dotty-stackchan/docker-compose.yml.template at main · BrettKinny/dotty-stackchan · GitHub

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
services:
  xiaozhi-esp32-server:
    build:
      context: .
      dockerfile: Dockerfile
    image: xiaozhi-esp32-server-piper:local
    container_name: xiaozhi-esp32-server
    restart: unless-stopped
    # --- BEGIN CUDA BLOCK (removed by `make setup` if no NVIDIA runtime) ---
    # GPU access for WhisperLocal ASR (faster-whisper float16 needs CUDA).
    # Requires nvidia-container-toolkit on the host.
    #
    # ASR ONLY — this has nothing to do with the LLM. A CPU-only host is fine:
    # `make setup` drops this block and selects FunASR (CPU) instead. Serving
    # models from a GPU on another machine (LM Studio, Ollama, llama-swap) needs
    # no GPU here at all — see docs/llm-backends.md.
    runtime: nvidia
    deploy:
      resources:
        reservations:
          devices:
            - driver: nvidia
              count: all
              capabilities: [gpu]
    # --- END CUDA BLOCK ---
    environment:
      # --- BEGIN CUDA ENV (removed by `make setup` if no NVIDIA runtime) ---
      - NVIDIA_VISIBLE_DEVICES=all
      - NVIDIA_DRIVER_CAPABILITIES=compute,utility
      # --- END CUDA ENV ---
      - TZ=<TZ_VALUE>
      # EDIT: set to the dotty-behaviour service URL for vision intent handling.
      # dotty-behaviour runs on the same host; use the host LAN IP, not loopback
      # (this container is on a bridge network, so 127.0.0.1 resolves to itself).
      - VISION_BRIDGE_URL=http://<XIAOZHI_HOST>:8090
      # Active persona file under ./personas/ (loaded by the LLM provider).
      # Currently fixed to "default" — smart.md exists for manual swap when you
      # want an adult-framed persona on the smart-mode cloud path.
      - PERSONA=default
      # kid_mode / smart_mode toggle state — receiveAudioHandle.py READS these
      # to paint the firmware LED pips on reconnect; the bridge dashboard WRITES
      # them. Both containers must point at the SAME files, so this path is
      # bind-mounted (below, read-only) from the bridge's state dir. Without
      # this the reader falls through to the dead /root/zeroclaw-bridge default
      # and the firmware pips desync from the dashboard on every reconnect.
      - DOTTY_KID_MODE_STATE=/var/lib/dotty-bridge/state/kid-mode
      - DOTTY_SMART_MODE_STATE=/var/lib/dotty-bridge/state/smart-mode
      # Shared admin-API secret (X-Admin-Token). Read from ./.env — `make
      # setup` generates it. Unset/empty = permissive (admin routes open);
      # set = this server enforces it AND every caller (bridge,
      # dotty-behaviour, dotty-pi) must carry the same value in its own
      # .env. See .env.example.
      - DOTTY_ADMIN_TOKEN=${DOTTY_ADMIN_TOKEN:-}
    ports:
      - "8000:8000"
      - "8003:8003"
    volumes:
      - ./data/.config.yaml:/opt/xiaozhi-esp32-server/data/.config.yaml:ro
      - ./models/SenseVoiceSmall:/opt/xiaozhi-esp32-server/models/SenseVoiceSmall
      - ./tmp:/opt/xiaozhi-esp32-server/tmp
      - ./custom-providers/openai_compat:/opt/xiaozhi-esp32-server/core/providers/llm/openai_compat
      - ./custom-providers/pi_voice:/opt/xiaozhi-esp32-server/core/providers/llm/pi_voice
      # PiVoiceLLM shells out to `docker exec -i dotty-pi pi --mode rpc ...`,
      # so this container needs the docker CLI binary AND the host docker
      # socket. WARNING: mounting /var/run/docker.sock gives this container
      # effective root on the docker host (it can `docker run --privileged
      # anything`). Acceptable for a single-purpose self-hosted appliance;
      # do not enable on a multi-tenant host. Required for `selected_module.LLM:
      # PiVoiceLLM` — comment both lines out if running OpenAICompat.
      - /var/run/docker.sock:/var/run/docker.sock
      - /usr/bin/docker:/usr/bin/docker:ro
      - ./personas:/opt/xiaozhi-esp32-server/personas:ro
      - ./custom-providers/edge_stream/edge_stream.py:/opt/xiaozhi-esp32-server/core/providers/tts/edge_stream.py:ro
      - ./custom-providers/piper_local/piper_local.py:/opt/xiaozhi-esp32-server/core/providers/tts/piper_local.py:ro
      - ./models/piper:/opt/xiaozhi-esp32-server/models/piper:ro
      - ./custom-providers/asr/fun_local.py:/opt/xiaozhi-esp32-server/core/providers/asr/fun_local.py:ro
      - ./custom-providers/asr/whisper_local.py:/opt/xiaozhi-esp32-server/core/providers/asr/whisper_local.py:ro
      - ./models/whisper-small.en-ct2:/opt/xiaozhi-esp32-server/models/whisper-small.en-ct2:ro
      - ./custom-providers/asr/sensevoice_onnx.py:/opt/xiaozhi-esp32-server/core/providers/asr/sensevoice_onnx.py:ro
      - ./models/SenseVoiceSmall-onnx:/opt/xiaozhi-esp32-server/models/SenseVoiceSmall-onnx:ro
      - ./receiveAudioHandle.py:/opt/xiaozhi-esp32-server/core/handle/receiveAudioHandle.py:ro
      - ./dances.py:/opt/xiaozhi-esp32-server/core/handle/dances.py:ro
      - ./custom-providers/textUtils.py:/opt/xiaozhi-esp32-server/core/utils/textUtils.py:ro
      # DOTTY DeviceCommand seam: MCP request ids + envelope + per-conn
      # serialized sends, shared by http_server.py + receiveAudioHandle.py
      - ./custom-providers/xiaozhi-patches/device_command.py:/opt/xiaozhi-esp32-server/core/utils/device_command.py:ro
      - ./songs:/opt/xiaozhi-esp32-server/config/assets/songs:ro
      # DOTTY portal: admin /inject-text route + active-connection registry
      - ./custom-providers/xiaozhi-patches/portal_bridge.py:/opt/xiaozhi-esp32-server/core/portal_bridge.py:ro
      - ./custom-providers/xiaozhi-patches/websocket_server.py:/opt/xiaozhi-esp32-server/core/websocket_server.py:ro
      - ./custom-providers/xiaozhi-patches/http_server.py:/opt/xiaozhi-esp32-server/core/http_server.py:ro
      # DOTTY ambient perception: relay firmware "event" frames to the bridge
      - ./custom-providers/xiaozhi-patches/textMessageHandlerRegistry.py:/opt/xiaozhi-esp32-server/core/handle/textMessageHandlerRegistry.py:ro
      # DOTTY OTA: patched ota_handler that composes the firmware download URL
      # directly from local_ip+http_port instead of doing a string-replace on
      # `vision_explain` (which only worked when vision lived at the canonical
      # `/mcp/vision/explain` path; our setup points it at the dotty-behaviour service).
      - ./custom-providers/xiaozhi-patches/ota_handler.py:/opt/xiaozhi-esp32-server/core/api/ota_handler.py:ro
      # OTA artifact directory — drop m5stack-stack-chan_<version>.bin here
      # and the device polls + self-flashes on next boot. See docs/ota-verification.md.
      - ./data/bin:/opt/xiaozhi-esp32-server/data/bin
      # Shared kid/smart-mode toggle state (READ-ONLY here — the bridge
      # dashboard owns writes). Host path MUST match the bridge container's
      # state mount (bridge/docker-compose.yml → /var/lib/dotty-bridge/state).
      # Override DOTTY_STATE_DIR if your bridge writes elsewhere. Keeps the
      # firmware kid/smart LED pips in sync with the dashboard across reconnects.
      - ${DOTTY_STATE_DIR:-/mnt/user/appdata/dotty-bridge/state}:/var/lib/dotty-bridge/state:ro