Merge pull request #16 from R0m1k3/claude/refonte-app-vocale-v91uz7

fix: liaison « dégradé », réponses vides, transcriptions parasites, v…
fix: liaison « dégradé », réponses vides, transcriptions parasites, voix masculine (v2.4.0.1) (#15 )
2026-10-11 17:29:03 +02:00 · 2026-09-04 07:53:40 +02:00 · 2026-09-04 07:44:07 +02:00 · 2026-09-04 05:43:41 +00:00 · 2026-09-04 00:47:34 +02:00 · 2026-09-03 20:46:26 +02:00
48 changed files with 2435 additions and 92 deletions
@@ -1,7 +1,7 @@
 # EveFlow 2 — Interface vocale JARVIS pour Hermes Agent

 [![Build](https://img.shields.io/github/actions/workflow/status/R0m1k3/EveFlow/windows-release.yml?style=flat-square)](https://github.com/R0m1k3/EveFlow/actions)
-[![Version](https://img.shields.io/badge/version-2.1.0-brightgreen.svg?style=flat-square)](https://github.com/R0m1k3/EveFlow/releases)
+[![Version](https://img.shields.io/badge/version-2.4.0-brightgreen.svg?style=flat-square)](https://github.com/R0m1k3/EveFlow/releases)
 [![License](https://img.shields.io/badge/license-MIT-lightgrey.svg?style=flat-square)](LICENSE)

 **EveFlow** est un compagnon de bureau Windows qui transforme [Hermes Agent](https://hermes-agent.nousresearch.com/) en assistant vocal à la JARVIS : un noyau holographique réactif au son, une conversation en streaming, les outils, sous-agents, approbations, crons, skills et sessions d'Hermes pilotés depuis un seul HUD.
@@ -23,7 +23,16 @@ La version 2 est une réécriture complète : plus de robot 3D, un pipeline voca
 * **Détection d'activité vocale** (seuil adaptatif, sensibilité et silence de fin réglables) : l'enregistrement s'arrête tout seul quand vous avez fini de parler.
 * **Mains libres** : le micro se réactive après chaque réponse.
 * **Modèles intégrés, hors ligne** (sherpa-onnx dans un processus séparé) : reconnaissance Whisper (base, small, large-v3 turbo) ou SenseVoice, synthèse Kokoro v1.0 (voix française Siwis et voix anglaises) ou Piper (Siwis, Tom, UPMC). Les modèles se téléchargent depuis **Paramètres → Modèles locaux** et tournent sur le processeur.
-* **Mot d'activation** en mains libres : seules les phrases commençant par « Jarvis » (configurable) partent vers Hermes, le reste est ignoré ; un « Jarvis » seul ouvre une fenêtre d'écoute.
+* **Fin de phrase neuronale** : en écoute permanente, Silero VAD (0,6 Mo, sherpa-onnx) décide du début et de la fin de la commande à la place du seuil d'énergie ; moins de faux départs sur le bruit, coupure plus nette. Repli automatique sur le VAD énergétique si le modèle n'est pas installé.
+* **Vision d'écran** : « Jarvis, regarde mon écran » (ou le bouton de la barre de commande) joint une capture de l'écran principal à la question envoyée à Hermes.
+* **Actions locales instantanées** : « verrouille la session », « monte le son », « coupe le son », « piste suivante », « ouvre Spotify », « ouvre github.com »… exécutées sur le PC sans passer par Hermes, résultat lu à voix haute. Liste blanche d'actions dans le processus principal, désactivable dans les paramètres.
+* **Serveur MCP intégré** : Hermes se connecte à `http://<pc>:7842/mcp` et obtient les outils du PC (capture d'écran renvoyée en image, verrouillage, applications, URL, touches média, presse-papiers, recherche de fichiers, voix, notifications, état du HUD, affichage dans le fil). Même port et même secret que le webhook ; en mode chat completions, les mêmes outils sont proposés directement au modèle.
+* **Heures calmes et priorités** : plage horaire pendant laquelle les messages poussés s'affichent sans être lus ni faire clignoter le noyau (badge « non lus » à la place), thème nuit automatique, mots prioritaires lus quand même, résumé vocal des rapports longs (les premières phrases seulement).
+* **Mode mission** : un bouton dans la barre de commande bascule sur un second modèle Hermes (plus puissant) pour les tâches longues ; le modèle rapide reste utilisé pour la conversation courante.
+* **Widget compact « glanceable »** : état (veille, écoute, réflexion, parle), dernière phrase de l'assistant, badge de non-lus, indicateurs heures calmes et mission.
+* **Voix masculine ou féminine** : un réglage unique (Paramètres → Voix) appliqué à tous les moteurs. En local, Piper Tom ou Pierre (UPMC) pour le masculin, téléchargé automatiquement si aucune voix masculine n'est installée ; onyx / nova pour les API compatibles OpenAI ; Paul / Hortense pour les voix Windows.
+* **Barge-in** : en mains libres, parler par-dessus l'assistant coupe sa voix ; le seuil est relevé pendant qu'il parle pour ignorer l'écho du haut-parleur.
+* **Écoute permanente** : un détecteur de mot-clé de 3 Mo (sherpa-onnx, keyword spotting) tourne en continu sur le micro, quasi gratuit en CPU. « Jarvis » (ou n'importe quel mot-clé) ouvre l'écoute, « Jarvis, allume… » envoie directement la commande, et le mot coupe la voix en cours. Alternative : filtre du mot après transcription en mains libres.
 * **STT externe** : n'importe quelle API `/v1/audio/transcriptions` compatible OpenAI (Qwen3-ASR, Whisper, Speaches, faster-whisper-server, LocalAI, OpenAI). Repli sur la reconnaissance Chromium.
 * **TTS externe** : API `/v1/audio/speech` compatible OpenAI, voix système Windows ou Google Translate. Lecture phrase par phrase pendant le streaming, préchargement du segment suivant, coupure instantanée.
 * Raccourcis globaux : `Ctrl+Shift+Espace` (micro), `Ctrl+Shift+J` (afficher/masquer), `Ctrl+Shift+Échap` (couper la voix).
@@ -89,6 +98,20 @@ Pour recevoir les résultats de crons ou le miroir d'autres canaux dans EveFlow,

 ---

+### Donner à Hermes les outils du PC (MCP)
+
+Dans `~/.hermes/config.yaml` côté Hermes :
+
+```yaml
+mcp_servers:
+  eveflow:
+    url: "http://<ip-du-pc>:7842/mcp"
+    headers:
+      Authorization: "Bearer <secret du webhook EveFlow>"
+```
+
+Sans secret, EveFlow n'écoute qu'en local (`127.0.0.1`) ; définissez un secret dans Paramètres → Webhook pour un Hermes distant. Outils exposés : `capture_screen`, `lock_session`, `open_app`, `open_url`, `media_key`, `clipboard_get`, `clipboard_set`, `find_files`, `speak_text`, `notify_user`, `set_hud_state`, `get_app_status`, `get_conversation_history`, `show_message`.
+
 ## Développement

 ```bash
@@ -9,8 +9,16 @@
 | HUD arc-reactor réactif au son | Fait | Canvas 2D optimisé (pas d'ombres, couleurs en cache, 30 fps en veille, arrêt fenêtre masquée) |
 | Reconnaissance vocale locale | Fait | Whisper base/small/turbo via sherpa-onnx dans un processus utilitaire |
 | Synthèse vocale locale | Fait | Kokoro v1.0 (voix française Siwis) et Piper fr |
-| Mot d'activation | Fait, mode « après transcription » | Filtre « Jarvis … » en mains libres, tolérant aux erreurs de transcription |
-| Détection de fin de phrase | Fait | VAD énergétique adaptatif, pré-roll 400 ms |
+| Mot d'activation permanent | Fait (2.2.0) | Keyword spotting sherpa-onnx en continu ; mot-clé libre encodé en BPE ; validé sur audio réel (détection, zéro faux positif sur le test anglais) |
+| Mot d'activation après transcription | Fait | Filtre « Jarvis … » en mains libres, tolérant aux erreurs de transcription |
+| Détection de fin de phrase | Fait (2.3.0) | Silero VAD neuronal dans le worker (segment renvoyé au renderer), repli sur le VAD énergétique si le modèle manque |
+| Vision d'écran | Fait (2.3.0) | Capture `desktopCapturer` jointe à la requête Hermes (bouton, ou « regarde mon écran ») |
+| Serveur MCP local (outils du PC pour Hermes) | Fait (2.4.0) | `/mcp` sur le serveur webhook, JSON-RPC Streamable HTTP, 14 outils dont la capture d'écran renvoyée en image |
+| Heures calmes, priorités, résumé vocal | Fait (2.4.0) | Paramètres → Notifications ; thème nuit automatique ; badge non-lus |
+| Mode mission (second modèle) | Fait (2.4.0) | Bouton dans la barre de commande ; modèle dédié aux tâches longues |
+| Widget compact glanceable | Fait (2.4.0) | État, dernière phrase, non-lus, indicateurs |
+| Barge-in | Fait (2.4.0) | Coupe la voix dès que l'utilisateur parle ; seuil relevé pendant la synthèse (à valider avec l'annulation d'écho Windows) |
+| Actions système locales | Fait (2.3.0) | Verrouillage, volume et touches média, ouvrir une application ou une URL, presse-papiers, recherche de fichiers ; intentions courtes exécutées sans passer par Hermes |
 | Hermes : runs, sessions, chat completions | Fait | Transport choisi selon `/v1/capabilities` |
 | Approbations, steer, stop | Fait | Modales, injection de consigne en cours de run |
 | Crons, skills, toolsets, sessions | Fait | Panneau Hermes Ops |
@@ -21,35 +29,41 @@

 Sources : [jarvis-desktop-ai](https://github.com/ccarloshenri/jarvis-desktop-ai), [JarvisAi](https://github.com/PanPenek/JarvisAi), [bertrandmbanwi/Jarvis](https://github.com/bertrandmbanwi/Jarvis), [InterGenJLU/jarvis](https://github.com/InterGenJLU/jarvis), [livekit-wakeword](https://livekit.com/blog/livekit-wakeword), [sherpa-onnx keyword spotting](https://k2-fsa.github.io/sherpa/onnx/kws/index.html), [Hermes Agent features](https://hermes-agent.nousresearch.com/docs/user-guide/features/overview).

-1. **Mot d'activation permanent, quasi gratuit en CPU.** Les projets de référence utilisent openWakeWord (« hey jarvis ») ou un modèle de keyword spotting qui écoute en continu, au lieu de transcrire chaque phrase. sherpa-onnx fournit un modèle KWS anglais de 3,3 Mo qui accepte n'importe quel mot-clé sans réentraînement ; « JARVIS » s'encode `▁JA R VI S` avec son modèle BPE (vérifié). C'est le prochain chantier prioritaire : streaming du micro vers le worker, détection en continu, puis capture de la commande.
-2. **VAD neuronal (Silero) au lieu du seuil d'énergie.** Fin de phrase plus nette (environ 500 ms gagnés) et beaucoup moins de faux départs sur le bruit ambiant. Silero est déjà livré dans sherpa-onnx (`silero_vad.onnx`, 0,6 Mo).
+1. **Mot d'activation permanent, quasi gratuit en CPU.** Les projets de référence utilisent openWakeWord (« hey jarvis ») ou un modèle de keyword spotting qui écoute en continu, au lieu de transcrire chaque phrase. sherpa-onnx fournit un modèle KWS anglais de 3,3 Mo qui accepte n'importe quel mot-clé sans réentraînement ; « JARVIS » s'encode `▁JA R VI S` avec son modèle BPE (vérifié). Livré en 2.2.0 (voir l'étape 1 ci-dessous).
+2. **VAD neuronal (Silero) au lieu du seuil d'énergie.** Fin de phrase plus nette (environ 500 ms gagnés) et beaucoup moins de faux départs sur le bruit ambiant. Silero est déjà livré dans sherpa-onnx (`silero_vad.onnx`, 0,6 Mo). Livré en 2.3.0.
 3. **Latence perçue sous la seconde.** Les références visent 1 s entre la fin de parole et le premier mot prononcé : STT rapide, premier token en streaming, TTS phrase par phrase (déjà en place), et un modèle Hermes rapide pour la conversation courante.
-4. **Vision d'écran.** Capture d'écran à la demande (« Jarvis, qu'est-ce que je regarde ? ») envoyée à Hermes comme image, ou lecture d'une fenêtre. Hermes accepte déjà les images inline.
-5. **Actions système locales.** Ouvrir une application, régler le volume, verrouiller la session, chercher un fichier ; ce sont des outils EveFlow côté client à exposer à Hermes (mode chat completions) ou un petit serveur MCP local que Hermes appelle.
+4. **Vision d'écran.** Capture d'écran à la demande (« Jarvis, qu'est-ce que je regarde ? ») envoyée à Hermes comme image, ou lecture d'une fenêtre. Hermes accepte déjà les images inline. Livré en 2.3.0 (transport chat completions pour les images).
+5. **Actions système locales.** Ouvrir une application, régler le volume, verrouiller la session, chercher un fichier. Livré en 2.3.0 sous forme d'intentions courtes exécutées localement avant Hermes ; reste à exposer les mêmes actions à Hermes via un serveur MCP local.
 6. **Proactivité.** Notifications parlées à l'arrivée d'un cron, rappel, événement webhook, avec un résumé plutôt que la lecture intégrale ; c'est en partie fait via le webhook, à enrichir avec des règles (heures calmes, priorité).
 7. **Mémoire et personnalisation.** Hermes gère la mémoire longue durée (`X-Hermes-Session-Key`) ; côté EveFlow, un profil (nom, préférences de voix, style de réponse) déjà transmis dans les instructions.

 ## Plan proposé

-### Étape 1 (courte) : écoute permanente
- Streamer l'audio du micro (16 kHz, blocs de 128 ms) du renderer vers le worker via IPC.
- Dans le worker : `KeywordSpotter` sherpa-onnx (modèle gigaspeech 3,3 Mo, mot-clé configurable encodé automatiquement avec `bpe.model` via un petit encodeur BPE côté Node ou un dictionnaire pré-encodé pour « jarvis », « eve », « hey jarvis », « ok jarvis »).
- À la détection : chime, capture de la commande avec Silero VAD, transcription, envoi.
- Consommation attendue : quelques pourcents d'un cœur, pas de transcription en continu.
+### Étape 1 : écoute permanente — livrée en 2.2.0
+- Le renderer garde un seul flux micro (AudioWorklet 16 kHz) et envoie des blocs de 256 ms au processus principal, qui alimente le `KeywordSpotter` sherpa-onnx dans le worker.
+- Mots-clés encodés en BPE (table SentencePiece pour les mots courants, repli glouton sur le vocabulaire du modèle), sensibilité réglable (seuil 0,45 → 0,12).
+- À la détection : chime, capture de la commande sur le même flux (VAD énergétique, pré-roll 400 ms), transcription locale ou API, envoi à Hermes ; retour automatique à l'écoute.
+- Fin de phrase Silero livrée en 2.3.0 : le renderer envoie des trames de 128 ms au worker pendant la commande, le worker renvoie le segment WAV complet ; VAD énergétique en repli.

-### Étape 2 : vision et actions locales
- Outil `capture_screen` (Electron `desktopCapturer`) qui joint une capture à la requête Hermes.
- Outils système : `open_app`, `set_volume`, `lock_session`, `find_file`, `clipboard` ; exposés en chat completions et via un serveur MCP local pour les transports runs/sessions.
+### Étape 2 : vision et actions locales — livrée en 2.3.0
+- Capture d'écran (`desktopCapturer`, JPEG 1600 px) jointe à la requête Hermes : bouton dans la barre de commande, ou phrase « regarde mon écran… » à l'oral comme à l'écrit.
+- Actions locales (processus principal, liste blanche) : verrouiller, volume/mute/lecture/piste, ouvrir une application connue ou une URL http(s), presse-papiers, recherche de fichiers dans Documents/Bureau/Téléchargements/Images.
+- Routeur d'intentions FR/EN (`src/services/localCommands.ts`) exécuté avant l'envoi à Hermes ; désactivable dans Paramètres → Micro. Les phrases composées (« ouvre X puis… ») partent à Hermes.
+- Exposition à Hermes via le serveur MCP local livrée en 2.4.0.

-### Étape 3 : conversation plus naturelle
- Barge-in réel : couper la voix dès que l'utilisateur parle (déjà préparé, à valider avec l'annulation d'écho Windows).
- Réponses courtes à l'oral, détails à l'écran : instruction Hermes dédiée déjà en place, à affiner avec des consignes de format (« deux phrases à l'oral, détails en Markdown »).
- Modèle rapide pour le bavardage, modèle puissant pour les missions (choix par transport ou par mot-clé).
+### Étape 3 : conversation plus naturelle — livrée en 2.4.0
+- Barge-in : l'utilisateur qui parle par-dessus coupe la voix ; seuil d'énergie relevé pendant la synthèse pour ignorer l'écho (à valider sur Windows avec l'annulation d'écho du micro).
+- Mode mission : second modèle Hermes choisi d'un clic pour les tâches longues.
+- Serveur MCP local : Hermes enchaîne lui-même les actions du PC dans ses runs.

-### Étape 4 : présence
- Widget compact « glanceable » : dernière phrase, état, badge d'alertes.
- Heures calmes, priorité des notifications, résumé vocal des crons.
- Thèmes et voix par contexte (nuit, travail).
+### Étape 4 : présence — livrée en 2.4.0
+- Widget compact glanceable : état, dernière phrase, non-lus, indicateurs heures calmes / mission.
+- Heures calmes, mots prioritaires, résumé vocal des messages entrants, thème nuit.
+
+### Pistes suivantes
+- Voix différente par contexte (nuit, travail) et profils de réponse.
+- Mémoire locale des préférences transmise à Hermes (`X-Hermes-Session-Key` déjà en place).
+- Validation du barge-in et des touches média sur Windows réel (retours utilisateurs).

 ## Résultats du test de bout en bout (Linux, Xvfb, 4 cœurs lents)

@@ -60,5 +74,10 @@ Sources : [jarvis-desktop-ai](https://github.com/ccarloshenri/jarvis-desktop-ai)
 | Kokoro (fr) → Whisper base, phrase courte de 2 s | synthèse 2,9 s, transcription 2,5 s, texte approximatif |
 | Kokoro (fr) → Whisper base, phrase de 7 s | transcription correcte à un mot près |
 | Piper (fr) → Whisper base | transcription exacte |
+| Silero VAD sur phrase Kokoro de 2,3 s (2.3.0) | un seul segment, début et fin détectés, 6,7 s d'audio traités en 190 ms |
+| Capture d'écran → Hermes (2.3.0) | JPEG de 107 ko reçu côté Hermes (mock chat completions) |
+| Intention locale « coupe le son » (2.3.0) | traitée sans Hermes, résultat affiché dans le fil |
+| Correctifs 2.4.0.1 | réponse vide en chat completions désormais expliquée (JSON non streamé ou erreur HTTP 200), la liaison ne passe plus en « dégradé » quand seule l'API des crons échoue, transcriptions parasites (« (cliquant) », « *Claire* ») ignorées, préférence de voix masculine/féminine |
+| Serveur MCP (2.4.0) | initialize, tools/list (14 outils), tools/call côté principal (presse-papiers, capture image) et côté renderer (état, message dans le fil) |

 Sur un PC à 28 cœurs les temps sont nettement plus courts. Whisper small est maintenant recommandé pour le français.
@@ -0,0 +1,184 @@
+/**
+ * Local "JARVIS" actions: screenshot for vision, and an allow-list of system actions
+ * (lock, open an application or URL, media keys, clipboard, file search).
+ */
+import { app, clipboard, desktopCapturer, ipcMain, screen, shell } from 'electron';
+import { execFile } from 'node:child_process';
+import fs from 'node:fs';
+import os from 'node:os';
+import path from 'node:path';
+import { IPC } from '../../shared/ipc';
+import type { SystemAction, SystemActionResult } from '../../shared/bridge';
+import { log } from '../logger';
+
+const MEDIA_KEYS: Record<string, number> = {
+  'volume-up': 175,
+  'volume-down': 174,
+  mute: 173,
+  'play-pause': 179,
+  next: 176,
+  previous: 177
+};
+
+/** Applications the assistant may launch by name (Windows aliases + common Linux/macOS names). */
+const APP_ALIASES: Record<string, string[]> = {
+  'bloc-notes': ['notepad.exe', 'gedit', 'TextEdit'],
+  notepad: ['notepad.exe', 'gedit', 'TextEdit'],
+  calculatrice: ['calc.exe', 'gnome-calculator', 'Calculator'],
+  calc: ['calc.exe', 'gnome-calculator', 'Calculator'],
+  explorateur: ['explorer.exe', 'nautilus', 'Finder'],
+  explorer: ['explorer.exe', 'nautilus', 'Finder'],
+  terminal: ['wt.exe', 'cmd.exe', 'gnome-terminal', 'Terminal'],
+  cmd: ['cmd.exe'],
+  powershell: ['powershell.exe'],
+  paint: ['mspaint.exe'],
+  chrome: ['chrome', 'google-chrome', 'Google Chrome'],
+  edge: ['msedge', 'microsoft-edge', 'Microsoft Edge'],
+  firefox: ['firefox', 'Firefox'],
+  vscode: ['code', 'Visual Studio Code'],
+  code: ['code', 'Visual Studio Code'],
+  spotify: ['spotify', 'Spotify'],
+  discord: ['discord', 'Discord'],
+  steam: ['steam', 'Steam'],
+  word: ['winword', 'Microsoft Word'],
+  excel: ['excel', 'Microsoft Excel'],
+  outlook: ['outlook', 'Microsoft Outlook'],
+  teams: ['ms-teams', 'Microsoft Teams'],
+  'task manager': ['taskmgr.exe'],
+  'gestionnaire des tâches': ['taskmgr.exe'],
+  paramètres: ['ms-settings:'],
+  settings: ['ms-settings:']
+};
+
+function run(cmd: string, args: string[], timeoutMs = 8000): Promise<{ code: number; out: string }> {
+  return new Promise((resolve) => {
+    execFile(cmd, args, { timeout: timeoutMs, windowsHide: true }, (err, stdout, stderr) => {
+      resolve({ code: err ? 1 : 0, out: `${stdout}${stderr}`.trim() });
+    });
+  });
+}
+
+async function openApp(name: string): Promise<SystemActionResult> {
+  const key = name.trim().toLowerCase();
+  if (!key || key.length > 40) return { ok: false, message: 'Nom d’application invalide' };
+  const candidates = APP_ALIASES[key] ?? [key.replace(/[^a-z0-9 ._-]/gi, '')];
+  if (process.platform === 'win32') {
+    for (const candidate of candidates) {
+      if (candidate.endsWith(':')) {
+        await shell.openExternal(candidate);
+        return { ok: true, message: `${name} ouvert` };
+      }
+      // `start` resolves App Paths, PATH and Start Menu names.
+      const result = await run('cmd.exe', ['/c', 'start', '', candidate]);
+      if (result.code === 0) return { ok: true, message: `${name} lancé` };
+    }
+    return { ok: false, message: `Impossible de lancer ${name}` };
+  }
+  for (const candidate of candidates) {
+    const result = process.platform === 'darwin' ? await run('open', ['-a', candidate]) : await run('sh', ['-c', `command -v ${JSON.stringify(candidate)} >/dev/null && (nohup ${JSON.stringify(candidate)} >/dev/null 2>&1 &)`]);
+    if (result.code === 0) return { ok: true, message: `${name} lancé` };
+  }
+  return { ok: false, message: `Application introuvable : ${name}` };
+}
+
+async function lockSession(): Promise<SystemActionResult> {
+  if (process.platform === 'win32') {
+    const r = await run('rundll32.exe', ['user32.dll,LockWorkStation']);
+    return { ok: r.code === 0, message: r.code === 0 ? 'Session verrouillée' : r.out };
+  }
+  if (process.platform === 'darwin') {
+    const r = await run('osascript', ['-e', 'tell application "System Events" to keystroke "q" using {command down, control down}']);
+    return { ok: r.code === 0, message: r.out };
+  }
+  const r = await run('sh', ['-c', 'loginctl lock-session || xdg-screensaver lock || gnome-screensaver-command -l']);
+  return { ok: r.code === 0, message: r.code === 0 ? 'Session verrouillée' : r.out };
+}
+
+async function mediaKey(key: string): Promise<SystemActionResult> {
+  const code = MEDIA_KEYS[key];
+  if (!code) return { ok: false, message: 'Touche inconnue' };
+  if (process.platform === 'win32') {
+    const script = `$s=Add-Type -MemberDefinition '[DllImport("user32.dll")] public static extern void keybd_event(byte b,byte s,uint f,UIntPtr e);' -Name K -Namespace W -PassThru; $s::keybd_event(${code},0,0,[UIntPtr]::Zero); $s::keybd_event(${code},0,2,[UIntPtr]::Zero)`;
+    const r = await run('powershell.exe', ['-NoProfile', '-NonInteractive', '-Command', script]);
+    return { ok: r.code === 0, message: r.code === 0 ? key : r.out };
+  }
+  const xdo: Record<string, string> = { 'volume-up': 'XF86AudioRaiseVolume', 'volume-down': 'XF86AudioLowerVolume', mute: 'XF86AudioMute', 'play-pause': 'XF86AudioPlay', next: 'XF86AudioNext', previous: 'XF86AudioPrev' };
+  const r = await run('sh', ['-c', `xdotool key ${xdo[key]}`]);
+  return { ok: r.code === 0, message: r.code === 0 ? key : 'xdotool indisponible' };
+}
+
+async function findFiles(query: string): Promise<SystemActionResult> {
+  const needle = query.trim().toLowerCase();
+  if (needle.length < 2) return { ok: false, message: 'Recherche trop courte' };
+  const roots = ['Documents', 'Desktop', 'Downloads', 'Pictures'].map((d) => path.join(os.homedir(), d));
+  const hits: string[] = [];
+  const deadline = Date.now() + 4000;
+  const walk = async (dir: string, depth: number) => {
+    if (depth > 4 || hits.length >= 25 || Date.now() > deadline) return;
+    let entries: fs.Dirent[] = [];
+    try {
+      entries = await fs.promises.readdir(dir, { withFileTypes: true });
+    } catch {
+      return;
+    }
+    for (const entry of entries) {
+      if (entry.name.startsWith('.') || entry.name === 'node_modules') continue;
+      const full = path.join(dir, entry.name);
+      if (entry.name.toLowerCase().includes(needle)) hits.push(full);
+      if (entry.isDirectory()) await walk(full, depth + 1);
+      if (hits.length >= 25) return;
+    }
+  };
+  for (const root of roots) await walk(root, 0);
+  return { ok: true, message: `${hits.length} résultat(s)`, data: hits };
+}
+
+export async function captureScreen(maxWidth = 1600): Promise<string> {
+  const display = screen.getPrimaryDisplay();
+  const scale = Math.min(1, maxWidth / display.size.width);
+  const sources = await desktopCapturer.getSources({
+    types: ['screen'],
+    thumbnailSize: { width: Math.round(display.size.width * scale), height: Math.round(display.size.height * scale) }
+  });
+  const primary = sources.find((s) => s.display_id === String(display.id)) ?? sources[0];
+  if (!primary) throw new Error('Aucun écran capturable');
+  return primary.thumbnail.toJPEG(82).toString('base64').replace(/^/, 'data:image/jpeg;base64,');
+}
+
+export async function runSystemAction(action: SystemAction): Promise<SystemActionResult> {
+  switch (action.type) {
+    case 'lock':
+      return lockSession();
+    case 'open-app':
+      return openApp(String(action.name ?? ''));
+    case 'open-url': {
+      const url = String(action.url ?? '');
+      if (!/^https?:\/\//i.test(url) || url.length > 2048) return { ok: false, message: 'URL refusée' };
+      await shell.openExternal(url);
+      return { ok: true, message: 'Ouvert dans le navigateur' };
+    }
+    case 'media':
+      return mediaKey(String(action.key));
+    case 'clipboard-read':
+      return { ok: true, data: (await clipboard.readText()).slice(0, 20_000) };
+    case 'clipboard-write':
+      await clipboard.writeText(String(action.text ?? '').slice(0, 100_000));
+      return { ok: true, message: 'Copié dans le presse-papiers' };
+    case 'find-files':
+      return findFiles(String(action.query ?? ''));
+    default:
+      return { ok: false, message: 'Action inconnue' };
+  }
+}
+
+export function registerSystemIpc(): void {
+  ipcMain.handle(IPC.screenCapture, (_e, maxWidth?: number) => captureScreen(Number(maxWidth) || 1600));
+  ipcMain.handle(IPC.systemAction, async (_e, action: SystemAction) => {
+    if (!action || typeof action.type !== 'string') throw new Error('Action invalide');
+    log('INFO', 'system', `action ${action.type}`);
+    const result = await runSystemAction(action);
+    if (!result.ok) log('WARN', 'system', `action ${action.type} failed: ${result.message}`);
+    return result;
+  });
+  void app;
+}
@@ -9,6 +9,7 @@ import { getSharedDirectory, registerFilesIpc } from './ipc/files';
 import { createMainWindow, getMainWindow, getWindowMode, setWindowMode, toggleWindowVisibility, windowEvents } from './window';
 import { getWebhookStatus, startWebhookServer, stopWebhookServer } from './webhook';
 import { registerVoiceIpc } from './voice/ipc';
+import { registerSystemIpc } from './ipc/system';
 import { stopEngine } from './voice/engine';

 // Audio playback must never be blocked behind a user gesture (TTS starts on incoming events).
@@ -159,6 +160,7 @@ async function boot(): Promise<void> {
  registerFilesIpc();
  registerCoreIpc();
  registerVoiceIpc();
+  registerSystemIpc();
  createMainWindow();
  createTray();
  registerShortcuts();
@@ -0,0 +1,191 @@
+/**
+ * Minimal MCP server (Streamable HTTP, JSON-RPC 2.0) mounted on the webhook HTTP server at /mcp.
+ * Hermes Agent connects to it as a remote MCP server and gains the PC-side tools: screen capture,
+ * system actions, voice, notifications and HUD state. No SDK: initialize / tools/list / tools/call
+ * are the only methods a client needs, and every reply is a plain JSON body.
+ */
+import { ipcMain, type BrowserWindow } from 'electron';
+import type { IncomingMessage, ServerResponse } from 'node:http';
+import { IPC, type McpToolRequest, type McpToolResponse } from '../shared/ipc';
+import type { SystemAction } from '../shared/bridge';
+import { log } from './logger';
+import { captureScreen, runSystemAction } from './ipc/system';
+
+export const MCP_PATH = '/mcp';
+const PROTOCOL = '2025-03-26';
+const RENDERER_TIMEOUT_MS = 20_000;
+
+interface ToolSpec {
+  name: string;
+  description: string;
+  inputSchema: Record<string, unknown>;
+  /** 'main' = executed here; 'renderer' = forwarded to the UI process. */
+  where: 'main' | 'renderer';
+}
+
+const obj = (properties: Record<string, unknown>, required: string[] = []) => ({ type: 'object', properties, required, additionalProperties: false });
+
+export const MCP_TOOLS: ToolSpec[] = [
+  { name: 'capture_screen', description: "Capture l'écran principal de l'utilisateur et renvoie l'image (JPEG). À utiliser quand l'utilisateur parle de ce qu'il voit ou demande de l'aide sur son écran.", inputSchema: obj({ max_width: { type: 'integer', description: 'Largeur maximale en pixels (défaut 1600)' } }), where: 'main' },
+  { name: 'lock_session', description: "Verrouille la session Windows/Linux/macOS de l'utilisateur.", inputSchema: obj({}), where: 'main' },
+  { name: 'open_app', description: "Lance une application sur le PC de l'utilisateur (bloc-notes, calculatrice, chrome, spotify, vscode, terminal, explorateur…).", inputSchema: obj({ name: { type: 'string' } }, ['name']), where: 'main' },
+  { name: 'open_url', description: "Ouvre une URL http(s) dans le navigateur par défaut de l'utilisateur.", inputSchema: obj({ url: { type: 'string' } }, ['url']), where: 'main' },
+  { name: 'media_key', description: 'Envoie une touche média : volume-up, volume-down, mute, play-pause, next, previous.', inputSchema: obj({ key: { type: 'string', enum: ['volume-up', 'volume-down', 'mute', 'play-pause', 'next', 'previous'] } }, ['key']), where: 'main' },
+  { name: 'clipboard_get', description: 'Lit le texte du presse-papiers.', inputSchema: obj({}), where: 'main' },
+  { name: 'clipboard_set', description: 'Place un texte dans le presse-papiers.', inputSchema: obj({ text: { type: 'string' } }, ['text']), where: 'main' },
+  { name: 'find_files', description: "Cherche des fichiers par nom dans Documents, Bureau, Téléchargements et Images de l'utilisateur (25 résultats max).", inputSchema: obj({ query: { type: 'string' } }, ['query']), where: 'main' },
+  { name: 'speak_text', description: "Fait prononcer un texte par la voix d'EveFlow, immédiatement.", inputSchema: obj({ text: { type: 'string' } }, ['text']), where: 'renderer' },
+  { name: 'notify_user', description: 'Affiche une notification système (titre + corps).', inputSchema: obj({ title: { type: 'string' }, body: { type: 'string' } }, ['body']), where: 'renderer' },
+  { name: 'set_hud_state', description: 'Change momentanément l’état visuel du HUD : neutral, happy, thinking, alert, error.', inputSchema: obj({ state: { type: 'string', enum: ['neutral', 'happy', 'thinking', 'alert', 'error'] } }, ['state']), where: 'renderer' },
+  { name: 'get_app_status', description: "État d'EveFlow : transport, liaison, nom de l'assistant, mains libres, voix en cours, heures calmes.", inputSchema: obj({}), where: 'renderer' },
+  { name: 'get_conversation_history', description: 'Derniers messages de la conversation EveFlow (n ≤ 30).', inputSchema: obj({ n: { type: 'integer' } }), where: 'renderer' },
+  { name: 'show_message', description: "Affiche un message dans le fil EveFlow (sans le prononcer) — pour les rapports longs ou les résultats de crons.", inputSchema: obj({ text: { type: 'string' }, title: { type: 'string' } }, ['text']), where: 'renderer' }
+];
+
+type Rec = Record<string, unknown>;
+interface RpcRequest {
+  jsonrpc?: string;
+  id?: number | string | null;
+  method?: string;
+  params?: Rec;
+}
+
+const pending = new Map<string, { resolve: (r: McpToolResponse) => void; timer: ReturnType<typeof setTimeout> }>();
+let seq = 0;
+let ipcRegistered = false;
+
+function ensureIpc(): void {
+  if (ipcRegistered) return;
+  ipcRegistered = true;
+  ipcMain.on(IPC.mcpResponse, (_e, res: McpToolResponse) => {
+    if (!res || typeof res.id !== 'string') return;
+    const entry = pending.get(res.id);
+    if (!entry) return;
+    clearTimeout(entry.timer);
+    pending.delete(res.id);
+    entry.resolve(res);
+  });
+}
+
+function askRenderer(win: BrowserWindow | null, name: string, args: Rec): Promise<McpToolResponse> {
+  ensureIpc();
+  if (!win || win.isDestroyed()) return Promise.resolve({ id: '', ok: false, error: 'EveFlow window unavailable' });
+  const id = `mcp-${++seq}-${Date.now()}`;
+  const req: McpToolRequest = { id, name, args };
+  return new Promise((resolve) => {
+    const timer = setTimeout(() => {
+      pending.delete(id);
+      resolve({ id, ok: false, error: 'renderer timeout' });
+    }, RENDERER_TIMEOUT_MS);
+    pending.set(id, { resolve, timer });
+    win.webContents.send(IPC.mcpRequest, req);
+  });
+}
+
+function text(value: unknown): { content: Array<Rec>; isError?: boolean } {
+  return { content: [{ type: 'text', text: typeof value === 'string' ? value : JSON.stringify(value) }] };
+}
+
+function failure(message: string): { content: Array<Rec>; isError: boolean } {
+  return { content: [{ type: 'text', text: message }], isError: true };
+}
+
+async function callTool(win: BrowserWindow | null, name: string, args: Rec): Promise<{ content: Array<Rec>; isError?: boolean }> {
+  const spec = MCP_TOOLS.find((t) => t.name === name);
+  if (!spec) return failure(`Outil inconnu : ${name}`);
+  if (spec.where === 'renderer') {
+    const res = await askRenderer(win, name, args);
+    return res.ok ? text(res.result ?? { ok: true }) : failure(res.error ?? 'échec');
+  }
+  const sys = async (action: SystemAction) => {
+    const r = await runSystemAction(action);
+    return r.ok ? text(r.data !== undefined ? { ok: true, message: r.message, data: r.data } : { ok: true, message: r.message }) : failure(r.message ?? 'échec');
+  };
+  switch (name) {
+    case 'capture_screen': {
+      const dataUrl = await captureScreen(Number(args.max_width) || 1600);
+      const base64 = dataUrl.slice(dataUrl.indexOf(',') + 1);
+      return { content: [{ type: 'image', data: base64, mimeType: 'image/jpeg' }, { type: 'text', text: `Capture de l'écran principal (${Math.round(base64.length * 0.75 / 1024)} ko).` }] };
+    }
+    case 'lock_session':
+      return sys({ type: 'lock' });
+    case 'open_app':
+      return sys({ type: 'open-app', name: String(args.name ?? '') });
+    case 'open_url':
+      return sys({ type: 'open-url', url: String(args.url ?? '') });
+    case 'media_key':
+      return sys({ type: 'media', key: String(args.key ?? '') as 'mute' });
+    case 'clipboard_get':
+      return sys({ type: 'clipboard-read' });
+    case 'clipboard_set':
+      return sys({ type: 'clipboard-write', text: String(args.text ?? '') });
+    case 'find_files':
+      return sys({ type: 'find-files', query: String(args.query ?? '') });
+    default:
+      return failure(`Outil non implémenté : ${name}`);
+  }
+}
+
+/** Handle one JSON-RPC message; returns the response object (or null for notifications). */
+export async function handleMcpMessage(win: BrowserWindow | null, msg: RpcRequest): Promise<Rec | null> {
+  const id = msg.id ?? null;
+  const reply = (result: unknown) => ({ jsonrpc: '2.0', id, result });
+  const error = (code: number, message: string) => ({ jsonrpc: '2.0', id, error: { code, message } });
+  if (!msg.method) return error(-32600, 'Invalid Request');
+  if (msg.method.startsWith('notifications/')) return null;
+  switch (msg.method) {
+    case 'initialize':
+      return reply({ protocolVersion: PROTOCOL, capabilities: { tools: { listChanged: false } }, serverInfo: { name: 'eveflow', version: process.env.npm_package_version ?? '2.4.0' }, instructions: "Outils du PC de l'utilisateur via EveFlow : écran, applications, volume, presse-papiers, voix et notifications." });
+    case 'ping':
+      return reply({});
+    case 'tools/list':
+      return reply({ tools: MCP_TOOLS.map(({ name, description, inputSchema }) => ({ name, description, inputSchema })) });
+    case 'tools/call': {
+      const params = (msg.params ?? {}) as Rec;
+      const name = String(params.name ?? '');
+      const args = (params.arguments && typeof params.arguments === 'object' ? params.arguments : {}) as Rec;
+      log('INFO', 'mcp', `tools/call ${name}`);
+      try {
+        return reply(await callTool(win, name, args));
+      } catch (err) {
+        return reply(failure((err as Error).message));
+      }
+    }
+    case 'resources/list':
+      return reply({ resources: [] });
+    case 'prompts/list':
+      return reply({ prompts: [] });
+    default:
+      return error(-32601, `Method not found: ${msg.method}`);
+  }
+}
+
+/** HTTP entry point: POST /mcp with one JSON-RPC message or a batch. */
+export async function handleMcpHttp(win: BrowserWindow | null, req: IncomingMessage, res: ServerResponse, body: string): Promise<void> {
+  const send = (code: number, payload: unknown) => {
+    if (payload === undefined) {
+      res.writeHead(code, { 'Cache-Control': 'no-store' });
+      res.end();
+      return;
+    }
+    res.writeHead(code, { 'Content-Type': 'application/json', 'Cache-Control': 'no-store' });
+    res.end(JSON.stringify(payload));
+  };
+  if (req.method === 'GET') return send(405, { error: 'SSE stream not supported; use POST' });
+  if (req.method === 'DELETE') return send(204, undefined);
+  if (req.method !== 'POST') return send(405, { error: 'Method Not Allowed' });
+  let parsed: unknown;
+  try {
+    parsed = body.trim() ? JSON.parse(body) : {};
+  } catch {
+    return send(400, { jsonrpc: '2.0', id: null, error: { code: -32700, message: 'Parse error' } });
+  }
+  const messages = Array.isArray(parsed) ? (parsed as RpcRequest[]) : [parsed as RpcRequest];
+  const replies: Rec[] = [];
+  for (const msg of messages) {
+    const out = await handleMcpMessage(win, msg && typeof msg === 'object' ? msg : {});
+    if (out) replies.push(out);
+  }
+  if (replies.length === 0) return send(202, undefined);
+  return send(200, Array.isArray(parsed) ? replies : replies[0]);
+}
@@ -13,8 +13,9 @@ import {
  type WebhookStatus,
  type WindowMode
 } from '../shared/ipc';
-import type { EveFlowBridge, Unsubscribe } from '../shared/bridge';
-import { VOICE_IPC, type SynthesizeRequest, type SynthesizeResult, type TranscribeRequest, type TranscribeResult, type VoiceDownloadProgress, type VoiceEngineStatus, type VoiceModelStatus } from '../shared/voice';
+import type { EveFlowBridge, SystemAction, SystemActionResult, Unsubscribe } from '../shared/bridge';
+import type { McpToolRequest, McpToolResponse } from '../shared/ipc';
+import { VOICE_IPC, type KwsDetection, type KwsStartRequest, type VadEvent, type VadStartRequest, type SynthesizeRequest, type SynthesizeResult, type TranscribeRequest, type TranscribeResult, type VoiceDownloadProgress, type VoiceEngineStatus, type VoiceModelStatus } from '../shared/voice';

 function subscribe<T>(channel: string, callback: (payload: T) => void): Unsubscribe {
  const listener = (_event: Electron.IpcRendererEvent, payload: T) => callback(payload);
@@ -43,7 +44,9 @@ const api: EveFlowBridge = {
  },
  system: {
    metrics: () => ipcRenderer.invoke(IPC.metrics) as Promise<SystemMetrics>,
-    appInfo: () => ipcRenderer.invoke(IPC.appInfo) as Promise<AppInfo>
+    appInfo: () => ipcRenderer.invoke(IPC.appInfo) as Promise<AppInfo>,
+    captureScreen: (maxWidth?: number) => ipcRenderer.invoke(IPC.screenCapture, maxWidth) as Promise<string>,
+    action: (action: SystemAction) => ipcRenderer.invoke(IPC.systemAction, action) as Promise<SystemActionResult>
  },
  files: {
    readLocal: (filePath: string) => ipcRenderer.invoke(IPC.readLocalFile, filePath) as Promise<string>,
@@ -55,6 +58,8 @@ const api: EveFlowBridge = {
  hermes: {
    onPush: (cb: (event: HermesPushEvent) => void) => subscribe<HermesPushEvent>(IPC.hermesPush, cb),
    webhookStatus: () => ipcRenderer.invoke(IPC.webhookStatus) as Promise<WebhookStatus>,
+    onToolRequest: (cb: (req: McpToolRequest) => void) => subscribe<McpToolRequest>(IPC.mcpRequest, cb),
+    toolResponse: (res: McpToolResponse) => ipcRenderer.send(IPC.mcpResponse, res),
    webhookRestart: () => ipcRenderer.invoke(IPC.webhookRestart) as Promise<WebhookStatus>
  },
  hotkeys: {
@@ -69,7 +74,15 @@ const api: EveFlowBridge = {
    onProgress: (cb: (progress: VoiceDownloadProgress) => void) => subscribe<VoiceDownloadProgress>(VOICE_IPC.modelsProgress, cb),
    transcribe: (req: TranscribeRequest) => ipcRenderer.invoke(VOICE_IPC.transcribe, req) as Promise<TranscribeResult>,
    synthesize: (req: SynthesizeRequest) => ipcRenderer.invoke(VOICE_IPC.synthesize, req) as Promise<SynthesizeResult>,
-    unload: (id?: string) => ipcRenderer.invoke(VOICE_IPC.unload, id) as Promise<unknown>
+    unload: (id?: string) => ipcRenderer.invoke(VOICE_IPC.unload, id) as Promise<unknown>,
+    kwsStart: (req: KwsStartRequest) => ipcRenderer.invoke(VOICE_IPC.kwsStart, req) as Promise<{ accepted: string[]; rejected: string[] }>,
+    kwsStop: () => ipcRenderer.invoke(VOICE_IPC.kwsStop) as Promise<void>,
+    kwsAudio: (pcm: Uint8Array, sampleRate: number) => ipcRenderer.send(VOICE_IPC.kwsAudio, pcm, sampleRate),
+    onKwsDetected: (cb: (detection: KwsDetection) => void) => subscribe<KwsDetection>(VOICE_IPC.kwsDetected, cb),
+    vadStart: (req: VadStartRequest) => ipcRenderer.invoke(VOICE_IPC.vadStart, req) as Promise<void>,
+    vadStop: () => ipcRenderer.invoke(VOICE_IPC.vadStop) as Promise<void>,
+    vadAudio: (pcm: Uint8Array, sampleRate: number) => ipcRenderer.send(VOICE_IPC.vadAudio, pcm, sampleRate),
+    onVadEvent: (cb: (event: VadEvent) => void) => subscribe<VadEvent>(VOICE_IPC.vadEvent, cb)
  }
 };

@@ -1,6 +1,7 @@
 import type { VoiceModelSpec, VoiceSpeaker } from '../../shared/voice';

 const ASR = 'https://github.com/k2-fsa/sherpa-onnx/releases/download/asr-models';
+const KWS = 'https://github.com/k2-fsa/sherpa-onnx/releases/download/kws-models';
 const TTS = 'https://github.com/k2-fsa/sherpa-onnx/releases/download/tts-models';

 const KOKORO_SPEAKERS: VoiceSpeaker[] = [
@@ -67,6 +68,37 @@ export const VOICE_CATALOG: VoiceModelSpec[] = [
    dir: 'sherpa-onnx-sense-voice-zh-en-ja-ko-yue-int8-2024-07-17',
    files: ['model.int8.onnx', 'tokens.txt']
  },
+  {
+    id: 'kws-en',
+    kind: 'kws',
+    engine: 'kws-transducer',
+    name: 'Détecteur de mot-clé (zipformer, 3 Mo)',
+    description: 'Écoute permanente du mot d’activation, quasi gratuite en CPU. Mot-clé libre (« jarvis », « hey jarvis »…).',
+    languages: ['en', 'fr'],
+    sizeMb: 4,
+    url: `${KWS}/sherpa-onnx-kws-zipformer-gigaspeech-3.3M-2024-01-01.tar.bz2`,
+    dir: 'sherpa-onnx-kws-zipformer-gigaspeech-3.3M-2024-01-01',
+    files: [
+      'encoder-epoch-12-avg-2-chunk-16-left-64.int8.onnx',
+      'decoder-epoch-12-avg-2-chunk-16-left-64.int8.onnx',
+      'joiner-epoch-12-avg-2-chunk-16-left-64.int8.onnx',
+      'tokens.txt'
+    ],
+    recommended: true
+  },
+  {
+    id: 'silero-vad',
+    kind: 'vad',
+    engine: 'silero',
+    name: 'Silero VAD (fin de phrase neuronale, 0,6 Mo)',
+    description: 'Détecte précisément le début et la fin de la parole pendant l’écoute permanente ; moins de faux départs sur le bruit.',
+    languages: ['multi'],
+    sizeMb: 1,
+    url: `${ASR}/silero_vad.onnx`,
+    dir: 'silero-vad',
+    files: ['silero_vad.onnx'],
+    recommended: true
+  },
  {
    id: 'kokoro-v1',
    kind: 'tts',
@@ -132,3 +164,12 @@ export const VOICE_CATALOG: VoiceModelSpec[] = [
 export function findModel(id: string): VoiceModelSpec | undefined {
  return VOICE_CATALOG.find((m) => m.id === id);
 }
+
+// Speaker gender from the label ("(homme, …)", "(femme, …)", known first names) so the UI and the
+// voice preference can pick a masculine or feminine voice without a lookup table per model.
+const MALE = /\b(homme|tom|pierre|adam|michael|eric|liam|george|lewis|daniel|fenrir|puck|onyx|echo|santa)\b/i;
+for (const spec of VOICE_CATALOG) {
+  for (const sp of spec.speakers ?? []) {
+    if (!sp.gender) sp.gender = /\b(femme|female)\b/i.test(sp.name) ? 'f' : MALE.test(sp.name) ? 'm' : /\bfemme\b/i.test(spec.name) ? 'f' : /\bhomme\b/i.test(spec.name) ? 'm' : undefined;
+  }
+}
@@ -2,13 +2,23 @@
 * Host side of the voice worker: spawns the utility process on demand, correlates
 * requests and responses, restarts the worker if it crashes.
 */
-import { utilityProcess, type UtilityProcess } from 'electron';
+import { utilityProcess, type UtilityProcess, type WebContents } from 'electron';
+import { createHash } from 'node:crypto';
+import fs from 'node:fs';
 import path from 'node:path';
-import type { SynthesizeRequest, SynthesizeResult, TranscribeRequest, TranscribeResult, VoiceEngineStatus } from '../../shared/voice';
+import { VOICE_IPC, type KwsDetection, type KwsStartRequest, type SynthesizeRequest, type SynthesizeResult, type TranscribeRequest, type TranscribeResult, type VadEvent, type VadStartRequest, type VoiceEngineStatus } from '../../shared/voice';
+import { buildKeywordsFile, parseTokens } from '../../shared/keywords';
 import { findModel } from './catalog';
 import { isInstalled, modelDir, modelsDir } from './models';
 import { log } from '../logger';

+/** Renderer that receives keyword detections while spotting is active. */
+let kwsSubscriber: WebContents | null = null;
+let kwsActive = false;
+let kwsRequest: KwsStartRequest | null = null;
+let vadSubscriber: WebContents | null = null;
+let vadActive = false;
+
 interface Pending {
  resolve: (value: unknown) => void;
  reject: (err: Error) => void;
@@ -34,7 +44,28 @@ function spawn(): UtilityProcess {
  child = proc;
  proc.stdout?.on('data', (d: Buffer) => log('DEBUG', 'voice-worker', d.toString().trim()));
  proc.stderr?.on('data', (d: Buffer) => log('WARN', 'voice-worker', d.toString().trim()));
-  proc.on('message', (msg: { id: number; ok: boolean; result?: unknown; error?: string }) => {
+  proc.on('message', (msg: { id?: number; type?: string; ok?: boolean; result?: unknown; error?: string; keyword?: string; at?: number; event?: { type: string; wav?: string; durationSec?: number } }) => {
+    if (msg.type === 'vad.event' && msg.event) {
+      if (vadSubscriber && !vadSubscriber.isDestroyed()) {
+        const ev = msg.event;
+        const payload: VadEvent =
+          ev.type === 'segment' && ev.wav
+            ? { type: 'segment', wav: new Uint8Array(Buffer.from(ev.wav, 'base64')), durationSec: ev.durationSec ?? 0 }
+            : ev.type === 'speech-start'
+              ? { type: 'speech-start' }
+              : { type: 'error', message: 'événement VAD inconnu' };
+        vadSubscriber.send(VOICE_IPC.vadEvent, payload);
+      }
+      return;
+    }
+    if (msg.type === 'kws.detected') {
+      if (kwsSubscriber && !kwsSubscriber.isDestroyed()) {
+        kwsSubscriber.send(VOICE_IPC.kwsDetected, { keyword: msg.keyword ?? '', at: msg.at ?? Date.now() } satisfies KwsDetection);
+      }
+      log('INFO', 'voice', `wake word detected: ${msg.keyword}`);
+      return;
+    }
+    if (typeof msg.id !== 'number') return;
    const p = pending.get(msg.id);
    if (!p) return;
    pending.delete(msg.id);
@@ -48,6 +79,8 @@ function spawn(): UtilityProcess {
    if (child !== proc) return;
    child = null;
    rejectAll('Le moteur vocal local s’est arrêté de façon inattendue.');
+    // Keyword spotting survives a worker restart: re-arm on the next audio frame.
+    if (kwsActive) kwsArmed = false;
  });
  log('INFO', 'voice', 'voice worker started');
  return proc;
@@ -108,6 +141,89 @@ export async function synthesize(req: SynthesizeRequest): Promise<SynthesizeResu
  return { ...result, wav: new Uint8Array(buffer.buffer, buffer.byteOffset, buffer.byteLength) };
 }

+let kwsArmed = false;
+const SENSITIVITY_THRESHOLD: Record<number, number> = { 1: 0.45, 2: 0.35, 3: 0.25, 4: 0.18, 5: 0.12 };
+const SENSITIVITY_SCORE: Record<number, number> = { 1: 1.0, 2: 1.0, 3: 1.2, 4: 1.5, 5: 2.0 };
+
+/** Start keyword spotting; the keywords file is derived from the phrases and the model vocabulary. */
+export async function kwsStart(req: KwsStartRequest, sender: WebContents): Promise<{ accepted: string[]; rejected: string[] }> {
+  const spec = findModel(req.modelId);
+  if (!spec || spec.kind !== 'kws') throw new Error('Modèle de détection introuvable');
+  if (!isInstalled(spec)) throw new Error('Détecteur de mot-clé non installé (Paramètres → Modèles locaux).');
+  const dir = modelDir(spec);
+  const vocab = parseTokens(fs.readFileSync(path.join(dir, 'tokens.txt'), 'utf8'));
+  const phrases = req.keywords.map((k) => String(k).slice(0, 40)).filter(Boolean).slice(0, 8);
+  const file = buildKeywordsFile(phrases, vocab);
+  if (file.accepted.length === 0) throw new Error(`Aucun mot d’activation encodable : ${file.rejected.join(', ')}`);
+  const hash = createHash('sha1').update(file.content).digest('hex').slice(0, 10);
+  const keywordsFile = path.join(modelsDir(), `keywords-${hash}.txt`);
+  fs.writeFileSync(keywordsFile, file.content, 'utf8');
+  const sensitivity = Math.min(5, Math.max(1, Math.round(req.sensitivity))) as 1 | 2 | 3 | 4 | 5;
+  kwsSubscriber = sender;
+  kwsRequest = req;
+  await request({ type: 'kws.start', model: { id: spec.id, engine: spec.engine, dir, files: spec.files }, keywordsFile, threshold: SENSITIVITY_THRESHOLD[sensitivity], score: SENSITIVITY_SCORE[sensitivity] }, 60_000);
+  kwsActive = true;
+  kwsArmed = true;
+  log('INFO', 'voice', `keyword spotting on: ${file.accepted.join(', ')} (threshold ${SENSITIVITY_THRESHOLD[sensitivity]})`);
+  return { accepted: file.accepted, rejected: file.rejected };
+}
+
+export async function kwsStop(): Promise<void> {
+  kwsActive = false;
+  kwsArmed = false;
+  kwsSubscriber = null;
+  if (child) await request({ type: 'kws.stop' }, 10_000).catch(() => undefined);
+}
+
+/** Feed 16-bit PCM from the renderer (fire-and-forget). */
+export function kwsFeed(pcm: Uint8Array, sampleRate: number): void {
+  if (!kwsActive) return;
+  const proc = spawn();
+  if (!kwsArmed) {
+    // Worker restarted: re-create the spotter before feeding audio.
+    if (kwsRequest && kwsSubscriber) {
+      kwsArmed = true;
+      kwsStart(kwsRequest, kwsSubscriber).catch((err) => log('WARN', 'voice', `kws re-arm failed: ${(err as Error).message}`));
+    }
+    return;
+  }
+  const b64 = Buffer.from(pcm.buffer, pcm.byteOffset, pcm.byteLength).toString('base64');
+  try {
+    proc.postMessage({ id: -1, type: 'kws.audio', pcm: b64, sampleRate });
+  } catch (err) {
+    log('WARN', 'voice', `kws feed failed: ${(err as Error).message}`);
+  }
+}
+
+/** Neural end-of-speech detection (Silero) on frames streamed by the renderer. */
+export async function vadStart(req: VadStartRequest, sender: WebContents): Promise<void> {
+  const spec = findModel(req.modelId);
+  if (!spec || spec.kind !== 'vad') throw new Error('Modèle VAD introuvable');
+  if (!isInstalled(spec)) throw new Error('Silero VAD non installé (Paramètres → Modèles locaux).');
+  vadSubscriber = sender;
+  await request(
+    { type: 'vad.start', model: { id: spec.id, engine: spec.engine, dir: modelDir(spec), files: spec.files }, silenceMs: req.silenceMs, threshold: req.threshold, maxUtteranceSec: req.maxUtteranceSec },
+    30_000
+  );
+  vadActive = true;
+}
+
+export async function vadStop(): Promise<void> {
+  vadActive = false;
+  vadSubscriber = null;
+  if (child) await request({ type: 'vad.stop' }, 10_000).catch(() => undefined);
+}
+
+export function vadFeed(pcm: Uint8Array, sampleRate: number): void {
+  if (!vadActive || !child) return;
+  const b64 = Buffer.from(pcm.buffer, pcm.byteOffset, pcm.byteLength).toString('base64');
+  try {
+    child.postMessage({ id: -1, type: 'vad.audio', pcm: b64, sampleRate });
+  } catch (err) {
+    log('WARN', 'voice', `vad feed failed: ${(err as Error).message}`);
+  }
+}
+
 /** Native model memory is only released deterministically by restarting the worker. */
 export function unload(_modelId?: string): Promise<unknown> {
  stopEngine();
@@ -1,6 +1,6 @@
 import { ipcMain } from 'electron';
-import { VOICE_IPC, type SynthesizeRequest, type TranscribeRequest } from '../../shared/voice';
-import { engineStatus, synthesize, transcribe, unload } from './engine';
+import { VOICE_IPC, type KwsStartRequest, type SynthesizeRequest, type TranscribeRequest, type VadStartRequest } from '../../shared/voice';
+import { engineStatus, kwsFeed, kwsStart, kwsStop, synthesize, transcribe, unload, vadFeed, vadStart, vadStop } from './engine';
 import { cancelDownload, downloadModel, listModels, removeModel } from './models';

 export function registerVoiceIpc(): void {
@@ -27,4 +27,22 @@ export function registerVoiceIpc(): void {
    return synthesize({ ...req, speaker: Number.isFinite(req.speaker) ? req.speaker : 0, speed: Number.isFinite(req.speed) ? req.speed : 1 });
  });
  ipcMain.handle(VOICE_IPC.unload, (_e, id?: string) => unload(id));
+  ipcMain.handle(VOICE_IPC.kwsStart, (event, req: KwsStartRequest) => {
+    if (!req || !Array.isArray(req.keywords) || typeof req.modelId !== 'string') throw new Error('Requête invalide');
+    return kwsStart({ ...req, keywords: req.keywords.filter((k) => typeof k === 'string'), sensitivity: Number(req.sensitivity) || 3 }, event.sender);
+  });
+  ipcMain.handle(VOICE_IPC.kwsStop, () => kwsStop());
+  ipcMain.handle(VOICE_IPC.vadStart, (event, req: VadStartRequest) => {
+    if (!req || typeof req.modelId !== 'string') throw new Error('Requête invalide');
+    return vadStart({ modelId: req.modelId, silenceMs: Number(req.silenceMs) || 700, threshold: Number(req.threshold) || 0.5, maxUtteranceSec: Number(req.maxUtteranceSec) || 30 }, event.sender);
+  });
+  ipcMain.handle(VOICE_IPC.vadStop, () => vadStop());
+  ipcMain.on(VOICE_IPC.vadAudio, (_e, pcm: unknown, sampleRate: unknown) => {
+    if (!(pcm instanceof Uint8Array) || pcm.byteLength === 0 || pcm.byteLength > 1024 * 1024) return;
+    vadFeed(pcm, typeof sampleRate === 'number' && sampleRate > 0 ? sampleRate : 16000);
+  });
+  ipcMain.on(VOICE_IPC.kwsAudio, (_e, pcm: unknown, sampleRate: unknown) => {
+    if (!(pcm instanceof Uint8Array) || pcm.byteLength === 0 || pcm.byteLength > 1024 * 1024) return;
+    kwsFeed(pcm, typeof sampleRate === 'number' && sampleRate > 0 ? sampleRate : 16000);
+  });
 }
@@ -18,7 +18,13 @@ type Request =
  | { id: number; type: 'status' }
  | { id: number; type: 'transcribe'; model: ModelRef; wav: Uint8Array | string; language: string }
  | { id: number; type: 'synthesize'; model: ModelRef; text: string; speaker: number; speed: number }
-  | { id: number; type: 'unload'; modelId?: string };
+  | { id: number; type: 'unload'; modelId?: string }
+  | { id: number; type: 'kws.start'; model: ModelRef; keywordsFile: string; threshold: number; score: number }
+  | { id: number; type: 'kws.audio'; pcm: string; sampleRate: number }
+  | { id: number; type: 'kws.stop' }
+  | { id: number; type: 'vad.start'; model: ModelRef; silenceMs: number; threshold: number; maxUtteranceSec: number }
+  | { id: number; type: 'vad.audio'; pcm: string; sampleRate: number }
+  | { id: number; type: 'vad.stop' };

 type Response = { id: number; ok: true; result: unknown } | { id: number; ok: false; error: string };

@@ -29,6 +35,23 @@ type Sherpa = {
    decode: (s: unknown) => void;
    getResult: (s: unknown) => { text: string; lang?: string };
  };
+  Vad: new (config: unknown, bufferSizeInSeconds: number) => {
+    acceptWaveform: (samples: Float32Array) => void;
+    isEmpty: () => boolean;
+    isDetected: () => boolean;
+    pop: () => void;
+    clear: () => void;
+    front: (enableExternalBuffer?: boolean) => { start: number; samples: Float32Array };
+    reset: () => void;
+    flush: () => void;
+  };
+  KeywordSpotter: new (config: unknown) => {
+    createStream: () => KwsStream;
+    isReady: (s: KwsStream) => boolean;
+    decode: (s: KwsStream) => void;
+    reset: (s: KwsStream) => void;
+    getResult: (s: KwsStream) => { keyword?: string };
+  };
  OfflineTts: new (config: unknown) => {
    numSpeakers: number;
    sampleRate: number;
@@ -37,7 +60,11 @@ type Sherpa = {
  version: string;
 };

+type KwsStream = { acceptWaveform: (w: { sampleRate: number; samples: Float32Array }) => void };
 let sherpa: Sherpa | null = null;
+let kws: { spotter: InstanceType<Sherpa['KeywordSpotter']>; stream: KwsStream } | null = null;
+let vad: { detector: InstanceType<Sherpa['Vad']>; speaking: boolean; windowSize: number; carry: Float32Array } | null = null;
+let notify: ((message: unknown) => void) | null = null;
 let loadError: string | null = null;

 function loadSherpa(): Sherpa {
@@ -225,13 +252,124 @@ function encodeWav(samples: Float32Array, sampleRate: number): Uint8Array {
  return new Uint8Array(buffer);
 }

+function startKws(model: ModelRef, keywordsFile: string, threshold: number, score: number): void {
+  const s = loadSherpa();
+  kws = null;
+  const p = (f: string) => path.join(model.dir, f);
+  const enc = model.files.find((f) => f.startsWith('encoder')) ?? 'encoder.int8.onnx';
+  const dec = model.files.find((f) => f.startsWith('decoder')) ?? 'decoder.int8.onnx';
+  const join = model.files.find((f) => f.startsWith('joiner')) ?? 'joiner.int8.onnx';
+  const spotter = new s.KeywordSpotter({
+    featConfig: { sampleRate: 16000, featureDim: 80 },
+    modelConfig: { transducer: { encoder: p(enc), decoder: p(dec), joiner: p(join) }, tokens: p('tokens.txt'), numThreads: 1, provider: 'cpu', debug: 0 },
+    maxActivePaths: 4,
+    numTrailingBlanks: 1,
+    keywordsScore: score,
+    keywordsThreshold: threshold,
+    keywordsFile
+  });
+  kws = { spotter, stream: spotter.createStream() };
+}
+
+function feedKws(pcmBase64: string, sampleRate: number): void {
+  if (!kws) return;
+  const bytes = Buffer.from(pcmBase64, 'base64');
+  const int16 = new Int16Array(bytes.buffer, bytes.byteOffset, Math.floor(bytes.byteLength / 2));
+  let samples: Float32Array = new Float32Array(int16.length);
+  for (let i = 0; i < int16.length; i++) samples[i] = int16[i] / 32768;
+  if (sampleRate !== 16000) samples = resampleTo16k(samples, sampleRate);
+  kws.stream.acceptWaveform({ sampleRate: 16000, samples });
+  while (kws.spotter.isReady(kws.stream)) {
+    kws.spotter.decode(kws.stream);
+    const result = kws.spotter.getResult(kws.stream);
+    if (result.keyword) {
+      kws.spotter.reset(kws.stream);
+      notify?.({ type: 'kws.detected', keyword: result.keyword, at: Date.now() });
+    }
+  }
+}
+
+function startVad(model: ModelRef, silenceMs: number, threshold: number, maxUtteranceSec: number): void {
+  const s = loadSherpa();
+  const windowSize = 512;
+  const detector = new s.Vad(
+    {
+      sileroVad: {
+        model: path.join(model.dir, 'silero_vad.onnx'),
+        threshold: Math.max(0.1, Math.min(0.95, threshold)),
+        minSilenceDuration: Math.max(0.15, silenceMs / 1000),
+        minSpeechDuration: 0.2,
+        windowSize,
+        maxSpeechDuration: Math.max(3, maxUtteranceSec)
+      },
+      sampleRate: 16000,
+      numThreads: 1,
+      provider: 'cpu',
+      debug: 0
+    },
+    Math.max(10, maxUtteranceSec + 5)
+  );
+  vad = { detector, speaking: false, windowSize, carry: new Float32Array(0) };
+}
+
+function feedVad(pcmBase64: string, sampleRate: number): void {
+  if (!vad) return;
+  const bytes = Buffer.from(pcmBase64, 'base64');
+  const int16 = new Int16Array(bytes.buffer, bytes.byteOffset, Math.floor(bytes.byteLength / 2));
+  let samples: Float32Array = new Float32Array(int16.length);
+  for (let i = 0; i < int16.length; i++) samples[i] = int16[i] / 32768;
+  if (sampleRate !== 16000) samples = resampleTo16k(samples, sampleRate);
+  // Silero expects fixed windows: keep the remainder for the next frame.
+  const merged = new Float32Array(vad.carry.length + samples.length);
+  merged.set(vad.carry, 0);
+  merged.set(samples, vad.carry.length);
+  const usable = merged.length - (merged.length % vad.windowSize);
+  for (let i = 0; i < usable; i += vad.windowSize) {
+    vad.detector.acceptWaveform(merged.subarray(i, i + vad.windowSize));
+    const detected = vad.detector.isDetected();
+    if (detected && !vad.speaking) {
+      vad.speaking = true;
+      notify?.({ type: 'vad.event', event: { type: 'speech-start' } });
+    }
+    while (!vad.detector.isEmpty()) {
+      const segment = vad.detector.front(false);
+      vad.detector.pop();
+      vad.speaking = false;
+      const wav = encodeWav(segment.samples, 16000);
+      notify?.({
+        type: 'vad.event',
+        event: { type: 'segment', wav: Buffer.from(wav.buffer, wav.byteOffset, wav.byteLength).toString('base64'), durationSec: segment.samples.length / 16000 }
+      });
+    }
+  }
+  vad.carry = merged.slice(usable);
+}
+
 // ── request handling ───────────────────────────────────────────────────────
 function handle(req: Request): unknown {
  switch (req.type) {
+    case 'vad.start':
+      startVad(req.model, req.silenceMs, req.threshold, req.maxUtteranceSec);
+      return { ok: true };
+    case 'vad.audio':
+      feedVad(req.pcm, req.sampleRate);
+      return { ok: true };
+    case 'vad.stop':
+      vad = null;
+      return { ok: true };
+    case 'kws.start':
+      startKws(req.model, req.keywordsFile, req.threshold, req.score);
+      return { ok: true };
+    case 'kws.audio':
+      feedKws(req.pcm, req.sampleRate);
+      return { ok: true };
+    case 'kws.stop':
+      kws = null;
+      return { ok: true };
    case 'status': {
      try {
        const s = loadSherpa();
-        return { available: true, version: s.version, loaded: [...recognizers.keys(), ...synthesizers.keys()] };
+        return { available: true, version: s.version, loaded: [...recognizers.keys(), ...synthesizers.keys(), ...(kws ? ['kws'] : [])] };
      } catch (err) {
        return { available: false, error: (err as Error).message, loaded: [] };
      }
@@ -271,6 +409,8 @@ function handle(req: Request): unknown {
      } else {
        recognizers.clear();
        synthesizers.clear();
+        kws = null;
+        vad = null;
      }
      return { ok: true };
    }
@@ -290,7 +430,9 @@ function respond(req: Request): Response {
 // Electron utility process transport, with a child_process fallback for tests.
 const parentPort = (process as unknown as { parentPort?: { on: (ev: 'message', cb: (e: { data: Request }) => void) => void; postMessage: (m: unknown) => void } }).parentPort;
 if (parentPort) {
+  notify = (m) => parentPort.postMessage(m);
  parentPort.on('message', (event) => parentPort.postMessage(respond(event.data)));
 } else if (process.send) {
+  notify = (m) => process.send!(m);
  process.on('message', (msg: Request) => process.send!(respond(msg)));
 }
@@ -8,6 +8,7 @@ import type { BrowserWindow } from 'electron';
 import { IPC, type WebhookStatus } from '../shared/ipc';
 import { normalizeHermesPush } from '../shared/hermesPush';
 import { log } from './logger';
+import { handleMcpHttp, MCP_PATH } from './mcp';

 export const WEBHOOK_PATH = '/eveflow/hook';
 const MAX_BODY_BYTES = 2 * 1024 * 1024;
@@ -57,9 +58,10 @@ export async function startWebhookServer(getWindow: () => BrowserWindow | null,
    };

    if (req.method === 'GET' && (req.url === '/health' || req.url === '/')) {
-      return json(200, { ok: true, app: 'eveflow', path: WEBHOOK_PATH });
+      return json(200, { ok: true, app: 'eveflow', path: WEBHOOK_PATH, mcp: MCP_PATH });
    }
-    if (req.method !== 'POST' || !req.url?.startsWith(WEBHOOK_PATH)) {
+    const isMcp = !!req.url && (req.url === MCP_PATH || req.url.startsWith(`${MCP_PATH}?`));
+    if (!isMcp && (req.method !== 'POST' || !req.url?.startsWith(WEBHOOK_PATH))) {
      return json(404, { error: 'Not Found' });
    }

@@ -91,6 +93,13 @@ export async function startWebhookServer(getWindow: () => BrowserWindow | null,
      if (tooLarge) return;
      // Concatenate before decoding so multibyte UTF-8 (accents, emoji) split across chunks stays intact.
      const body = Buffer.concat(chunks).toString('utf8');
+      if (isMcp) {
+        handleMcpHttp(getWindow(), req, res, body).catch((err: Error) => {
+          log('ERROR', 'mcp', err.message);
+          if (!res.headersSent) json(500, { error: err.message });
+        });
+        return;
+      }
      try {
        const raw: unknown = body.trim() ? JSON.parse(body) : {};
        const events = normalizeHermesPush(raw);
@@ -1,7 +1,7 @@
 {
  "name": "eveflow",
-  "version": "2.1.0",
-  "releaseVersion": "2.1.0.1",
+  "version": "2.4.0",
+  "releaseVersion": "2.4.0.1",
  "description": "JARVIS-style desktop HUD for Hermes Agent: voice, streaming runs, scheduled jobs, skills and telemetry",
  "main": "dist-electron/main.js",
  "private": true,
@@ -11,7 +11,12 @@ import type {
  WebhookStatus,
  WindowMode
 } from './ipc';
+import type { McpToolRequest, McpToolResponse } from './ipc';
 import type {
+  KwsDetection,
+  KwsStartRequest,
+  VadEvent,
+  VadStartRequest,
  SynthesizeRequest,
  SynthesizeResult,
  TranscribeRequest,
@@ -44,10 +49,6 @@ export interface EveFlowBridge {
    streamAbort: (id: string) => void;
    onStreamEvent: (cb: (event: HttpStreamEvent) => void) => Unsubscribe;
  };
-  system: {
-    metrics: () => Promise<SystemMetrics>;
-    appInfo: () => Promise<AppInfo>;
-  };
  files: {
    readLocal: (filePath: string) => Promise<string>;
    writeShared: (filename: string, content: string, isBase64?: boolean) => Promise<{ path: string; url: string }>;
@@ -57,6 +58,9 @@ export interface EveFlowBridge {
  hermes: {
    onPush: (cb: (event: HermesPushEvent) => void) => Unsubscribe;
    webhookStatus: () => Promise<WebhookStatus>;
+    /** Tool calls from Hermes via the local MCP endpoint that need the renderer. */
+    onToolRequest: (cb: (req: McpToolRequest) => void) => Unsubscribe;
+    toolResponse: (res: McpToolResponse) => void;
    webhookRestart: () => Promise<WebhookStatus>;
  };
  hotkeys: {
@@ -72,5 +76,37 @@ export interface EveFlowBridge {
    transcribe: (req: TranscribeRequest) => Promise<TranscribeResult>;
    synthesize: (req: SynthesizeRequest) => Promise<SynthesizeResult>;
    unload: (id?: string) => Promise<unknown>;
+    kwsStart: (req: KwsStartRequest) => Promise<{ accepted: string[]; rejected: string[] }>;
+    kwsStop: () => Promise<void>;
+    /** Fire-and-forget 16-bit PCM frames for the keyword spotter. */
+    kwsAudio: (pcm: Uint8Array, sampleRate: number) => void;
+    onKwsDetected: (cb: (detection: KwsDetection) => void) => Unsubscribe;
+    vadStart: (req: VadStartRequest) => Promise<void>;
+    vadStop: () => Promise<void>;
+    vadAudio: (pcm: Uint8Array, sampleRate: number) => void;
+    onVadEvent: (cb: (event: VadEvent) => void) => Unsubscribe;
+  };
+  system: {
+    metrics: () => Promise<SystemMetrics>;
+    appInfo: () => Promise<AppInfo>;
+    /** Screenshot of the primary display as a JPEG data URL. */
+    captureScreen: (maxWidth?: number) => Promise<string>;
+    /** Allow-listed local actions (lock session, open app/url, media keys, clipboard). */
+    action: (action: SystemAction) => Promise<SystemActionResult>;
  };
 }
+
+export type SystemAction =
+  | { type: 'lock' }
+  | { type: 'open-app'; name: string }
+  | { type: 'open-url'; url: string }
+  | { type: 'media'; key: 'volume-up' | 'volume-down' | 'mute' | 'play-pause' | 'next' | 'previous' }
+  | { type: 'clipboard-read' }
+  | { type: 'clipboard-write'; text: string }
+  | { type: 'find-files'; query: string };
+
+export interface SystemActionResult {
+  ok: boolean;
+  message?: string;
+  data?: unknown;
+}
@@ -111,6 +111,10 @@ export const IPC = {
  httpStreamEvent: 'http:stream:event',
  metrics: 'system:metrics',
  appInfo: 'app:info',
+  screenCapture: 'system:screen-capture',
+  systemAction: 'system:action',
+  mcpRequest: 'mcp:request',
+  mcpResponse: 'mcp:response',
  readLocalFile: 'files:read-local',
  writeSharedFile: 'files:write-shared',
  openPath: 'files:open-path',
@@ -122,3 +126,17 @@ export const IPC = {
 } as const;

 export type HotkeyEvent = 'ptt-toggle' | 'toggle-window' | 'stop-speaking';
+
+/** A tool call arriving from Hermes through the local MCP endpoint that needs the renderer (UI, voice, chat). */
+export interface McpToolRequest {
+  id: string;
+  name: string;
+  args: Record<string, unknown>;
+}
+
+export interface McpToolResponse {
+  id: string;
+  ok: boolean;
+  result?: unknown;
+  error?: string;
+}
@@ -0,0 +1,108 @@
+/**
+ * Encodes wake phrases into the BPE token sequences expected by the sherpa-onnx keyword
+ * spotter (gigaspeech BPE-500 model). Common phrases use sequences produced by the real
+ * SentencePiece model; anything else falls back to a greedy longest-match over the vocabulary,
+ * which is a close approximation for short words.
+ */
+
+const KNOWN: Record<string, string> = {
+  'JARVIS': '▁JA R VI S',
+  'HEY JARVIS': '▁HE Y ▁JA R VI S',
+  'OK JARVIS': '▁O K ▁JA R VI S',
+  'EVE': '▁E VE',
+  'HEY EVE': '▁HE Y ▁E VE',
+  'COMPUTER': '▁COMP U TER',
+  'HEY COMPUTER': '▁HE Y ▁COMP U TER',
+  'FRIDAY': '▁F RI DAY',
+  'HEY FRIDAY': '▁HE Y ▁F RI DAY',
+  'ALFRED': '▁A L F RE D',
+  'HERMES': '▁HER ME S',
+  'HEY HERMES': '▁HE Y ▁HER ME S',
+  'ASSISTANT': '▁AS S IST ANT',
+  'OK GOOGLE': '▁O K ▁GO O G LE',
+  'ALEXA': '▁A LE X A',
+  'SIRI': '▁S I RI',
+  'HEY SIRI': '▁HE Y ▁S I RI',
+  'NOVA': '▁NO V A',
+  'ATLAS': '▁AT LA S',
+  'HAL': '▁HA L'
+};
+
+export function normalizeKeyword(phrase: string): string {
+  return phrase
+    .normalize('NFD')
+    .replace(/[̀-ͯ]/g, '')
+    .toUpperCase()
+    .replace(/[^A-Z ]+/g, ' ')
+    .replace(/\s+/g, ' ')
+    .trim();
+}
+
+/** Parse a sherpa tokens.txt ("piece id" per line) into the set of pieces. */
+export function parseTokens(tokensFile: string): Set<string> {
+  const pieces = new Set<string>();
+  for (const line of tokensFile.split(/\r?\n/)) {
+    const piece = line.trim().split(/\s+/)[0];
+    if (piece && !piece.startsWith('<')) pieces.add(piece);
+  }
+  return pieces;
+}
+
+function greedyWord(word: string, vocab: Set<string>): string[] | null {
+  const out: string[] = [];
+  let i = 0;
+  let first = true;
+  while (i < word.length) {
+    let matched = '';
+    for (let len = word.length - i; len >= 1; len--) {
+      const candidate = (first ? '▁' : '') + word.slice(i, i + len);
+      if (vocab.has(candidate)) {
+        matched = candidate;
+        break;
+      }
+    }
+    if (!matched && first) {
+      // No word-initial piece: use a bare "▁" if available, then continue without the prefix.
+      if (vocab.has('▁')) out.push('▁');
+      first = false;
+      continue;
+    }
+    if (!matched) return null;
+    out.push(matched);
+    i += matched.length - (first ? 1 : 0);
+    first = false;
+  }
+  return out;
+}
+
+/** Encode a phrase into space-separated BPE pieces, or null when it cannot be represented. */
+export function encodeKeyword(phrase: string, vocab: Set<string>): string | null {
+  const normalized = normalizeKeyword(phrase);
+  if (!normalized) return null;
+  if (KNOWN[normalized]) return KNOWN[normalized];
+  const pieces: string[] = [];
+  for (const word of normalized.split(' ')) {
+    const encoded = greedyWord(word, vocab);
+    if (!encoded) return null;
+    pieces.push(...encoded);
+  }
+  return pieces.join(' ');
+}
+
+/** Build the keywords file content: one line per phrase, with the readable label. */
+export function buildKeywordsFile(phrases: string[], vocab: Set<string>): { content: string; accepted: string[]; rejected: string[] } {
+  const lines: string[] = [];
+  const accepted: string[] = [];
+  const rejected: string[] = [];
+  for (const phrase of phrases) {
+    const encoded = encodeKeyword(phrase, vocab);
+    const label = normalizeKeyword(phrase).toLowerCase().replace(/ /g, '_');
+    if (!encoded || !label) {
+      rejected.push(phrase);
+      continue;
+    }
+    lines.push(`${encoded} @${label}`);
+    accepted.push(label);
+  }
+  return { content: lines.join('\n') + '\n', accepted, rejected };
+}
@@ -1,12 +1,14 @@
 /** Local voice engine contract (sherpa-onnx in a utility process). Shared by main and renderer. */

-export type VoiceModelKind = 'stt' | 'tts';
-export type VoiceEngineKind = 'whisper' | 'sense-voice' | 'nemo-transducer' | 'kokoro' | 'piper';
+export type VoiceModelKind = 'stt' | 'tts' | 'kws' | 'vad';
+export type VoiceEngineKind = 'whisper' | 'sense-voice' | 'nemo-transducer' | 'kokoro' | 'piper' | 'kws-transducer' | 'silero';

 export interface VoiceSpeaker {
  id: number;
  name: string;
  lang: string;
+  /** m = masculine, f = feminine (from the catalog label). */
+  gender?: 'm' | 'f';
 }

 export interface VoiceModelSpec {
@@ -69,6 +71,33 @@ export interface SynthesizeResult {
  audioSec: number;
 }

+export interface KwsStartRequest {
+  modelId: string;
+  /** Wake phrases in plain text (e.g. "jarvis", "hey jarvis"). */
+  keywords: string[];
+  /** 1 (strict) .. 5 (eager) */
+  sensitivity: number;
+}
+
+export interface KwsDetection {
+  keyword: string;
+  at: number;
+}
+
+export interface VadStartRequest {
+  modelId: string;
+  /** Silence that ends an utterance, in ms. */
+  silenceMs: number;
+  /** Detection threshold 0..1 (0.5 default). */
+  threshold: number;
+  maxUtteranceSec: number;
+}
+
+export type VadEvent =
+  | { type: 'speech-start' }
+  | { type: 'segment'; wav: Uint8Array; durationSec: number }
+  | { type: 'error'; message: string };
+
 export interface VoiceEngineStatus {
  available: boolean;
  error?: string;
@@ -86,5 +115,13 @@ export const VOICE_IPC = {
  modelsProgress: 'voice:models:progress',
  transcribe: 'voice:transcribe',
  synthesize: 'voice:synthesize',
-  unload: 'voice:unload'
+  unload: 'voice:unload',
+  kwsStart: 'voice:kws:start',
+  kwsStop: 'voice:kws:stop',
+  kwsAudio: 'voice:kws:audio',
+  kwsDetected: 'voice:kws:detected',
+  vadStart: 'voice:vad:start',
+  vadStop: 'voice:vad:stop',
+  vadAudio: 'voice:vad:audio',
+  vadEvent: 'voice:vad:event'
 } as const;
@@ -3,7 +3,10 @@ import { AlertTriangle, X } from 'lucide-react';
 import type { WindowMode } from '../shared/ipc';
 import { bridge } from './lib/bridge';
 import { Log } from './lib/log';
-import { handlePush, stopGeneration } from './services/conversation';
+import { handlePush, localToolContext, stopGeneration } from './services/conversation';
+import { initMcpBridge } from './services/mcpBridge';
+import { ensurePreferredVoice } from './services/voice/voicePreference';
+import { isQuietTime } from './lib/quietHours';
 import { speech } from './services/voice/speech';
 import { voiceController } from './services/voice/voiceController';
 import { useChat } from './state/chat';
@@ -43,8 +46,21 @@ function useBoot(): boolean {
      }
      if (api) {
        disposers.push(useVoiceModels.getState().subscribe());
-        void useVoiceModels.getState().refresh();
+        void useVoiceModels.getState().refresh().then(() => ensurePreferredVoice());
+        if (settings.voice.wakeMode === 'kws') void voiceController.startWakeMode();
+        let lastWake = JSON.stringify([settings.voice.wakeMode, settings.voice.wakeWord, settings.voice.kwsSensitivity, settings.voice.micDeviceId]);
+        disposers.push(
+          useSettings.subscribe((s) => {
+            const v = s.settings.voice;
+            const key = JSON.stringify([v.wakeMode, v.wakeWord, v.kwsSensitivity, v.micDeviceId]);
+            if (key === lastWake) return;
+            lastWake = key;
+            if (v.wakeMode === 'kws') void voiceController.restartWakeMode();
+            else void voiceController.stopWakeMode();
+          })
+        );
        disposers.push(api.hermes.onPush(handlePush));
+        initMcpBridge(localToolContext);
        disposers.push(
          api.hotkeys.on((event) => {
            if (event === 'ptt-toggle') voiceController.toggle();
@@ -83,6 +99,24 @@ function useBoot(): boolean {
 function useThemeSync(): void {
  const theme = useSettings((s) => s.settings.theme);
  const reduce = useSettings((s) => s.settings.ui.reduceMotion);
+  // Quiet hours: computed every 30 s; dims the HUD (night theme) and silences pushes.
+  const notif = useSettings((s) => s.settings.notifications);
+  useEffect(() => {
+    const tick = () => {
+      const quiet = notif.quietEnabled && isQuietTime(notif.quietStart, notif.quietEnd);
+      useChat.getState().setQuiet(quiet);
+      document.documentElement.toggleAttribute('data-night', quiet && notif.nightTheme);
+    };
+    tick();
+    const timer = setInterval(tick, 30_000);
+    return () => clearInterval(timer);
+  }, [notif]);
+  useEffect(() => {
+    const onVisible = () => { if (!document.hidden) useChat.getState().markRead(); };
+    document.addEventListener('visibilitychange', onVisible);
+    return () => document.removeEventListener('visibilitychange', onVisible);
+  }, []);
+
  useEffect(() => {
    document.documentElement.dataset.theme = theme;
    document.documentElement.dataset.motion = reduce ? 'reduce' : 'full';
@@ -1,6 +1,6 @@
 import { useEffect, useRef, useState } from 'react';
-import { Mic, Paperclip, Send, Square, X, Radio, Loader2, Volume2, VolumeX, Navigation } from 'lucide-react';
-import { sendMessage, steer, stopGeneration } from '../../services/conversation';
+import { Mic, Paperclip, Send, Square, X, Radio, Loader2, Volume2, VolumeX, Navigation, ScanEye, Crosshair } from 'lucide-react';
+import { sendMessage, steer, stopGeneration , captureScreen } from '../../services/conversation';
 import { speech } from '../../services/voice/speech';
 import { voiceController } from '../../services/voice/voiceController';
 import { useChat } from '../../state/chat';
@@ -50,6 +50,8 @@ export function CommandBar({ compact }: Props) {
  const autoSpeak = useSettings((s) => s.settings.speech.autoSpeak);
  const update = useSettings((s) => s.update);
  const [images, setImages] = useState<string[]>([]);
+  const missionMode = useChat((s) => s.missionMode);
+  const missionModel = useSettings((s) => s.settings.hermes.missionModel);
  const [steerMode, setSteerMode] = useState(false);
  const fileRef = useRef<HTMLInputElement>(null);
  const inputRef = useRef<HTMLTextAreaElement>(null);
@@ -167,9 +169,20 @@ export function CommandBar({ compact }: Props) {
          spellCheck={false}
        />
        {!compact && (
-          <button className="icon-btn" title="Joindre une image" onClick={() => fileRef.current?.click()} style={{ height: 46, width: 40 }}>
-            <Paperclip size={18} />
-          </button>
+          <>
+            <button
+              className="icon-btn"
+              title="Joindre une capture de l’écran (Hermes la voit)"
+              aria-label="Capturer l’écran"
+              onClick={() => void captureScreen().then((shot) => { if (shot) setImages((prev) => [...prev, shot].slice(-4)); })}
+              style={{ height: 46, width: 40 }}
+            >
+              <ScanEye size={18} />
+            </button>
+            <button className="icon-btn" title="Joindre une image" onClick={() => fileRef.current?.click()} style={{ height: 46, width: 40 }}>
+              <Paperclip size={18} />
+            </button>
+          </>
        )}
        <input ref={fileRef} type="file" accept="image/*" multiple hidden onChange={(e) => void onFiles(e.target.files)} />
        {isSending && !steerMode ? (
@@ -187,6 +200,9 @@ export function CommandBar({ compact }: Props) {
          <button className={`toggle-btn${handsFree ? ' on' : ''}`} onClick={() => voiceController.setHandsFree(!handsFree)} title="Écoute continue : le micro se réactive après chaque réponse">
            <Radio size={12} /> mains libres
          </button>
+          <button className={`toggle-btn${missionMode ? ' on' : ''}`} onClick={() => useChat.getState().setMissionMode(!missionMode)} title={missionModel ? `Mode mission : modèle ${missionModel} pour les tâches longues` : 'Mode mission (définissez un modèle « mission » dans Paramètres → Hermes)'}>
+            <Crosshair size={14} /> Mission
+          </button>
          <button className={`toggle-btn${autoSpeak && speechProvider !== 'off' ? ' on' : ''}`} onClick={() => update({ speech: { autoSpeak: !autoSpeak } })} title="Lire les réponses à voix haute">
            {autoSpeak && speechProvider !== 'off' ? <Volume2 size={12} /> : <VolumeX size={12} />} voix
          </button>
@@ -1,13 +1,37 @@
-import { Maximize2, X, Minus } from 'lucide-react';
+import { Maximize2, X, Minus, Moon, Crosshair } from 'lucide-react';
+import { useShallow } from 'zustand/react/shallow';
 import { bridge } from '../../lib/bridge';
 import { useSettings } from '../../state/settings';
+import { useChat } from '../../state/chat';
+import { useVoice } from '../../state/voice';
 import { CoreStage } from '../hud/CoreStage';
 import { ChatPanel } from '../chat/ChatPanel';

+const STATE_LABEL: Record<string, string> = {
+  idle: 'veille',
+  listening: 'écoute',
+  thinking: 'réflexion',
+  speaking: 'parle',
+  alert: 'alerte',
+  error: 'erreur',
+  success: 'terminé'
+};
+
+/** Compact mode: a glanceable strip (state, unread, last sentence) above the mini chat. */
 export function CompactWidget() {
  const assistantName = useSettings((s) => s.settings.assistantName);
  const compactOpacity = useSettings((s) => s.settings.ui.compactOpacity);
+  const { hud, hudOverride, unread, missionMode, quiet, last } = useChat(
+    useShallow((s) => {
+      const lastAssistant = [...s.messages].reverse().find((m) => m.role === 'assistant' && m.content.trim());
+      return { hud: s.hud, hudOverride: s.hudOverride, unread: s.unread, missionMode: s.missionMode, quiet: s.quiet, last: lastAssistant?.content ?? '' };
+    })
+  );
+  const wake = useVoice((s) => s.wake);
  const api = bridge();
+  const state = hudOverride ?? hud;
+  const label = state === 'idle' && wake === 'spotting' ? 'à l’écoute du mot-clé' : STATE_LABEL[state] ?? state;
+  const sentence = last.replace(/```[\s\S]*?```/g, ' ').replace(/[#*_>`|]/g, '').replace(/\s+/g, ' ').trim().slice(0, 140);
  return (
    <div className="compact-root" style={{ ['--compact-opacity' as string]: compactOpacity }}>
      <div className="compact-head">
@@ -21,6 +45,15 @@ export function CompactWidget() {
        )}
      </div>
      <CoreStage compact />
+      <div className="compact-glance" onClick={() => useChat.getState().markRead()} title={sentence}>
+        <span className={`state${state === 'alert' || state === 'error' ? ' alert' : ''}`}>{label}</span>
+        <span className="last">{sentence || 'Aucun message pour l’instant.'}</span>
+        <span className="badges">
+          {missionMode && <Crosshair size={12} className="mission" aria-label="mode mission" />}
+          {quiet && <Moon size={12} aria-label="heures calmes" />}
+          {unread > 0 && <span className="unread" aria-label={`${unread} non lus`}>{unread}</span>}
+        </span>
+      </div>
      <div className="compact-body">
        <ChatPanel compact />
      </div>
@@ -35,6 +35,8 @@ export function CoreStage({ compact }: Props) {
  const transcript = useVoice((s) => s.lastTranscript);
  const interim = useVoice((s) => s.interim);
  const voiceError = useVoice((s) => s.error);
+  const wake = useVoice((s) => s.wake);
+  const wakeKeywords = useVoice((s) => s.wakeKeywords);
  const assistantName = useSettings((s) => s.settings.assistantName);
  const reduceMotion = useSettings((s) => s.settings.ui.reduceMotion);
  const transport = useHermes((s) => s.transport);
@@ -57,6 +59,7 @@ export function CoreStage({ compact }: Props) {
  else if (hud === 'error') sub = previewText(error ?? voiceError ?? 'erreur', 90);
  else if (hud === 'speaking') sub = 'synthèse vocale';
  else if (transcript) sub = `« ${previewText(transcript, 80)} »`;
+  else if (wake === 'spotting') sub = `dites « ${wakeKeywords[0] ?? 'jarvis'} »${link === 'online' ? ` · ${transport}` : link === 'offline' ? ' · Hermes hors ligne' : ''}`;
  else sub = link === 'online' ? `liaison Hermes · ${transport}` : link === 'offline' ? 'Hermes hors ligne' : '';

  return (
@@ -1,6 +1,7 @@
 import { Fragment, useEffect, useState } from 'react';
 import { useShallow } from 'zustand/react/shallow';
 import {
+  AlertTriangle,
  Wrench, CalendarClock, Sparkles, Layers, History, Play, Pause, Trash2, Pencil, Plus, Save, X, RefreshCw,
  CheckCircle2, XCircle, Loader2, Bot, Cpu, GitBranch, Webhook, FolderOpen
 } from 'lucide-react';
@@ -41,6 +42,7 @@ function ActivityTab() {

 function JobsTab() {
  const jobs = useHermes((s) => s.jobs);
+  const jobsError = useHermes((s) => s.jobsError);
  const busy = useHermes((s) => s.busy);
  const lastSyncAt = useHermes((s) => s.lastSyncAt);
  const refreshJobs = useHermes((s) => s.refreshJobs);
@@ -95,8 +97,9 @@ function JobsTab() {
          </div>
        </div>
      )}
+      {jobsError && <div className="test-result fail" style={{ marginBottom: 8 }}><AlertTriangle size={13} /> Crons indisponibles : {jobsError}</div>}
      {jobs.length === 0 ? (
-        <div className="empty">Aucun cron Hermes. Créez une mission planifiée en langage naturel.</div>
+        <div className="empty">{jobsError ? 'La liaison Hermes reste utilisable ; seule la planification est indisponible.' : 'Aucun cron Hermes. Créez une mission planifiée en langage naturel.'}</div>
      ) : (
        <div className="list">
          {jobs.map((job) => {
@@ -1,5 +1,5 @@
 import { useEffect } from 'react';
-import { Download, Trash2, X, CheckCircle2, Cpu, Mic, Volume2, AlertTriangle, RefreshCw } from 'lucide-react';
+import { Download, Trash2, X, CheckCircle2, Cpu, Mic, Volume2, AlertTriangle, RefreshCw, Ear, Activity } from 'lucide-react';
 import type { VoiceModelStatus } from '../../../shared/voice';
 import { bridge } from '../../lib/bridge';
 import { useShallow } from 'zustand/react/shallow';
@@ -19,20 +19,25 @@ function ModelRow({ model }: { model: VoiceModelStatus }) {
  const isActive = model.kind === 'stt' ? activeStt === model.id : activeTts === model.id;
  const busy = !!progress && (progress.phase === 'download' || progress.phase === 'extract');

+  const wakeMode = useSettings((s) => s.settings.voice.wakeMode);
  const activate = () => {
    if (model.kind === 'stt') update({ voice: { localModel: model.id, provider: 'local' } });
+    else if (model.kind === 'kws') update({ voice: { wakeMode: 'kws' } });
+    else if (model.kind === 'vad') update({ voice: { neuralVad: true } });
    else update({ speech: { localModel: model.id, provider: 'local', localSpeaker: model.speakers?.[0]?.id ?? 0 } });
  };
+  const neuralVad = useSettings((s) => s.settings.voice.neuralVad);
+  const activeNow = model.kind === 'kws' ? wakeMode === 'kws' : model.kind === 'vad' ? neuralVad : isActive;

  return (
-    <div className="row-item" style={isActive && model.installed ? { borderColor: 'var(--accent)' } : undefined}>
+    <div className="row-item" style={activeNow && model.installed ? { borderColor: 'var(--accent)' } : undefined}>
      <div className="main">
        <div className="row wrap" style={{ gap: 6 }}>
          <span className="name">{model.name}</span>
          {model.recommended && <span className="status-pill ok">recommandé</span>}
          {model.installed && <span className="status-pill online">installé · {formatMb(model.installedBytes)}</span>}
          {!model.installed && !busy && <span className="status-pill">{model.sizeMb} Mo</span>}
-          {isActive && model.installed && <span className="status-pill active">actif</span>}
+          {activeNow && model.installed && <span className="status-pill active">actif</span>}
        </div>
        <span className="desc">{model.description}</span>
        <span className="meta">langues : {model.languages.join(', ')}{model.speakers ? ` · ${model.speakers.length} voix` : ''}</span>
@@ -49,7 +54,7 @@ function ModelRow({ model }: { model: VoiceModelStatus }) {
          <button className="icon-btn danger" title="Annuler" onClick={() => void cancel(model.id)}><X size={13} /></button>
        ) : model.installed ? (
          <>
-            {!isActive && <button className="btn small" onClick={activate}>Utiliser</button>}
+            {!activeNow && <button className="btn small" onClick={activate}>Utiliser</button>}
            <button className="icon-btn danger" title="Supprimer du disque" onClick={() => { if (confirm(`Supprimer « ${model.name} » ?`)) void remove(model.id); }}><Trash2 size={13} /></button>
          </>
        ) : (
@@ -72,6 +77,8 @@ export function ModelsSection() {

  const stt = models.filter((m) => m.kind === 'stt');
  const tts = models.filter((m) => m.kind === 'tts');
+  const kws = models.filter((m) => m.kind === 'kws');
+  const vad = models.filter((m) => m.kind === 'vad');
  return (
    <>
      <div className="card">
@@ -91,6 +98,15 @@ export function ModelsSection() {
          <button className="btn small ghost" onClick={() => { void refresh(); void checkEngine(); }}><RefreshCw size={13} /> Actualiser</button>
        </div>
      </div>
+      <div className="card">
+        <div className="section-title"><Ear size={12} /> Mot d’activation</div>
+        <div className="list">{kws.map((m) => <ModelRow key={m.id} model={m} />)}</div>
+      </div>
+      <div className="card">
+        <div className="section-title"><Activity size={12} /> Fin de phrase</div>
+        <div className="list">{vad.map((m) => <ModelRow key={m.id} model={m} />)}</div>
+        <span className="hint" style={{ display: 'block', marginTop: 6 }}>Utilisé automatiquement en écoute permanente dès qu’il est installé ; sinon le VAD énergétique prend le relais.</span>
+      </div>
      <div className="card">
        <div className="section-title"><Mic size={12} /> Reconnaissance vocale</div>
        <div className="list">{stt.map((m) => <ModelRow key={m.id} model={m} />)}</div>
@@ -4,6 +4,7 @@ import { bridge } from '../../lib/bridge';
 import { HermesClient } from '../../services/hermes/client';
 import { listSystemVoices } from '../../services/voice/tts';
 import { speech } from '../../services/voice/speech';
+import { ensurePreferredVoice } from '../../services/voice/voicePreference';
 import { listMicrophones } from '../../services/voice/capture';
 import { transcribeWav } from '../../services/voice/stt';
 import { encodeWav } from '../../services/voice/wav';
@@ -13,7 +14,7 @@ import { useVoice } from '../../state/voice';
 import { installedModels, useVoiceModels } from '../../state/voiceModels';
 import { ModelsSection } from './ModelsSection';

-type Section = 'general' | 'hermes' | 'voice' | 'speech' | 'models' | 'webhook' | 'ui';
+type Section = 'general' | 'hermes' | 'voice' | 'speech' | 'models' | 'webhook' | 'notifications' | 'ui';

 interface Props {
  onClose: () => void;
@@ -33,6 +34,33 @@ function Toggle({ on, onChange, label, hint }: { on: boolean; onChange: (v: bool

 type TestState = { status: 'idle' | 'running' | 'ok' | 'fail'; message: string };

+function WakeStatus() {
+  const wake = useVoice((s) => s.wake);
+  const keywords = useVoice((s) => s.wakeKeywords);
+  const error = useVoice((s) => s.error);
+  const models = useVoiceModels((s) => s.models);
+  const download = useVoiceModels((s) => s.download);
+  const progress = useVoiceModels((s) => s.progress['kws-en']);
+  const installed = models.find((m) => m.id === 'kws-en')?.installed;
+  if (installed === false) {
+    return (
+      <div className="test-result fail" style={{ marginTop: 8 }}>
+        <XCircle size={13} />
+        <span>Détecteur non installé.</span>
+        <button className="btn small primary" style={{ marginLeft: 'auto' }} disabled={!!progress} onClick={() => void download('kws-en')}>
+          {progress ? `${progress.percent}%` : 'Télécharger (4 Mo)'}
+        </button>
+      </div>
+    );
+  }
+  return (
+    <div className={`test-result ${wake === 'spotting' ? 'ok' : wake === 'error' ? 'fail' : ''}`} style={{ marginTop: 8 }}>
+      {wake === 'spotting' ? <CheckCircle2 size={13} /> : wake === 'error' ? <XCircle size={13} /> : <Loader2 size={13} className="spin" />}
+      <span>{wake === 'spotting' ? `à l’écoute de « ${keywords.join(' », « ')} »` : wake === 'error' ? error ?? 'erreur' : 'démarrage…'}</span>
+    </div>
+  );
+}
+
 function TestResult({ t }: { t: TestState }) {
  if (t.status === 'idle') return null;
  return (
@@ -124,6 +152,7 @@ export function SettingsDrawer({ onClose }: Props) {
    ['speech', 'Voix / TTS'],
    ['models', 'Modèles locaux'],
    ['webhook', 'Webhook'],
+    ['notifications', 'Notifications'],
    ['general', 'Général'],
    ['ui', 'Interface']
  ];
@@ -188,6 +217,11 @@ export function SettingsDrawer({ onClose }: Props) {
                </div>
              </div>
              <div className="field">
+              <label>Modèle « mission » (tâches longues)</label>
+              <input className="input" list="hermes-models" value={settings.hermes.missionModel} placeholder="(même modèle)" onChange={(e) => update({ hermes: { missionModel: e.target.value } })} />
+              <span className="hint">Utilisé quand le mode Mission est activé dans la barre de commande. Laissez vide pour garder le modèle principal.</span>
+            </div>
+            <div className="field">
                <label>Instructions EveFlow (superposées au prompt Hermes)</label>
                <textarea className="textarea" value={settings.hermes.instructions} onChange={(e) => update({ hermes: { instructions: e.target.value } })} />
              </div>
@@ -278,13 +312,43 @@ export function SettingsDrawer({ onClose }: Props) {
            </div>
            <Toggle on={settings.voice.handsFree} onChange={(v) => { update({ voice: { handsFree: v } }); useVoice.getState().setHandsFree(v); }} label="Mains libres au démarrage" hint="Le micro se réactive automatiquement après chaque réponse." />
            <Toggle on={settings.voice.wakeChime} onChange={(v) => update({ voice: { wakeChime: v } })} label="Signal sonore d’écoute" />
-            <Toggle on={settings.voice.wakeWordEnabled} onChange={(v) => update({ voice: { wakeWordEnabled: v } })} label="Mot d’activation en mains libres" hint="Seules les phrases commençant par ce mot sont envoyées à Hermes ; le reste est ignoré. Recommandé avec la reconnaissance locale." />
-            {settings.voice.wakeWordEnabled && (
-              <div className="field" style={{ marginTop: 8 }}>
-                <label>Mot d’activation</label>
-                <input className="input" value={settings.voice.wakeWord} placeholder="jarvis" onChange={(e) => update({ voice: { wakeWord: e.target.value.toLowerCase() } })} />
+            <div className="field" style={{ marginTop: 10 }}>
+              <label>Mot d’activation</label>
+              <select className="select" value={settings.voice.wakeMode} onChange={(e) => update({ voice: { wakeMode: e.target.value as typeof settings.voice.wakeMode, wakeWordEnabled: e.target.value === 'transcript' } })}>
+                <option value="off">Désactivé (bouton, raccourci ou mains libres)</option>
+                <option value="kws">Écoute permanente : le micro reste ouvert et réagit au mot-clé (recommandé)</option>
+                <option value="transcript">Filtre après transcription (mains libres) : les phrases sans le mot sont ignorées</option>
+              </select>
+              <span className="hint">
+                {settings.voice.wakeMode === 'kws'
+                  ? 'Détection locale par un modèle de 3 Mo, quasi gratuite en CPU. Dire le mot seul ouvre l’écoute ; dire le mot puis la commande envoie directement. Le mot coupe aussi la voix en cours.'
+                  : settings.voice.wakeMode === 'transcript'
+                    ? 'Chaque phrase est transcrite puis filtrée : plus coûteux, à réserver à la reconnaissance locale.'
+                    : 'Le micro s’active avec le bouton, Ctrl+Shift+Espace ou la boucle mains libres.'}
+              </span>
+            </div>
+            {settings.voice.wakeMode !== 'off' && (
+              <div className="grid-2">
+                <div className="field">
+                  <label>Mot-clé</label>
+                  <input className="input" value={settings.voice.wakeWord} placeholder="jarvis" onChange={(e) => update({ voice: { wakeWord: e.target.value.toLowerCase().slice(0, 40) } })} />
+                  <span className="hint">Prononciation anglaise conseillée (« jarvis », « hey jarvis », « computer », « friday »…).</span>
+                </div>
+                {settings.voice.wakeMode === 'kws' && (
+                  <div className="field">
+                    <label>Sensibilité du mot-clé : {settings.voice.kwsSensitivity}/5</label>
+                    <input className="range" type="range" min={1} max={5} step={1} value={settings.voice.kwsSensitivity} onChange={(e) => update({ voice: { kwsSensitivity: Number(e.target.value) } })} />
+                  </div>
+                )}
              </div>
            )}
+            {settings.voice.wakeMode === 'kws' && (
+              <>
+                <WakeStatus />
+                <Toggle on={settings.voice.neuralVad} onChange={(v) => update({ voice: { neuralVad: v } })} label="Fin de phrase neuronale (Silero)" hint="Coupe l’écoute au bon moment, même avec du bruit de fond. Nécessite le modèle Silero VAD (0,6 Mo) dans Modèles locaux." />
+              </>
+            )}
+            <Toggle on={settings.voice.localCommands} onChange={(v) => update({ voice: { localCommands: v } })} label="Commandes locales instantanées" hint="« Verrouille la session », « monte le son », « ouvre Spotify », « regarde mon écran »… exécutées sur ce PC sans passer par Hermes." />
            <div className="row" style={{ marginTop: 10 }}>
              <button className="btn small" onClick={() => void testStt()} disabled={settings.voice.provider === 'browser'}><Mic size={13} /> Tester la reconnaissance</button>
            </div>
@@ -294,6 +358,14 @@ export function SettingsDrawer({ onClose }: Props) {

        {section === 'speech' && (
          <div className="card">
+            <div className="field">
+              <label>Voix</label>
+              <div className="segmented">
+                <button className={(settings.speech.voiceGender ?? 'male') === 'male' ? 'active' : ''} onClick={() => { update({ speech: { voiceGender: 'male' } }); void ensurePreferredVoice().then((m) => m && setTtsTest({ status: 'ok', message: m })); }}>Masculine</button>
+                <button className={settings.speech.voiceGender === 'female' ? 'active' : ''} onClick={() => { update({ speech: { voiceGender: 'female' } }); void ensurePreferredVoice().then((m) => m && setTtsTest({ status: 'ok', message: m })); }}>Féminine</button>
+              </div>
+              <span className="hint">S’applique à tous les moteurs : voix locale (Piper Tom ou Pierre pour le masculin, téléchargée automatiquement si besoin), voix OpenAI (onyx / nova), voix système Windows (Paul / Hortense).</span>
+            </div>
            <div className="field">
              <label>Moteur de synthèse</label>
              <select className="select" value={settings.speech.provider} onChange={(e) => update({ speech: { provider: e.target.value as typeof settings.speech.provider } })}>
@@ -414,6 +486,14 @@ export function SettingsDrawer({ onClose }: Props) {
  -d '{"role":"assistant","text":"Rapport terminé","source":"telegram"}'`}</pre>
              <span className="hint">Formats acceptés : {'{role,text}'}, {'{event:"run.completed",input,output}'}, {'{event:"job.completed",job:{name},output,status}'}, {'{type:"message",payload:{text}}'}.</span>
            </div>
+            <div className="field">
+              <label>Serveur MCP pour Hermes (outils du PC : écran, applications, volume, presse-papiers, voix)</label>
+              <pre className="input" style={{ whiteSpace: 'pre-wrap', fontSize: 11.5, userSelect: 'text' }}>{`# ~/.hermes/config.yaml côté Hermes
+mcp_servers:
+  eveflow:
+    url: "http://<ip-de-ce-pc>:${settings.webhook.port}/mcp"${settings.webhook.secret ? '\n    headers:\n      Authorization: "Bearer <secret>"' : ''}`}</pre>
+              <span className="hint">Même port et même secret que le webhook. Sans secret, EveFlow n’écoute qu’en local (127.0.0.1) : définissez un secret pour un Hermes distant.</span>
+            </div>
            <div className="row">
              <button className="btn primary small" onClick={() => void applyWebhook()} disabled={!bridge()}>Appliquer et redémarrer</button>
              {hermesWebhook && <span className={`status-pill ${hermesWebhook.listening ? 'online' : 'offline'}`}>{hermesWebhook.listening ? `port ${hermesWebhook.port}` : hermesWebhook.error ?? 'inactif'}</span>}
@@ -421,6 +501,35 @@ export function SettingsDrawer({ onClose }: Props) {
          </div>
        )}

+        {section === 'notifications' && (
+          <div className="card">
+            <Toggle on={settings.notifications.quietEnabled} onChange={(v) => update({ notifications: { quietEnabled: v } })} label="Heures calmes" hint="Les messages poussés (crons, Telegram) s’affichent sans être lus à voix haute ni faire clignoter le noyau ; le HUD passe en mode nuit." />
+            <div className="grid-2">
+              <div className="field">
+                <label>Début</label>
+                <input className="input" type="time" value={settings.notifications.quietStart} onChange={(e) => update({ notifications: { quietStart: e.target.value } })} />
+              </div>
+              <div className="field">
+                <label>Fin</label>
+                <input className="input" type="time" value={settings.notifications.quietEnd} onChange={(e) => update({ notifications: { quietEnd: e.target.value } })} />
+              </div>
+            </div>
+            <Toggle on={settings.notifications.nightTheme} onChange={(v) => update({ notifications: { nightTheme: v } })} label="Thème nuit pendant les heures calmes" />
+            <div className="field" style={{ marginTop: 10 }}>
+              <label>Mots prioritaires (lus même en heures calmes)</label>
+              <input className="input" value={settings.notifications.priorityKeywords} onChange={(e) => update({ notifications: { priorityKeywords: e.target.value } })} placeholder="urgent, alerte, panne" />
+              <span className="hint">Séparés par des virgules ; comparés au texte et au nom du cron. Les échecs de crons sont toujours prioritaires.</span>
+            </div>
+            <Toggle on={settings.notifications.summarizeIncoming} onChange={(v) => update({ notifications: { summarizeIncoming: v } })} label="Résumé vocal des messages entrants" hint="Seules les premières phrases sont lues ; le message complet reste dans le fil." />
+            {settings.notifications.summarizeIncoming && (
+              <div className="field">
+                <label>Phrases lues : {settings.notifications.summarySentences}</label>
+                <input className="range" type="range" min={1} max={5} step={1} value={settings.notifications.summarySentences} onChange={(e) => update({ notifications: { summarySentences: Number(e.target.value) } })} />
+              </div>
+            )}
+          </div>
+        )}
+
        {section === 'general' && (
          <div className="card">
            <div className="grid-2">
@@ -0,0 +1,43 @@
+/** Quiet-hours and spoken-summary helpers (pure, unit tested). */
+
+function minutesOf(hhmm: string): number | null {
+  const m = /^(\d{1,2}):(\d{2})$/.exec(hhmm.trim());
+  if (!m) return null;
+  const h = Number(m[1]);
+  const min = Number(m[2]);
+  if (h > 23 || min > 59) return null;
+  return h * 60 + min;
+}
+
+/** True when `now` falls inside [start, end), with ranges crossing midnight ("22:30" → "07:30"). */
+export function isQuietTime(start: string, end: string, now: Date = new Date()): boolean {
+  const s = minutesOf(start);
+  const e = minutesOf(end);
+  if (s === null || e === null || s === e) return false;
+  const cur = now.getHours() * 60 + now.getMinutes();
+  return s < e ? cur >= s && cur < e : cur >= s || cur < e;
+}
+
+/** True when the text or job name contains one of the comma-separated priority words. */
+export function isPriority(text: string, keywords: string, jobName = ''): boolean {
+  const words = keywords
+    .split(/[,;\n]/)
+    .map((w) => w.trim().toLowerCase())
+    .filter(Boolean);
+  if (!words.length) return false;
+  const hay = `${jobName} ${text}`.toLowerCase();
+  return words.some((w) => hay.includes(w));
+}
+
+/** First `count` sentences of a text, for a short spoken summary of a long push. */
+export function summarize(text: string, count: number): string {
+  const clean = text
+    .replace(/```[\s\S]*?```/g, ' ')
+    .replace(/[#*_>`|]/g, '')
+    .replace(/\s+/g, ' ')
+    .trim();
+  if (!clean) return '';
+  const sentences = clean.match(/[^.!?…]+[.!?…]+["»)]?\s*|[^.!?…]+$/g) ?? [clean];
+  const picked = sentences.slice(0, Math.max(1, count)).join('').trim();
+  return picked.length < clean.length ? picked : clean;
+}
@@ -154,3 +154,27 @@ export function formatDuration(seconds: number): string {
  if (h > 0) return `${h}h ${m.toString().padStart(2, '0')}m`;
  return `${m}m ${(s % 60).toString().padStart(2, '0')}s`;
 }
+
+const NOISE_PHRASES = [
+  'sous-titres réalisés par la communauté d\'amara.org',
+  'sous-titrage société radio-canada',
+  'merci d\'avoir regardé',
+  'abonnez-vous',
+  'thank you for watching',
+  'thanks for watching',
+  '...'
+];
+
+/**
+ * Whisper-style hallucinations on silence or clicks: "(cliquant)", "*Claire*", "[Musique]",
+ * "Sous-titres réalisés par…". Such transcripts must not be sent to Hermes.
+ */
+export function isTranscriptNoise(text: string): boolean {
+  const t = text.trim();
+  if (!t) return true;
+  if (/^[\s\p{P}\p{S}]*$/u.test(t)) return true;
+  // whole transcript wrapped in brackets/asterisks: a sound description
+  if (/^[(\[*«"'\s]+[^()\[\]*]{0,60}[)\]*»"'\s]+$/u.test(t) && !/[a-zà-ÿ]{3,}\s+[a-zà-ÿ]{3,}\s+[a-zà-ÿ]{3,}/i.test(t)) return true;
+  const lower = t.toLowerCase().replace(/[.!?…\s]+$/u, '');
+  return NOISE_PHRASES.some((p) => lower === p.replace(/[.!?…\s]+$/u, ''));
+}
@@ -0,0 +1,66 @@
+/** Voice gender preference applied to every TTS provider (pure helpers, unit tested). */
+import type { VoiceModelStatus } from '../../shared/voice';
+
+export type VoiceGender = 'male' | 'female';
+
+const MALE_HINTS = /\b(homme|male|masculin|paul|thomas|claude|henri|guillaume|mathieu|antoine|nicolas|denis|pierre|tom|adam|michael|eric|liam|george|lewis|daniel|fenrir|puck|onyx|echo|david|mark|richard|james|ryan|guy)\b/i;
+const FEMALE_HINTS = /\b(femme|female|f[ée]minin|hortense|julie|denise|eloise|am[ée]lie|audrey|siwis|jessica|heart|bella|sarah|nicole|sky|alloy|nova|shimmer|zira|aria|jenny|emma|isabella|sophie|charlotte|vivienne|coral|sage)\b/i;
+
+/** Best-effort gender from a voice or speaker label ("Piper Tom (homme, français)", "Microsoft Paul", "am_adam"). */
+export function inferGender(name: string): VoiceGender | undefined {
+  const n = name.replace(/_/g, ' ');
+  if (/^(am|bm|em|hm|im|jm|pm|zm)\b/i.test(n) || MALE_HINTS.test(n)) return 'male';
+  if (/^(af|bf|ef|ff|hf|if|jf|pf|zf)\b/i.test(n) || FEMALE_HINTS.test(n)) return 'female';
+  return undefined;
+}
+
+/** OpenAI-compatible default voice for the preferred gender (used when the user left the field empty). */
+export function defaultOpenAiVoice(gender: VoiceGender): string {
+  return gender === 'male' ? 'onyx' : 'nova';
+}
+
+/** Rank system voices: language first, then gender, then quality hints. */
+export function rankSystemVoice(v: { name: string; lang: string; localService?: boolean }, lang: string, gender: VoiceGender): number {
+  const name = v.name.toLowerCase();
+  let score = v.lang.toLowerCase().startsWith(lang) ? 100 : 0;
+  const g = inferGender(v.name);
+  if (g === gender) score += 50;
+  else if (g && g !== gender) score -= 30;
+  if (name.includes('natural') || name.includes('neural') || name.includes('online')) score += 20;
+  if (name.includes('google')) score += 10;
+  if (name.includes('microsoft')) score += 5;
+  if (v.localService) score += 2;
+  return score;
+}
+
+export interface LocalVoiceChoice {
+  modelId: string;
+  speaker: number;
+}
+
+/** Installed local voice matching language + gender, or null. */
+export function findLocalVoice(models: VoiceModelStatus[], lang: string, gender: VoiceGender): LocalVoiceChoice | null {
+  const l = lang.toLowerCase().split('-')[0];
+  const installed = models.filter((m) => m.kind === 'tts' && m.installed);
+  for (const model of installed) {
+    for (const sp of model.speakers ?? []) {
+      const spLang = (sp.lang || '').toLowerCase();
+      if (spLang && spLang !== l && spLang !== 'multi') continue;
+      const g = (sp as { gender?: string }).gender === 'm' ? 'male' : (sp as { gender?: string }).gender === 'f' ? 'female' : inferGender(sp.name);
+      if (g === gender) return { modelId: model.id, speaker: sp.id };
+    }
+  }
+  return null;
+}
+
+/** Catalog model to download when nothing installed offers the preferred gender (French). */
+export function suggestedDownload(models: VoiceModelStatus[], lang: string, gender: VoiceGender): string | null {
+  const l = lang.toLowerCase().split('-')[0];
+  if (l !== 'fr') return null;
+  const wanted = gender === 'male' ? ['piper-fr-tom', 'piper-fr-upmc'] : ['kokoro-v1', 'piper-fr-siwis'];
+  for (const id of wanted) {
+    const m = models.find((x) => x.id === id);
+    if (m && !m.installed) return id;
+  }
+  return null;
+}
@@ -9,9 +9,12 @@ import { previewText } from '../lib/text';
 import { useChat, type PendingRequest } from '../state/chat';
 import { useHermes } from '../state/hermes';
 import { useSettings } from '../state/settings';
-import { executeLocalTool, LOCAL_TOOL_DEFINITIONS } from './hermes/localTools';
+import { executeLocalTool, LOCAL_TOOL_DEFINITIONS, type LocalToolContext } from './hermes/localTools';
+import { isPriority, summarize } from '../lib/quietHours';
 import type { HermesStreamEvent, SendHandle } from './hermes/types';
 import { speech } from './voice/speech';
+import { bridge } from '../lib/bridge';
+import { parseLocalIntent, runLocalIntent } from './localCommands';

 let active: SendHandle | null = null;
 let activeMessageId: string | null = null;
@@ -46,7 +49,7 @@ export function pingCore(): void {
  useChat.getState().ping();
 }

-function localToolContext() {
+export function localToolContext(): LocalToolContext {
  const chat = useChat.getState();
  const settings = useSettings.getState().settings;
  return {
@@ -58,13 +61,20 @@ function localToolContext() {
      setTimeout(() => useChat.getState().setHudOverride(null), 6000);
    },
    speak: (text: string) => speech.say(text),
+    showMessage: (text: string, title?: string) => {
+      if (!text.trim()) return;
+      useChat.getState().addMessage({ role: 'assistant', content: text, source: 'hermes', jobName: title, status: 'done' });
+    },
    getStatus: () => ({
      transport: useHermes.getState().transport,
      link: useHermes.getState().link,
      assistant: settings.assistantName,
      messages: chat.messages.length,
      speaking: speech.isSpeaking(),
-      handsFree: settings.voice.handsFree
+      handsFree: settings.voice.handsFree,
+      quietHours: chat.quiet,
+      missionMode: chat.missionMode,
+      wakeMode: settings.voice.wakeMode
    }),
    getHistory: (n: number) => chat.messages.slice(-n).map((m) => ({ role: m.role, content: m.content })),
    notify: (title: string, body: string) => {
@@ -165,9 +175,49 @@ function addPending(request: Omit<PendingRequest, 'id' | 'createdAt'>): void {
  speech.say(request.kind === 'approval' ? 'Autorisation requise.' : request.kind === 'clarify' ? request.description : 'Saisie requise.', { interrupt: false });
 }

+/** Screenshot of the primary display as a data URL (Electron only). */
+export async function captureScreen(): Promise<string | null> {
+  const api = bridge();
+  if (!api) return null;
+  try {
+    return await api.system.captureScreen(1600);
+  } catch (err) {
+    useChat.getState().setError(`Capture d’écran : ${(err as Error).message}`);
+    return null;
+  }
+}
+
+/** Short system intents handled on the machine without a Hermes round trip. Returns true when consumed. */
+async function tryLocalIntent(text: string, source: string): Promise<{ handled: boolean; images?: string[]; text?: string }> {
+  const settings = useSettings.getState().settings;
+  if (!settings.voice.localCommands || !bridge()) return { handled: false };
+  const intent = parseLocalIntent(text);
+  if (!intent) return { handled: false };
+  if (intent.kind === 'screenshot') {
+    const shot = await captureScreen();
+    if (!shot) return { handled: false };
+    return { handled: false, images: [shot], text: intent.question || text };
+  }
+  const chat = useChat.getState();
+  chat.addMessage({ role: 'user', content: text, source, status: 'done' });
+  const result = await runLocalIntent(intent);
+  chat.addMessage({ role: 'assistant', content: result.message, source: 'local', status: result.ok ? 'done' : 'error' });
+  chat.setHud(result.ok ? 'idle' : 'error');
+  if (settings.speech.autoSpeak) speech.say(result.message, { interrupt: true });
+  return { handled: true };
+}
+
 export async function sendMessage(text: string, images: string[] = [], source = 'eveflow'): Promise<void> {
-  const trimmed = text.trim();
+  let trimmed = text.trim();
  if ((!trimmed && images.length === 0) || active) return;
+  if (trimmed && images.length === 0) {
+    const local = await tryLocalIntent(trimmed, source);
+    if (local.handled) return;
+    if (local.images) {
+      images = local.images;
+      trimmed = (local.text ?? trimmed).trim();
+    }
+  }
  const chat = useChat.getState();
  const hermes = useHermes.getState();
  const settings = useSettings.getState().settings;
@@ -189,7 +239,9 @@ export async function sendMessage(text: string, images: string[] = [], source =
  const timing = { firstToken: null as number | null, startedAt };
  const runIdRef = { id: null as string | null };

-  const client = hermes.client();
+  const mission = chat.missionMode && settings.hermes.missionModel.trim();
+  const client = hermes.client(mission ? settings.hermes.missionModel : undefined);
+  if (mission) Log.info('hermes', `mission mode → model ${settings.hermes.missionModel}`);
  const handle = client.send(
    {
      text: trimmed || 'Analyse cette image.',
@@ -321,11 +373,17 @@ export function handlePush(event: HermesPushEvent): void {
    chat.pushActivity({ kind: 'job', name: event.jobName || 'cron', status: event.status?.includes('fail') ? 'error' : 'done', detail: previewText(event.text, 160) });
    void useHermes.getState().refreshJobs();
  }
-  if (event.role !== 'user' && settings.speech.speakIncoming) {
-    speech.say(event.jobName ? `Résultat de ${event.jobName}. ${event.text}` : event.text);
+  const n = settings.notifications;
+  const priority = isPriority(event.text, n.priorityKeywords, event.jobName) || !!event.status?.includes('fail');
+  const silenced = chat.quiet && !priority;
+  if (document.hidden || silenced) chat.incUnread();
+  if (event.role !== 'user' && settings.speech.speakIncoming && !silenced) {
+    const body = n.summarizeIncoming ? summarize(event.text, n.summarySentences) : event.text;
+    speech.say(event.jobName ? `Résultat de ${event.jobName}. ${body}` : body);
  }
+  if (silenced) Log.info('webhook', 'quiet hours: push shown silently');
  chat.setHud(event.status?.includes('fail') ? 'alert' : 'success');
-  pingCore();
+  if (!silenced) pingCore();
  setTimeout(() => {
    const s = useChat.getState();
    if (s.hud === 'success' || s.hud === 'alert') s.setHud(speech.isSpeaking() ? 'speaking' : 'idle');
@@ -430,6 +430,7 @@ export class HermesClient {
      let iterationText = '';
      const errorChunks: string[] = [];
      let ready = false;
+      let rawBody = '';
      const parser = new SseParser((message) => {
        if (message.data === '[DONE]') return;
        const data = tryParseJson<unknown>(message.data);
@@ -451,7 +452,13 @@ export class HermesClient {
      if (sessionId) headers['X-Hermes-Session-Id'] = sessionId;
      const handle = await httpStream(
        { url: `${this.base}/v1/chat/completions`, method: 'POST', headers, body: JSON.stringify(payload), timeoutMs: 10 * 60_000 },
-        { onChunk: (text) => (ready ? parser.feed(text) : errorChunks.push(text)) }
+        {
+          onChunk: (text) => {
+            if (rawBody.length < 512_000) rawBody += text;
+            if (ready) parser.feed(text);
+            else errorChunks.push(text);
+          }
+        }
      );
      currentHandle = handle;
      if (!handle.start.ok) {
@@ -467,11 +474,24 @@ export class HermesClient {
        throw new HttpError(handle.start.status, detail);
      }
      ready = true;
+      // Chunks that raced ahead of the start event were buffered as potential error bodies: replay them.
+      for (const chunk of errorChunks.splice(0)) parser.feed(chunk);
      const rotated = handle.start.headers['x-hermes-session-id'];
      if (rotated && rotated !== options.sessionId) onEvent({ kind: 'session', sessionId: rotated });

      await handle.done;
      parser.end();
+      if (!iterationText && toolCalls.size === 0 && !aborted()) {
+        // Nothing streamed: the server may have answered with a plain JSON completion (stream ignored)
+        // or with a 200 carrying an error object. Surface it instead of an empty bubble.
+        const recovered = recoverCompletion(rawBody);
+        if (recovered.text) {
+          iterationText = recovered.text;
+          onEvent({ kind: 'delta', text: recovered.text });
+        } else {
+          throw new Error(recovered.error ?? `Réponse vide de Hermes (${rawBody.length} octets reçus${rawBody ? ` : ${rawBody.slice(0, 160).replace(/\s+/g, ' ')}` : ''})`);
+        }
+      }
      fullText += iterationText;

      if (finishReason === 'tool_calls' && toolsAllowed && toolCalls.size > 0 && !aborted()) {
@@ -496,6 +516,30 @@ export class HermesClient {
  }
 }

+/** Extract text or an error message from a non-streamed chat completion body. */
+export function recoverCompletion(raw: string): { text?: string; error?: string } {
+  const body = raw.trim();
+  if (!body) return {};
+  const candidates = body.startsWith('data:') || body.startsWith('event:')
+    ? body.split(/\n+/).filter((l) => l.startsWith('data:')).map((l) => l.replace(/^data:\s*/, '')).filter((l) => l && l !== '[DONE]')
+    : [body];
+  let text = '';
+  for (const c of candidates) {
+    const data = tryParseJson<Rec>(c);
+    if (!data || !isRec(data)) continue;
+    if (data.error) {
+      const e = data.error as Rec | string;
+      return { error: `Hermes : ${typeof e === 'string' ? e : String((e as Rec).message ?? JSON.stringify(e))}` };
+    }
+    const choice = Array.isArray(data.choices) && isRec(data.choices[0]) ? (data.choices[0] as Rec) : null;
+    const msg = choice && isRec(choice.message) ? (choice.message as Rec) : choice && isRec(choice.delta) ? (choice.delta as Rec) : null;
+    if (msg && typeof msg.content === 'string') text += msg.content;
+    else if (typeof data.output === 'string') text += data.output;
+    else if (typeof data.text === 'string') text += data.text;
+  }
+  return text ? { text } : {};
+}
+
 /** Session ids created by the sessions transport carry an `hs:` prefix; other endpoints get the bare id. */
 export function plainSession(id: string): string {
  return id.startsWith('hs:') ? id.slice(3) : id;
@@ -3,16 +3,35 @@
 * With the runs/sessions transports Hermes uses its own server-side toolsets instead.
 */
 import { bridge } from '../../lib/bridge';
+import type { SystemAction } from '../../../shared/bridge';

 export interface LocalToolContext {
  setEmotion: (emotion: string) => void;
+  showMessage?: (text: string, title?: string) => void;
  speak: (text: string) => void;
  getStatus: () => Record<string, unknown>;
  getHistory: (n: number) => Array<{ role: string; content: string }>;
  notify: (title: string, body: string) => void;
 }

+const fn = (name: string, description: string, properties: Record<string, unknown> = {}, required: string[] = []) => ({
+  type: 'function',
+  function: { name, description, parameters: { type: 'object', properties, required } }
+});
+
+/** Tools executed on the user's PC through the Electron main process (allow-listed). */
+export const SYSTEM_TOOL_DEFINITIONS = [
+  fn('lock_session', "Verrouille la session de l'utilisateur."),
+  fn('open_app', "Lance une application du PC de l'utilisateur (bloc-notes, calculatrice, chrome, spotify, vscode, terminal, explorateur…).", { name: { type: 'string' } }, ['name']),
+  fn('open_url', "Ouvre une URL http(s) dans le navigateur de l'utilisateur.", { url: { type: 'string' } }, ['url']),
+  fn('media_key', 'Touche média : volume-up, volume-down, mute, play-pause, next, previous.', { key: { type: 'string', enum: ['volume-up', 'volume-down', 'mute', 'play-pause', 'next', 'previous'] } }, ['key']),
+  fn('clipboard_get', 'Lit le texte du presse-papiers.'),
+  fn('clipboard_set', 'Place un texte dans le presse-papiers.', { text: { type: 'string' } }, ['text']),
+  fn('find_files', "Cherche des fichiers par nom dans Documents, Bureau, Téléchargements et Images (25 max).", { query: { type: 'string' } }, ['query'])
+];
+
 export const LOCAL_TOOL_DEFINITIONS = [
+  ...SYSTEM_TOOL_DEFINITIONS,
  {
    type: 'function',
    function: {
@@ -95,6 +114,29 @@ export async function executeLocalTool(name: string, rawArgs: string, ctx: Local
      case 'notify_user':
        ctx.notify(String(args.title ?? 'EveFlow'), String(args.body ?? ''));
        return JSON.stringify({ ok: true });
+      case 'show_message':
+        ctx.showMessage?.(String(args.text ?? ''), typeof args.title === 'string' ? args.title : undefined);
+        return JSON.stringify({ ok: true });
+      case 'lock_session':
+      case 'open_app':
+      case 'open_url':
+      case 'media_key':
+      case 'clipboard_get':
+      case 'clipboard_set':
+      case 'find_files': {
+        const api = bridge();
+        if (!api) return JSON.stringify({ error: 'system actions unavailable outside Electron' });
+        const action: SystemAction =
+          name === 'lock_session' ? { type: 'lock' }
+          : name === 'open_app' ? { type: 'open-app', name: String(args.name ?? '') }
+          : name === 'open_url' ? { type: 'open-url', url: String(args.url ?? '') }
+          : name === 'media_key' ? { type: 'media', key: String(args.key ?? '') as 'mute' }
+          : name === 'clipboard_get' ? { type: 'clipboard-read' }
+          : name === 'clipboard_set' ? { type: 'clipboard-write', text: String(args.text ?? '') }
+          : { type: 'find-files', query: String(args.query ?? '') };
+        const result = await api.system.action(action);
+        return JSON.stringify(result.ok ? { ok: true, message: result.message, data: result.data } : { error: result.message ?? 'failed' });
+      }
      case 'write_shared_file': {
        const api = bridge();
        if (!api) return JSON.stringify({ error: 'file system unavailable outside Electron' });
@@ -10,6 +10,8 @@ export interface HermesConfig {
  reasoningEffort: '' | 'low' | 'medium' | 'high';
  /** Extra instructions layered on top of the Hermes system prompt. */
  instructions: string;
+  /** Model used in "mission" mode (long tasks); empty = same as `model`. */
+  missionModel: string;
  /** Expose EveFlow client tools in chat-completions mode. */
  localTools: boolean;
 }
@@ -0,0 +1,75 @@
+/**
+ * Instant local commands: short French/English intents handled on the machine without a
+ * round trip to Hermes (lock, open app/url, volume, media, screenshot to Hermes).
+ * Anything unmatched goes to Hermes as usual.
+ */
+import type { SystemAction } from '../../shared/bridge';
+import { bridge } from '../lib/bridge';
+import { Log } from '../lib/log';
+
+export interface LocalIntent {
+  kind: 'action' | 'screenshot';
+  action?: SystemAction;
+  /** Spoken confirmation. */
+  reply: string;
+  /** For screenshot: the question to send to Hermes with the image. */
+  question?: string;
+}
+
+const norm = (t: string) =>
+  t
+    .normalize('NFD')
+    .replace(/[̀-ͯ]/g, '')
+    .toLowerCase()
+    .replace(/[’']/g, ' ')
+    .replace(/[^a-z0-9 :/._-]+/g, ' ')
+    .replace(/\s+/g, ' ')
+    .trim();
+
+const LOCK = /^(verrouille|verrouiller|bloque|lock)( (la |ma )?(session|le pc|l ordinateur|the (pc|computer|screen)))?$/;
+const VOL_UP = /^(monte|augmente|hausse|plus fort|volume plus|turn up|raise)( (le|the)? ?(son|volume))?( de \d+)?$/;
+const VOL_DOWN = /^(baisse|diminue|moins fort|volume moins|turn down|lower)( (le|the)? ?(son|volume))?( de \d+)?$/;
+const MUTE = /^(coupe|couper|mute|silence|desactive)( (le|the)? ?(son|volume|audio))?$/;
+const PLAY = /^(pause|play|lecture|reprends|reprendre|mets en pause|met en pause|stop la musique|arrete la musique)$/;
+const NEXT = /^(suivant|suivante|piste suivante|musique suivante|next|skip)$/;
+const PREV = /^(precedent|precedente|piste precedente|previous)$/;
+const OPEN_APP = /^(ouvre|ouvrir|lance|lancer|demarre|open|launch|start) ((l application|l appli|le logiciel|le programme|the app|le|la|les|l|un|une|the|moi) )*(.+)$/;
+const OPEN_URL = /^(ouvre|ouvrir|va sur|open|go to) (https?:\/\/\S+|(www\.)?[a-z0-9-]+\.[a-z]{2,}(\/\S*)?)$/;
+const SCREEN = /(regarde|regardes|analyse|decris|decrit|lis|explique|qu est ce qu il y a sur|que vois tu sur|what is on|look at|read) (mon |l |the )?(ecran|screen)|capture (d )?ecran|screenshot/;
+
+export function parseLocalIntent(text: string): LocalIntent | null {
+  const t = norm(text).replace(/^(jarvis|hey jarvis|ok jarvis)[ ,]*/, '');
+  if (!t) return null;
+  if (LOCK.test(t)) return { kind: 'action', action: { type: 'lock' }, reply: 'Session verrouillée.' };
+  if (MUTE.test(t)) return { kind: 'action', action: { type: 'media', key: 'mute' }, reply: 'Son coupé.' };
+  if (VOL_UP.test(t)) return { kind: 'action', action: { type: 'media', key: 'volume-up' }, reply: 'Volume augmenté.' };
+  if (VOL_DOWN.test(t)) return { kind: 'action', action: { type: 'media', key: 'volume-down' }, reply: 'Volume baissé.' };
+  if (PLAY.test(t)) return { kind: 'action', action: { type: 'media', key: 'play-pause' }, reply: 'Lecture.' };
+  if (NEXT.test(t)) return { kind: 'action', action: { type: 'media', key: 'next' }, reply: 'Piste suivante.' };
+  if (PREV.test(t)) return { kind: 'action', action: { type: 'media', key: 'previous' }, reply: 'Piste précédente.' };
+  if (SCREEN.test(t)) return { kind: 'screenshot', reply: 'Je regarde votre écran.', question: text.trim() };
+  const url = OPEN_URL.exec(t);
+  if (url) {
+    const target = url[2].startsWith('http') ? url[2] : `https://${url[2]}`;
+    return { kind: 'action', action: { type: 'open-url', url: target }, reply: `J’ouvre ${url[2]}.` };
+  }
+  const app = OPEN_APP.exec(t);
+  if (app) {
+    const name = app[app.length - 1].trim();
+    if (name.length <= 40 && !/\s(et|puis|and)\s/.test(name)) return { kind: 'action', action: { type: 'open-app', name }, reply: `J’ouvre ${name}.` };
+  }
+  return null;
+}
+
+/** Execute an action intent through the bridge. Returns the spoken outcome. */
+export async function runLocalIntent(intent: LocalIntent): Promise<{ ok: boolean; message: string }> {
+  const api = bridge();
+  if (!api || !intent.action) return { ok: false, message: 'Actions locales indisponibles hors Electron.' };
+  try {
+    const result = await api.system.action(intent.action);
+    Log.info('local', `${intent.action.type}: ${result.ok ? 'ok' : result.message}`);
+    return { ok: result.ok, message: result.ok ? intent.reply : result.message || 'Échec de la commande locale.' };
+  } catch (err) {
+    return { ok: false, message: (err as Error).message };
+  }
+}
@@ -0,0 +1,42 @@
+/**
+ * Answers tool calls that Hermes sends through the local MCP endpoint and that need the renderer
+ * (voice, notifications, HUD, chat). System tools are executed in the main process directly.
+ */
+import type { McpToolRequest } from '../../shared/ipc';
+import { bridge } from '../lib/bridge';
+import { Log } from '../lib/log';
+import { useChat } from '../state/chat';
+import { executeLocalTool, type LocalToolContext } from './hermes/localTools';
+
+let unsubscribe: (() => void) | null = null;
+
+export function initMcpBridge(context: () => LocalToolContext): void {
+  const api = bridge();
+  if (!api || unsubscribe) return;
+  unsubscribe = api.hermes.onToolRequest((req: McpToolRequest) => {
+    void handle(req, context)
+      .then((result) => api.hermes.toolResponse({ id: req.id, ok: true, result }))
+      .catch((err: Error) => api.hermes.toolResponse({ id: req.id, ok: false, error: err.message }));
+  });
+}
+
+async function handle(req: McpToolRequest, context: () => LocalToolContext): Promise<unknown> {
+  Log.info('mcp', `tool ${req.name}`);
+  const chat = useChat.getState();
+  if (req.name === 'show_message') {
+    const text = String(req.args.text ?? '');
+    if (!text.trim()) throw new Error('texte vide');
+    chat.addMessage({ role: 'assistant', content: text, source: 'mcp', jobName: typeof req.args.title === 'string' ? req.args.title : undefined, status: 'done' });
+    chat.pushActivity({ kind: 'system', name: 'hermes → eveflow', status: 'done', detail: text.slice(0, 160) });
+    return { ok: true };
+  }
+  const raw = await executeLocalTool(req.name, JSON.stringify(req.args ?? {}), context());
+  try {
+    const parsed = JSON.parse(raw) as { error?: string };
+    if (parsed && typeof parsed === 'object' && parsed.error) throw new Error(parsed.error);
+    return parsed;
+  } catch (err) {
+    if ((err as Error).message && !(err instanceof SyntaxError)) throw err;
+    return raw;
+  }
+}
@@ -4,6 +4,7 @@
 * OpenAI-compatible /v1/audio/speech, system voices, and the legacy Google Translate endpoint.
 */
 import { Log } from '../../lib/log';
+import { defaultOpenAiVoice, rankSystemVoice } from '../../lib/voicePreference';
 import { bridge } from '../../lib/bridge';
 import { chunkForSpeech, cleanForSpeech, extractSentences } from '../../lib/text';
 import { httpFetch } from '../../lib/transport';
@@ -26,6 +27,8 @@ export interface TtsConfig {
  localModel: string;
  /** Speaker id inside the local model. */
  localSpeaker: number;
+  /** Preferred voice gender, applied to every provider's default voice. */
+  voiceGender?: 'male' | 'female';
 }

 export type TtsState = 'idle' | 'loading' | 'speaking';
@@ -201,7 +204,7 @@ export class TtsEngine {
      body: JSON.stringify({
        model: model || 'tts-1',
        input: text,
-        voice: voice || 'alloy',
+        voice: voice || defaultOpenAiVoice(this.config.voiceGender ?? 'male'),
        speed: Math.max(0.5, Math.min(2, speed || 1)),
        response_format: format || 'mp3'
      }),
@@ -312,8 +315,8 @@ export class TtsEngine {
      if (exact) return exact;
    }
    const lang = (this.config.language || 'fr').toLowerCase().split('-')[0];
-    const candidates = voices.filter((v) => v.lang.toLowerCase().startsWith(lang));
-    return candidates.sort((a, b) => scoreVoice(b) - scoreVoice(a))[0] ?? voices[0];
+    const gender = this.config.voiceGender ?? 'male';
+    return [...voices].sort((a, b) => rankSystemVoice(b, lang, gender) - rankSystemVoice(a, lang, gender))[0];
  }
 }

@@ -2,6 +2,7 @@
 * Voice controller: microphone → VAD → STT → conversation, plus hands-free loop and barge-in.
 */
 import { Log } from '../../lib/log';
+import { isTranscriptNoise } from '../../lib/text';
 import { useChat } from '../../state/chat';
 import { useSettings } from '../../state/settings';
 import { useVoice } from '../../state/voice';
@@ -10,7 +11,9 @@ import { audioBus } from './audioBus';
 import { listMicrophones, MicCapture } from './capture';
 import { speech } from './speech';
 import { BrowserRecognizer, transcribeWav } from './stt';
+import { WakeListener } from './wakeListener';
 import type { WavResult } from './wav';
+import { bridge } from '../../lib/bridge';

 const SENSITIVITY_RATIO: Record<number, number> = { 1: 4.5, 2: 3.4, 3: 2.6, 4: 2.0, 5: 1.6 };
 const SENSITIVITY_MIN_RMS: Record<number, number> = { 1: 0.03, 2: 0.02, 3: 0.012, 4: 0.008, 5: 0.005 };
@@ -46,6 +49,8 @@ export function matchWakeWord(text: string, wakeWord: string): { matched: boolea

 class VoiceController {
  private capture = new MicCapture();
+  private wake = new WakeListener();
+  private unsubscribeKws: (() => void) | null = null;
  /** After a bare wake word, the next utterance is accepted without the wake word. */
  private attentionUntil = 0;
  private browser = new BrowserRecognizer();
@@ -74,17 +79,122 @@ class VoiceController {
  dispose(): void {
    this.unsubscribeVoice?.();
    this.stop();
+    void this.stopWakeMode();
  }

  get isListening(): boolean {
    return useVoice.getState().phase !== 'off';
  }

+  get wakeActive(): boolean {
+    return this.wake.isActive;
+  }
+
  toggle(): void {
+    if (this.wake.isActive) {
+      // Always-on mode: the button starts/cancels a command capture on the shared microphone.
+      if (this.wake.currentPhase === 'command' || this.wake.currentPhase === 'speech') this.wake.cancelCommand();
+      else this.onWakeDetected('manuel');
+      return;
+    }
    if (this.isListening) this.stop();
    else void this.start();
  }

+  /** Always-on keyword spotting: the microphone stays open and the main process spots the wake word. */
+  async startWakeMode(): Promise<void> {
+    const api = bridge();
+    const voice = useVoice.getState();
+    if (!api || this.wake.isActive) return;
+    const settings = useSettings.getState().settings.voice;
+    voice.setWake('starting');
+    try {
+      const phrases = [settings.wakeWord || 'jarvis', `hey ${settings.wakeWord || 'jarvis'}`];
+      const result = await api.voice.kwsStart({ modelId: 'kws-en', keywords: phrases, sensitivity: settings.kwsSensitivity });
+      this.unsubscribeKws?.();
+      this.unsubscribeKws = api.voice.onKwsDetected((d) => this.onWakeDetected(d.keyword));
+      const sensitivity = Math.min(5, Math.max(1, Math.round(settings.sensitivity))) as 1 | 2 | 3 | 4 | 5;
+      let neuralVad = false;
+      if (settings.neuralVad) {
+        try {
+          const installed = (await api.voice.listModels()).some((m) => m.id === 'silero-vad' && m.installed);
+          if (installed) {
+            await api.voice.vadStart({ modelId: 'silero-vad', silenceMs: Math.max(300, Math.min(1500, settings.silenceMs - 200)), threshold: 0.5, maxUtteranceSec: 25 });
+            neuralVad = true;
+          }
+        } catch (err) {
+          Log.warn('voice', `silero unavailable, energy VAD fallback: ${(err as Error).message}`);
+        }
+      }
+      await this.wake.start({
+        deviceId: settings.micDeviceId || undefined,
+        neuralVad,
+        vad: { silenceMs: settings.silenceMs, speechRatio: SENSITIVITY_RATIO[sensitivity], minRms: SENSITIVITY_MIN_RMS[sensitivity] },
+        callbacks: {
+          onPhase: (phase) => {
+            const v = useVoice.getState();
+            if (phase === 'spotting') {
+              v.setPhase('off');
+              const chat = useChat.getState();
+              if (chat.hud === 'listening') chat.setHud(chat.isSending ? 'thinking' : 'idle');
+            } else if (phase === 'command') {
+              v.setPhase('listening');
+              useChat.getState().setHud('listening');
+            } else if (phase === 'speech') {
+              v.setPhase('speech');
+              useChat.getState().ping();
+            }
+          },
+          onLevel: (level) => {
+            const q = Math.round(level * 20) / 20;
+            if (q !== useVoice.getState().inputLevel) useVoice.getState().setInputLevel(q);
+          },
+          onUtterance: (wav) => void this.transcribe(wav, true),
+          onNoSpeech: () => useVoice.getState().setInputLevel(0),
+          onError: (message) => useVoice.getState().setError(message)
+        }
+      });
+      voice.setWake('spotting', result.accepted.map((k) => k.replace(/_/g, ' ')));
+      voice.setNeuralVad(neuralVad);
+      Log.info('voice', `wake mode on: ${result.accepted.join(', ')}${result.rejected.length ? ` (rejetés : ${result.rejected.join(', ')})` : ''}${neuralVad ? ' · Silero' : ''}`);
+    } catch (err) {
+      const message = (err as Error).message;
+      voice.setWake('error');
+      voice.setError(message);
+      useChat.getState().setError(`Mot d’activation : ${message}`);
+      Log.error('voice', `wake mode failed: ${message}`);
+      await api.voice.kwsStop().catch(() => undefined);
+      await api.voice.vadStop().catch(() => undefined);
+    }
+  }
+
+  async stopWakeMode(): Promise<void> {
+    this.unsubscribeKws?.();
+    this.unsubscribeKws = null;
+    this.wake.stop();
+    useVoice.getState().setWake('off');
+    useVoice.getState().setNeuralVad(false);
+    const api = bridge();
+    await api?.voice.kwsStop().catch(() => undefined);
+    await api?.voice.vadStop().catch(() => undefined);
+  }
+
+  /** Restart spotting with the current settings (wake word, sensitivity, microphone). */
+  async restartWakeMode(): Promise<void> {
+    await this.stopWakeMode();
+    if (useSettings.getState().settings.voice.wakeMode === 'kws') await this.startWakeMode();
+  }
+
+  private onWakeDetected(keyword: string): void {
+    if (!this.wake.isActive || this.wake.currentPhase !== 'spotting') return;
+    if (useChat.getState().isSending) return;
+    Log.info('voice', `wake: ${keyword}`);
+    speech.stop();
+    useChat.getState().ping();
+    this.playChime(true);
+    this.wake.beginCommand();
+  }
+
  setHandsFree(on: boolean): void {
    useVoice.getState().setHandsFree(on);
    useSettings.getState().update({ voice: { handsFree: on } });
@@ -113,6 +223,11 @@ class VoiceController {
  async start(auto = false): Promise<void> {
    const voice = useVoice.getState();
    if (voice.phase !== 'off') return;
+    if (this.wake.isActive) {
+      // The always-on listener owns the microphone: open a command capture on it instead.
+      if (!auto) this.onWakeDetected('manuel');
+      return;
+    }
    const settings = useSettings.getState().settings.voice;
    const seq = ++this.startSeq;
    voice.setError(null);
@@ -128,6 +243,8 @@ class VoiceController {
    }

    const sensitivity = Math.min(5, Math.max(1, Math.round(settings.sensitivity))) as 1 | 2 | 3 | 4 | 5;
+    // Barge-in: while the assistant talks, demand a clearly louder signal so the speaker echo cannot trigger.
+    const barging = auto && settings.bargeIn && speech.isSpeaking();
    try {
      await audioBus.resume();
      await this.capture.start({
@@ -135,8 +252,8 @@ class VoiceController {
        mode: settings.captureMode,
        vad: {
          silenceMs: settings.silenceMs,
-          speechRatio: SENSITIVITY_RATIO[sensitivity],
-          minRms: SENSITIVITY_MIN_RMS[sensitivity],
+          speechRatio: barging ? Math.min(0.9, SENSITIVITY_RATIO[sensitivity] + 0.15) : SENSITIVITY_RATIO[sensitivity],
+          minRms: barging ? SENSITIVITY_MIN_RMS[sensitivity] * 2.5 : SENSITIVITY_MIN_RMS[sensitivity],
          noSpeechTimeoutMs: useVoice.getState().handsFree ? Number.POSITIVE_INFINITY : 9_000
        },
        callbacks: {
@@ -145,6 +262,8 @@ class VoiceController {
            if (q !== useVoice.getState().inputLevel) useVoice.getState().setInputLevel(q);
          },
          onSpeechStart: () => {
+            // The user started talking over the assistant: cut the voice (barge-in).
+            if (speech.isSpeaking()) speech.stop();
            useVoice.getState().setPhase('speech');
            useChat.getState().ping();
          },
@@ -218,7 +337,7 @@ class VoiceController {
    }
  }

-  private async transcribe(wav: WavResult): Promise<void> {
+  private async transcribe(wav: WavResult, fromWake = false): Promise<void> {
    const voice = useVoice.getState();
    voice.setPhase('transcribing');
    useChat.getState().setHud('thinking');
@@ -226,10 +345,15 @@ class VoiceController {
    const settings = useSettings.getState().settings.voice;
    try {
      let text = await transcribeWav(wav.bytes, settings);
+      if (isTranscriptNoise(text)) {
+        Log.debug('voice', `transcript ignored as noise: ${text}`);
+        text = '';
+      }
      voice.setTranscript(text);
      voice.setPhase('off');
      Log.info('voice', `transcript (${wav.durationSec.toFixed(1)}s): ${text}`);
-      if (voice.handsFree && settings.wakeWordEnabled && Date.now() > this.attentionUntil) {
+      const filterByTranscript = settings.wakeMode === 'transcript' || (settings.wakeMode === 'off' && settings.wakeWordEnabled);
+      if (!fromWake && voice.handsFree && filterByTranscript && Date.now() > this.attentionUntil) {
        const { matched, rest } = matchWakeWord(text, settings.wakeWord || 'jarvis');
        if (!matched) {
          Log.debug('voice', 'utterance ignored: no wake word');
@@ -0,0 +1,48 @@
+/**
+ * Applies the preferred voice gender to the active TTS provider: switches the local model/speaker,
+ * downloads a matching French voice when none is installed, and logs what it did.
+ */
+import { Log } from '../../lib/log';
+import { findLocalVoice, inferGender, suggestedDownload } from '../../lib/voicePreference';
+import { useSettings } from '../../state/settings';
+import { useVoiceModels } from '../../state/voiceModels';
+
+let downloading: string | null = null;
+
+/** Make the local voice match `speech.voiceGender`. Returns a short status message. */
+export async function ensurePreferredVoice(): Promise<string> {
+  const { settings, update } = useSettings.getState();
+  const { speech } = settings;
+  const gender = speech.voiceGender ?? 'male';
+  if (speech.provider !== 'local') return '';
+  const models = useVoiceModels.getState().models;
+  if (!models.length) return '';
+  const lang = speech.language || 'fr-FR';
+  const current = models.find((m) => m.id === speech.localModel);
+  const currentSpeaker = current?.speakers?.find((s) => s.id === speech.localSpeaker);
+  const currentGender = currentSpeaker ? (currentSpeaker as { gender?: string }).gender === 'm' ? 'male' : (currentSpeaker as { gender?: string }).gender === 'f' ? 'female' : inferGender(currentSpeaker.name) : undefined;
+  if (current?.installed && currentGender === gender) return '';
+
+  const choice = findLocalVoice(models, lang, gender);
+  if (choice) {
+    update({ speech: { localModel: choice.modelId, localSpeaker: choice.speaker } });
+    const name = models.find((m) => m.id === choice.modelId)?.speakers?.find((s) => s.id === choice.speaker)?.name ?? choice.modelId;
+    Log.info('tts', `voice preference ${gender}: ${name}`);
+    return `Voix ${gender === 'male' ? 'masculine' : 'féminine'} : ${name}`;
+  }
+  const download = suggestedDownload(models, lang, gender);
+  if (!download || downloading === download) return download ? 'Téléchargement de la voix en cours…' : '';
+  downloading = download;
+  Log.info('tts', `no ${gender} voice installed, downloading ${download}`);
+  try {
+    await useVoiceModels.getState().download(download);
+    const after = findLocalVoice(useVoiceModels.getState().models, lang, gender);
+    if (after) update({ speech: { localModel: after.modelId, localSpeaker: after.speaker } });
+    return after ? `Voix ${gender === 'male' ? 'masculine' : 'féminine'} installée.` : 'Voix téléchargée.';
+  } catch (err) {
+    Log.warn('tts', `voice download failed: ${(err as Error).message}`);
+    return `Téléchargement impossible : ${(err as Error).message}`;
+  } finally {
+    downloading = null;
+  }
+}
@@ -0,0 +1,273 @@
+/**
+ * Always-on microphone for the keyword spotter. Streams 16 kHz PCM to the main process
+ * (which feeds the sherpa-onnx spotter); once the wake word is detected it captures the
+ * following utterance with the energy VAD and hands back a WAV, then resumes spotting.
+ * One microphone stream, no re-opening between phrases.
+ */
+import { bridge } from '../../lib/bridge';
+import { Log } from '../../lib/log';
+import { audioBus } from './audioBus';
+import { DEFAULT_VAD, EnergyVad, type VadOptions } from './vad';
+import { buildWav16k, rms, type WavResult } from './wav';
+import type { VadEvent } from '../../../shared/voice';
+
+const WORKLET_SOURCE = `
+class EveFlowWakeProcessor extends AudioWorkletProcessor {
+  constructor() { super(); this.buffer = new Float32Array(2048); this.offset = 0; }
+  process(inputs) {
+    const channel = inputs[0] && inputs[0][0];
+    if (!channel) return true;
+    let i = 0;
+    while (i < channel.length) {
+      const n = Math.min(channel.length - i, this.buffer.length - this.offset);
+      this.buffer.set(channel.subarray(i, i + n), this.offset);
+      this.offset += n; i += n;
+      if (this.offset === this.buffer.length) {
+        this.port.postMessage(this.buffer, [this.buffer.buffer]);
+        this.buffer = new Float32Array(2048); this.offset = 0;
+      }
+    }
+    return true;
+  }
+}
+registerProcessor('eveflow-wake', EveFlowWakeProcessor);
+`;
+
+export type WakePhase = 'off' | 'spotting' | 'command' | 'speech';
+
+export interface WakeCallbacks {
+  onPhase: (phase: WakePhase) => void;
+  onLevel?: (level: number) => void;
+  onUtterance: (wav: WavResult) => void;
+  onNoSpeech: () => void;
+  onError: (message: string) => void;
+}
+
+let workletUrl: string | null = null;
+
+export class WakeListener {
+  private ctx: AudioContext | null = null;
+  private stream: MediaStream | null = null;
+  private node: AudioWorkletNode | null = null;
+  private phase: WakePhase = 'off';
+  private vad: EnergyVad | null = null;
+  private chunks: Float32Array[] = [];
+  private totalSamples = 0;
+  private sampleRate = 16_000;
+  private pending: Float32Array[] = [];
+  private pendingSamples = 0;
+  private callbacks: WakeCallbacks | null = null;
+  private vadOptions: Partial<VadOptions> = {};
+  private noSpeechTimer: ReturnType<typeof setTimeout> | null = null;
+  /** Neural end-of-speech (Silero in the main process) instead of the energy VAD. */
+  private neural = false;
+  private unsubscribeVad: (() => void) | null = null;
+  private vadPending: Float32Array[] = [];
+  private vadPendingSamples = 0;
+
+  get isActive(): boolean {
+    return this.phase !== 'off';
+  }
+
+  get currentPhase(): WakePhase {
+    return this.phase;
+  }
+
+  async start(options: { deviceId?: string; vad?: Partial<VadOptions>; neuralVad?: boolean; callbacks: WakeCallbacks }): Promise<void> {
+    if (this.phase !== 'off') return;
+    this.callbacks = options.callbacks;
+    this.vadOptions = options.vad ?? {};
+    this.neural = !!options.neuralVad;
+    if (this.neural) {
+      this.unsubscribeVad?.();
+      this.unsubscribeVad = bridge()?.voice.onVadEvent((event) => this.onVadEvent(event)) ?? null;
+    }
+    this.stream = await navigator.mediaDevices.getUserMedia({
+      audio: {
+        deviceId: options.deviceId ? { exact: options.deviceId } : undefined,
+        echoCancellation: true,
+        noiseSuppression: true,
+        autoGainControl: true,
+        channelCount: 1
+      }
+    });
+    try {
+      this.ctx = new AudioContext({ sampleRate: 16_000, latencyHint: 'playback' });
+    } catch {
+      this.ctx = new AudioContext({ latencyHint: 'playback' });
+    }
+    if (this.ctx.state === 'suspended') await this.ctx.resume().catch(() => undefined);
+    this.sampleRate = this.ctx.sampleRate;
+    const source = this.ctx.createMediaStreamSource(this.stream);
+    const analyser = this.ctx.createAnalyser();
+    analyser.fftSize = 512;
+    source.connect(analyser);
+    audioBus.setInputAnalyser(analyser);
+    if (!workletUrl) workletUrl = URL.createObjectURL(new Blob([WORKLET_SOURCE], { type: 'application/javascript' }));
+    await this.ctx.audioWorklet.addModule(workletUrl);
+    this.node = new AudioWorkletNode(this.ctx, 'eveflow-wake', { numberOfInputs: 1, numberOfOutputs: 0, channelCount: 1 });
+    this.node.port.onmessage = (event: MessageEvent<Float32Array>) => this.onSamples(event.data);
+    source.connect(this.node);
+    this.setPhase('spotting');
+    Log.info('wake', `listener started (${this.sampleRate} Hz, fin de phrase ${this.neural ? 'Silero' : 'énergie'})`);
+  }
+
+  private onVadEvent(event: VadEvent): void {
+    if (!this.neural || (this.phase !== 'command' && this.phase !== 'speech')) return;
+    if (event.type === 'speech-start') {
+      if (this.phase === 'command') {
+        this.setPhase('speech');
+        if (this.noSpeechTimer) clearTimeout(this.noSpeechTimer);
+        this.noSpeechTimer = null;
+      }
+    } else if (event.type === 'segment') {
+      const wav: WavResult = { bytes: event.wav, sampleRate: 16_000, durationSec: event.durationSec };
+      this.resumeSpotting();
+      if (wav.durationSec >= 0.25) this.callbacks?.onUtterance(wav);
+      else this.callbacks?.onNoSpeech();
+    } else if (event.type === 'error') {
+      this.callbacks?.onError(event.message);
+      this.resumeSpotting();
+    }
+  }
+
+  /** Called by the controller when the main process reports the wake word (or on manual trigger). */
+  beginCommand(): void {
+    if (this.phase === 'off' || this.phase === 'command' || this.phase === 'speech') return;
+    this.vad = this.neural ? null : new EnergyVad({ ...DEFAULT_VAD, ...this.vadOptions, noSpeechTimeoutMs: Number.POSITIVE_INFINITY });
+    this.vadPending = [];
+    this.vadPendingSamples = 0;
+    this.chunks = [];
+    this.totalSamples = 0;
+    this.setPhase('command');
+    if (this.noSpeechTimer) clearTimeout(this.noSpeechTimer);
+    this.noSpeechTimer = setTimeout(() => {
+      if (this.phase === 'command') {
+        this.resumeSpotting();
+        this.callbacks?.onNoSpeech();
+      }
+    }, 8000);
+  }
+
+  /** Abort a command capture and go back to spotting. */
+  cancelCommand(): void {
+    if (this.phase === 'command' || this.phase === 'speech') this.resumeSpotting();
+  }
+
+  stop(): void {
+    if (this.phase === 'off') return;
+    if (this.noSpeechTimer) clearTimeout(this.noSpeechTimer);
+    this.noSpeechTimer = null;
+    this.unsubscribeVad?.();
+    this.unsubscribeVad = null;
+    audioBus.setInputAnalyser(null);
+    if (this.node) {
+      this.node.port.onmessage = null;
+      this.node.disconnect();
+      this.node = null;
+    }
+    this.stream?.getTracks().forEach((t) => t.stop());
+    this.stream = null;
+    void this.ctx?.close().catch(() => undefined);
+    this.ctx = null;
+    this.chunks = [];
+    this.pending = [];
+    this.pendingSamples = 0;
+    this.vadPending = [];
+    this.vadPendingSamples = 0;
+    this.setPhase('off');
+    Log.info('wake', 'listener stopped');
+  }
+
+  private setPhase(phase: WakePhase): void {
+    this.phase = phase;
+    this.callbacks?.onPhase(phase);
+  }
+
+  private resumeSpotting(): void {
+    if (this.noSpeechTimer) clearTimeout(this.noSpeechTimer);
+    this.noSpeechTimer = null;
+    this.vad = null;
+    this.chunks = [];
+    this.totalSamples = 0;
+    this.setPhase('spotting');
+  }
+
+  private onSamples(samples: Float32Array): void {
+    if (this.phase === 'off') return;
+    const level = rms(samples);
+    this.callbacks?.onLevel?.(Math.min(1, level * 6));
+
+    if (this.phase === 'spotting') {
+      // Batch ~256 ms of audio per IPC message for the keyword spotter.
+      this.pending.push(samples);
+      this.pendingSamples += samples.length;
+      if (this.pendingSamples >= this.sampleRate * 0.25) this.flushToSpotter();
+      return;
+    }
+
+    if (this.neural) {
+      // command / speech with Silero: stream ~128 ms frames to the main process, which returns the segment.
+      this.vadPending.push(samples);
+      this.vadPendingSamples += samples.length;
+      if (this.vadPendingSamples >= this.sampleRate * 0.128) this.flushToVad();
+      return;
+    }
+
+    // command / speech: collect the utterance
+    this.chunks.push(samples);
+    this.totalSamples += samples.length;
+    if (this.vad && this.phase === 'command') {
+      // keep only a short pre-roll before speech starts
+      while (this.chunks.length > 1 && this.totalSamples - this.chunks[0].length > this.sampleRate * 0.4) {
+        this.totalSamples -= this.chunks.shift()!.length;
+      }
+    }
+    if (!this.vad) return;
+    const signal = this.vad.feed(level, (samples.length / this.sampleRate) * 1000);
+    if (signal === 'speech-start') {
+      this.setPhase('speech');
+      if (this.noSpeechTimer) clearTimeout(this.noSpeechTimer);
+      this.noSpeechTimer = null;
+    } else if (signal === 'speech-end' || signal === 'max-length') {
+      const wav = buildWav16k(this.chunks, this.sampleRate);
+      this.resumeSpotting();
+      if (wav.durationSec >= 0.25) this.callbacks?.onUtterance(wav);
+      else this.callbacks?.onNoSpeech();
+    } else if (signal === 'too-short') {
+      this.resumeSpotting();
+      this.callbacks?.onNoSpeech();
+    }
+  }
+
+  private static toInt16(chunks: Float32Array[], total: number): Uint8Array {
+    const merged = new Int16Array(total);
+    let offset = 0;
+    for (const chunk of chunks) {
+      for (let i = 0; i < chunk.length; i++) {
+        const s = Math.max(-1, Math.min(1, chunk[i]));
+        merged[offset + i] = s < 0 ? s * 0x8000 : s * 0x7fff;
+      }
+      offset += chunk.length;
+    }
+    return new Uint8Array(merged.buffer);
+  }
+
+  private flushToSpotter(): void {
+    const api = bridge();
+    if (!api) return;
+    const bytes = WakeListener.toInt16(this.pending, this.pendingSamples);
+    this.pending = [];
+    this.pendingSamples = 0;
+    api.voice.kwsAudio(bytes, this.sampleRate);
+  }
+
+  private flushToVad(): void {
+    const api = bridge();
+    if (!api) return;
+    const bytes = WakeListener.toInt16(this.vadPending, this.vadPendingSamples);
+    this.vadPending = [];
+    this.vadPendingSamples = 0;
+    api.voice.vadAudio(bytes, this.sampleRate);
+  }
+}
@@ -57,6 +57,16 @@ interface ChatStore {
  pingCount: number;
  /** Time to first token of the current reply, in ms. */
  latencyMs: number | null;
+  /** Pushes received while the window was hidden / compact, cleared when the user looks. */
+  unread: number;
+  /** Mission mode routes the next messages to the "mission" model (long tasks). */
+  missionMode: boolean;
+  /** True while quiet hours apply (computed by App). */
+  quiet: boolean;
+  incUnread: () => void;
+  markRead: () => void;
+  setMissionMode: (on: boolean) => void;
+  setQuiet: (quiet: boolean) => void;

  addMessage: (message: Omit<ChatMessage, 'id' | 'timestamp'> & Partial<Pick<ChatMessage, 'id' | 'timestamp'>>) => string;
  updateMessage: (id: string, patch: Partial<ChatMessage> | ((m: ChatMessage) => Partial<ChatMessage>)) => void;
@@ -91,6 +101,13 @@ export const useChat = create<ChatStore>((set, get) => ({
  draft: '',
  pingCount: 0,
  latencyMs: null,
+  unread: 0,
+  missionMode: false,
+  quiet: false,
+  incUnread: () => set((s) => ({ unread: Math.min(99, s.unread + 1) })),
+  markRead: () => set((s) => (s.unread ? { unread: 0 } : {})),
+  setMissionMode: (missionMode) => set({ missionMode }),
+  setQuiet: (quiet) => set((s) => (s.quiet === quiet ? {} : { quiet })),

  addMessage: (message) => {
    const id = message.id ?? uid('msg');
@@ -38,12 +38,14 @@ interface HermesStore {
  sessions: HermesSession[];
  jobs: HermesJob[];
  jobRuns: JobRun[];
+  /** Last cron sync failure (the chat link stays independent of it). */
+  jobsError: string | null;
  transport: ResolvedTransport;
  lastSyncAt: number | null;
  webhook: WebhookStatus | null;
  busy: boolean;

-  client: () => HermesClient;
+  client: (modelOverride?: string) => HermesClient;
  connect: () => Promise<void>;
  refreshJobs: () => Promise<void>;
  refreshSessions: () => Promise<void>;
@@ -88,15 +90,16 @@ export const useHermes = create<HermesStore>((set, get) => ({
  sessions: [],
  jobs: [],
  jobRuns: [],
+  jobsError: null,
  transport: 'completions',
  lastSyncAt: null,
  webhook: null,
  busy: false,

-  client: () => {
+  client: (modelOverride) => {
    const config = useSettings.getState().settings.hermes;
    // Without an explicit model, use the alias advertised by /v1/models (Hermes rejects unknown names).
-    const model = config.model.trim() || get().models[0]?.id || '';
+    const model = (modelOverride ?? '').trim() || config.model.trim() || get().models[0]?.id || '';
    return new HermesClient({ ...config, model });
  },

@@ -119,8 +122,8 @@ export const useHermes = create<HermesStore>((set, get) => ({
        Log.warn('hermes', `capabilities unavailable: ${(err as Error).message}`);
      }
      const transport = resolveTransport(config, capabilities);
-      const degraded = String(health.status ?? 'ok').toLowerCase() !== 'ok';
-      set({ health, capabilities, transport, link: degraded ? 'degraded' : 'online', linkDetail: degraded ? `état ${health.status}` : '' });
+      const degraded = !isHealthyStatus(health.status);
+      set({ health, capabilities, transport, link: degraded ? 'degraded' : 'online', linkDetail: degraded ? describeHealth(health) : '' });
      Log.info('hermes', `connected (${transport})`, { status: health.status, model: capabilities?.model });
      void get().refreshCatalog();
      void get().refreshJobs();
@@ -165,7 +168,7 @@ export const useHermes = create<HermesStore>((set, get) => ({
      const merged = [...incoming, ...previous.filter((r) => !incoming.some((i) => i.id === r.id))]
        .sort((a, b) => new Date(b.at).getTime() - new Date(a.at).getTime())
        .slice(0, 100);
-      set({ jobs, jobRuns: merged, lastSyncAt: Date.now(), link: get().link === 'offline' || get().link === 'degraded' ? 'online' : get().link });
+      set({ jobs, jobRuns: merged, jobsError: null, lastSyncAt: Date.now(), link: get().link === 'offline' ? 'online' : get().link });
      const snapshot = JSON.stringify({ jobs, jobRuns: merged });
      if (snapshot !== lastCacheSnapshot) {
        lastCacheSnapshot = snapshot;
@@ -174,9 +177,8 @@ export const useHermes = create<HermesStore>((set, get) => ({
    } catch (err) {
      const message = (err as Error).message;
      Log.warn('hermes', `jobs sync failed: ${message}`);
-      if (/HTTP 404/.test(message)) return; // jobs API disabled on this server
-      // A failing jobs poll does not mean chat is down: degrade, and let the next health probe decide.
-      set({ linkDetail: message, link: get().link === 'online' ? 'degraded' : get().link });
+      // The cron API can be absent or restricted on a given Hermes: the chat link is unaffected.
+      set({ jobsError: /HTTP 404/.test(message) ? 'API des crons absente sur ce serveur Hermes' : message });
    }
  },

@@ -233,3 +235,31 @@ export const useHermes = create<HermesStore>((set, get) => ({
 }));

 export { jobStatus };
+
+const HEALTHY = new Set(['ok', 'healthy', 'up', 'alive', 'pass', 'ready', 'running', 'online', 'true']);
+
+/** Hermes /health reports "ok"; /health/detailed may report "healthy", "degraded" or "unhealthy". */
+export function isHealthyStatus(status: unknown): boolean {
+  if (status === undefined || status === null || status === '') return true;
+  return HEALTHY.has(String(status).toLowerCase());
+}
+
+/** Human summary of a degraded health payload: failing checks by name. */
+export function describeHealth(health: Record<string, unknown>): string {
+  const failing: string[] = [];
+  const visit = (node: unknown, prefix: string, depth: number) => {
+    if (!node || typeof node !== 'object' || depth > 3) return;
+    for (const [key, value] of Object.entries(node as Record<string, unknown>)) {
+      if (key === 'status') continue;
+      if (value && typeof value === 'object') {
+        const rec = value as Record<string, unknown>;
+        const st = rec.status ?? rec.ok ?? rec.healthy;
+        if (st !== undefined && !isHealthyStatus(st)) failing.push(prefix + key);
+        else visit(value, `${prefix}${key}.`, depth + 1);
+      } else if (typeof value === 'boolean' && !value && /ok|healthy|ready|connected|available/i.test(key)) failing.push(prefix + key);
+    }
+  };
+  visit(health, '', 0);
+  const base = `état ${String(health.status)}`;
+  return failing.length ? `${base} · ${failing.slice(0, 4).join(', ')}` : base;
+}
@@ -19,6 +19,13 @@ export interface VoiceSettings extends SttConfig {
  /** Hands-free: only react to utterances starting with this word (local STT recommended). */
  wakeWordEnabled: boolean;
  wakeWord: string;
+  /** off = push-to-talk / hands-free; transcript = filter after transcription; kws = always-on keyword spotting. */
+  wakeMode: 'off' | 'transcript' | 'kws';
+  kwsSensitivity: number; // 1..5
+  /** Use Silero VAD for end-of-speech in always-on mode when the model is installed. */
+  neuralVad: boolean;
+  /** Execute short system intents locally (lock, volume, open app…) instead of asking Hermes. */
+  localCommands: boolean;
 }

 export interface SpeechSettings extends TtsConfig {
@@ -32,6 +39,20 @@ export interface WebhookSettings {
  secret: string;
 }

+export interface NotificationSettings {
+  /** Quiet hours: no spoken pushes, no chime, dimmed HUD (24 h "HH:MM"). */
+  quietEnabled: boolean;
+  quietStart: string;
+  quietEnd: string;
+  /** Pushes whose text or job name contains one of these words are spoken even during quiet hours. */
+  priorityKeywords: string;
+  /** Speak only the first sentences of incoming pushes (cron reports can be long). */
+  summarizeIncoming: boolean;
+  summarySentences: number;
+  /** Dim the HUD (night theme) during quiet hours. */
+  nightTheme: boolean;
+}
+
 export interface Settings {
  version: 2;
  assistantName: string;
@@ -42,6 +63,7 @@ export interface Settings {
  voice: VoiceSettings;
  speech: SpeechSettings;
  webhook: WebhookSettings;
+  notifications: NotificationSettings;
  ui: {
    showTelemetry: boolean;
    showReasoning: boolean;
@@ -66,7 +88,8 @@ export const DEFAULT_SETTINGS: Settings = {
    reasoningEffort: '',
    instructions:
      "Tu es l'interface vocale EveFlow (style JARVIS). Réponds en français, de façon concise et orale quand la question est simple; utilise le Markdown uniquement pour le contenu structuré (code, listes, tableaux). Les images doivent être des URL http(s) ou des fichiers du dossier partagé.",
-    localTools: true
+    localTools: true,
+    missionModel: ''
  },
  voice: {
    provider: 'openai-compatible',
@@ -83,6 +106,10 @@ export const DEFAULT_SETTINGS: Settings = {
    wakeChime: true,
    wakeWordEnabled: false,
    wakeWord: 'jarvis',
+    wakeMode: 'off',
+    kwsSensitivity: 3,
+    neuralVad: true,
+    localCommands: true,
    localModel: 'whisper-base'
  },
  speech: {
@@ -98,10 +125,12 @@ export const DEFAULT_SETTINGS: Settings = {
    volume: 1,
    localModel: 'kokoro-v1',
    localSpeaker: 30,
+    voiceGender: 'male',
    autoSpeak: true,
    speakIncoming: true
  },
  webhook: { enabled: true, port: 7842, secret: '' },
+  notifications: { quietEnabled: false, quietStart: '22:30', quietEnd: '07:30', priorityKeywords: 'urgent, alerte, alarme, panne', summarizeIncoming: true, summarySentences: 2, nightTheme: true },
  ui: { showTelemetry: true, showReasoning: false, reduceMotion: false, compactOpacity: 0.92 },
  hermesSessionId: ''
 };
@@ -181,6 +210,7 @@ export const useSettings = create<SettingsStore>((set, get) => ({
    }
    const merged = merge(DEFAULT_SETTINGS, saved);
    if (!merged.hermesSessionId) merged.hermesSessionId = uid('eveflow');
+    if (saved && saved.voice && saved.voice.wakeMode === undefined && saved.voice.wakeWordEnabled) merged.voice.wakeMode = 'transcript';
    set({ settings: merged, loaded: true });
    persistSet(STORAGE_KEY, merged);
  },
@@ -2,6 +2,7 @@ import { create } from 'zustand';
 import type { TtsState } from '../services/voice/tts';

 export type ListenPhase = 'off' | 'arming' | 'listening' | 'speech' | 'transcribing';
+export type WakeState = 'off' | 'starting' | 'spotting' | 'error';

 interface VoiceStore {
  phase: ListenPhase;
@@ -12,6 +13,11 @@ interface VoiceStore {
  interim: string;
  error: string | null;
  micDevices: Array<{ deviceId: string; label: string }>;
+  wake: WakeState;
+  wakeKeywords: string[];
+  neuralVad: boolean;
+  setNeuralVad: (on: boolean) => void;
+  setWake: (state: WakeState, keywords?: string[]) => void;
  setPhase: (phase: ListenPhase) => void;
  setInputLevel: (level: number) => void;
  setTts: (state: TtsState) => void;
@@ -31,6 +37,11 @@ export const useVoice = create<VoiceStore>((set) => ({
  interim: '',
  error: null,
  micDevices: [],
+  wake: 'off',
+  wakeKeywords: [],
+  neuralVad: false,
+  setNeuralVad: (neuralVad) => set({ neuralVad }),
+  setWake: (wake, wakeKeywords) => set(wakeKeywords ? { wake, wakeKeywords } : { wake }),
  setPhase: (phase) => set({ phase }),
  setInputLevel: (inputLevel) => set({ inputLevel }),
  setTts: (tts) => set({ tts }),
@@ -67,3 +67,54 @@
  font-size: 13px;
  padding: 8px 11px;
 }
+
+/* Glanceable strip: state, unread badge, last sentence. */
+.compact-glance {
+  display: grid;
+  grid-template-columns: auto 1fr auto;
+  align-items: center;
+  gap: 8px;
+  padding: 4px 12px 8px;
+  font-family: var(--font-mono);
+  font-size: 10.5px;
+  letter-spacing: 0.12em;
+  text-transform: uppercase;
+  color: var(--ink-2);
+}
+.compact-glance .state {
+  color: var(--accent);
+}
+.compact-glance .state.alert {
+  color: var(--danger);
+}
+.compact-glance .last {
+  font-family: var(--font-body);
+  font-size: 12.5px;
+  letter-spacing: 0;
+  text-transform: none;
+  color: var(--ink-1);
+  white-space: nowrap;
+  overflow: hidden;
+  text-overflow: ellipsis;
+}
+.compact-glance .badges {
+  display: flex;
+  gap: 6px;
+  align-items: center;
+}
+.compact-glance .unread {
+  min-width: 18px;
+  height: 18px;
+  padding: 0 5px;
+  border-radius: 9px;
+  background: var(--accent);
+  color: #03101c;
+  font-weight: 700;
+  font-size: 11px;
+  display: inline-flex;
+  align-items: center;
+  justify-content: center;
+}
+.compact-glance .mission {
+  color: var(--warn, #f5c451);
+}
@@ -159,3 +159,15 @@ textarea {
 .kv dd {
  user-select: text;
 }
+
+/* Quiet hours: dimmer HUD, softer glow. */
+:root[data-night] {
+  --accent-glow: rgba(var(--accent-rgb), 0.18);
+}
+:root[data-night] .hud-root,
+:root[data-night] .compact-root {
+  filter: brightness(0.72) saturate(0.85);
+}
+:root[data-night] .core-stage canvas {
+  opacity: 0.8;
+}
@@ -0,0 +1,27 @@
+import { describe, expect, it } from 'vitest';
+import { recoverCompletion } from '../src/services/hermes/client';
+import { describeHealth, isHealthyStatus } from '../src/state/hermes';
+
+describe('health status', () => {
+  it('accepts the usual healthy words', () => {
+    for (const s of ['ok', 'OK', 'healthy', 'ready', undefined]) expect(isHealthyStatus(s)).toBe(true);
+    for (const s of ['degraded', 'unhealthy', 'error']) expect(isHealthyStatus(s)).toBe(false);
+  });
+  it('names the failing checks', () => {
+    expect(describeHealth({ status: 'degraded', checks: { memory: { status: 'ok' }, sessions_db: { status: 'error' } } })).toBe('état degraded · checks.sessions_db');
+    expect(describeHealth({ status: 'degraded' })).toBe('état degraded');
+  });
+});
+
+describe('recoverCompletion', () => {
+  it('reads a plain JSON completion when the server ignored streaming', () => {
+    expect(recoverCompletion(JSON.stringify({ choices: [{ message: { role: 'assistant', content: 'Bonjour.' } }] }))).toEqual({ text: 'Bonjour.' });
+  });
+  it('surfaces an error object returned with HTTP 200', () => {
+    expect(recoverCompletion(JSON.stringify({ error: { message: 'model not found' } })).error).toContain('model not found');
+  });
+  it('handles SSE bodies and empty input', () => {
+    expect(recoverCompletion('data: {"choices":[{"delta":{"content":"A"}}]}\n\ndata: {"choices":[{"delta":{"content":"B"}}]}\n\ndata: [DONE]\n\n')).toEqual({ text: 'AB' });
+    expect(recoverCompletion('')).toEqual({});
+  });
+});
@@ -0,0 +1,25 @@
+import { describe, expect, it } from 'vitest';
+import { buildKeywordsFile, encodeKeyword, normalizeKeyword, parseTokens } from '../shared/keywords';
+
+const vocab = parseTokens(['<blk> 0', '▁ 3', '▁JA 4', 'R 5', 'VI 6', 'S 7', '▁HE 8', 'Y 9', '▁NO 10', 'V 11', 'A 12', '▁MA 13', 'X 14', '▁T 15', 'ON 16', 'Y 17'].join('\n'));
+
+describe('keywords', () => {
+  it('normalises accents, case and punctuation', () => {
+    expect(normalizeKeyword('  Hé, Jarvis ! ')).toBe('HE JARVIS');
+  });
+  it('uses the known SentencePiece encodings', () => {
+    expect(encodeKeyword('jarvis', vocab)).toBe('▁JA R VI S');
+    expect(encodeKeyword('Hey Jarvis', vocab)).toBe('▁HE Y ▁JA R VI S');
+  });
+  it('falls back to greedy longest match', () => {
+    expect(encodeKeyword('max', vocab)).toBe('▁MA X');
+    expect(encodeKeyword('tony', vocab)).toBe('▁T ON Y');
+    expect(encodeKeyword('zzz', vocab)).toBeNull();
+  });
+  it('builds a keywords file with labels', () => {
+    const file = buildKeywordsFile(['jarvis', 'hey jarvis', 'zzz'], vocab);
+    expect(file.content).toBe('▁JA R VI S @jarvis\n▁HE Y ▁JA R VI S @hey_jarvis\n');
+    expect(file.accepted).toEqual(['jarvis', 'hey_jarvis']);
+    expect(file.rejected).toEqual(['zzz']);
+  });
+});
@@ -0,0 +1,24 @@
+import { describe, expect, it } from 'vitest';
+import { parseLocalIntent } from '../src/services/localCommands';
+
+describe('parseLocalIntent', () => {
+  it('recognises system actions in French and English', () => {
+    expect(parseLocalIntent('Jarvis, verrouille la session')?.action).toEqual({ type: 'lock' });
+    expect(parseLocalIntent('monte le son')?.action).toEqual({ type: 'media', key: 'volume-up' });
+    expect(parseLocalIntent('Baisse le volume')?.action).toEqual({ type: 'media', key: 'volume-down' });
+    expect(parseLocalIntent('coupe le son')?.action).toEqual({ type: 'media', key: 'mute' });
+    expect(parseLocalIntent('piste suivante')?.action).toEqual({ type: 'media', key: 'next' });
+    expect(parseLocalIntent('ouvre le bloc-notes')?.action).toEqual({ type: 'open-app', name: 'bloc-notes' });
+    expect(parseLocalIntent('open spotify')?.action).toEqual({ type: 'open-app', name: 'spotify' });
+    expect(parseLocalIntent('ouvre github.com')?.action).toEqual({ type: 'open-url', url: 'https://github.com' });
+  });
+  it('detects screen questions', () => {
+    expect(parseLocalIntent('Jarvis, regarde mon écran et dis-moi ce que tu vois')?.kind).toBe('screenshot');
+    expect(parseLocalIntent('fais une capture d’écran')?.kind).toBe('screenshot');
+  });
+  it('leaves everything else to Hermes', () => {
+    expect(parseLocalIntent('Quelle est la météo à Paris demain ?')).toBeNull();
+    expect(parseLocalIntent('ouvre le fichier puis envoie-le à Marc')).toBeNull();
+    expect(parseLocalIntent('')).toBeNull();
+  });
+});
@@ -0,0 +1,37 @@
+import { describe, expect, it } from 'vitest';
+import { isPriority, isQuietTime, summarize } from '../src/lib/quietHours';
+
+const at = (h: number, m = 0) => new Date(2026, 0, 1, h, m);
+
+describe('isQuietTime', () => {
+  it('handles ranges crossing midnight', () => {
+    expect(isQuietTime('22:30', '07:30', at(23))).toBe(true);
+    expect(isQuietTime('22:30', '07:30', at(3, 15))).toBe(true);
+    expect(isQuietTime('22:30', '07:30', at(7, 30))).toBe(false);
+    expect(isQuietTime('22:30', '07:30', at(12))).toBe(false);
+  });
+  it('handles same-day ranges and invalid input', () => {
+    expect(isQuietTime('13:00', '14:00', at(13, 30))).toBe(true);
+    expect(isQuietTime('13:00', '14:00', at(14))).toBe(false);
+    expect(isQuietTime('bad', '14:00', at(13))).toBe(false);
+    expect(isQuietTime('10:00', '10:00', at(10))).toBe(false);
+  });
+});
+
+describe('isPriority', () => {
+  it('matches keywords in text or job name', () => {
+    expect(isPriority('Serveur en panne depuis 5 min', 'urgent, panne')).toBe(true);
+    expect(isPriority('Rapport quotidien', 'urgent', 'Alerte disque')).toBe(false);
+    expect(isPriority('Rapport quotidien', 'alerte', 'Alerte disque')).toBe(true);
+    expect(isPriority('x', '')).toBe(false);
+  });
+});
+
+describe('summarize', () => {
+  it('keeps the first sentences and strips markdown', () => {
+    const text = '## Rapport\n\nTrois annonces **majeures** aujourd’hui. Le marché monte de 2 %. Détails ci-dessous :\n\n```\ncode\n```\n- point 1';
+    expect(summarize(text, 2)).toBe('Rapport Trois annonces majeures aujourd’hui. Le marché monte de 2 %.');
+    expect(summarize('Une seule phrase', 2)).toBe('Une seule phrase');
+    expect(summarize('', 2)).toBe('');
+  });
+});
@@ -1,5 +1,5 @@
 import { describe, expect, it } from 'vitest';
-import { chunkForSpeech, cleanForSpeech, extractSentences, preprocessMedia } from '../src/lib/text';
+import { chunkForSpeech, cleanForSpeech, extractSentences, isTranscriptNoise, preprocessMedia } from '../src/lib/text';

 describe('text', () => {
  it('cleans markdown for speech', () => {
@@ -25,3 +25,18 @@ describe('text', () => {
    expect(preprocessMedia('MEDIA:/tmp/a.png')).toContain('![image](/tmp/a.png)');
  });
 });
+
+describe('isTranscriptNoise', () => {
+  it('drops Whisper hallucinations on silence', () => {
+    expect(isTranscriptNoise('(cliquant)')).toBe(true);
+    expect(isTranscriptNoise('*Claire*')).toBe(true);
+    expect(isTranscriptNoise('[Musique]')).toBe(true);
+    expect(isTranscriptNoise('...')).toBe(true);
+    expect(isTranscriptNoise("Sous-titres réalisés par la communauté d'Amara.org")).toBe(true);
+  });
+  it('keeps real sentences', () => {
+    expect(isTranscriptNoise('Jarvis, allume la lumière du salon.')).toBe(false);
+    expect(isTranscriptNoise('(Jarvis) quelle heure est-il maintenant ?')).toBe(false);
+    expect(isTranscriptNoise('Oui')).toBe(false);
+  });
+});
@@ -0,0 +1,49 @@
+import { describe, expect, it } from 'vitest';
+import { defaultOpenAiVoice, findLocalVoice, inferGender, rankSystemVoice, suggestedDownload } from '../src/lib/voicePreference';
+import type { VoiceModelStatus } from '../shared/voice';
+
+const model = (id: string, installed: boolean, speakers: Array<{ id: number; name: string; lang: string }>): VoiceModelStatus =>
+  ({ id, kind: 'tts', engine: 'piper', name: id, description: '', languages: ['fr'], sizeMb: 1, url: '', dir: id, files: [], speakers, installed, installedBytes: 0 }) as unknown as VoiceModelStatus;
+
+describe('inferGender', () => {
+  it('reads catalog labels, system voices and kokoro ids', () => {
+    expect(inferGender('Piper Tom (homme, français)')).toBe('male');
+    expect(inferGender('Siwis (femme, français)')).toBe('female');
+    expect(inferGender('Microsoft Paul - French (France)')).toBe('male');
+    expect(inferGender('Microsoft Hortense - French (France)')).toBe('female');
+    expect(inferGender('am_adam')).toBe('male');
+    expect(inferGender('Voix 3')).toBeUndefined();
+  });
+});
+
+describe('local voice selection', () => {
+  const models = [
+    model('kokoro-v1', true, [{ id: 30, name: 'Siwis (femme, français)', lang: 'fr' }, { id: 4, name: 'Adam (homme, anglais US)', lang: 'en' }]),
+    model('piper-fr-tom', false, [{ id: 0, name: 'Tom', lang: 'fr' }]),
+    model('piper-fr-upmc', true, [{ id: 0, name: 'Jessica (femme)', lang: 'fr' }, { id: 1, name: 'Pierre (homme)', lang: 'fr' }])
+  ];
+  it('prefers an installed speaker of the wanted gender in the right language', () => {
+    expect(findLocalVoice(models, 'fr-FR', 'male')).toEqual({ modelId: 'piper-fr-upmc', speaker: 1 });
+    expect(findLocalVoice(models, 'fr-FR', 'female')).toEqual({ modelId: 'kokoro-v1', speaker: 30 });
+    expect(findLocalVoice(models.slice(0, 2), 'fr', 'male')).toBeNull();
+  });
+  it('suggests a French download when nothing matches', () => {
+    expect(suggestedDownload(models, 'fr', 'male')).toBe('piper-fr-tom');
+    expect(suggestedDownload(models, 'en', 'male')).toBeNull();
+  });
+});
+
+describe('provider defaults', () => {
+  it('picks onyx for male and nova for female on OpenAI-compatible APIs', () => {
+    expect(defaultOpenAiVoice('male')).toBe('onyx');
+    expect(defaultOpenAiVoice('female')).toBe('nova');
+  });
+  it('ranks system voices by language then gender', () => {
+    const paul = { name: 'Microsoft Paul - French (France)', lang: 'fr-FR', localService: true };
+    const hortense = { name: 'Microsoft Hortense - French (France)', lang: 'fr-FR', localService: true };
+    const david = { name: 'Microsoft David - English (US)', lang: 'en-US', localService: true };
+    expect(rankSystemVoice(paul, 'fr', 'male')).toBeGreaterThan(rankSystemVoice(hortense, 'fr', 'male'));
+    expect(rankSystemVoice(hortense, 'fr', 'female')).toBeGreaterThan(rankSystemVoice(paul, 'fr', 'female'));
+    expect(rankSystemVoice(paul, 'fr', 'male')).toBeGreaterThan(rankSystemVoice(david, 'fr', 'male'));
+  });
+});
Author	SHA1	Message	Date
LogiFlow	c75c1d674c	Merge pull request #16 from R0m1k3/claude/refonte-app-vocale-v91uz7 fix: liaison « dégradé », réponses vides, transcriptions parasites, v…	2026-09-04 07:53:40 +02:00
LogiFlowandClaude	6d3073b083	fix: liaison « dégradé », réponses vides, transcriptions parasites, voix masculine (v2.4.0.1) (#15 ) - Liaison Hermes : l'échec de l'API des crons (/api/jobs absente ou refusée) ne fait plus passer la liaison en « dégradé » ; l'onglet Crons affiche la raison. Les états « healthy / ready / ok » sont reconnus et un état « degraded » détaille les contrôles en échec. - Chat completions : une réponse sans fragment n'affiche plus « … » ; le corps est relu (JSON non streamé, erreur renvoyée en HTTP 200) et sinon l'erreur explicite « Réponse vide de Hermes » est affichée avec le début du corps. - Voix : transcriptions parasites de Whisper (« (cliquant) », « Claire », « [Musique] », génériques de sous-titres) ignorées au lieu d'être envoyées. - Préférence de voix masculine / féminine (Paramètres → Voix), appliquée à tous les moteurs : Piper Tom / Pierre en local (téléchargement automatique), onyx / nova en API OpenAI, Paul / Hortense en voix Windows. Genre des voix du catalogue déduit des libellés. Tests : 42 tests unitaires (santé, récupération de réponse, filtre de bruit, préférence de voix) ; e2e Electron inchangé et vert. Claude-Session: https://claude.ai/code/session_017Wn5VX9HNbJ7N54hR24u9Y Co-authored-by: Claude <noreply@anthropic.com>	2026-09-04 07:44:07 +02:00
Claude	31659500ab	fix: liaison « dégradé », réponses vides, transcriptions parasites, voix masculine (v2.4.0.1) - Liaison Hermes : l'échec de l'API des crons (/api/jobs absente ou refusée) ne fait plus passer la liaison en « dégradé » ; l'onglet Crons affiche la raison. Les états « healthy / ready / ok » sont reconnus et un état « degraded » détaille les contrôles en échec. - Chat completions : une réponse sans fragment n'affiche plus « … » ; le corps est relu (JSON non streamé, erreur renvoyée en HTTP 200) et sinon l'erreur explicite « Réponse vide de Hermes » est affichée avec le début du corps. - Voix : transcriptions parasites de Whisper (« (cliquant) », « Claire », « [Musique] », génériques de sous-titres) ignorées au lieu d'être envoyées. - Préférence de voix masculine / féminine (Paramètres → Voix), appliquée à tous les moteurs : Piper Tom / Pierre en local (téléchargement automatique), onyx / nova en API OpenAI, Paul / Hortense en voix Windows. Genre des voix du catalogue déduit des libellés. Tests : 42 tests unitaires (santé, récupération de réponse, filtre de bruit, préférence de voix) ; e2e Electron inchangé et vert. Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_017Wn5VX9HNbJ7N54hR24u9Y	2026-09-04 05:43:41 +00:00
LogiFlowandClaude	6fc38d335b	feat: serveur MCP, heures calmes, mode mission, widget glanceable, barge-in (v2.4.0) (#14 ) Serveur MCP local - Endpoint /mcp (JSON-RPC 2.0, Streamable HTTP, réponses JSON) sur le serveur webhook : initialize, tools/list, tools/call, ping. Même secret que le webhook. 14 outils : capture_screen (image MCP), lock_session, open_app, open_url, media_key, clipboard_get/set, find_files, speak_text, notify_user, set_hud_state, get_app_status, get_conversation_history, show_message. Les outils UI transitent par IPC vers le renderer. - Les mêmes actions système sont proposées au modèle en chat completions. Notifications - Heures calmes (plage horaire, franchissement de minuit), mots prioritaires, échecs de crons toujours lus, résumé vocal des messages entrants (n premières phrases), thème nuit automatique, compteur de non-lus. Conversation - Mode mission : bouton dans la barre de commande, second modèle Hermes (hermes.missionModel) pour les tâches longues. - Barge-in : la parole de l'utilisateur coupe la voix ; seuil d'énergie et ratio relevés pendant la synthèse pour ignorer l'écho. Widget compact - Bandeau glanceable : état, dernière phrase, non-lus, indicateurs heures calmes et mission. Tests : 28 tests unitaires (heures calmes, priorité, résumé), e2e Electron sous Xvfb : MCP initialize/tools/list/tools/call (presse-papiers, capture en image, état et message via le renderer), bandeau compact, aucune erreur. Claude-Session: https://claude.ai/code/session_017Wn5VX9HNbJ7N54hR24u9Y Co-authored-by: Claude <noreply@anthropic.com>	2026-09-04 00:47:34 +02:00
LogiFlowandClaude	af8bdc5e9d	feat: fin de phrase Silero, vision d'écran et actions locales (v2.3.0) (#13 ) Écoute permanente - Silero VAD dans le worker sherpa-onnx (vad.start/audio/stop) ; le renderer envoie des trames de 128 ms pendant la commande et reçoit le segment WAV complet (speech-start / segment). Repli sur le VAD énergétique si le modèle n'est pas installé ou si l'option est désactivée. - Modèle « silero-vad » (0,6 Mo) dans le catalogue, section « Fin de phrase » dans Modèles locaux, réglage dans Paramètres → Micro. Vision d'écran - IPC system:screen-capture (desktopCapturer, JPEG 1600 px) ; bouton dans la barre de commande ; « regarde mon écran… » joint la capture à la question envoyée à Hermes (transport chat completions pour les images). Actions locales - IPC system:action à liste blanche : verrouiller la session, touches média (volume, mute, lecture, piste), ouvrir une application connue ou une URL http(s), presse-papiers, recherche de fichiers dans les dossiers utilisateur. - Routeur d'intentions FR/EN exécuté avant Hermes (« coupe le son », « ouvre Spotify », « verrouille la session »…), résultat affiché et lu ; désactivable. Correctif - Chat completions : les premiers fragments SSE arrivés avant l'événement de démarrage étaient perdus (premier mot manquant) ; ils sont maintenant rejoués. Tests : 26 tests unitaires (intentions locales ajoutées), e2e Electron sous Xvfb avec les vrais modèles (Silero : un segment par phrase, 6,7 s d'audio traités en 190 ms ; capture 109 ko reçue par Hermes ; intention locale traitée sans Hermes). Claude-Session: https://claude.ai/code/session_017Wn5VX9HNbJ7N54hR24u9Y Co-authored-by: Claude <noreply@anthropic.com>	2026-09-03 20:46:26 +02:00
LogiFlow	4f6074183f	feat: écoute permanente par détection de mot-clé (v2.2.0) (#12 ) feat: écoute permanente par détection de mot-clé (v2.2.0)	2026-09-03 20:22:00 +02:00
Claude	0d284fea9e	feat: always-on wake word with sherpa-onnx keyword spotting (v2.2.0) - Catalog: 3.3 MB zipformer keyword-spotting model (kws-en). - shared/keywords: BPE encoding of wake phrases (SentencePiece table for common words, greedy longest-match fallback over the model vocabulary) and keywords file builder. - Worker: KeywordSpotter stream fed with 16-bit PCM, detections pushed as unsolicited messages; engine derives the keywords file, maps sensitivity to threshold/score, forwards detections to the renderer and re-arms after a worker restart. - Renderer: WakeListener keeps one microphone stream, batches 256 ms frames to the spotter and captures the command on the same stream after detection (pre-roll, VAD), then resumes spotting; the wake word also interrupts speech. Settings: wake mode (off / always-on / transcript filter), keyword, sensitivity, status and one-click model download; HUD caption shows the active keyword. - Validated: detection in the worker (fork) and through the real Electron IPC path on Kokoro audio, no false positive on an English recording; 23 unit tests. Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_017Wn5VX9HNbJ7N54hR24u9Y	2026-09-03 18:21:35 +00:00