version: "3.9"

services:
  speech-edit-encoder:
    container_name: speech-edit-encoder
    build:
      context: .
      dockerfile: Dockerfile
      args:
        PIP_MIRROR: ${PIP_MIRROR:-https://mirrors.cloud.tencent.com/pypi/simple}
    image: speech-edit-grpc-v2:latest
    restart: unless-stopped
    working_dir: /workspace
    environment:
      PYTHONUNBUFFERED: "1"
      PYTHONPATH: /workspace
      VIITORVOICE_GRPC_HOST: 0.0.0.0
      VIITORVOICE_DEVICE_ID: ${VIITORVOICE_DEVICE_ID:-0}
      VIITORVOICE_WARMUP_ON_START: ${VIITORVOICE_WARMUP_ON_START:-true}
      VIITORVOICE_V2_REQUEST_TIMEOUT_SEC: ${VIITORVOICE_V2_REQUEST_TIMEOUT_SEC:-900}
      VIITORVOICE_REQUEST_TIMEOUT_SEC: ${VIITORVOICE_REQUEST_TIMEOUT_SEC:-900}
      VIITORVOICE_V2_LOG_JSON: ${VIITORVOICE_V2_LOG_JSON:-true}
      VIITORVOICE_LOCAL_MODELS: /workspace/local_models
      VIITORVOICE_CODEC_DUALCODEC_PATH: ${VIITORVOICE_CODEC_DUALCODEC_PATH:-/workspace/local_models/dualcodec/dualcodec_ckpts}
      VIITORVOICE_CODEC_ENCODER_MODEL_ID: ${VIITORVOICE_CODEC_ENCODER_MODEL_ID:-25hz_v1}
      VIITORVOICE_CODEC_ENCODER_BACKEND: ${VIITORVOICE_CODEC_ENCODER_BACKEND:-torch}
      VIITORVOICE_CODEC_ENCODER_PRECISION: ${VIITORVOICE_CODEC_ENCODER_PRECISION:-bf16}
      no_proxy: "127.0.0.1,localhost,0.0.0.0,speech-edit-encoder,speech-edit-llm,speech-edit-decoder,speech-edit-orchestrator,speech-edit-http,speech-edit-provider"
      NO_PROXY: "127.0.0.1,localhost,0.0.0.0,speech-edit-encoder,speech-edit-llm,speech-edit-decoder,speech-edit-orchestrator,speech-edit-http,speech-edit-provider"
      http_proxy: ""
      https_proxy: ""
      all_proxy: ""
      HTTP_PROXY: ""
      HTTPS_PROXY: ""
      ALL_PROXY: ""
    ports:
      - "${VIITORVOICE_V2_ENCODER_HOST_PORT:-51051}:51051"
    volumes:
      - .:/workspace
    command:
      - /opt/viitorvoice-runtime/.venv/bin/python
      - -m
      - viitorvoice.grpc_server.encoder.server
      - --host
      - 0.0.0.0
      - --port
      - "51051"
      - --log-level
      - ${VIITORVOICE_V2_LOG_LEVEL:-INFO}
    deploy:
      resources:
        reservations:
          devices:
            - driver: nvidia
              count: all
              capabilities: [gpu]

  speech-edit-llm:
    container_name: speech-edit-llm
    image: speech-edit-grpc-v2:latest
    restart: unless-stopped
    working_dir: /workspace
    environment:
      PYTHONUNBUFFERED: "1"
      PYTHONPATH: /workspace
      VIITORVOICE_GRPC_HOST: 0.0.0.0
      VIITORVOICE_DEVICE_ID: ${VIITORVOICE_DEVICE_ID:-0}
      VIITORVOICE_WARMUP_ON_START: ${VIITORVOICE_WARMUP_ON_START:-true}
      VIITORVOICE_V2_REQUEST_TIMEOUT_SEC: ${VIITORVOICE_V2_REQUEST_TIMEOUT_SEC:-900}
      VIITORVOICE_REQUEST_TIMEOUT_SEC: ${VIITORVOICE_REQUEST_TIMEOUT_SEC:-900}
      VIITORVOICE_V2_LOG_JSON: ${VIITORVOICE_V2_LOG_JSON:-true}
      VIITORVOICE_LOCAL_MODELS: /workspace/local_models
      VIITORVOICE_LLM_CHECKPOINT: ${VIITORVOICE_LLM_CHECKPOINT:-/workspace/local_models/llm/0p6_emotion}
      VIITORVOICE_LLM_ONNX: ${VIITORVOICE_LLM_ONNX:-/workspace/local_models/llm/0p6_emotion/.cache/onnx_backbone_fp32/llm_backbone_dynamic.onnx}
      VIITORVOICE_LLM_BACKEND: ${VIITORVOICE_LLM_BACKEND:-onnx-trt}
      VIITORVOICE_LLM_PRECISION: ${VIITORVOICE_LLM_PRECISION:-fp32}
      VIITORVOICE_LLM_TRT_CACHE: ${VIITORVOICE_LLM_TRT_CACHE:-/workspace/local_models/llm/0p6_emotion/.cache/trt_cache}
      no_proxy: "127.0.0.1,localhost,0.0.0.0,speech-edit-encoder,speech-edit-llm,speech-edit-decoder,speech-edit-orchestrator,speech-edit-http,speech-edit-provider"
      NO_PROXY: "127.0.0.1,localhost,0.0.0.0,speech-edit-encoder,speech-edit-llm,speech-edit-decoder,speech-edit-orchestrator,speech-edit-http,speech-edit-provider"
      http_proxy: ""
      https_proxy: ""
      all_proxy: ""
      HTTP_PROXY: ""
      HTTPS_PROXY: ""
      ALL_PROXY: ""
    ports:
      - "${VIITORVOICE_V2_LLM_HOST_PORT:-51052}:51052"
    volumes:
      - .:/workspace
    command:
      - /opt/viitorvoice-runtime/.venv/bin/python
      - -m
      - viitorvoice.grpc_server.llm.server
      - --host
      - 0.0.0.0
      - --port
      - "51052"
      - --log-level
      - ${VIITORVOICE_V2_LOG_LEVEL:-INFO}
    deploy:
      resources:
        reservations:
          devices:
            - driver: nvidia
              count: all
              capabilities: [gpu]

  speech-edit-decoder:
    container_name: speech-edit-decoder
    image: speech-edit-grpc-v2:latest
    restart: unless-stopped
    working_dir: /workspace
    environment:
      PYTHONUNBUFFERED: "1"
      PYTHONPATH: /workspace
      VIITORVOICE_GRPC_HOST: 0.0.0.0
      VIITORVOICE_DEVICE_ID: ${VIITORVOICE_DEVICE_ID:-0}
      VIITORVOICE_WARMUP_ON_START: ${VIITORVOICE_WARMUP_ON_START:-true}
      VIITORVOICE_V2_REQUEST_TIMEOUT_SEC: ${VIITORVOICE_V2_REQUEST_TIMEOUT_SEC:-900}
      VIITORVOICE_REQUEST_TIMEOUT_SEC: ${VIITORVOICE_REQUEST_TIMEOUT_SEC:-900}
      VIITORVOICE_V2_LOG_JSON: ${VIITORVOICE_V2_LOG_JSON:-true}
      VIITORVOICE_LOCAL_MODELS: /workspace/local_models
      VIITORVOICE_CODEC_DUALCODEC_PATH: ${VIITORVOICE_CODEC_DUALCODEC_PATH:-/workspace/local_models/dualcodec/dualcodec_ckpts}
      VIITORVOICE_CODEC_DECODER_BACKEND: ${VIITORVOICE_CODEC_DECODER_BACKEND:-onnx-trt}
      no_proxy: "127.0.0.1,localhost,0.0.0.0,speech-edit-encoder,speech-edit-llm,speech-edit-decoder,speech-edit-orchestrator,speech-edit-http,speech-edit-provider"
      NO_PROXY: "127.0.0.1,localhost,0.0.0.0,speech-edit-encoder,speech-edit-llm,speech-edit-decoder,speech-edit-orchestrator,speech-edit-http,speech-edit-provider"
      http_proxy: ""
      https_proxy: ""
      all_proxy: ""
      HTTP_PROXY: ""
      HTTPS_PROXY: ""
      ALL_PROXY: ""
    ports:
      - "${VIITORVOICE_V2_DECODER_HOST_PORT:-51053}:51053"
    volumes:
      - .:/workspace
    command:
      - /opt/viitorvoice-runtime/.venv/bin/python
      - -m
      - viitorvoice.grpc_server.decoder.server
      - --host
      - 0.0.0.0
      - --port
      - "51053"
      - --log-level
      - ${VIITORVOICE_V2_LOG_LEVEL:-INFO}
    deploy:
      resources:
        reservations:
          devices:
            - driver: nvidia
              count: all
              capabilities: [gpu]

  speech-edit-orchestrator:
    container_name: speech-edit-orchestrator
    image: speech-edit-grpc-v2:latest
    restart: unless-stopped
    working_dir: /workspace
    depends_on:
      - speech-edit-encoder
      - speech-edit-llm
      - speech-edit-decoder
    environment:
      PYTHONUNBUFFERED: "1"
      PYTHONPATH: /workspace
      VIITORVOICE_GRPC_HOST: 0.0.0.0
      VIITORVOICE_DEVICE_ID: ${VIITORVOICE_DEVICE_ID:-0}
      VIITORVOICE_WARMUP_ON_START: ${VIITORVOICE_WARMUP_ON_START:-true}
      VIITORVOICE_V2_REQUEST_TIMEOUT_SEC: ${VIITORVOICE_V2_REQUEST_TIMEOUT_SEC:-900}
      VIITORVOICE_REQUEST_TIMEOUT_SEC: ${VIITORVOICE_REQUEST_TIMEOUT_SEC:-900}
      VIITORVOICE_V2_LOG_JSON: ${VIITORVOICE_V2_LOG_JSON:-true}
      VIITORVOICE_LOCAL_MODELS: /workspace/local_models
      VIITORVOICE_ALIGNER_MODEL: ${VIITORVOICE_ALIGNER_MODEL:-/workspace/local_models/aligner/Qwen3-ForcedAligner-0.6B}
      VIITORVOICE_ALIGNER_DEVICE: ${VIITORVOICE_ALIGNER_DEVICE:-cuda:0}
      VIITORVOICE_ALIGNER_LANGUAGE: ${VIITORVOICE_ALIGNER_LANGUAGE:-zh}
      VIITORVOICE_V2_ENCODER_TARGET: speech-edit-encoder:51051
      VIITORVOICE_V2_LLM_TARGET: speech-edit-llm:51052
      VIITORVOICE_V2_DECODER_TARGET: speech-edit-decoder:51053
      no_proxy: "127.0.0.1,localhost,0.0.0.0,speech-edit-encoder,speech-edit-llm,speech-edit-decoder,speech-edit-orchestrator,speech-edit-http,speech-edit-provider"
      NO_PROXY: "127.0.0.1,localhost,0.0.0.0,speech-edit-encoder,speech-edit-llm,speech-edit-decoder,speech-edit-orchestrator,speech-edit-http,speech-edit-provider"
      http_proxy: ""
      https_proxy: ""
      all_proxy: ""
      HTTP_PROXY: ""
      HTTPS_PROXY: ""
      ALL_PROXY: ""
    ports:
      - "${VIITORVOICE_V2_ORCH_HOST_PORT:-50051}:50051"
    volumes:
      - .:/workspace
    command:
      - /opt/viitorvoice-runtime/.venv/bin/python
      - -m
      - viitorvoice.grpc_server.orchestrator.server
      - --host
      - 0.0.0.0
      - --port
      - "50051"
      - --log-level
      - ${VIITORVOICE_V2_LOG_LEVEL:-INFO}
    deploy:
      resources:
        reservations:
          devices:
            - driver: nvidia
              count: all
              capabilities: [gpu]

  speech-edit-provider:
    container_name: speech-edit-provider
    image: speech-edit-grpc-v2:latest
    restart: unless-stopped
    working_dir: /workspace
    depends_on:
      - speech-edit-orchestrator
    environment:
      PYTHONUNBUFFERED: "1"
      PYTHONPATH: /workspace
      VIITORVOICE_GRPC_HOST: 0.0.0.0
      VIITORVOICE_V2_REQUEST_TIMEOUT_SEC: ${VIITORVOICE_V2_REQUEST_TIMEOUT_SEC:-900}
      VIITORVOICE_REQUEST_TIMEOUT_SEC: ${VIITORVOICE_REQUEST_TIMEOUT_SEC:-900}
      VIITORVOICE_V2_LOG_JSON: ${VIITORVOICE_V2_LOG_JSON:-true}
      SPEECH_EDIT_ORCHESTRATOR_TARGET: speech-edit-orchestrator:50051
      no_proxy: "127.0.0.1,localhost,0.0.0.0,speech-edit-encoder,speech-edit-llm,speech-edit-decoder,speech-edit-orchestrator,speech-edit-http,speech-edit-provider"
      NO_PROXY: "127.0.0.1,localhost,0.0.0.0,speech-edit-encoder,speech-edit-llm,speech-edit-decoder,speech-edit-orchestrator,speech-edit-http,speech-edit-provider"
      http_proxy: ""
      https_proxy: ""
      all_proxy: ""
      HTTP_PROXY: ""
      HTTPS_PROXY: ""
      ALL_PROXY: ""
    ports:
      - "${VIITORVOICE_V2_PROVIDER_HOST_PORT:-50062}:50062"
    volumes:
      - .:/workspace
    command:
      - /opt/viitorvoice-runtime/.venv/bin/python
      - -m
      - viitorvoice.grpc_server.provider.server
      - --host
      - 0.0.0.0
      - --port
      - "50062"
      - --orchestrator-target
      - speech-edit-orchestrator:50051
      - --log-level
      - ${VIITORVOICE_V2_LOG_LEVEL:-INFO}

  speech-edit-http:
    container_name: speech-edit-http
    image: speech-edit-grpc-v2:latest
    restart: unless-stopped
    working_dir: /workspace
    depends_on:
      - speech-edit-orchestrator
    environment:
      PYTHONUNBUFFERED: "1"
      PYTHONPATH: /workspace
      VIITORVOICE_GRPC_HOST: 0.0.0.0
      VIITORVOICE_V2_HTTP_TARGET: speech-edit-orchestrator:50051
      no_proxy: "127.0.0.1,localhost,0.0.0.0,speech-edit-encoder,speech-edit-llm,speech-edit-decoder,speech-edit-orchestrator,speech-edit-http,speech-edit-provider"
      NO_PROXY: "127.0.0.1,localhost,0.0.0.0,speech-edit-encoder,speech-edit-llm,speech-edit-decoder,speech-edit-orchestrator,speech-edit-http,speech-edit-provider"
      http_proxy: ""
      https_proxy: ""
      all_proxy: ""
      HTTP_PROXY: ""
      HTTPS_PROXY: ""
      ALL_PROXY: ""
    ports:
      - "${VIITORVOICE_V2_HTTP_HOST_PORT:-7861}:7861"
    volumes:
      - .:/workspace
    command:
      - /opt/viitorvoice-runtime/.venv/bin/python
      - -m
      - viitorvoice.grpc_server.http.server
      - --host
      - 0.0.0.0
      - --port
      - "7861"
      - --grpc-target
      - speech-edit-orchestrator:50051
      - --log-level
      - ${VIITORVOICE_V2_LOG_LEVEL:-INFO}