Funktionierender STand!!!

docker/entrypoint_clone.sh

@@ -2,33 +2,43 @@
 set -e
 
 MODEL="${QWEN3_TTS_CLONE_MODEL:-}"
-PORT="${QWEN3_TTS_CLONE_PORT:-8091}"
+PORT="${QWEN3_TTS_CLONE_PORT:-8091}"   # external-facing proxy port
+VLLM_PORT=8095                          # internal vllm-omni port
 
 if [ -z "$MODEL" ]; then
   echo "[clone] QWEN3_TTS_CLONE_MODEL ist nicht gesetzt." >&2
   exit 2
 fi
 
-echo "[clone] starte Clone-Modell: ${MODEL} auf Port ${PORT}"
+echo "[clone] starte Clone-Modell: ${MODEL} auf Port ${PORT} (vllm intern: ${VLLM_PORT})"
 
 if [ -f /patch_qwen3_tts_runtime.py ]; then
   python3 /patch_qwen3_tts_runtime.py
 fi
 
+# Proxy auf client-facing PORT; leitet an internen VLLM_PORT weiter.
+if [ -f /ws_log_proxy.py ]; then
+  python3 /ws_log_proxy.py \
+    --host 0.0.0.0 \
+    --port "$PORT" \
+    --upstream "ws://localhost:${VLLM_PORT}/v1/audio/speech/stream" &
+  echo "[ws-proxy] Proxy läuft auf Port ${PORT} -> vllm-omni Port ${VLLM_PORT}"
+fi
+
 vllm-omni serve "$MODEL" \
   --omni \
   --host 0.0.0.0 \
-  --port "$PORT" \
+  --port "$VLLM_PORT" \
   --deploy-config /deploy/qwen3_tts_clone.yaml \
   --trust-remote-code &
 SERVER_PID=$!
 
 (
   for _ in $(seq 1 150); do
-    if curl -sf "http://localhost:${PORT}/health" >/dev/null 2>&1; then
+    if curl -sf "http://localhost:${VLLM_PORT}/health" >/dev/null 2>&1; then
       echo "[warmup] Clone gesund — sende Warmup-Request (kompiliert CUDA-Graphen)..."
       t0=$SECONDS
-      curl -sf -X POST "http://localhost:${PORT}/v1/audio/speech" \
+      curl -sf -X POST "http://localhost:${VLLM_PORT}/v1/audio/speech" \
         -H 'Content-Type: application/json' \
         --max-time 180 \
         -d "{\"model\":\"${MODEL}\",\"input\":\"System wird aufgewärmt.\",\"language\":\"German\",\"response_format\":\"wav\",\"task_type\":\"VoiceDesign\",\"instructions\":\"Eine ruhige Stimme.\"}" \