@chat21/chat21-web-widget 5.1.33 → 5.1.34-rc1

package/CHANGELOG.md

@@ -6,17 +6,122 @@
 ### **Copyrigth**: 
 *Tiledesk SRL*
 
-# 5.1.33
-- **bug fixed**: widget not loaded because blob block loading in lauch.js
 
-# 5.1.31
-- **bug fixed**: bug fix disabled user-typing with human and user-typing with human is not available
+# 5.1.34-rc1
+- **added**: onPageChangeVisibilityDesktop:'open' and 'onPageChangeVisibilityMobile: 'open' in chatbot-panel.html file
+
+# 5.1.33-rc12
+- **bug-fixed**: if last message is ulr_preview shows previous message buttons
+
+# 5.1.33-rc11
+- **changed**: **Stream audio** — updated the streaming/voice-mode footer icon.
+- **added**: **Stream audio** — tooltip on the stream button (i18n key `STREAM_AUDIO`, e.g. “Use voice mode”) via SVG `<title>` and `aria-label`.
+
+# 5.1.33-rc10
+- **bug-fixed**: fixed bug with knowledge base json sources without URLs
+
+# 5.1.33-rc9
+- **changed**: **Conversation footer** — accessibility-focused markup (ARIA roles/labels, live regions, semantic send control), stream-audio layout (wrapper + voice mode: hide attach/emoji while streaming, inline status, stream button + spectrum), optional **Close chat** action when `closeChatInConversation` is enabled; emoji restriction alert uses assertive live region semantics.
+- **bug-fixed**: **`getConversationDetail` Tiledesk fallback** — when `getMyRequests()` rejects, the conversation is treated as archived (`isConversationArchived = true`) and the handler returns immediately instead of resetting state from an empty fallback payload.
+- **bug-fixed**: **`VoiceService`** — skip Web Audio `createMediaStreamSource` when the `MediaStream` has no audio tracks (avoids `InvalidStateError` under mocked `getUserMedia` in unit tests and edge browsers).
+- **bug-fixed**: **`TtsAudioPlaybackCoordinator`** — avoid emitting a duplicate `isTTSPlaying$` `true` when preempting an already-playing owner (keeps emission sequence stable for consumers).
+- **bug-fixed**: **Unit tests** — aligned `BubbleMessageComponent` specs with `calcImageSize` + DI mocks; `AudioSyncComponent` spec module setup (`declarations` + `CommonModule`) and TTS streaming call expectations; `VoiceService` specs with realistic mic streams and updated expectations for proxy `listening` / `barge_in` (no `setAudioMuted` on those paths); `TtsAudioPlaybackCoordinator` `stopAll` test asserts `_stopAll$` broadcast via spy; `ConversationComponent` spec for Tiledesk error path.
+
+# 5.1.33-rc7
+- **added**: added more URL source types in kb_json_sources
+
+# 5.1.33-rc5
+- **added**: cssSource tiledeskSettings property to manage and override widget style 
+
+# 5.1.33-rc4
+- **bug fixed**: bug fixed extractUrlsFromText
+
+# 5.1.32-rc18
+- **bug fixed**: bug fix css kb_json_sources
+
+# 5.1.32-rc17
+- **bug fixed**: empty message in preview URLs
+
+# 5.1.32-rc16
+- **added**: added chat-json-sources to preview URLs
+
+# 5.1.32-rc15
+- **changed**: redemptionMs: 800
+
+# 5.1.32-rc14
+- **changed**: minor ui fixed
+
+# 5.1.32-rc13
+- **added**: VAD speech state events (`speechStart$`, `speechEnd$`) to improve UI/state transitions around user speech
+- **changed**: stream audio footer UI — stream button expands into a “Terminate” pill with animated level bars driven by mic intensity; recorder icon hidden while streaming; textarea width adjusted while streaming
+- **changed**: `StreamAudioSpectrum` — consolidated stream spectrum + stream button visuals into a single component with improved silence vs speaking handling and volume-driven bar heights
+- **changed**: conversation layout while streaming — adjusted received bubble sizing (`fullSizeMessage`) and loading spinner spacing (`fullSize`) for full-width stream mode
+
+
+
+# 5.1.32-rc12
+- **changed**: voice acquisition blocking during TTS response — pause VAD after user speech ends until the TTS response cycle completes; added safety timeout and `isAcquisitionBlocked$` to drive UI (e.g. greyed spectrum)
+- **chore**: version bump to `5.1.32-rc12`
+
+# 5.1.32-rc11
+- **added**: global TTS stop — `TtsAudioPlaybackCoordinator.stopAll()` + `stopAllPlayback$` to abort current and queued TTS playback and reveal full message text
+- **changed**: stop TTS playback when closing stream audio
+- **chore**: version bump to `5.1.32-rc11`
+
+# 5.1.32-rc10
+- **added**: TTS playback state (`isTTSPlaying$`) to coordinate voice UI and suppress mic segment emission while the bot is speaking
+- **changed**: stream spectrum theme color turns grey while TTS is playing
+
+
+# 5.1.32-rc9
+- **added**: mic-triggered TTS interruption — when VAD detects user speech, stop current TTS playback, clear the queue, and reveal the full message text
+- **added**: global TTS stop API (`TtsAudioPlaybackCoordinator.stopAll()` + `stopAllPlayback$`) to stop current + queued TTS playback from UI/events (e.g. close stream)
+- **changed**: `chat-audio-sync` TTS playback now streams audio via authenticated POST to `message.metadata.src`, sending `voiceSettings` + `text` and `streaming: true`
+- **changed**: stream UI spectrum — removed circular orb and stretched the spectrum line to fill the `#streamAudioAlert` width with 10px side padding
+- **changed**: conversation content layout while streaming — adjusted received bubble left margin and loading spinner margins for full-size mode
+
+
+# 5.1.32-rc8
+- **changed**: updated the dev environment defaults to align with the stage setup (remote config URL, API endpoints, logging level, storage prefix, and related settings)
+
+# 5.1.32-rc7
+- **added**: `StreamAudioSpectrum` component for audio visualization in the streaming footer UI
+- **added**: TTS playback coordinator queue — ensures TTS messages play sequentially without interrupting the previous one
+- **changed**: `chat-audio-sync` — updated TTS audio handling to support streaming playback and improved autoplay/animation timing
+- **changed**: iframe loader (`launch.js`, `launch_template.js`) — streamlined loading logic and improved error handling, with fixes for localhost environments
+
+# 5.1.32-rc4
+- **added**: “Close stream” control (`.close-stream-button`) — content and sheet bottom offset in fullscreen using `--chat-footer-stream-button-height` only while the stream is listening (`isStreamAudioActive`); variables in `_variables.scss`.
+- **added**: `VoiceService.discardCurrentRecordingSegment()` — when a message arrives from another sender during streaming, the current WebM segment is discarded (no upload) without stopping mic/VAD; `interruptStreamDueToPeerMessage()` in the footer no longer clears `isStreamAudioActive`.
+- **changed**: `#streamAudioAlert` — band above the footer with a frosted-glass look (`backdrop-filter`, semi-transparent `color-mix`).
+
+# 5.1.32-rc3
+- **changed**: `nginx.conf` (Docker image) — explicit MIME types for `.mjs`, `.wasm`, `.onnx` and `default_type` at `http` level (avoids `text/plain` on ONNX/VAD modules behind containerized deploys).
+- **chore**: removed deprecated Amazon beta/prod deploy scripts from the repository.
+
+# 5.1.32-rc2
+- **bug fixed**: minor streaming icon UI fixed
+- **changed**: Refactor stream audio button UI in the conversation footer (layout / classes).
+
+# 5.1.32-rc1
+- **added**: Voice pipeline — VAD (`@ricky0123/vad-web`) with ONNX Runtime WASM served from `/assets/onnx` (`copy-onnx-wasm`), `VoiceService` with `audioSegment$` (WebM segments) and optional STT/TTS via unified OpenAI provider using `HttpClient`, transcript / error fields on segment payloads.
+- **added**: Stream audio UI in conversation footer — toggle, real-time volume stream and animated waveform (`volume$`); mic session lifecycle wired to upload segments on speech end.
+- **added**: `MessageModel.isJustRecived` — set when ingesting messages (MQTT/Firebase `addCommandMessage` for `command.type === "message"`, and default for non-command messages in `addedMessage` / `addedNew`) to distinguish “new in session” vs history.
+- **added**: `chat-audio-sync` for TTS messages — karaoke-style word sync to audio, full `message` input, typography aligned with text bubbles; skips animation when `isJustRecived === false`; after playback ends sets `message.isJustRecived = false` so replays show full text without re-animating.
+- **bug fixed**: `AnalyserNode.getByteFrequencyData` TypeScript error — `Uint8Array` created from an explicit `ArrayBuffer` for correct DOM typings.
+- **bug fixed**: `isStreamAudioActive` no longer derived from per-frame mic level (`volume > 1`), which caused the stream button / active state to flash continuously while listening.
 
 # 5.1.30
 - **bug fixed**: startHidden is not working properly
 
-# 5.1.28
-- **bug fixed**: header option menu is deactivated on mobile
+# 5.1.30-rc3
+- **bug fixed**: bug fix user-typing with human is not available
+
+# 5.1.30-rc2
+- **bug fixed**: bug fix disabled user-typing with human
+
+# 5.1.30-rc1
+- **bug fixed**: startHidden is not working properly
 
 # 5.1.28
 - **bug fixed**: fixed Bot/Human conversation detection by correctly classifying bot replies
@@ -28,8 +133,22 @@
 - **changed**: Set the default autoStart value to false
 - **added**: Added the open widget loading spinner
 - **changed**: Load the widget without authentication and display the speech bubble
+
+# 5.1.27-rc3
+- **bug fixed**: fixed Bot/Human conversation detection by correctly classifying bot replies
+
+# 5.1.27-rc2
+- **bug fixed**: centralized fullscreen management on mobile and handled the case of the closed widget that remained fullscreen
+
+# 5.1.27-rc1
+- **added**: closeChatInConversation parameters
+- **added**: close chat button under textarea footer component
+
+# 5.1.26-rc6
 - **changed**: mobile always opens fullscreen and ignores legacy stored size”.
 - **changed**: changed user-typing 
+
+# 5.1.26-rc5
 - **changed**: Hide the resize-widget button when on mobile
 - **added**: added "I'm thinking" when the bot responds
 
@@ -91,6 +210,11 @@
 - **bug-fixed**: check showEmojiFooterButton to enable/disable emojii
 - **bug-fixed**: markdown is fired as an emojii and blocked by isEmojii check fn
 
+# 5.1.7-rc7
+- **bug-fixed**: button new_conversation always appear. added subscription to conversationAdded
+
+# 5.1.7-rc6
+- **added**: Added MAX_ATTACHMENT_ERROR error message when uploading a file larger than 10 MB
 
 # 5.1.7-rc5
 - **bug-fixed**: bug fixed BUTTON STYLES

package/Dockerfile

@@ -1,7 +1,7 @@
 ### STAGE 1: Build ###
 
 # We label our stage as ‘builder’
-FROM node:20.12.2-alpine3.19 as builder
+FROM --platform=$BUILDPLATFORM node:20.12.2-alpine3.19 as builder
 
 COPY package.json package-lock.json ./
 
@@ -15,12 +15,11 @@ COPY . .
 
 ## Build the angular app in production mode and store the artifacts in dist folder
 
-RUN npx ng build --configuration="prod" --output-path=dist --base-href=./ --output-hashing=none
 
+RUN npx ng build --configuration="prod" --output-path=dist --base-href=./ --output-hashing=none
 
 ### STAGE 2: Setup ###
-
-FROM nginx:1.14.1-alpine
+FROM --platform=$BUILDPLATFORM nginx:1.14.1-alpine
 
 ## Copy our default nginx config
 COPY nginx.conf /etc/nginx/nginx.conf
@@ -33,4 +32,4 @@ COPY --from=builder /ng-app/dist/browser /usr/share/nginx/html
 
 RUN echo "Chat21 Web Widget Started!!"
 
-CMD ["/bin/sh",  "-c",  "envsubst < /usr/share/nginx/html/widget-config-template.json > /usr/share/nginx/html/widget-config.json && exec nginx -g 'daemon off;'"]
+CMD ["/bin/sh", "-c", "envsubst < /usr/share/nginx/html/widget-config-template.json > /usr/share/nginx/html/widget-config.json && exec nginx -g 'daemon off;'"]

package/angular.json

@@ -44,7 +44,9 @@
               "src/environments/real_data/widget-config-docker.json",
               "src/environments/real_data/widget-config-native-mqtt.json",
               "src/environments/real_data/widget-config-native-prod.json",
-              "src/environments/real_data/widget-config-aws-stage.json"
+              "src/environments/real_data/widget-config-aws-stage.json",
+              "src/environments/real_data/widget-config-aws-aruba.json",
+              "src/environments/real_data/widget-config-regione-puglia.json"
             ],
             "styles": [
               "src/app/sass/styles.scss"
@@ -59,7 +61,9 @@
               "idb",
               "accept-language-parser",
               "file-saver",
-              "dayjs"
+              "dayjs",
+              "onnxruntime-web",
+              "@ricky0123/vad-web"
             ],
             "sourceMap": true,
             "optimization": false,
@@ -158,7 +162,19 @@
               "src/styles.scss"
             ],
             "scripts": [],
-            "codeCoverage": true
+            "codeCoverage": true,
+            "codeCoverageExclude": [
+              "src/chat21-core/providers/firebase/**/*.ts",
+              "src/chat21-core/providers/mqtt/**/*.ts",
+              "src/main.ts",
+              "src/polyfills.ts",
+              "src/zone-flag.ts",
+              "src/models/**/*.ts",
+              "src/chat21-core/models/**/*.ts",
+              "src/chat21-core/providers/native/**/*.ts",
+              "src/chat21-core/providers/logger/customLogger.ts",
+              "src/chat21-core/providers/logger/loggerInstance.ts"
+            ]
           }
         },
         "e2e": {
@@ -180,5 +196,9 @@
           "defaultConfiguration": "development"
         }
       }
-    }}
+    }
+  },
+  "cli": {
+    "analytics": false
+  }
 }

package/docs/changelog/this-branch.md

@@ -45,3 +45,39 @@ Questo branch migliora il feedback in conversazione e rende il comportamento del
 - Il callout non compare quando il widget e' aperto o quando la preview nuovo messaggio e' attiva.
 - La UI della conversazione indica chiaramente se l'ultimo responder e' bot o umano.
 - "Sto pensando..." compare solo nelle conversazioni bot e ha un comportamento prevedibile.
+# This branch: identificazione bot o umano
+
+## Obiettivo
+
+In questo branch e' stata introdotta una logica esplicita per capire, all'apertura della conversazione, se l'ultimo responder lato server e' un **bot** oppure un **umano**.
+
+## Come viene fatta l'identificazione
+
+- La valutazione parte dai messaggi gia' caricati in conversazione.
+- Viene cercato l'**ultimo messaggio ricevuto dal server** (non inviato dal client corrente).
+- Quel messaggio viene classificato con una funzione dedicata (`classifyMessageSenderKind`) che usa piu' segnali:
+  - `attributes.flowAttributes.chatbot_id` (quando presente indica bot)
+  - pattern del mittente (es. `senderId` con prefisso bot, quando applicabile)
+  - informazioni del mittente (`sender_fullname` e metadati associati)
+
+## Regola speciale per messaggi di sistema
+
+Se l'ultimo messaggio utile e' di tipo `system`, viene fatto un controllo aggiuntivo:
+
+- se in `attributes` e' presente un evento con `messagelabel.key = MEMBER_JOINED_GROUP`
+- e rappresenta il passaggio della conversazione a un operatore
+
+allora la conversazione viene forzata a **Umano** anche se altri indizi potrebbero suggerire bot.
+
+## Risultato in UI
+
+- In apertura conversazione viene mostrato un badge con stato:
+  - `Bot`
+  - `Umano`
+- Questo stato viene ricalcolato al variare dei messaggi ricevuti.
+
+## Effetto sui feedback utente
+
+- Il messaggio temporaneo `"sto pensando..."` viene mostrato solo quando la conversazione risulta di tipo **Bot**.
+- Alla ricezione della prima risposta dal server, `"sto pensando..."` viene nascosto **immediatamente**.
+- Non e' previsto alcun tempo minimo di visualizzazione del messaggio.

package/env.sample

@@ -19,5 +19,6 @@ API_URL=CHANGEIT
 API_BASEIMAGE_URL=CHANGEIT
 ENBED_JS=true
 
-
-
+# Speech proxy WebSocket base URL (ws:// or wss://).
+# Used by the voice streaming feature; maps to voiceProxyWsBaseUrl in widget-config.json.
+VOICE_PROXY_WS_BASE_URL=CHANGEIT

package/mocks/voice-websocket-mock/server.cjs

@@ -0,0 +1,245 @@
+/**
+ * Mock proxy voce WSS: salva l'audio in `temp/` e simula gli eventi del protocollo reale.
+ *
+ * Eventi JSON inviati al client (`msg.event`):
+ *   session_started | listening | transcript | thinking | speaking | done | error
+ * Frame binari: piccolo WAV silenzio (TTS simulato), dopo `speaking`.
+ *
+ * Uso: `npm run voice-mock`
+ * Env: VOICE_MOCK_PORT, VOICE_MOCK_PATH, VOICE_MOCK_SILENCE_MS (debounce fine parlato, default 700),
+ *      VOICE_MOCK_SEND_ERROR=1 (invia subito `error`)
+ */
+/* eslint-disable no-console */
+const fs = require('fs');
+const path = require('path');
+const http = require('http');
+const { WebSocketServer } = require('ws');
+
+const PORT = parseInt(process.env.VOICE_MOCK_PORT || '4587', 10);
+const PATH = process.env.VOICE_MOCK_PATH || '/ws/voice';
+const SILENCE_MS = parseInt(process.env.VOICE_MOCK_SILENCE_MS || '700', 10);
+const SEND_ERROR = process.env.VOICE_MOCK_SEND_ERROR === '1' || process.env.VOICE_MOCK_SEND_ERROR === 'true';
+
+const here = __dirname;
+const outDir = path.join(here, 'temp');
+const DEFAULT_TOKEN = 'mock-token-ok';
+const DEFAULT_PROJECT = 'mock-project';
+
+if (!fs.existsSync(outDir)) {
+  fs.mkdirSync(outDir, { recursive: true });
+}
+
+/** WAV PCM 16-bit mono, silenzio (decodificabile da Web Audio). */
+function buildSilenceWav(durationSec = 0.2) {
+  const sampleRate = 8000;
+  const bitsPerSample = 16;
+  const channels = 1;
+  const blockAlign = (channels * bitsPerSample) / 8;
+  const byteRate = sampleRate * blockAlign;
+  const numSamples = Math.floor(sampleRate * durationSec);
+  const dataSize = numSamples * blockAlign;
+  const buffer = Buffer.alloc(44 + dataSize);
+  buffer.write('RIFF', 0);
+  buffer.writeUInt32LE(36 + dataSize, 4);
+  buffer.write('WAVE', 8);
+  buffer.write('fmt ', 12);
+  buffer.writeUInt32LE(16, 16);
+  buffer.writeUInt16LE(1, 20);
+  buffer.writeUInt16LE(channels, 22);
+  buffer.writeUInt32LE(sampleRate, 24);
+  buffer.writeUInt32LE(byteRate, 28);
+  buffer.writeUInt16LE(blockAlign, 32);
+  buffer.writeUInt16LE(bitsPerSample, 34);
+  buffer.write('data', 36);
+  buffer.writeUInt32LE(dataSize, 40);
+  return buffer;
+}
+
+const TTS_MOCK_CHUNK = buildSilenceWav(0.25);
+
+function sendJson(ws, obj) {
+  if (ws.readyState !== 1) {
+    return;
+  }
+  try {
+    ws.send(JSON.stringify(obj));
+  } catch (e) {
+    console.error('[voice-mock] sendJson failed', e.message);
+  }
+}
+
+const server = http.createServer((_req, res) => {
+  res.writeHead(200, { 'Content-Type': 'text/plain' });
+  res.end(
+    'Tiledesk voice-websocket-mock. Events: session_started, listening, transcript, thinking, speaking, TTS binary, done, error.\n',
+  );
+});
+
+const wss = new WebSocketServer({ noServer: true });
+
+server.on('upgrade', (req, socket, head) => {
+  const host = (req.headers.host || 'localhost').split(':')[0];
+  const u = new URL(req.url, 'http://' + host);
+  if (u.pathname !== PATH) {
+    socket.destroy();
+    return;
+  }
+
+  wss.handleUpgrade(req, socket, head, (ws) => {
+    wss.emit('connection', ws, req, u);
+  });
+});
+
+wss.on('connection', (ws, _req, urlObj) => {
+  const q = urlObj.searchParams;
+  const token = q.get('token') || DEFAULT_TOKEN;
+  const projectId = q.get('projectId') || DEFAULT_PROJECT;
+  const qRequestId = q.get('requestId') || '';
+  const mime = q.get('mimeType') || 'application/octet-stream';
+  const ext = mime.indexOf('webm') >= 0 ? 'webm' : 'bin';
+
+  const convId =
+    qRequestId && qRequestId !== 'new'
+      ? qRequestId
+      : `mock-req-${Date.now()}-${Math.random().toString(36).slice(2, 8)}`;
+
+  const safeTs = new Date().toISOString().replace(/[:.]/g, '-');
+  const outFile = path.join(
+    outDir,
+    `voice-${safeTs}-p${String(projectId).replace(/[^a-zA-Z0-9._-]/g, '_')}-r${String(convId).replace(/[^a-zA-Z0-9._-]/g, '_')}.${ext}`,
+  );
+  const fileStream = fs.createWriteStream(outFile, { flags: 'a' });
+  const tokenOk = String(token).length > 0;
+
+  let chunkCount = 0;
+  let silenceTimer = null;
+  let finalPipelineStarted = false;
+
+  function clearSilenceTimer() {
+    if (silenceTimer) {
+      clearTimeout(silenceTimer);
+      silenceTimer = null;
+    }
+  }
+
+  /** Dopo debounce senza nuovi chunk audio: chiude il turno come farebbe il proxy reale. */
+  function runAfterUtteranceEnd() {
+    if (finalPipelineStarted || ws.readyState !== 1) {
+      return;
+    }
+    finalPipelineStarted = true;
+    clearSilenceTimer();
+
+    sendJson(ws, {
+      event: 'transcript',
+      text: '[mock] Trascrizione finale simulata.',
+      isFinal: true,
+    });
+
+    setTimeout(() => {
+      sendJson(ws, { event: 'thinking' });
+    }, 150);
+
+    setTimeout(() => {
+      sendJson(ws, { event: 'speaking' });
+      try {
+        if (ws.readyState === 1) {
+          ws.send(TTS_MOCK_CHUNK, { binary: true });
+        }
+      } catch (e) {
+        console.error('[voice-mock] TTS binary send failed', e.message);
+      }
+    }, 450);
+
+    setTimeout(() => {
+      sendJson(ws, {
+        event: 'done',
+        url: 'https://example.com/mock-voice-session-complete',
+      });
+    }, 900);
+  }
+
+  function scheduleUtteranceEndCheck() {
+    clearSilenceTimer();
+    silenceTimer = setTimeout(runAfterUtteranceEnd, SILENCE_MS);
+  }
+
+  console.log('[voice-mock] client connected', {
+    projectId,
+    requestId: convId,
+    mimeType: mime,
+    outFile,
+    token: tokenOk ? 'present' : 'missing',
+  });
+
+  if (!tokenOk) {
+    sendJson(ws, { event: 'error', message: 'Missing or invalid token' });
+    ws.close();
+    return;
+  }
+
+  if (SEND_ERROR) {
+    sendJson(ws, { event: 'error', message: 'VOICE_MOCK_SEND_ERROR is set' });
+    setTimeout(() => ws.close(), 100);
+    return;
+  }
+
+  sendJson(ws, { event: 'session_started', requestId: convId });
+
+  setTimeout(() => {
+    sendJson(ws, { event: 'listening' });
+  }, 80);
+
+  ws.on('message', (data, isBinary) => {
+    if (isBinary || Buffer.isBuffer(data)) {
+      const buf = Buffer.isBuffer(data) ? data : Buffer.from(data);
+      fileStream.write(buf);
+      chunkCount += 1;
+
+      if (chunkCount === 1) {
+        sendJson(ws, {
+          event: 'transcript',
+          text: '[mock] Utterance…',
+          isFinal: false,
+        });
+      } else if (chunkCount === 4) {
+        sendJson(ws, {
+          event: 'transcript',
+          text: '[mock] Utterance parziale aggiornata.',
+          isFinal: false,
+        });
+      }
+
+      finalPipelineStarted = false;
+      scheduleUtteranceEndCheck();
+      return;
+    }
+
+    const line = 'TEXT ' + (typeof data === 'string' ? data : data.toString()) + '\n';
+    fileStream.write(Buffer.from(line, 'utf8'));
+  });
+
+  ws.on('close', () => {
+    clearSilenceTimer();
+    fileStream.end();
+    const stat = fs.existsSync(outFile) ? fs.statSync(outFile) : null;
+    console.log('[voice-mock] client closed, bytes on disk:', stat ? stat.size : 0, outFile);
+  });
+
+  ws.on('error', (e) => {
+    console.error('[voice-mock] socket error', e);
+    clearSilenceTimer();
+    try {
+      fileStream.end();
+    } catch {
+      // ignore
+    }
+  });
+});
+
+server.listen(PORT, '0.0.0.0', () => {
+  console.log(
+    '[voice-mock] listening on ws://127.0.0.1:' + PORT + PATH + ' — output dir: ' + outDir,
+  );
+  console.log('[voice-mock] silence debounce:', SILENCE_MS, 'ms (fire transcript→thinking→speaking→TTS wav→done)');
+});

package/nginx.conf

@@ -5,18 +5,38 @@ events {
 }
 
 http {
+    include /etc/nginx/mime.types;
+    default_type application/octet-stream;
+
     server {
         listen 80;
         server_name  localhost;
 
         root   /usr/share/nginx/html;
         index  index.html index.htm;
-        include /etc/nginx/mime.types;
 
         gzip on;
         gzip_min_length 1000;
         gzip_proxied expired no-cache no-store private auth;
-        gzip_types text/plain text/css application/json application/javascript application/x-javascript text/xml application/xml application/xml+rss text/javascript;
+        gzip_types text/plain text/css application/json application/javascript application/x-javascript text/xml application/xml application/xml+rss text/javascript application/wasm;
+
+        # ONNX Runtime (.mjs), WASM e modelli VAD: il browser rifiuta moduli ES con Content-Type: text/plain
+        location ~* \.mjs$ {
+            root /usr/share/nginx/html;
+            default_type application/javascript;
+            charset utf-8;
+            add_header Cache-Control "public, max-age=31536000, immutable";
+        }
+        location ~* \.wasm$ {
+            root /usr/share/nginx/html;
+            default_type application/wasm;
+            add_header Cache-Control "public, max-age=31536000, immutable";
+        }
+        location ~* \.onnx$ {
+            root /usr/share/nginx/html;
+            default_type application/octet-stream;
+            add_header Cache-Control "public, max-age=31536000, immutable";
+        }
 
         location / {
             try_files $uri $uri/ /index.html;

package/package.json

@@ -1,7 +1,7 @@
 {
   "name": "@chat21/chat21-web-widget",
   "author": "Tiledesk SRL",
-  "version": "5.1.33",
+  "version": "5.1.34-rc1",
   "license": "MIT",
   "homepage": "https://www.tiledesk.com",
   "repository": {
@@ -10,12 +10,15 @@
   },
   "scripts": {
     "ng": "ng",
+    "copy-onnx-wasm": "mkdir -p src/assets/onnx && cp node_modules/onnxruntime-web/dist/ort-wasm-simd-threaded.mjs node_modules/onnxruntime-web/dist/ort-wasm-simd-threaded.wasm src/assets/onnx/",
     "start": "ng serve",
     "build": "ng build",
     "test": "ng test",
+    "test:e2e": "playwright test",
     "lint": "ng lint",
     "e2e": "ng e2e",
-    "source-map-explorer": "source-map-explorer dist/*.js"
+    "source-map-explorer": "source-map-explorer dist/*.js",
+    "voice-mock": "node mocks/voice-websocket-mock/server.cjs"
   },
   "private": false,
   "dependencies": {
@@ -32,6 +35,7 @@
     "@ctrl/ngx-emoji-mart": "^9.2.0",
     "@ngx-translate/core": "^16.0.4",
     "@ngx-translate/http-loader": "^16.0.1",
+    "@ricky0123/vad-web": "^0.0.30",
     "accept-language-parser": "^1.5.0",
     "bootstrap": "^5.1.3",
     "dayjs": "^1.11.7",
@@ -40,6 +44,7 @@
     "humanize-duration-ts": "^2.1.1",
     "marked": "^16.3.0",
     "ngx-logger": "^5.0.11",
+    "onnxruntime-web": "^1.24.3",
     "replace": "^1.2.2",
     "rxjs": "^7.8.2",
     "source-map-explorer": "^2.5.3",
@@ -52,6 +57,7 @@
     "@angular/cli": "^18.2.19",
     "@angular/compiler-cli": "^18.2.13",
     "@angular/language-service": "^18.2.13",
+    "@playwright/test": "^1.59.1",
     "@types/jasmine": "^3.6.11",
     "@types/jasminewd2": "~2.0.3",
     "@types/marked": "^6.0.0",
@@ -68,6 +74,7 @@
     "protractor": "~7.0.0",
     "ts-node": "~7.0.0",
     "tslint": "~6.1.0",
-    "typescript": "~5.4"
+    "typescript": "~5.4",
+    "ws": "^8.18.0"
   }
 }