feat(flux): Modell-Wahl per Diagnostic + raw/switch-Keywords + Download-Hinweis

Diagnostic-Einstellungen fuer FLUX:
- Default-Modell (dev | schnell) — wird via RVS gepusht, flux-bridge
  hot-swappt die Pipeline aus dem HF-Cache (~15-30s)
- Raw-Keyword (Default 'flux') — Pipe-Modus, Brain leitet Stefans Text
  1:1 als prompt durch, kein Rewriting/Beautify
- Switch-Keyword (Default 'fix') — zwingt das ANDERE Modell als Default

Brain-Tool flux_generate um model + raw erweitert, System-Prompt-Block
mit den aktuellen Diagnostic-Settings + Whisper-Toleranz-Hinweis.

Kein eager Bootstrap-Load: flux-bridge wartet auf config oder ersten
Request. Bei erstem HF-Download zeigt Banner "laedt erstmalig runter"
mit Pfeil-Icon, Toast in der App wenn fertig.

FLUX_MODEL aus der .env entfernt (Steuerung jetzt komplett ueber
Diagnostic). HF_TOKEN-Kommentar erklaert warum trotz lokaler Inference
noetig (HF Gate-Mechanismus fuer FLUX.1-dev).

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>

diagnostic/index.html

@@ -583,6 +583,48 @@
           </div>
         </details>
 
+        <!-- FLUX Bildgenerierung -->
+        <details style="background:#0D0D1A;border:1px solid #2A2A3E;border-radius:6px;padding:10px 12px;margin-bottom:12px;">
+          <summary style="color:#8888AA;font-size:12px;cursor:pointer;">FLUX Bildgenerierung (Modell + Keywords)</summary>
+          <div style="margin-top:10px;display:flex;flex-direction:column;gap:8px;">
+            <div style="color:#8888AA;font-size:11px;">
+              Steuerung der Image-Generation (flux-bridge auf der Gamebox).
+              Default-Modell wird via RVS gepusht — Wechsel triggert Pipeline-Reload (15-30s).
+              Keywords nutzt ARIAs Brain im System-Prompt.
+            </div>
+
+            <label style="color:#8888AA;font-size:12px;">Default-Modell:</label>
+            <select id="diag-flux-default-model" onchange="sendVoiceConfig()"
+              style="background:#1E1E2E;color:#fff;border:1px solid #2A2A3E;border-radius:6px;padding:6px 10px;font-size:13px;">
+              <option value="dev">FLUX.1-dev (hoechste Qualitaet, 20-90s)</option>
+              <option value="schnell">FLUX.1-schnell (4-step, 5-15s)</option>
+            </select>
+
+            <label style="color:#8888AA;font-size:12px;">
+              Raw-Keyword — Pipe-Modus, ARIA leitet den Prompt 1:1 durch (kein Rewriting):
+            </label>
+            <input type="text" id="diag-flux-keyword-raw"
+              placeholder="flux"
+              style="background:#1E1E2E;color:#fff;border:1px solid #2A2A3E;border-radius:6px;padding:6px 10px;font-size:13px;">
+
+            <label style="color:#8888AA;font-size:12px;">
+              Switch-Keyword — zwingt das ANDERE Modell als das Default fuer diesen Request:
+            </label>
+            <input type="text" id="diag-flux-keyword-switch"
+              placeholder="fix"
+              style="background:#1E1E2E;color:#fff;border:1px solid #2A2A3E;border-radius:6px;padding:6px 10px;font-size:13px;">
+
+            <div style="display:flex;gap:8px;align-items:center;margin-top:6px;">
+              <button class="btn primary" onclick="sendVoiceConfig()" style="padding:6px 14px;font-size:12px;">
+                Anwenden
+              </button>
+              <div style="color:#666680;font-size:10px;">
+                Beide Modelle = volle Qualitaet, schnell ist nur ein 4-Step-Distillat (Apache-2.0).
+              </div>
+            </div>
+          </div>
+        </details>
+
         <!-- Voice Cloning -->
         <div style="background:#1E1E2E;border-radius:8px;padding:12px;margin-top:8px;">
           <div style="color:#0096FF;font-size:13px;font-weight:600;margin-bottom:8px;">Stimme klonen</div>
@@ -1339,6 +1381,10 @@
           setIfPresent('diag-f5tts-vocab', msg.f5ttsVocabFile);
           setIfPresent('diag-f5tts-cfg', msg.f5ttsCfgStrength);
           setIfPresent('diag-f5tts-nfe', msg.f5ttsNfeStep);
+          // FLUX-Settings wiederherstellen
+          setIfPresent('diag-flux-default-model', msg.fluxDefaultModel);
+          setIfPresent('diag-flux-keyword-raw', msg.fluxKeywordRaw);
+          setIfPresent('diag-flux-keyword-switch', msg.fluxKeywordSwitch);
           return;
         }
 
@@ -2129,12 +2175,17 @@
         row.style.cssText = 'display:flex;align-items:center;gap:6px;';
         let dot = '⚫', color = '#666680', text = '';
         if (info.state === 'loading') {
-          dot = '⏳'; color = '#FFD60A'; anyLoading = true;
-          text = `${labels[s] || s}: laedt${info.model ? ' ' + info.model : ''}...`;
+          dot = info.downloading ? '⬇' : '⏳';
+          color = '#FFD60A'; anyLoading = true;
+          const action = info.downloading
+            ? 'laedt erstmalig runter (mehrere GB, kann dauern)'
+            : 'laedt';
+          text = `${labels[s] || s}: ${action}${info.model ? ' ' + info.model : ''}...`;
         } else if (info.state === 'ready') {
-          dot = '✅'; color = '#34C759';
+          dot = info.freshlyDownloaded ? '🎉' : '✅'; color = '#34C759';
           const sec = info.loadSeconds ? ` (${info.loadSeconds.toFixed(1)}s)` : '';
-          text = `${labels[s] || s}: bereit${info.model ? ' ' + info.model : ''}${sec}`;
+          const downloadedHint = info.freshlyDownloaded ? ' — Download fertig!' : '';
+          text = `${labels[s] || s}: bereit${info.model ? ' ' + info.model : ''}${sec}${downloadedHint}`;
         } else if (info.state === 'error') {
           dot = '❌'; color = '#FF3B30'; anyError = true;
           text = `${labels[s] || s}: Fehler ${info.error || ''}`;
@@ -2649,11 +2700,15 @@
       const f5ttsNfeRaw = document.getElementById('diag-f5tts-nfe')?.value || '';
       const f5ttsCfgStrength = f5ttsCfgRaw ? parseFloat(f5ttsCfgRaw) : undefined;
       const f5ttsNfeStep = f5ttsNfeRaw ? parseInt(f5ttsNfeRaw, 10) : undefined;
+      const fluxDefaultModel = document.getElementById('diag-flux-default-model')?.value || undefined;
+      const fluxKeywordRaw = document.getElementById('diag-flux-keyword-raw')?.value;
+      const fluxKeywordSwitch = document.getElementById('diag-flux-keyword-switch')?.value;
       send({
         action: 'send_voice_config',
         ttsEnabled, xttsVoice, whisperModel,
         f5ttsModel, f5ttsCkptFile, f5ttsVocabFile,
         f5ttsCfgStrength, f5ttsNfeStep,
+        fluxDefaultModel, fluxKeywordRaw, fluxKeywordSwitch,
       });
       const statusEl = document.getElementById('voice-status');
       if (statusEl && xttsVoice) {

diagnostic/server.js

@@ -1945,6 +1945,18 @@ wss.on("connection", (ws) => {
         if (msg.f5ttsNfeStep !== undefined && !isNaN(msg.f5ttsNfeStep)) {
           voiceConfig.f5ttsNfeStep = msg.f5ttsNfeStep;
         }
+        // FLUX-Settings (Default-Modell + User-Keywords). flux-bridge nutzt
+        // fluxDefaultModel zum Hot-Swap, Brain liest die Keywords direkt aus
+        // /shared/config/voice_config.json fuer den System-Prompt.
+        if (msg.fluxDefaultModel !== undefined) {
+          voiceConfig.fluxDefaultModel = (msg.fluxDefaultModel === "schnell") ? "schnell" : "dev";
+        }
+        if (msg.fluxKeywordRaw !== undefined) {
+          voiceConfig.fluxKeywordRaw = String(msg.fluxKeywordRaw || "").trim().toLowerCase() || "flux";
+        }
+        if (msg.fluxKeywordSwitch !== undefined) {
+          voiceConfig.fluxKeywordSwitch = String(msg.fluxKeywordSwitch || "").trim().toLowerCase() || "fix";
+        }
         try {
           fs.mkdirSync("/shared/config", { recursive: true });
           fs.writeFileSync("/shared/config/voice_config.json", JSON.stringify(voiceConfig, null, 2));