful in 1m5s Details Release / Check ui (push) Successful in 45s Details Release / Test backend (push) Successful in 1m29s Details CI / Docker / ui (pull_request) Successful in 1m28s Details Release / Docker / backend (push) Successful in 3m14s Details CI / Test backend (pull_request) Failing after 42s Details CI / Docker / backend (pull_request) Has been skipped Details CI / Docker / runner (pull_request) Has been skipped Details Release / Docker / caddy (push) Successful in 6m48s Details Release / Docker / ui (push) Successful in 2m8s Details Release / Docker / runner (push) Successful in 2m51s Details Release / Upload source maps (push) Failing after 53s Details Release / Gitea Release (push) Has been skipped Details Introduce a Redis-backed Asynq task queue so the runner consumes TTS jobs pushed by the backend instead of polling PocketBase. - backend/internal/asynqqueue: Producer and Consumer wrappers - backend/internal/runner: AsynqRunner mux, per-instance Prometheus registry (fixes duplicate-collector panic in tests), redisConnOpt - backend/internal/config: REDIS_ADDR / REDIS_PASSWORD env vars - backend/cmd/{backend,runner}/main.go: wire Redis when env set; fall back to legacy poll mode when unset - Caddyfile: caddy-l4 TCP proxy for redis.libnovel.cc:6380 → homelab - caddy/Dockerfile: add --with github.com/mholt/caddy-l4 - docker-compose.yml: Caddy exposes 6380, backend/runner get Redis env - homelab/runner/docker-compose.yml: Redis sidecar, runner depends_on - homelab/otel/grafana: Grafana dashboards (backend, catalogue, runner) and alerting rules / contact-points provisioning	2026-03-28 14:32:40 +05:00
Admin	98e4a87432	feat(tts): dual-engine voice list (kokoro + pocket-tts) Expose all available voices from both TTS engines via the /api/voices endpoint. AudioPlayer and profile voice-selector now group voices by engine and show a labelled optgroup. Voice type carries an engine field so the chapter-reader can route synthesis to the correct backend.	2026-03-28 14:32:06 +05:00