FROM faster-qwen3-tts-dgx-spark:v4

ENV PYTHONUNBUFFERED=1 \
    QWEN_TTS_MODEL=/models/Qwen3-TTS \
    QWEN_TTS_VOICES=/config/voices.json \
    QWEN_TTS_MAX_SEQ_LEN=4096

EXPOSE 8000

CMD ["/bin/bash", "-lc", "python3 /config/generate_voices.py && python3 /config/run_server.py --model ${QWEN_TTS_MODEL} --voices ${QWEN_TTS_VOICES} --port 8000 --max-seq-len ${QWEN_TTS_MAX_SEQ_LEN}"]