Transcription on mac

quick cosmetic fix
Dictation through AI
2026-03-11 03:08:27 +01:00 · 2026-03-11 00:44:36 +01:00 · 2026-03-11 00:26:49 +01:00 · 2026-03-10 22:36:21 +01:00 · 2026-03-10 04:13:32 +01:00 · 2026-03-10 03:27:11 +01:00
27 changed files with 2411 additions and 131 deletions
--- a/README.md
+++ b/README.md
@@ -71,6 +71,13 @@ The repo also includes a multiplatform SwiftUI client in `apple-client/` for mac
 - Generate the Xcode project with `xcodegen generate --spec apple-client/project.yml --project-root apple-client`.
 - A build of the Apple app automatically rebuilds the Angular client into `apple-client/WebApp/` before bundling it.

+The backend also exposes the latest Angular browser build through an API that native clients can sync into their local `WKWebView` bundle cache:
+
+- `GET /api/web-app/manifest`: Returns a bundle manifest with a stable `bundleId`, latest `generatedAt` timestamp, and a file list containing relative paths, SHA-256 hashes, MIME types, sizes, and download URLs.
+- `GET /api/web-app/files/<relative-path>`: Streams an individual file from `client/dist/client/browser` with `ETag` and `Last-Modified` headers for native caching.
+
+If the Angular build does not exist yet, those endpoints return `404`.
+
 ## Backend environment

 The backend accepts these environment variables:
--- a/client/angular.json
+++ b/client/angular.json
@@ -68,8 +68,8 @@
                },
                {
                  "type": "anyComponentStyle",
-                  "maximumWarning": "4kB",
-                  "maximumError": "8kB"
+                  "maximumWarning": "10kB",
+                  "maximumError": "12kB"
                }
              ],
              "outputHashing": "all"
--- a/client/public/SymphonyDing.mp3
+++ b/client/public/SymphonyDing.mp3
--- a/client/public/apple-touch-icon.png
+++ b/client/public/apple-touch-icon.png
--- a/client/public/icon-source.svg
+++ b/client/public/icon-source.svg
@@ -0,0 +1,21 @@
+<svg width="512" height="512" viewBox="0 0 512 512" fill="none" xmlns="http://www.w3.org/2000/svg">
+  <defs>
+    <linearGradient id="bg" x1="64" y1="40" x2="448" y2="472" gradientUnits="userSpaceOnUse">
+      <stop stop-color="#10233B"/>
+      <stop offset="1" stop-color="#06111D"/>
+    </linearGradient>
+    <linearGradient id="accent" x1="124" y1="124" x2="389" y2="389" gradientUnits="userSpaceOnUse">
+      <stop stop-color="#84F4D6"/>
+      <stop offset="1" stop-color="#56ABFF"/>
+    </linearGradient>
+  </defs>
+  <rect width="512" height="512" rx="120" fill="url(#bg)"/>
+  <circle cx="154" cy="148" r="96" fill="#8DF0DF" fill-opacity="0.16"/>
+  <circle cx="394" cy="118" r="78" fill="#58ABFF" fill-opacity="0.16"/>
+  <path d="M152 164C152 132.967 177.167 107.8 208.2 107.8H303.8C334.833 107.8 360 132.967 360 164V227.7C360 258.733 334.833 283.9 303.8 283.9H257.8L198.6 335.3C190.343 342.468 177.4 336.601 177.4 325.666V283.9H208.2C177.167 283.9 152 258.733 152 227.7V164Z" fill="url(#accent)"/>
+  <rect x="195" y="154" width="122" height="18" rx="9" fill="#062039" fill-opacity="0.9"/>
+  <rect x="195" y="196" width="86" height="18" rx="9" fill="#062039" fill-opacity="0.9"/>
+  <path d="M354.8 334.9C354.8 379.013 319.046 414.767 274.933 414.767C255.288 414.767 237.299 407.666 223.396 395.888L172.572 410.4C163.669 412.942 155.453 404.726 157.995 395.823L172.507 344.999C160.729 331.096 153.628 313.107 153.628 293.462C153.628 249.349 189.382 213.595 233.495 213.595C277.608 213.595 313.362 249.349 313.362 293.462C313.362 304.056 311.3 314.171 307.553 323.426L344.213 360.086C350.981 366.854 354.8 376.033 354.8 385.604V334.9Z" fill="#0F2540"/>
+  <circle cx="233.495" cy="293.462" r="52.895" fill="#E8F3FF"/>
+  <path d="M233.495 258.246C252.941 258.246 268.711 274.016 268.711 293.462C268.711 312.908 252.941 328.678 233.495 328.678C214.049 328.678 198.279 312.908 198.279 293.462C198.279 274.016 214.049 258.246 233.495 258.246Z" fill="#56ABFF"/>
+</svg>
--- a/client/public/manifest.webmanifest
+++ b/client/public/manifest.webmanifest
@@ -0,0 +1,40 @@
+{
+  "id": "/",
+  "name": "PrivateChat",
+  "short_name": "PrivateChat",
+  "description": "Private peer-to-peer chat with Angular, Fastify, and WebRTC.",
+  "lang": "en",
+  "dir": "ltr",
+  "start_url": "/",
+  "scope": "/",
+  "display": "standalone",
+  "orientation": "portrait",
+  "background_color": "#08111d",
+  "theme_color": "#08111d",
+  "icons": [
+    {
+      "src": "pwa-192x192.png",
+      "sizes": "192x192",
+      "type": "image/png",
+      "purpose": "any"
+    },
+    {
+      "src": "pwa-512x512.png",
+      "sizes": "512x512",
+      "type": "image/png",
+      "purpose": "any"
+    },
+    {
+      "src": "maskable-192x192.png",
+      "sizes": "192x192",
+      "type": "image/png",
+      "purpose": "maskable"
+    },
+    {
+      "src": "maskable-512x512.png",
+      "sizes": "512x512",
+      "type": "image/png",
+      "purpose": "maskable"
+    }
+  ]
+}
--- a/client/public/maskable-192x192.png
+++ b/client/public/maskable-192x192.png
--- a/client/public/maskable-512x512.png
+++ b/client/public/maskable-512x512.png
--- a/client/public/pwa-192x192.png
+++ b/client/public/pwa-192x192.png
--- a/client/public/pwa-512x512.png
+++ b/client/public/pwa-512x512.png
--- a/client/public/service-worker.js
+++ b/client/public/service-worker.js
@@ -0,0 +1,77 @@
+const APP_SHELL_CACHE = 'privatechat-app-shell-v1';
+const APP_SHELL_FILES = [
+  '/',
+  '/index.html',
+  '/manifest.webmanifest',
+  '/favicon.ico',
+  '/apple-touch-icon.png',
+  '/pwa-192x192.png',
+  '/pwa-512x512.png',
+  '/maskable-192x192.png',
+  '/maskable-512x512.png',
+];
+
+self.addEventListener('install', (event) => {
+  event.waitUntil(
+    caches.open(APP_SHELL_CACHE).then((cache) => cache.addAll(APP_SHELL_FILES)),
+  );
+  self.skipWaiting();
+});
+
+self.addEventListener('activate', (event) => {
+  event.waitUntil(
+    caches.keys().then((cacheNames) => Promise.all(
+      cacheNames
+        .filter((cacheName) => cacheName !== APP_SHELL_CACHE)
+        .map((cacheName) => caches.delete(cacheName)),
+    )),
+  );
+  self.clients.claim();
+});
+
+self.addEventListener('fetch', (event) => {
+  const { request } = event;
+
+  if (request.method !== 'GET') {
+    return;
+  }
+
+  const url = new URL(request.url);
+
+  if (url.origin !== self.location.origin || url.pathname.startsWith('/api/') || url.pathname === '/ws') {
+    return;
+  }
+
+  if (request.mode === 'navigate') {
+    event.respondWith(
+      fetch(request)
+        .then((response) => {
+          const responseCopy = response.clone();
+          void caches.open(APP_SHELL_CACHE).then((cache) => cache.put('/index.html', responseCopy));
+          return response;
+        })
+        .catch(async () => {
+          const cache = await caches.open(APP_SHELL_CACHE);
+          return cache.match('/index.html') || Response.error();
+        }),
+    );
+    return;
+  }
+
+  event.respondWith(
+    caches.match(request).then((cachedResponse) => {
+      const networkFetch = fetch(request)
+        .then((response) => {
+          if (response.ok) {
+            const responseCopy = response.clone();
+            void caches.open(APP_SHELL_CACHE).then((cache) => cache.put(request, responseCopy));
+          }
+
+          return response;
+        })
+        .catch(() => cachedResponse || Response.error());
+
+      return cachedResponse || networkFetch;
+    }),
+  );
+});
--- a/client/src/app/chat-page.component.html
+++ b/client/src/app/chat-page.component.html
@@ -7,6 +7,36 @@
        [title]="(peer()?.displayName ?? 'Peer') + ' webcam'"
        (closeRequested)="closeRemoteVideoModal()"
      ></app-peer-video-modal>
+      <audio #callAudioElement hidden autoplay playsinline></audio>
+
+      @if (incomingVoiceCallPeer(); as callingPeer) {
+        <div class="call-modal-backdrop">
+          <section class="panel p-4" style="width:min(100%,24rem)" (click)="$event.stopPropagation()">
+            <div class="mb-3">
+              <div>
+                <h2 class="h5 mb-1">Incoming voice call</h2>
+                <p class="small mb-0">{{ callingPeer.displayName }} is calling you.</p>
+              </div>
+            </div>
+            <div class="d-flex flex-wrap gap-2 justify-content-end">
+              <button
+                class="btn btn-success"
+                type="button"
+                (click)="acceptIncomingVoiceCall(callingPeer.id)"
+              >
+                Accept
+              </button>
+              <button
+                class="btn btn-outline-secondary"
+                type="button"
+                (click)="rejectIncomingVoiceCall(callingPeer.id)"
+              >
+                Reject
+              </button>
+            </div>
+          </section>
+        </div>
+      }

      <div class="chat-header d-flex flex-column flex-lg-row justify-content-between align-items-start align-items-lg-center gap-3 mb-4">
        <div>
@@ -49,7 +79,11 @@
            }

            @for (connectedPeer of session.peers(); track connectedPeer.id) {
-              <article class="peer-tile" [class.peer-tile-active]="connectedPeer.id === peerId()">
+              <article
+                class="peer-tile"
+                [class.peer-tile-active]="connectedPeer.id === peerId()"
+                [class.peer-tile-unread]="isPeerUnread(connectedPeer.id)"
+              >
                <button
                  class="peer-tile-main text-start"
                  type="button"
@@ -189,8 +223,28 @@
                      }
                    </div>
                  }
+                  @case ('voice') {
+                    <div class="voice-bubble">
+                      <div class="voice-bubble-label">Voice message</div>
+                      @if (entry.downloadUrl) {
+                        <audio
+                          class="voice-player"
+                          [src]="entry.downloadUrl"
+                          controls
+                          preload="metadata"
+                        ></audio>
+                      }
+                    </div>
+                  }
                  @default {
-                    <p class="mb-0">{{ entry.text }}</p>
+                    @if (entry.showSpinner) {
+                      <div class="bubble-system-status">
+                        <span class="bubble-spinner" aria-hidden="true"></span>
+                        <p class="mb-0">{{ entry.text }}</p>
+                      </div>
+                    } @else {
+                      <p class="mb-0">{{ entry.text }}</p>
+                    }
                  }
                }
              </article>
@@ -198,39 +252,6 @@
          </div>

          <div class="composer">
-            @if (peer(); as selectedPeer) {
-              <div class="composer-actions">
-                <button
-                  class="composer-camera"
-                  type="button"
-                  [disabled]="selectedPeer.channelState !== 'open' && !isStreamingCameraToSelectedPeer()"
-                  (click)="toggleCameraStream(selectedPeer.id)"
-                  [title]="isStreamingCameraToSelectedPeer() ? 'Stop webcam' : 'Start webcam'"
-                  [attr.aria-label]="isStreamingCameraToSelectedPeer() ? 'Stop webcam' : 'Start webcam'"
-                >
-                  {{ isStreamingCameraToSelectedPeer() ? '🛑' : '📹' }}
-                </button>
-
-              <input
-                #fileInput
-                class="composer-file-input"
-                type="file"
-                [disabled]="selectedPeer.channelState !== 'open'"
-                (change)="sendFile(selectedPeer.id, fileInput)"
-              />
-              <button
-                class="composer-plus"
-                type="button"
-                [disabled]="selectedPeer.channelState !== 'open'"
-                (click)="fileInput.click()"
-                title="Send file"
-                aria-label="Send file"
-              >
-                +
-              </button>
-              </div>
-            }
-
            <textarea
              #composerTextarea
              class="form-control composer-textarea"
@@ -245,7 +266,97 @@
              placeholder="Write a text message to your peer"
            ></textarea>

-            <div class="composer-send">
+            <div class="composer-toolbar">
+              @if (peer(); as selectedPeer) {
+                <button
+                  class="composer-call"
+                  type="button"
+                  [disabled]="!canStartSelectedVoiceCall()"
+                  (click)="startVoiceCall(selectedPeer.id)"
+                  title="Start voice call"
+                  aria-label="Start voice call"
+                >
+                  📞
+                </button>
+
+                @if (canEndSelectedVoiceCall()) {
+                  <button
+                    class="composer-hangup"
+                    type="button"
+                    (click)="endVoiceCall(selectedPeer.id)"
+                    title="End voice call"
+                    aria-label="End voice call"
+                  >
+                    🛑
+                  </button>
+                }
+
+                <button
+                  class="composer-camera"
+                  type="button"
+                  [disabled]="selectedPeer.channelState !== 'open' && !isStreamingCameraToSelectedPeer()"
+                  (click)="toggleCameraStream(selectedPeer.id)"
+                  [title]="isStreamingCameraToSelectedPeer() ? 'Stop webcam' : 'Start webcam'"
+                  [attr.aria-label]="isStreamingCameraToSelectedPeer() ? 'Stop webcam' : 'Start webcam'"
+                >
+                  {{ isStreamingCameraToSelectedPeer() ? '🛑' : '📹' }}
+                </button>
+
+                <button
+                  class="composer-voice"
+                  type="button"
+                  [disabled]="selectedPeer.channelState !== 'open' && !isRecordingVoice()"
+                  (click)="toggleVoiceRecording()"
+                  [title]="isRecordingVoice() ? 'Stop and send voice message' : 'Record voice message'"
+                  [attr.aria-label]="isRecordingVoice() ? 'Stop and send voice message' : 'Record voice message'"
+                  [class.composer-voice-recording]="isRecordingVoice()"
+                >
+                  {{ isRecordingVoice() ? '⏹️' : '🎙️' }}
+                </button>
+
+                <button
+                  class="composer-dictation"
+                  type="button"
+                  [disabled]="!session.isSelectedPeerReady() || session.signalingState() !== 'connected' || isTranscribingDictation()"
+                  (click)="toggleDictation(composerTextarea)"
+                  [title]="
+                    isDictating()
+                      ? 'Stop dictation and transcribe'
+                      : isTranscribingDictation()
+                        ? 'Transcribing dictated audio'
+                        : 'Start dictation'
+                  "
+                  [attr.aria-label]="
+                    isDictating()
+                      ? 'Stop dictation and transcribe'
+                      : isTranscribingDictation()
+                        ? 'Transcribing dictated audio'
+                        : 'Start dictation'
+                  "
+                  [class.composer-dictation-active]="isDictating() || isTranscribingDictation()"
+                >
+                  {{ isDictating() ? '🛑' : isTranscribingDictation() ? '⏳' : '🗣️' }}
+                </button>
+
+                <input
+                  #fileInput
+                  class="composer-file-input"
+                  type="file"
+                  [disabled]="selectedPeer.channelState !== 'open'"
+                  (change)="sendFile(selectedPeer.id, fileInput)"
+                />
+                <button
+                  class="composer-plus"
+                  type="button"
+                  [disabled]="selectedPeer.channelState !== 'open'"
+                  (click)="fileInput.click()"
+                  title="Send file"
+                  aria-label="Send file"
+                >
+                  +
+                </button>
+              }
+
              <button
                class="composer-image-generate"
                type="button"
--- a/client/src/app/chat-page.component.scss
+++ b/client/src/app/chat-page.component.scss
@@ -16,6 +16,23 @@
  box-shadow: 0 20px 60px var(--shadow-color);
 }

+.chat-page {
+  width: min(100%, 800px);
+  margin-inline: auto;
+  overflow-x: hidden;
+}
+
+.call-modal-backdrop {
+  position: fixed;
+  inset: 0;
+  z-index: 1250;
+  display: grid;
+  place-items: center;
+  padding: 1.5rem;
+  background: rgba(3, 8, 14, 0.52);
+  backdrop-filter: blur(8px);
+}
+
 .back-link {
  color: var(--link-color);
  text-decoration: none;
@@ -77,12 +94,12 @@

 .chat-layout {
  display: grid;
-  grid-template-columns: minmax(15rem, 19rem) minmax(0, 1fr);
-  gap: 1.25rem;
+  grid-template-columns: minmax(10rem, 13rem) minmax(0, 1fr);
+  gap:1.25rem;
 }

 .peer-sidebar {
-  padding: 1rem;
+  padding:1rem;
  border-radius: 1.3rem;
  border: 1px solid var(--surface-border-soft);
  background: var(--panel-soft-background);
@@ -151,6 +168,11 @@
  background: var(--surface-hover-background);
 }

+.peer-tile-unread {
+  border-color: #c62828;
+  box-shadow: inset 0 0 0 2px #c62828;
+}
+
 .peer-tile-row {
  display: flex;
  align-items: center;
@@ -289,32 +311,40 @@
  opacity: 0.7;
 }

-.bubble-author {
-  font-weight: 600;
-}
-
 .bubble-time {
  display: block;
 }

+.bubble-system-status {
+  display: inline-flex;
+  align-items: center;
+  gap: 0.7rem;
+}
+
+.bubble-spinner {
+  width: 1rem;
+  height: 1rem;
+  flex: 0 0 auto;
+  border: 0.15rem solid currentColor;
+  border-right-color: transparent;
+  border-radius: 999px;
+  opacity: 0.8;
+  animation: bubble-spin 700ms linear infinite;
+}
+
 .composer {
  display: grid;
-  grid-template-columns: auto minmax(0, 1fr) auto;
-  gap: 0.9rem;
-  align-items: end;
+  gap: 0.85rem;
  padding-top: 1rem;
  margin-top: 1rem;
  border-top: 1px solid var(--surface-border-soft);
 }

-.composer-actions {
-  display: grid;
-  gap: 0.6rem;
-}
-
-.composer-send {
-  display: grid;
+.composer-toolbar {
+  display: flex;
+  flex-wrap: wrap;
  gap: 0.6rem;
+  align-items: center;
 }

 .composer-emoji-picker-shell {
@@ -326,6 +356,10 @@
 }

 .composer-camera,
+.composer-call,
+.composer-dictation,
+.composer-hangup,
+.composer-voice,
 .composer-image-generate,
 .composer-emoji-trigger,
 .composer-plus,
@@ -349,26 +383,48 @@
  color: var(--placeholder-color);
 }

-.composer-camera {
+.composer-textarea {
+  min-height: 7rem;
+}
+
+.composer-call {
+  color: var(--page-text);
+  background: linear-gradient(135deg, #bfe9ff, #96c3ff);
+}
+
+.composer-camera,
+.composer-emoji-trigger,
+.composer-plus {
  color: var(--page-text);
  background: var(--badge-background);
 }

+.composer-dictation {
+  color: var(--page-text);
+  background: linear-gradient(135deg, #f6d8ff, #ffcadb);
+}
+
+.composer-dictation-active,
+.composer-hangup,
+.composer-voice-recording {
+  color: #fff;
+  background: linear-gradient(135deg, #ff7d63, #dc3e5d);
+}
+
+.composer-voice {
+  color: var(--page-text);
+  background: linear-gradient(135deg, #ffd8bf, #ff9b8a);
+}
+
+.composer-voice-recording {
+  box-shadow: 0 0 0 0.2rem rgba(220, 62, 93, 0.18);
+}
+
 .composer-image-generate {
  color: var(--page-text);
  background: linear-gradient(135deg, #ffe6b0, #ffc8a8);
 }

-.composer-emoji-trigger {
-  color: var(--page-text);
-  background: var(--badge-background);
-}
-
-.composer-plus {
-  color: var(--page-text);
-  background: var(--badge-background);
-}
-
 .send-emoji {
  background: linear-gradient(135deg, #def7dd, #9bd5ff);
 }
@@ -408,26 +464,36 @@
  background: var(--surface-hover-background);
 }

-.bubble-image {
-  width: 200px;
-  max-width: 100%;
-  height: auto;
-  border-radius: 1rem;
-  display: block;
+.bubble-author,
+.bubble-download,
+.voice-bubble-label {
+  font-weight: 600;
 }

+.bubble-image,
 .bubble-video {
  width: 200px;
  max-width: 100%;
  height: auto;
  display: block;
  border-radius: 1rem;
-  background: #000;
 }

-.bubble-download {
-  color: inherit;
-  font-weight: 600;
+.bubble-video {
+  background: #000;
+}
+.bubble-download { color: inherit; }
+
+.voice-bubble {
+  display: grid;
+  gap: 0.65rem;
+}
+
+.voice-bubble-label { font-size: 0.88rem; }
+
+.voice-player {
+  display: block;
+  width: min(100%, 18rem);
 }

 .bubble-json {
@@ -467,6 +533,12 @@
  }
 }

+@keyframes bubble-spin {
+  to {
+    transform: rotate(360deg);
+  }
+}
+
@media (max-width: 767.98px) {
  .chat-layout {
    grid-template-columns: 1fr;
@@ -479,4 +551,8 @@
  .bubble {
    max-width: 88%;
  }
+
+  .composer-toolbar {
+    justify-content: flex-start;
+  }
 }
--- a/client/src/app/chat-page.component.ts
+++ b/client/src/app/chat-page.component.ts
@@ -1,5 +1,5 @@
 import { CommonModule } from '@angular/common';
-import { Component, computed, effect, inject, signal } from '@angular/core';
+import { Component, computed, effect, ElementRef, inject, NgZone, OnDestroy, signal, ViewChild } from '@angular/core';
 import { toSignal } from '@angular/core/rxjs-interop';
 import { FormsModule } from '@angular/forms';
 import { ActivatedRoute, Router, RouterLink } from '@angular/router';
@@ -15,18 +15,41 @@ import type { ChatEntry, ConnectionState, PeerSummary } from './models';
  templateUrl: './chat-page.component.html',
  styleUrl: './chat-page.component.scss',
 })
-export class ChatPageComponent {
+export class ChatPageComponent implements OnDestroy {
  private readonly route = inject(ActivatedRoute);
  private readonly router = inject(Router);
+  private readonly ngZone = inject(NgZone);
  private readonly routeParamMap = toSignal(this.route.paramMap, {
    initialValue: this.route.snapshot.paramMap,
  });
  private composerSelectionStart = 0;
  private composerSelectionEnd = 0;
+  private voiceRecorder: MediaRecorder | null = null;
+  private voiceStream: MediaStream | null = null;
+  private voiceChunks: Blob[] = [];
+  private discardRecordedVoice = false;
+  private recordingPeerId: string | null = null;
+  private dictationRecorder: MediaRecorder | null = null;
+  private dictationStream: MediaStream | null = null;
+  private dictationChunks: Blob[] = [];
+  private dictationBaseText = '';
+  private discardRecordedDictation = false;
+  private dictationCompletionPromise: Promise<void> | null = null;
+  private resolveDictationCompletion: (() => void) | null = null;
+  private dictationApplyToken = 0;
+  @ViewChild('callAudioElement')
+  set callAudioElementRef(value: ElementRef<HTMLAudioElement> | undefined) {
+    this.callAudioElement = value;
+    this.syncCallAudioSource();
+  }
+  private callAudioElement?: ElementRef<HTMLAudioElement>;

  messageText = '';
  readonly forwardingEntryId = signal<string | null>(null);
  readonly emojiPickerOpen = signal(false);
+  readonly isRecordingVoice = signal(false);
+  readonly isDictating = signal(false);
+  readonly isTranscribingDictation = signal(false);
  readonly emojiOptions = [
    '😀', '😁', '😂', '🤣', '😊',
    '😉', '😍', '😘', '😎', '🤔',
@@ -40,15 +63,65 @@ export class ChatPageComponent {
  readonly peerId = computed(() => this.routeParamMap().get('peerId') ?? '');
  readonly peer = computed(() => this.session.peers().find((item) => item.id === this.peerId()) ?? null);
  readonly currentUser = computed(() => this.session.currentUser());
+  readonly incomingVoiceCallPeer = computed(() => {
+    const peerId = this.session.incomingVoiceCallPeerId();
+
+    return peerId ? this.session.peers().find((peer) => peer.id === peerId) ?? null : null;
+  });
  readonly conversation = computed(() =>
    this.session
      .messages()
      .filter((entry) => entry.peerId === this.peerId()),
  );
  readonly remoteVideoStream = computed(() => this.session.remoteVideoStreamForPeer(this.peerId()));
+  readonly remoteCallAudioStream = computed(() =>
+    this.session.remoteAudioStreamForPeer(this.session.activeVoiceCallPeerId() ?? ''),
+  );
  readonly remoteVideoModalVisible = computed(
    () => this.session.remoteVideoModalPeerId() === this.peerId() && !!this.remoteVideoStream(),
  );
+  readonly selectedPeerVoiceCallState = computed<'idle' | 'incoming' | 'outgoing' | 'active'>(() => {
+    const peerId = this.peerId();
+
+    if (!peerId) {
+      return 'idle';
+    }
+
+    if (this.session.activeVoiceCallPeerId() === peerId) {
+      return 'active';
+    }
+
+    if (this.session.outgoingVoiceCallPeerId() === peerId) {
+      return 'outgoing';
+    }
+
+    if (this.session.incomingVoiceCallPeerId() === peerId) {
+      return 'incoming';
+    }
+
+    return 'idle';
+  });
+  readonly canStartSelectedVoiceCall = computed(() => {
+    const selectedPeer = this.peer();
+
+    if (!selectedPeer || selectedPeer.channelState !== 'open') {
+      return false;
+    }
+
+    const activePeerId = this.session.activeVoiceCallPeerId();
+    const outgoingPeerId = this.session.outgoingVoiceCallPeerId();
+    const incomingPeerId = this.session.incomingVoiceCallPeerId();
+
+    return !activePeerId && !outgoingPeerId && !incomingPeerId;
+  });
+  readonly canEndSelectedVoiceCall = computed(() => {
+    const peerId = this.peerId();
+
+    return !!peerId && (
+      this.session.activeVoiceCallPeerId() === peerId ||
+      this.session.outgoingVoiceCallPeerId() === peerId
+    );
+  });
  readonly webRtcState = computed<ConnectionState>(() => {
    const selectedPeer = this.peer();

@@ -82,6 +155,17 @@ export class ChatPageComponent {
      this.session.selectPeer(peerId);
      void this.session.connectToPeer(peerId);
    });
+
+    effect(() => {
+      this.remoteCallAudioStream();
+      this.syncCallAudioSource();
+    });
+  }
+
+  ngOnDestroy(): void {
+    void this.stopDictation(true);
+    this.stopVoiceRecording(true);
+    this.detachCallAudioSource();
  }

  async ensureConnection(): Promise<void> {
@@ -102,6 +186,7 @@ export class ChatPageComponent {
      return;
    }

+    await this.stopDictation(false);
    await this.session.sendText(peerId, this.messageText);
    this.messageText = '';
    this.emojiPickerOpen.set(false);
@@ -116,7 +201,18 @@ export class ChatPageComponent {
      return;
    }

-    await this.session.requestGeneratedImage(peerId, this.messageText);
+    await this.stopDictation(false);
+    const requested = await this.session.requestGeneratedImage(peerId, this.messageText);
+
+    if (!requested) {
+      return;
+    }
+
+    this.messageText = '';
+    this.handleMessageTextChange('');
+    this.emojiPickerOpen.set(false);
+    this.composerSelectionStart = 0;
+    this.composerSelectionEnd = 0;
  }

  handleComposerEnter(event: Event): void {
@@ -180,6 +276,162 @@ export class ChatPageComponent {
    input.value = '';
  }

+  async toggleDictation(textarea: HTMLTextAreaElement): Promise<void> {
+    if (this.isDictating()) {
+      await this.stopDictation(false);
+      return;
+    }
+
+    if (this.isTranscribingDictation()) {
+      return;
+    }
+    const peerId = this.peerId();
+
+    if (!peerId) {
+      return;
+    }
+
+    if (typeof MediaRecorder === 'undefined' || typeof navigator === 'undefined') {
+      this.session.error.set('This browser does not support dictation recording.');
+      return;
+    }
+
+    if (typeof navigator.mediaDevices?.getUserMedia !== 'function') {
+      this.session.error.set('This browser cannot access the microphone for dictation.');
+      return;
+    }
+
+    this.dictationBaseText = this.messageText;
+    this.discardRecordedDictation = false;
+    this.dictationApplyToken += 1;
+
+    try {
+      const stream = await navigator.mediaDevices.getUserMedia({ audio: true });
+      const preferredMimeType = this.preferredVoiceMimeType();
+      const recorder = preferredMimeType
+        ? new MediaRecorder(stream, { mimeType: preferredMimeType })
+        : new MediaRecorder(stream);
+      const applyToken = this.dictationApplyToken;
+
+      this.dictationChunks = [];
+      this.dictationStream = stream;
+      this.dictationRecorder = recorder;
+      this.dictationCompletionPromise = new Promise<void>((resolve) => {
+        this.resolveDictationCompletion = resolve;
+      });
+
+      recorder.ondataavailable = (event) => {
+        if (event.data.size > 0) {
+          this.dictationChunks.push(event.data);
+        }
+      };
+
+      recorder.onerror = () => {
+        this.ngZone.run(() => {
+          this.session.error.set('Could not record dictation audio.');
+          this.cleanupDictationRecorder();
+          this.finishDictationCompletion();
+        });
+      };
+
+      recorder.onstop = () => {
+        const shouldDiscard = this.discardRecordedDictation;
+        const mimeType = recorder.mimeType || preferredMimeType || 'audio/webm';
+        const blob = new Blob(this.dictationChunks, { type: mimeType });
+
+        this.ngZone.run(() => {
+          this.cleanupDictationRecorder();
+
+          if (shouldDiscard || blob.size === 0) {
+            this.finishDictationCompletion();
+            return;
+          }
+
+          this.isTranscribingDictation.set(true);
+          void this.transcribeDictation(blob, textarea, applyToken);
+        });
+      };
+
+      recorder.start();
+      this.isDictating.set(true);
+      this.session.error.set(null);
+    } catch {
+      this.session.error.set('Could not start dictation recording.');
+      this.cleanupDictationRecorder();
+      this.finishDictationCompletion();
+    }
+  }
+
+  async toggleVoiceRecording(): Promise<void> {
+    if (this.isRecordingVoice()) {
+      this.stopVoiceRecording(false);
+      return;
+    }
+
+    const peerId = this.peerId();
+
+    if (!peerId) {
+      return;
+    }
+
+    if (typeof MediaRecorder === 'undefined' || typeof navigator === 'undefined') {
+      this.session.error.set('This browser does not support voice recording.');
+      return;
+    }
+
+    if (typeof navigator.mediaDevices?.getUserMedia !== 'function') {
+      this.session.error.set('This browser cannot access the microphone.');
+      return;
+    }
+
+    try {
+      const stream = await navigator.mediaDevices.getUserMedia({ audio: true });
+      const preferredMimeType = this.preferredVoiceMimeType();
+      const recorder = preferredMimeType
+        ? new MediaRecorder(stream, { mimeType: preferredMimeType })
+        : new MediaRecorder(stream);
+
+      this.voiceChunks = [];
+      this.discardRecordedVoice = false;
+      this.recordingPeerId = peerId;
+      this.voiceStream = stream;
+      this.voiceRecorder = recorder;
+
+      recorder.ondataavailable = (event) => {
+        if (event.data.size > 0) {
+          this.voiceChunks.push(event.data);
+        }
+      };
+
+      recorder.onerror = () => {
+        this.session.error.set('Could not record voice message.');
+        this.cleanupVoiceRecording();
+      };
+
+      recorder.onstop = () => {
+        const shouldDiscard = this.discardRecordedVoice;
+        const targetPeerId = this.recordingPeerId;
+        const mimeType = recorder.mimeType || preferredMimeType || 'audio/webm';
+        const blob = new Blob(this.voiceChunks, { type: mimeType });
+
+        this.cleanupVoiceRecording();
+
+        if (shouldDiscard || !targetPeerId || blob.size === 0) {
+          return;
+        }
+
+        void this.session.sendVoiceMessage(targetPeerId, blob, mimeType);
+      };
+
+      recorder.start();
+      this.isRecordingVoice.set(true);
+      this.session.error.set(null);
+    } catch {
+      this.session.error.set('Could not start microphone recording.');
+      this.cleanupVoiceRecording();
+    }
+  }
+
  async deleteMessage(entry: ChatEntry): Promise<void> {
    await this.session.deleteMessage(entry);
  }
@@ -231,6 +483,34 @@ export class ChatPageComponent {
    await this.session.startCameraStream(peerId);
  }

+  async startVoiceCall(peerId: string): Promise<void> {
+    await this.session.startVoiceCall(peerId);
+  }
+
+  async endVoiceCall(peerId: string): Promise<void> {
+    await this.session.endVoiceCall(peerId);
+  }
+
+  async acceptIncomingVoiceCall(peerId: string): Promise<void> {
+    if (!peerId) {
+      return;
+    }
+
+    if (peerId !== this.peerId()) {
+      await this.router.navigate(['/chat', peerId]);
+    }
+
+    await this.session.acceptVoiceCall(peerId);
+  }
+
+  rejectIncomingVoiceCall(peerId: string): void {
+    if (!peerId) {
+      return;
+    }
+
+    this.session.rejectVoiceCall(peerId);
+  }
+
  isImageEntry(entry: ChatEntry): boolean {
    return entry.kind === 'file' && !!entry.downloadUrl && (entry.fileMimeType?.startsWith('image/') ?? false);
  }
@@ -261,6 +541,10 @@ export class ChatPageComponent {
    return this.session.typingPeerIds().includes(peerId);
  }

+  isPeerUnread(peerId: string): boolean {
+    return this.session.unreadPeerIds().includes(peerId);
+  }
+
  indicatorTone(state: ConnectionState): 'ok' | 'connecting' | 'offline' {
    if (state === 'connected') {
      return 'ok';
@@ -298,9 +582,186 @@ export class ChatPageComponent {
      return;
    }

+    await this.stopDictation(true);
+    this.stopVoiceRecording(true);
    this.forwardingEntryId.set(null);
    this.emojiPickerOpen.set(false);
    this.session.selectPeer(peerId);
    await this.router.navigate(['/chat', peerId]);
  }
+
+  private stopVoiceRecording(discard: boolean): void {
+    const recorder = this.voiceRecorder;
+
+    if (!recorder) {
+      this.discardRecordedVoice = discard;
+      this.cleanupVoiceRecording();
+      return;
+    }
+
+    this.discardRecordedVoice = discard;
+
+    if (recorder.state !== 'inactive') {
+      recorder.stop();
+      return;
+    }
+
+    this.cleanupVoiceRecording();
+  }
+
+  private cleanupVoiceRecording(): void {
+    if (this.voiceStream) {
+      for (const track of this.voiceStream.getTracks()) {
+        track.stop();
+      }
+    }
+
+    this.voiceRecorder = null;
+    this.voiceStream = null;
+    this.voiceChunks = [];
+    this.recordingPeerId = null;
+    this.isRecordingVoice.set(false);
+  }
+
+  private preferredVoiceMimeType(): string {
+    if (typeof MediaRecorder === 'undefined' || typeof MediaRecorder.isTypeSupported !== 'function') {
+      return '';
+    }
+
+    const candidates = ['audio/webm;codecs=opus', 'audio/webm', 'audio/ogg;codecs=opus', 'audio/ogg'];
+
+    return candidates.find((candidate) => MediaRecorder.isTypeSupported(candidate)) ?? '';
+  }
+
+  private async stopDictation(discard: boolean): Promise<void> {
+    const completion = this.dictationCompletionPromise;
+
+    if (discard) {
+      this.dictationApplyToken += 1;
+      this.messageText = this.dictationBaseText || this.messageText;
+      this.handleMessageTextChange(this.messageText);
+      this.isTranscribingDictation.set(false);
+    } else {
+      this.dictationBaseText = this.messageText;
+    }
+
+    if (this.dictationRecorder) {
+      this.discardRecordedDictation = discard;
+
+      if (this.dictationRecorder.state !== 'inactive') {
+        this.dictationRecorder.stop();
+      } else {
+        this.cleanupDictationRecorder();
+        this.finishDictationCompletion();
+      }
+    } else if (!completion) {
+      this.dictationBaseText = '';
+    }
+
+    if (completion) {
+      await completion;
+    }
+  }
+
+  private cleanupDictationRecorder(): void {
+    if (this.dictationStream) {
+      for (const track of this.dictationStream.getTracks()) {
+        track.stop();
+      }
+    }
+
+    this.dictationRecorder = null;
+    this.dictationStream = null;
+    this.dictationChunks = [];
+    this.discardRecordedDictation = false;
+    this.isDictating.set(false);
+  }
+
+  private finishDictationCompletion(): void {
+    this.resolveDictationCompletion?.();
+    this.resolveDictationCompletion = null;
+    this.dictationCompletionPromise = null;
+    this.dictationBaseText = '';
+  }
+
+  private async transcribeDictation(blob: Blob, textarea: HTMLTextAreaElement, applyToken: number): Promise<void> {
+    try {
+      const transcript = await this.session.requestSpeechTranscription(blob);
+
+      if (applyToken !== this.dictationApplyToken) {
+        return;
+      }
+
+      this.applyDictatedText(this.mergeDictatedText(this.dictationBaseText, transcript), textarea);
+    } catch {
+      if (applyToken === this.dictationApplyToken) {
+        this.session.error.set('Dictation transcription failed.');
+      }
+    } finally {
+      if (applyToken === this.dictationApplyToken) {
+        this.isTranscribingDictation.set(false);
+      }
+
+      this.finishDictationCompletion();
+    }
+  }
+
+  private mergeDictatedText(baseText: string, transcript: string): string {
+    const trimmedTranscript = transcript.trim();
+
+    if (!trimmedTranscript) {
+      return baseText;
+    }
+
+    if (!baseText.trim()) {
+      return trimmedTranscript;
+    }
+
+    return `${baseText.trimEnd()} ${trimmedTranscript}`;
+  }
+
+  private applyDictatedText(text: string, textarea: HTMLTextAreaElement): void {
+    this.messageText = text;
+    textarea.value = text;
+    this.composerSelectionStart = text.length;
+    this.composerSelectionEnd = text.length;
+    this.handleMessageTextChange(text);
+
+    queueMicrotask(() => {
+      textarea.focus();
+      textarea.setSelectionRange(text.length, text.length);
+    });
+  }
+
+  private syncCallAudioSource(): void {
+    const audio = this.callAudioElement?.nativeElement;
+
+    if (!audio) {
+      return;
+    }
+
+    const stream = this.remoteCallAudioStream();
+
+    audio.srcObject = stream;
+
+    if (stream) {
+      void audio.play().catch(() => {
+        // Autoplay may wait for a browser gesture.
+      });
+      return;
+    }
+
+    audio.pause();
+  }
+
+  private detachCallAudioSource(): void {
+    const audio = this.callAudioElement?.nativeElement;
+
+    if (!audio) {
+      return;
+    }
+
+    audio.pause();
+    audio.srcObject = null;
+  }
 }
--- a/client/src/app/chat-session.service.ts
+++ b/client/src/app/chat-session.service.ts
--- a/client/src/app/home-page.component.scss
+++ b/client/src/app/home-page.component.scss
@@ -22,7 +22,8 @@
  border-radius: 2rem;
 }

-.panel {
+.panel,
+.session-card {
  border-radius: 1.5rem;
 }

@@ -77,15 +78,12 @@

 .theme-toggle:hover,
 .theme-toggle:focus-visible {
-  transform: translateY(-1px);
  border-color: color-mix(in srgb, var(--accent-color) 35%, var(--surface-border));
  background: var(--surface-hover-background);
+  transform: translateY(-1px);
 }

-.session-card {
-  min-width: min(100%, 18rem);
-  border-radius: 1.5rem;
-}
+.session-card { min-width: min(100%, 18rem); }

 .status-pill {
  display: inline-flex;
@@ -94,15 +92,19 @@
  background: var(--badge-background);
 }

-.btn-accent {
+.btn-accent,
+.btn-accent:hover,
+.btn-accent:focus-visible {
  color: #06111d;
  border: 0;
+}
+
+.btn-accent {
  background: var(--accent-gradient);
 }

 .btn-accent:hover,
 .btn-accent:focus-visible {
-  color: #06111d;
  background: var(--accent-gradient-hover);
 }

--- a/client/src/app/json-file-viewer.component.scss
+++ b/client/src/app/json-file-viewer.component.scss
@@ -5,8 +5,6 @@

 .json-viewer-shell {
  width: min(95%, 480px);
-  max-width: min(95%, 480px);
-  min-width: 0;
  overflow: hidden;
  border-radius: 0.9rem;
  background: rgba(255, 255, 255, 0.06);
--- a/client/src/app/models.ts
+++ b/client/src/app/models.ts
@@ -94,9 +94,10 @@ export interface ChatEntry {
  id: string;
  peerId: string;
  direction: 'incoming' | 'outgoing' | 'system';
-  kind: 'text' | 'json' | 'file' | 'system';
+  kind: 'text' | 'json' | 'file' | 'voice' | 'system';
  createdAt: number;
  authorLabel: string;
+  showSpinner?: boolean;
  text?: string;
  payload?: unknown;
  fileName?: string;
@@ -129,6 +130,16 @@ export type ServerEvent =
      peerId: string;
      message: string;
    }
+  | {
+      type: 'speech-transcribed';
+      requestId: string;
+      text: string;
+    }
+  | {
+      type: 'speech-transcription-error';
+      requestId: string;
+      message: string;
+    }
  | { type: 'pong' }
  | { type: 'error'; message: string };

@@ -155,6 +166,7 @@ export type DataEnvelope =
      name: string;
      mimeType: string;
      size: number;
+      attachmentKind?: 'file' | 'voice';
      authorId: string;
      authorName: string;
      sentAt: number;
@@ -170,4 +182,14 @@ export type DataEnvelope =
  | {
      type: 'camera-state';
      active: boolean;
+    }
+  | {
+      type: 'voice-call-offer';
+    }
+  | {
+      type: 'voice-call-response';
+      accepted: boolean;
+    }
+  | {
+      type: 'voice-call-ended';
    };
--- a/client/src/index.html
+++ b/client/src/index.html
@@ -5,6 +5,13 @@
  <title>PrivateChat</title>
  <base href="/">
  <meta name="viewport" content="width=device-width, initial-scale=1">
+  <meta name="theme-color" content="#08111d">
+  <meta name="mobile-web-app-capable" content="yes">
+  <meta name="apple-mobile-web-app-capable" content="yes">
+  <meta name="apple-mobile-web-app-status-bar-style" content="black-translucent">
+  <meta name="apple-mobile-web-app-title" content="PrivateChat">
+  <link rel="manifest" href="manifest.webmanifest">
+  <link rel="apple-touch-icon" href="apple-touch-icon.png">
  <link rel="preconnect" href="https://fonts.googleapis.com">
  <link rel="preconnect" href="https://fonts.gstatic.com" crossorigin>
  <link href="https://fonts.googleapis.com/css2?family=Space+Grotesk:wght@400;500;700&display=swap" rel="stylesheet">
--- a/client/src/main.ts
+++ b/client/src/main.ts
@@ -2,5 +2,13 @@ import { bootstrapApplication } from '@angular/platform-browser';
 import { appConfig } from './app/app.config';
 import { App } from './app/app';

+if (typeof window !== 'undefined' && 'serviceWorker' in navigator && window.isSecureContext) {
+  window.addEventListener('load', () => {
+    void navigator.serviceWorker.register('/service-worker.js').catch((error) => {
+      console.error('Service worker registration failed.', error);
+    });
+  });
+}
+
 bootstrapApplication(App, appConfig)
  .catch((err) => console.error(err));
--- a/client/src/styles.scss
+++ b/client/src/styles.scss
@@ -101,10 +101,6 @@
  color-scheme: dark;
 }

-:root[data-theme='light'] {
-  color-scheme: light;
-}
-
 html,
 body {
  min-height: 100dvh;
@@ -138,27 +134,30 @@ textarea {
  background: var(--badge-background) !important;
 }

-.btn-outline-light {
+.btn-outline-light,
+.btn-outline-light:hover,
+.btn-outline-light:focus-visible {
  color: var(--page-text);
  border-color: var(--surface-border);
 }

 .btn-outline-light:hover,
 .btn-outline-light:focus-visible {
-  color: var(--page-text);
-  border-color: var(--surface-border);
  background: var(--panel-soft-background);
 }

+.btn-outline-light,
+.btn-outline-secondary {
+  border-color: var(--surface-border);
+}
+
 .btn-outline-secondary {
  color: var(--page-text-muted);
-  border-color: var(--surface-border);
 }

 .btn-outline-secondary:hover,
 .btn-outline-secondary:focus-visible {
  color: var(--page-text);
-  border-color: var(--surface-border);
  background: var(--panel-soft-background);
 }

--- a/server/dist/index.js
+++ b/server/dist/index.js
@@ -13,6 +13,7 @@ import { generateAuthenticationOptions, generateRegistrationOptions, verifyAuthe
 import Fastify from 'fastify';
 import { Redis } from 'ioredis';
 import { z } from 'zod';
+import { SpeechTranscriber } from './speech-transcriber.js';
 dotenv.config({ path: fileURLToPath(new URL('../../.env', import.meta.url)) });
 const projectRootPath = fileURLToPath(new URL('../../', import.meta.url));
 const registerSchema = z.object({
@@ -43,6 +44,9 @@ const approvePendingUserParamsSchema = z.object({
 const adminDeleteUserParamsSchema = z.object({
    userId: z.string().min(1),
 });
+const webBundleFileParamsSchema = z.object({
+    '*': z.string().min(1),
+});
 const wsQuerySchema = z.object({
    token: z.string().min(1),
 });
@@ -78,6 +82,12 @@ const signalMessageSchema = z.discriminatedUnion('type', [
    z.object({
        type: z.literal('ping'),
    }),
+    z.object({
+        type: z.literal('speech-transcription'),
+        requestId: z.string().uuid(),
+        mimeType: z.string().trim().min(1).max(128),
+        audioBase64: z.string().min(1).max(32_000_000),
+    }),
 ]);
 const app = Fastify({ logger: true, trustProxy: true });
 const approvalAdminUsername = 'ladparis';
@@ -88,6 +98,9 @@ const frontendDistPath = resolveProjectPath(process.env.PRIVATECHAT_WEB_DIST_DIR
 const ollamaServerUrl = (process.env.PRIVATECHAT_OLLAMA_URL ?? 'http://192.168.1.19:11434').replace(/\/+$/, '');
 const ollamaImageModel = process.env.PRIVATECHAT_OLLAMA_IMAGE_MODEL ?? 'x/z-image-turbo:latest';
 const ollamaImageSize = process.env.PRIVATECHAT_OLLAMA_IMAGE_SIZE ?? '1024x1024';
+const speechTranscriptionServiceUrl = process.env.PRIVATECHAT_TRANSCRIPTION_WS_URL ?? 'ws://192.168.1.19:8080';
+const speechTranscriptionLanguage = process.env.PRIVATECHAT_TRANSCRIPTION_LANGUAGE ?? 'auto';
+const speechTranscriptionTimeoutMs = Number(process.env.PRIVATECHAT_TRANSCRIPTION_TIMEOUT_MS ?? 120_000);
 const sessionTtlSeconds = Number(process.env.SESSION_TTL_SECONDS ?? 60 * 60 * 12);
 const webAuthnChallengeTtlSeconds = Number(process.env.WEBAUTHN_CHALLENGE_TTL_SECONDS ?? 5 * 60);
 const allowedCorsOrigins = parseAllowedOrigins(process.env.CORS_ORIGIN);
@@ -98,6 +111,11 @@ const webAuthnRpName = process.env.WEBAUTHN_RP_NAME ?? 'PrivateChat';
 const webAuthnUserVerification = resolveWebAuthnUserVerification(process.env.WEBAUTHN_USER_VERIFICATION);
 const frontendIndexPath = path.join(frontendDistPath, 'index.html');
 const hasFrontendBuild = fs.existsSync(frontendIndexPath);
+const speechTranscriber = new SpeechTranscriber({
+    serviceUrl: speechTranscriptionServiceUrl,
+    language: speechTranscriptionLanguage,
+    requestTimeoutMs: speechTranscriptionTimeoutMs,
+}, app.log);
 fs.mkdirSync(path.dirname(sqlitePath), { recursive: true });
 fs.mkdirSync(path.dirname(masterKeyPath), { recursive: true });
 const encryptionKey = deriveEncryptionKey(loadOrCreateMasterKey(masterKeyPath));
@@ -240,6 +258,45 @@ else {
    app.log.warn({ frontendDistPath }, 'Angular frontend build not found. Build the client before serving it from the backend.');
 }
 app.get('/api/health', async () => ({ ok: true }));
+app.get('/api/web-app/manifest', async (request, reply) => {
+    const manifest = getFrontendBundleManifest();
+    if (!manifest) {
+        return reply.code(404).send({
+            message: 'Angular frontend build not found.',
+            frontendDistPath,
+        });
+    }
+    const etag = `"${manifest.bundleId}"`;
+    reply.header('Cache-Control', 'no-cache');
+    reply.header('ETag', etag);
+    if (requestMatchesEtag(request.headers['if-none-match'], etag)) {
+        return reply.code(304).send();
+    }
+    return manifest;
+});
+app.get('/api/web-app/files/*', async (request, reply) => {
+    const parsed = webBundleFileParamsSchema.safeParse(request.params);
+    if (!parsed.success) {
+        return reply.code(400).send({
+            message: 'Invalid web bundle asset path.',
+            issues: parsed.error.flatten(),
+        });
+    }
+    const asset = resolveFrontendBundleAsset(parsed.data['*']);
+    if (!asset) {
+        return reply.code(404).send({ message: 'Frontend asset not found.' });
+    }
+    const etag = `W/"${asset.etag}"`;
+    reply.header('Cache-Control', 'public, max-age=300');
+    reply.header('ETag', etag);
+    reply.header('Last-Modified', new Date(asset.lastModifiedMs).toUTCString());
+    if (requestMatchesEtag(request.headers['if-none-match'], etag)) {
+        return reply.code(304).send();
+    }
+    reply.header('Content-Length', String(asset.size));
+    reply.type(asset.contentType);
+    return reply.send(fs.createReadStream(asset.absolutePath));
+});
 app.post('/api/auth/register', async (request, reply) => {
    const parsed = registerSchema.safeParse(request.body);
    if (!parsed.success) {
@@ -677,6 +734,25 @@ async function handleSocketMessage(userId, sessionId, socket, rawMessage) {
        }
        return;
    }
+    if (parsed.type === 'speech-transcription') {
+        try {
+            const text = await transcribeAudioPayload(parsed.requestId, parsed.audioBase64, parsed.mimeType);
+            send(socket, {
+                type: 'speech-transcribed',
+                requestId: parsed.requestId,
+                text,
+            });
+        }
+        catch (error) {
+            app.log.warn({ err: error, userId }, 'Speech transcription failed');
+            send(socket, {
+                type: 'speech-transcription-error',
+                requestId: parsed.requestId,
+                message: error instanceof Error ? error.message : 'Speech transcription failed.',
+            });
+        }
+        return;
+    }
    let delivered = 0;
    const recipientSockets = socketsByUserId.get(parsed.to);
    if (recipientSockets) {
@@ -1053,12 +1129,23 @@ function parseClientMessage(rawMessage) {
            prompt: parsed.data.prompt,
        };
    }
+    if (parsed.data.type === 'speech-transcription') {
+        return {
+            type: 'speech-transcription',
+            requestId: parsed.data.requestId,
+            mimeType: parsed.data.mimeType,
+            audioBase64: parsed.data.audioBase64,
+        };
+    }
    return {
        type: 'signal',
        to: parsed.data.to,
        signal: normalizeSignal(parsed.data.signal),
    };
 }
+async function transcribeAudioPayload(requestId, audioBase64, mimeType) {
+    return await speechTranscriber.transcribe(requestId, audioBase64, mimeType);
+}
 async function generateImageFromPrompt(prompt) {
    const abortController = new AbortController();
    const timeoutId = setTimeout(() => abortController.abort(), 120_000);
@@ -1123,6 +1210,133 @@ function inferImageMimeType(imageBuffer) {
    }
    return 'application/octet-stream';
 }
+function getFrontendBundleManifest() {
+    if (!fs.existsSync(frontendIndexPath)) {
+        return null;
+    }
+    const files = listBundleFilePaths(frontendDistPath).map((absolutePath) => {
+        const relativePath = toBundleRelativePath(path.relative(frontendDistPath, absolutePath));
+        const stats = fs.statSync(absolutePath);
+        const sha256 = crypto.createHash('sha256').update(fs.readFileSync(absolutePath)).digest('hex');
+        return {
+            path: relativePath,
+            size: stats.size,
+            sha256,
+            lastModified: stats.mtime.toISOString(),
+            contentType: detectBundleContentType(relativePath),
+            href: bundleAssetHref(relativePath),
+        };
+    });
+    files.sort((left, right) => left.path.localeCompare(right.path));
+    const generatedAt = files.reduce((latest, file) => (file.lastModified > latest ? file.lastModified : latest), new Date(0).toISOString());
+    const bundleId = files.reduce((hash, file) => {
+        hash.update(file.path);
+        hash.update(file.sha256);
+        hash.update(String(file.size));
+        return hash;
+    }, crypto.createHash('sha256')).digest('hex');
+    return {
+        bundleId,
+        generatedAt,
+        indexPath: 'index.html',
+        files,
+    };
+}
+function resolveFrontendBundleAsset(relativeAssetPath) {
+    if (!fs.existsSync(frontendIndexPath) || !fs.existsSync(frontendDistPath)) {
+        return null;
+    }
+    const normalizedPath = toBundleRelativePath(relativeAssetPath);
+    if (normalizedPath.length === 0 ||
+        normalizedPath === '.' ||
+        normalizedPath.startsWith('../') ||
+        normalizedPath.startsWith('/')) {
+        return null;
+    }
+    const absolutePath = path.resolve(frontendDistPath, normalizedPath);
+    const relativeToRoot = path.relative(frontendDistPath, absolutePath);
+    if (relativeToRoot.startsWith('..') || path.isAbsolute(relativeToRoot) || !fs.existsSync(absolutePath)) {
+        return null;
+    }
+    const stats = fs.statSync(absolutePath);
+    if (!stats.isFile()) {
+        return null;
+    }
+    return {
+        absolutePath,
+        contentType: detectBundleContentType(normalizedPath),
+        size: stats.size,
+        lastModifiedMs: stats.mtimeMs,
+        etag: `${stats.size}-${Math.floor(stats.mtimeMs)}`,
+    };
+}
+function listBundleFilePaths(rootPath) {
+    return fs.readdirSync(rootPath, { withFileTypes: true }).flatMap((entry) => {
+        const entryPath = path.join(rootPath, entry.name);
+        if (entry.isDirectory()) {
+            return listBundleFilePaths(entryPath);
+        }
+        if (!entry.isFile()) {
+            return [];
+        }
+        return [entryPath];
+    });
+}
+function bundleAssetHref(relativePath) {
+    return `/api/web-app/files/${relativePath.split('/').map((segment) => encodeURIComponent(segment)).join('/')}`;
+}
+function toBundleRelativePath(inputPath) {
+    return path.posix.normalize(inputPath.replaceAll('\\', '/'));
+}
+function detectBundleContentType(assetPath) {
+    const extension = path.extname(assetPath).toLowerCase();
+    switch (extension) {
+        case '.mp3':
+            return 'audio/mpeg';
+        case '.m4a':
+            return 'audio/mp4';
+        case '.css':
+            return 'text/css; charset=utf-8';
+        case '.html':
+            return 'text/html; charset=utf-8';
+        case '.ico':
+            return 'image/x-icon';
+        case '.jpeg':
+        case '.jpg':
+            return 'image/jpeg';
+        case '.js':
+            return 'text/javascript; charset=utf-8';
+        case '.json':
+            return 'application/json; charset=utf-8';
+        case '.map':
+            return 'application/json; charset=utf-8';
+        case '.png':
+            return 'image/png';
+        case '.svg':
+            return 'image/svg+xml; charset=utf-8';
+        case '.txt':
+            return 'text/plain; charset=utf-8';
+        case '.webp':
+            return 'image/webp';
+        case '.webmanifest':
+            return 'application/manifest+json; charset=utf-8';
+        case '.woff':
+            return 'font/woff';
+        case '.woff2':
+            return 'font/woff2';
+        default:
+            return 'application/octet-stream';
+    }
+}
+function requestMatchesEtag(headerValue, etag) {
+    if (!headerValue) {
+        return false;
+    }
+    const incomingEtags = Array.isArray(headerValue)
+        ? headerValue
+        : headerValue.split(',').map((value) => value.trim());
+    return incomingEtags.includes(etag) || incomingEtags.includes('*');
+}
 function normalizeSignal(signal) {
    if (signal.type === 'sdp') {
        return {
--- a/server/dist/speech-transcriber.js
+++ b/server/dist/speech-transcriber.js
@@ -0,0 +1,124 @@
+import WebSocket from 'ws';
+export class SpeechTranscriber {
+    options;
+    logger;
+    constructor(options, logger) {
+        this.options = options;
+        this.logger = logger;
+    }
+    async transcribe(requestId, audioBase64, mimeType) {
+        const audio = this.normalizeAudioPayload(audioBase64, mimeType);
+        return await new Promise((resolve, reject) => {
+            let settled = false;
+            const socket = new WebSocket(this.options.serviceUrl);
+            const finish = (handler) => {
+                if (settled) {
+                    return;
+                }
+                settled = true;
+                clearTimeout(timeout);
+                socket.removeAllListeners();
+                if (socket.readyState === WebSocket.CONNECTING || socket.readyState === WebSocket.OPEN) {
+                    socket.close();
+                }
+                handler();
+            };
+            const timeout = setTimeout(() => {
+                finish(() => {
+                    reject(new Error(`The transcription service timed out after ${this.options.requestTimeoutMs}ms.`));
+                });
+            }, this.options.requestTimeoutMs);
+            socket.on('open', () => {
+                try {
+                    socket.send(JSON.stringify({
+                        type: 'transcribe',
+                        id: requestId,
+                        language: this.options.language,
+                        audio,
+                    }));
+                }
+                catch (error) {
+                    finish(() => {
+                        reject(error instanceof Error ? error : new Error('Could not send transcription request.'));
+                    });
+                }
+            });
+            socket.on('message', (payload) => {
+                const event = this.parseEvent(payload);
+                if (!event) {
+                    return;
+                }
+                if (event.id && event.id !== requestId) {
+                    this.logger.warn({ requestId, event }, 'Ignored transcription event for another request');
+                    return;
+                }
+                if (event.type === 'start') {
+                    this.logger.info({ requestId, model: event.model, language: event.language }, 'Speech transcription started');
+                    return;
+                }
+                if (event.type === 'delta') {
+                    return;
+                }
+                if (event.type === 'done') {
+                    finish(() => {
+                        resolve(event.text.trim());
+                    });
+                    return;
+                }
+                finish(() => {
+                    reject(new Error(event.message));
+                });
+            });
+            socket.on('error', (error) => {
+                finish(() => {
+                    reject(error instanceof Error ? error : new Error('The transcription service connection failed.'));
+                });
+            });
+            socket.on('close', (code, reasonBuffer) => {
+                if (settled) {
+                    return;
+                }
+                const reason = reasonBuffer.toString().trim();
+                const detail = reason
+                    ? `The transcription service closed the connection unexpectedly (code=${code}, reason=${reason}).`
+                    : `The transcription service closed the connection unexpectedly (code=${code}).`;
+                finish(() => {
+                    reject(new Error(detail));
+                });
+            });
+        });
+    }
+    normalizeAudioPayload(audioBase64, mimeType) {
+        const trimmedAudio = audioBase64.trim();
+        if (trimmedAudio.startsWith('data:')) {
+            return trimmedAudio;
+        }
+        const normalizedMimeType = mimeType.trim() || 'audio/webm';
+        return `data:${normalizedMimeType};base64,${trimmedAudio}`;
+    }
+    parseEvent(payload) {
+        const message = this.rawDataToString(payload).trim();
+        if (!message) {
+            return null;
+        }
+        try {
+            return JSON.parse(message);
+        }
+        catch {
+            this.logger.warn({ transcriptionPayload: message }, 'Ignored non-JSON transcription event');
+            return null;
+        }
+    }
+    rawDataToString(payload) {
+        if (typeof payload === 'string') {
+            return payload;
+        }
+        if (payload instanceof ArrayBuffer) {
+            return Buffer.from(payload).toString('utf8');
+        }
+        if (Array.isArray(payload)) {
+            return Buffer.concat(payload).toString('utf8');
+        }
+        return payload.toString('utf8');
+    }
+}
--- a/server/package-lock.json
+++ b/server/package-lock.json
@@ -16,6 +16,7 @@
        "dotenv": "^17.3.1",
        "fastify": "^5.8.2",
        "ioredis": "^5.10.0",
+        "ws": "^8.19.0",
        "zod": "^4.3.6"
      },
      "devDependencies": {
--- a/server/package.json
+++ b/server/package.json
@@ -17,6 +17,7 @@
    "dotenv": "^17.3.1",
    "fastify": "^5.8.2",
    "ioredis": "^5.10.0",
+    "ws": "^8.19.0",
    "zod": "^4.3.6"
  },
  "devDependencies": {
--- a/server/src/index.ts
+++ b/server/src/index.ts
@@ -23,6 +23,8 @@ import { Redis } from 'ioredis';
 import type WebSocket from 'ws';
 import { z } from 'zod';

+import { SpeechTranscriber } from './speech-transcriber.js';
+
 dotenv.config({ path: fileURLToPath(new URL('../../.env', import.meta.url)) });

 const projectRootPath = fileURLToPath(new URL('../../', import.meta.url));
@@ -120,6 +122,12 @@ type ClientMessage =
    }
  | {
      type: 'ping';
+    }
+  | {
+      type: 'speech-transcription';
+      requestId: string;
+      mimeType: string;
+      audioBase64: string;
    };

 type ServerMessage =
@@ -142,6 +150,16 @@ type ServerMessage =
      peerId: string;
      message: string;
    }
+  | {
+      type: 'speech-transcribed';
+      requestId: string;
+      text: string;
+    }
+  | {
+      type: 'speech-transcription-error';
+      requestId: string;
+      message: string;
+    }
  | { type: 'pong' }
  | { type: 'error'; message: string };

@@ -171,6 +189,22 @@ type AccessKeySummary = {
  createdAt: string;
 };

+type WebBundleFileEntry = {
+  path: string;
+  size: number;
+  sha256: string;
+  lastModified: string;
+  contentType: string;
+  href: string;
+};
+
+type WebBundleManifest = {
+  bundleId: string;
+  generatedAt: string;
+  indexPath: string;
+  files: WebBundleFileEntry[];
+};
+
 type PendingRegistration = {
  challenge: string;
  label: string;
@@ -233,6 +267,10 @@ const adminDeleteUserParamsSchema = z.object({
  userId: z.string().min(1),
 });

+const webBundleFileParamsSchema = z.object({
+  '*': z.string().min(1),
+});
+
 const wsQuerySchema = z.object({
  token: z.string().min(1),
 });
@@ -269,6 +307,12 @@ const signalMessageSchema = z.discriminatedUnion('type', [
  z.object({
    type: z.literal('ping'),
  }),
+  z.object({
+    type: z.literal('speech-transcription'),
+    requestId: z.string().uuid(),
+    mimeType: z.string().trim().min(1).max(128),
+    audioBase64: z.string().min(1).max(32_000_000),
+  }),
 ]);

 const app = Fastify({ logger: true, trustProxy: true });
@@ -287,6 +331,9 @@ const frontendDistPath = resolveProjectPath(
 const ollamaServerUrl = (process.env.PRIVATECHAT_OLLAMA_URL ?? 'http://192.168.1.19:11434').replace(/\/+$/, '');
 const ollamaImageModel = process.env.PRIVATECHAT_OLLAMA_IMAGE_MODEL ?? 'x/z-image-turbo:latest';
 const ollamaImageSize = process.env.PRIVATECHAT_OLLAMA_IMAGE_SIZE ?? '1024x1024';
+const speechTranscriptionServiceUrl = process.env.PRIVATECHAT_TRANSCRIPTION_WS_URL ?? 'ws://192.168.1.19:8080';
+const speechTranscriptionLanguage = process.env.PRIVATECHAT_TRANSCRIPTION_LANGUAGE ?? 'auto';
+const speechTranscriptionTimeoutMs = Number(process.env.PRIVATECHAT_TRANSCRIPTION_TIMEOUT_MS ?? 120_000);
 const sessionTtlSeconds = Number(process.env.SESSION_TTL_SECONDS ?? 60 * 60 * 12);
 const webAuthnChallengeTtlSeconds = Number(process.env.WEBAUTHN_CHALLENGE_TTL_SECONDS ?? 5 * 60);
 const allowedCorsOrigins = parseAllowedOrigins(process.env.CORS_ORIGIN);
@@ -300,6 +347,15 @@ const webAuthnUserVerification = resolveWebAuthnUserVerification(
 const frontendIndexPath = path.join(frontendDistPath, 'index.html');
 const hasFrontendBuild = fs.existsSync(frontendIndexPath);

+const speechTranscriber = new SpeechTranscriber(
+  {
+    serviceUrl: speechTranscriptionServiceUrl,
+    language: speechTranscriptionLanguage,
+    requestTimeoutMs: speechTranscriptionTimeoutMs,
+  },
+  app.log,
+);
+
 fs.mkdirSync(path.dirname(sqlitePath), { recursive: true });
 fs.mkdirSync(path.dirname(masterKeyPath), { recursive: true });

@@ -461,6 +517,57 @@ if (hasFrontendBuild) {

 app.get('/api/health', async () => ({ ok: true }));

+app.get('/api/web-app/manifest', async (request, reply) => {
+  const manifest = getFrontendBundleManifest();
+
+  if (!manifest) {
+    return reply.code(404).send({
+      message: 'Angular frontend build not found.',
+      frontendDistPath,
+    });
+  }
+
+  const etag = `"${manifest.bundleId}"`;
+  reply.header('Cache-Control', 'no-cache');
+  reply.header('ETag', etag);
+
+  if (requestMatchesEtag(request.headers['if-none-match'], etag)) {
+    return reply.code(304).send();
+  }
+
+  return manifest;
+});
+
+app.get('/api/web-app/files/*', async (request, reply) => {
+  const parsed = webBundleFileParamsSchema.safeParse(request.params);
+
+  if (!parsed.success) {
+    return reply.code(400).send({
+      message: 'Invalid web bundle asset path.',
+      issues: parsed.error.flatten(),
+    });
+  }
+
+  const asset = resolveFrontendBundleAsset(parsed.data['*']);
+
+  if (!asset) {
+    return reply.code(404).send({ message: 'Frontend asset not found.' });
+  }
+
+  const etag = `W/"${asset.etag}"`;
+  reply.header('Cache-Control', 'public, max-age=300');
+  reply.header('ETag', etag);
+  reply.header('Last-Modified', new Date(asset.lastModifiedMs).toUTCString());
+
+  if (requestMatchesEtag(request.headers['if-none-match'], etag)) {
+    return reply.code(304).send();
+  }
+
+  reply.header('Content-Length', String(asset.size));
+  reply.type(asset.contentType);
+  return reply.send(fs.createReadStream(asset.absolutePath));
+});
+
 app.post('/api/auth/register', async (request, reply) => {
  const parsed = registerSchema.safeParse(request.body);

@@ -1056,6 +1163,27 @@ async function handleSocketMessage(
    return;
  }

+  if (parsed.type === 'speech-transcription') {
+    try {
+      const text = await transcribeAudioPayload(parsed.requestId, parsed.audioBase64, parsed.mimeType);
+
+      send(socket, {
+        type: 'speech-transcribed',
+        requestId: parsed.requestId,
+        text,
+      });
+    } catch (error) {
+      app.log.warn({ err: error, userId }, 'Speech transcription failed');
+      send(socket, {
+        type: 'speech-transcription-error',
+        requestId: parsed.requestId,
+        message: error instanceof Error ? error.message : 'Speech transcription failed.',
+      });
+    }
+
+    return;
+  }
+
  let delivered = 0;
  const recipientSockets = socketsByUserId.get(parsed.to);

@@ -1597,6 +1725,15 @@ function parseClientMessage(rawMessage: string): ClientMessage | null {
    };
  }

+  if (parsed.data.type === 'speech-transcription') {
+    return {
+      type: 'speech-transcription',
+      requestId: parsed.data.requestId,
+      mimeType: parsed.data.mimeType,
+      audioBase64: parsed.data.audioBase64,
+    };
+  }
+
  return {
    type: 'signal',
    to: parsed.data.to,
@@ -1604,6 +1741,10 @@ function parseClientMessage(rawMessage: string): ClientMessage | null {
  };
 }

+async function transcribeAudioPayload(requestId: string, audioBase64: string, mimeType: string): Promise<string> {
+  return await speechTranscriber.transcribe(requestId, audioBase64, mimeType);
+}
+
 async function generateImageFromPrompt(prompt: string): Promise<{ imageBase64: string; mimeType: string }> {
  const abortController = new AbortController();
  const timeoutId = setTimeout(() => abortController.abort(), 120_000);
@@ -1685,6 +1826,169 @@ function inferImageMimeType(imageBuffer: Buffer): string {
  return 'application/octet-stream';
 }

+function getFrontendBundleManifest(): WebBundleManifest | null {
+  if (!fs.existsSync(frontendIndexPath)) {
+    return null;
+  }
+
+  const files = listBundleFilePaths(frontendDistPath).map((absolutePath) => {
+    const relativePath = toBundleRelativePath(path.relative(frontendDistPath, absolutePath));
+    const stats = fs.statSync(absolutePath);
+    const sha256 = crypto.createHash('sha256').update(fs.readFileSync(absolutePath)).digest('hex');
+
+    return {
+      path: relativePath,
+      size: stats.size,
+      sha256,
+      lastModified: stats.mtime.toISOString(),
+      contentType: detectBundleContentType(relativePath),
+      href: bundleAssetHref(relativePath),
+    } satisfies WebBundleFileEntry;
+  });
+
+  files.sort((left, right) => left.path.localeCompare(right.path));
+
+  const generatedAt = files.reduce(
+    (latest, file) => (file.lastModified > latest ? file.lastModified : latest),
+    new Date(0).toISOString(),
+  );
+  const bundleId = files.reduce((hash, file) => {
+    hash.update(file.path);
+    hash.update(file.sha256);
+    hash.update(String(file.size));
+    return hash;
+  }, crypto.createHash('sha256')).digest('hex');
+
+  return {
+    bundleId,
+    generatedAt,
+    indexPath: 'index.html',
+    files,
+  };
+}
+
+function resolveFrontendBundleAsset(relativeAssetPath: string): {
+  absolutePath: string;
+  contentType: string;
+  size: number;
+  lastModifiedMs: number;
+  etag: string;
+} | null {
+  if (!fs.existsSync(frontendIndexPath) || !fs.existsSync(frontendDistPath)) {
+    return null;
+  }
+
+  const normalizedPath = toBundleRelativePath(relativeAssetPath);
+
+  if (
+    normalizedPath.length === 0 ||
+    normalizedPath === '.' ||
+    normalizedPath.startsWith('../') ||
+    normalizedPath.startsWith('/')
+  ) {
+    return null;
+  }
+
+  const absolutePath = path.resolve(frontendDistPath, normalizedPath);
+  const relativeToRoot = path.relative(frontendDistPath, absolutePath);
+
+  if (relativeToRoot.startsWith('..') || path.isAbsolute(relativeToRoot) || !fs.existsSync(absolutePath)) {
+    return null;
+  }
+
+  const stats = fs.statSync(absolutePath);
+
+  if (!stats.isFile()) {
+    return null;
+  }
+
+  return {
+    absolutePath,
+    contentType: detectBundleContentType(normalizedPath),
+    size: stats.size,
+    lastModifiedMs: stats.mtimeMs,
+    etag: `${stats.size}-${Math.floor(stats.mtimeMs)}`,
+  };
+}
+
+function listBundleFilePaths(rootPath: string): string[] {
+  return fs.readdirSync(rootPath, { withFileTypes: true }).flatMap((entry) => {
+    const entryPath = path.join(rootPath, entry.name);
+
+    if (entry.isDirectory()) {
+      return listBundleFilePaths(entryPath);
+    }
+
+    if (!entry.isFile()) {
+      return [];
+    }
+
+    return [entryPath];
+  });
+}
+
+function bundleAssetHref(relativePath: string): string {
+  return `/api/web-app/files/${relativePath.split('/').map((segment) => encodeURIComponent(segment)).join('/')}`;
+}
+
+function toBundleRelativePath(inputPath: string): string {
+  return path.posix.normalize(inputPath.replaceAll('\\', '/'));
+}
+
+function detectBundleContentType(assetPath: string): string {
+  const extension = path.extname(assetPath).toLowerCase();
+
+  switch (extension) {
+    case '.mp3':
+      return 'audio/mpeg';
+    case '.m4a':
+      return 'audio/mp4';
+    case '.css':
+      return 'text/css; charset=utf-8';
+    case '.html':
+      return 'text/html; charset=utf-8';
+    case '.ico':
+      return 'image/x-icon';
+    case '.jpeg':
+    case '.jpg':
+      return 'image/jpeg';
+    case '.js':
+      return 'text/javascript; charset=utf-8';
+    case '.json':
+      return 'application/json; charset=utf-8';
+    case '.map':
+      return 'application/json; charset=utf-8';
+    case '.png':
+      return 'image/png';
+    case '.svg':
+      return 'image/svg+xml; charset=utf-8';
+    case '.txt':
+      return 'text/plain; charset=utf-8';
+    case '.webp':
+      return 'image/webp';
+    case '.webmanifest':
+      return 'application/manifest+json; charset=utf-8';
+    case '.woff':
+      return 'font/woff';
+    case '.woff2':
+      return 'font/woff2';
+    default:
+      return 'application/octet-stream';
+  }
+}
+
+function requestMatchesEtag(headerValue: string | string[] | undefined, etag: string): boolean {
+  if (!headerValue) {
+    return false;
+  }
+
+  const incomingEtags = Array.isArray(headerValue)
+    ? headerValue
+    : headerValue.split(',').map((value) => value.trim());
+
+  return incomingEtags.includes(etag) || incomingEtags.includes('*');
+}
+
 function normalizeSignal(signal: Extract<ClientMessage, { type: 'signal' }>['signal']): SignalPayload {
  if (signal.type === 'sdp') {
    return {
--- a/server/src/speech-transcriber.ts
+++ b/server/src/speech-transcriber.ts
@@ -0,0 +1,173 @@
+import WebSocket, { type RawData } from 'ws';
+
+type LoggerLike = {
+  info: (payload: unknown, message?: string) => void;
+  warn: (payload: unknown, message?: string) => void;
+  error: (payload: unknown, message?: string) => void;
+};
+
+type SpeechTranscriberOptions = {
+  serviceUrl: string;
+  language: string;
+  requestTimeoutMs: number;
+};
+
+type ServiceEvent =
+  | { type: 'start'; id: string; model: string; language: string }
+  | { type: 'delta'; id: string; text: string; fullText: string }
+  | { type: 'done'; id: string; text: string }
+  | { type: 'error'; id?: string; message: string };
+
+export class SpeechTranscriber {
+  constructor(
+    private readonly options: SpeechTranscriberOptions,
+    private readonly logger: LoggerLike,
+  ) {}
+
+  async transcribe(requestId: string, audioBase64: string, mimeType: string): Promise<string> {
+    const audio = this.normalizeAudioPayload(audioBase64, mimeType);
+
+    return await new Promise<string>((resolve, reject) => {
+      let settled = false;
+      const socket = new WebSocket(this.options.serviceUrl);
+
+      const finish = (handler: () => void): void => {
+        if (settled) {
+          return;
+        }
+
+        settled = true;
+        clearTimeout(timeout);
+        socket.removeAllListeners();
+
+        if (socket.readyState === WebSocket.CONNECTING || socket.readyState === WebSocket.OPEN) {
+          socket.close();
+        }
+
+        handler();
+      };
+
+      const timeout = setTimeout(() => {
+        finish(() => {
+          reject(new Error(`The transcription service timed out after ${this.options.requestTimeoutMs}ms.`));
+        });
+      }, this.options.requestTimeoutMs);
+
+      socket.on('open', () => {
+        try {
+          socket.send(
+            JSON.stringify({
+              type: 'transcribe',
+              id: requestId,
+              language: this.options.language,
+              audio,
+            }),
+          );
+        } catch (error) {
+          finish(() => {
+            reject(error instanceof Error ? error : new Error('Could not send transcription request.'));
+          });
+        }
+      });
+
+      socket.on('message', (payload) => {
+        const event = this.parseEvent(payload);
+
+        if (!event) {
+          return;
+        }
+
+        if (event.id && event.id !== requestId) {
+          this.logger.warn({ requestId, event }, 'Ignored transcription event for another request');
+          return;
+        }
+
+        if (event.type === 'start') {
+          this.logger.info(
+            { requestId, model: event.model, language: event.language },
+            'Speech transcription started',
+          );
+          return;
+        }
+
+        if (event.type === 'delta') {
+          return;
+        }
+
+        if (event.type === 'done') {
+          finish(() => {
+            resolve(event.text.trim());
+          });
+          return;
+        }
+
+        finish(() => {
+          reject(new Error(event.message));
+        });
+      });
+
+      socket.on('error', (error) => {
+        finish(() => {
+          reject(error instanceof Error ? error : new Error('The transcription service connection failed.'));
+        });
+      });
+
+      socket.on('close', (code, reasonBuffer) => {
+        if (settled) {
+          return;
+        }
+
+        const reason = reasonBuffer.toString().trim();
+        const detail = reason
+          ? `The transcription service closed the connection unexpectedly (code=${code}, reason=${reason}).`
+          : `The transcription service closed the connection unexpectedly (code=${code}).`;
+
+        finish(() => {
+          reject(new Error(detail));
+        });
+      });
+    });
+  }
+
+  private normalizeAudioPayload(audioBase64: string, mimeType: string): string {
+    const trimmedAudio = audioBase64.trim();
+
+    if (trimmedAudio.startsWith('data:')) {
+      return trimmedAudio;
+    }
+
+    const normalizedMimeType = mimeType.trim() || 'audio/webm';
+    return `data:${normalizedMimeType};base64,${trimmedAudio}`;
+  }
+
+  private parseEvent(payload: RawData): ServiceEvent | null {
+    const message = this.rawDataToString(payload).trim();
+
+    if (!message) {
+      return null;
+    }
+
+    try {
+      return JSON.parse(message) as ServiceEvent;
+    } catch {
+      this.logger.warn({ transcriptionPayload: message }, 'Ignored non-JSON transcription event');
+      return null;
+    }
+  }
+
+  private rawDataToString(payload: RawData): string {
+    if (typeof payload === 'string') {
+      return payload;
+    }
+
+    if (payload instanceof ArrayBuffer) {
+      return Buffer.from(payload).toString('utf8');
+    }
+
+    if (Array.isArray(payload)) {
+      return Buffer.concat(payload).toString('utf8');
+    }
+
+    return payload.toString('utf8');
+  }
+}
Author	SHA1	Message	Date
Laurent Dubertrand	f0e2b60f43	Transcription on mac	2026-03-11 03:08:27 +01:00
Laurent Dubertrand	0da98bfd96	quick cosmetic fix	2026-03-11 00:44:36 +01:00
Laurent Dubertrand	189f989c0d	Dictation through AI	2026-03-11 00:26:49 +01:00
Laurent Dubertrand	d2c4152ea7	Many features	2026-03-10 22:36:21 +01:00
Laurent Dubertrand	df309d088c	PWA	2026-03-10 04:13:32 +01:00
Laurent Dubertrand	506a824401	minor fixes and improvments	2026-03-10 03:27:11 +01:00