Jarvis2.O/docker-compose.yml at main · AISocietyIITJ/Jarvis2.O · 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
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
# See NOTE comments for places to modify.
services:
  traefik:
    image: traefik:v3.3.1
    command:
      # Swarm provider configuration
      - "--providers.docker=true"
      - "--providers.docker.exposedbydefault=false"

      # This is set up for HTTP. If you want HTTPS support for production, use Docker Swarm
      # (check out swarm-deploy.yml) or ask ChatGPT to modify this file for you.
      - "--entrypoints.web.address=:80"
    ports:
      - "80:80"
    volumes:
      - "/var/run/docker.sock:/var/run/docker.sock:ro"

  frontend:
    image: unmute-frontend:latest
    build:
      context: frontend/
      dockerfile: hot-reloading.Dockerfile
    volumes:
      - ./frontend/src:/app/src
    labels:
      - "traefik.enable=true"
      - "traefik.http.routers.frontend.rule=PathPrefix(`/`)"
      - "traefik.http.routers.frontend.entrypoints=web"
      - "traefik.http.services.frontend.loadbalancer.server.port=3000"
      - "traefik.http.routers.frontend.priority=10" # lowest priority

  backend:
    image: unmute-backend:latest
    build:
      context: ./
      target: hot-reloading
    volumes:
      - ./unmute:/app/unmute
    environment:
      - KYUTAI_STT_URL=ws://stt:8080
      - KYUTAI_TTS_URL=ws://tts:8080
      - KYUTAI_LLM_URL=http://llm:8000
      - NEWSAPI_API_KEY=$NEWSAPI_API_KEY
    labels:
      - "traefik.enable=true"
      - "traefik.http.routers.backend.rule=PathPrefix(`/api`)"
      - "traefik.http.routers.backend.middlewares=strip-api"
      - "traefik.http.middlewares.strip-api.replacepathregex.regex=^/api/(.*)"
      - "traefik.http.middlewares.strip-api.replacepathregex.replacement=/$$1"
      - "traefik.http.routers.backend.entrypoints=web"
      - "traefik.http.services.backend.loadbalancer.server.port=80"
      - "traefik.http.routers.backend.priority=100" # higher priority than frontend
      - "prometheus-port=80"

  tts:
    image: moshi-server:latest
    command: ["worker", "--config", "configs/tts.toml"]
    build:
      context: services/moshi-server
      dockerfile: public.Dockerfile
    environment:
      - HUGGING_FACE_HUB_TOKEN=$HUGGING_FACE_HUB_TOKEN
    volumes:
      - ./volumes/hf-cache:/root/.cache/huggingface
      - ./volumes/cargo-registry-tts:/root/.cargo/registry
      - ./volumes/tts-target:/app/target
      - ./volumes/uv-cache:/root/.cache/uv
      - /tmp/models/:/models
      - ./volumes/tts-logs:/logs
    deploy:
      resources:
        reservations:
          devices:
            - driver: nvidia
              count: 1
              capabilities: [gpu]

  stt:
    image: moshi-server:latest
    command: ["worker", "--config", "configs/stt.toml"]
    build:
      context: services/moshi-server
      dockerfile: public.Dockerfile
    environment:
      - HUGGING_FACE_HUB_TOKEN=$HUGGING_FACE_HUB_TOKEN
    volumes:
      - ./volumes/hf-cache:/root/.cache/huggingface
      - ./volumes/cargo-registry-stt:/root/.cargo/registry
      - ./volumes/stt-target:/app/target
      - ./volumes/uv-cache:/root/.cache/uv
      - /tmp/models/:/models
      - ./volumes/stt-logs:/logs
    deploy:
      resources:
        reservations:
          devices:
            - driver: nvidia
              count: 1
              capabilities: [gpu]

  llm:
    image: vllm/vllm-openai:v0.9.1
    command:
      [
        # NOTE: Change the LLM here if you want.
        # (caution: gemma-3-1b-it also exists but it's slow on vLLM: https://github.com/vllm-project/vllm/issues/19575)
        "--model=meta-llama/Llama-3.2-1B-Instruct",
        # NOTE: You can adapt this based on your GPU memory.
        # A higher value takes more memory but supports longer conversations.
        "--max-model-len=1536",
        "--dtype=bfloat16",
        # NOTE: Change this based on your GPU memory.
        # A higher value can make inference faster.
        "--gpu-memory-utilization=0.4",
      ]
    volumes:
      - ./volumes/hf-cache:/root/.cache/huggingface
      - ./volumes/vllm-cache:/root/.cache/vllm
    environment:
      - HUGGING_FACE_HUB_TOKEN=$HUGGING_FACE_HUB_TOKEN
    deploy:
      resources:
        reservations:
          devices:
            - driver: nvidia
              count: 1
              capabilities: [gpu]

networks:
  default: