docker-compose.yml

version: "3.9"

# =============================================================================
# Agent-Q3 — 8-container stack
# Ollama Cloud model stack (nicholasjmcleod namespace)
# =============================================================================
#
#  Model → role mapping:
#    nicholasjmcleod/kimi-linear:q6_k        reasoner    /v1/instruct /v1/chat (multimodal)
#    nicholasjmcleod/harmonic-hermes:q6_k    tandem      /v1/tandem stage-2 + /v1/monitor analysis
#    nicholasjmcleod/ultracoder:q4_k_m       coder       /v1/code /v1/tandem stage-3 + monitor remediation
#    nicholasjmcleod/qwopus:q8_0             fallback    /v1/fallback
#    nicholasjmcleod/qwen3-coder-53b:q6_k    coder_ded   /v1/coder
#    nicholasjmcleod/genstruct:q6_k          genstruct   /v1/generate (structure middleware)
#    nicholasjmcleod/infinity-parser:latest  parser      /v1/parse (parse middleware)
#
#  Containers:
#    ollama       — Ollama Cloud proxy (nicholasjmcleod signed-in)
#    redis        — shared memory (cache, sessions, events)
#    mcp-bridge   — :8004  unified MCP gateway (.mcp.json)
#    multimodal   — :8000  reasoner + tandem + coder + fallback + genstruct + parser
#    coder        — :8001  coder_dedicated (Qwen3-Coder-53B)
#    research     — :8002  Perplexity/Exa/Tavily/Free/Polymarket + synthesis
#    monitor      — :8003  Harmonic-Hermes (analysis) + Ultra-Coder (remediation)
#    open-webui   — :3000  GUI
# =============================================================================

x-app-base: &app-base
  build:
    context: .
    dockerfile: Dockerfile
  env_file:
    - .env
  environment:
    OLLAMA_BASE_URL:          "http://ollama:11434"
    REDIS_URL:                "redis://redis:6379/0"
    COMPUTE_STRATEGY:         "cloud_priority"
    REQUEST_TIMEOUT_SECS:     "300"
    OLLAMA_CLOUD_ENABLED:     "true"
    REASONER_MODEL:           "nicholasjmcleod/kimi-linear:q6_k"
    TANDEM_MODEL:             "nicholasjmcleod/harmonic-hermes:q6_k"
    CODER_MODEL:              "nicholasjmcleod/ultracoder:q4_k_m"
    FALLBACK_MODEL:           "nicholasjmcleod/qwopus:q8_0"
    CODER_DEDICATED_MODEL:    "nicholasjmcleod/qwen3-coder-53b:q6_k"
    GENSTRUCT_MODEL:          "nicholasjmcleod/genstruct:q6_k"
    INFINITY_PARSER_MODEL:    "nicholasjmcleod/infinity-parser:latest"
    MONITOR_MODEL:            "nicholasjmcleod/harmonic-hermes:q6_k"
    MONITOR_CODER_MODEL:      "nicholasjmcleod/ultracoder:q4_k_m"
  depends_on:
    redis:
      condition: service_healthy
    ollama:
      condition: service_healthy
  volumes:
    - ./skills:/app/skills:ro
    - ./plugins:/app/plugins:ro
    - ./.mcp.json:/app/.mcp.json:ro
  restart: unless-stopped

services:

  # ── Shared memory ─────────────────────────────────────────────────────────
  redis:
    image: redis:7-alpine
    container_name: agent-q3-redis
    ports:
      - "6379:6379"
    volumes:
      - redis_data:/data
    healthcheck:
      test: ["CMD", "redis-cli", "ping"]
      interval: 10s
      timeout: 3s
      retries: 10
    restart: unless-stopped

  # ── Ollama Cloud proxy ─────────────────────────────────────────────────────
  # Device key at /root/.ollama/id_ed25519 enables cloud model pull.
  # Models pulled from nicholasjmcleod namespace on Ollama Cloud.
  ollama:
    build:
      context: .
      dockerfile: Dockerfile.ollama
    container_name: agent-q3-ollama
    ports:
      - "11434:11434"
    volumes:
      - ~/.ollama:/root/.ollama
    environment:
      OLLAMA_HOST:              "0.0.0.0"
      OLLAMA_ORIGINS:           "*"
      OLLAMA_KEEP_ALIVE:        "24h"
      OLLAMA_MAX_LOADED_MODELS: "7"
    restart: unless-stopped
    healthcheck:
      test: ["CMD", "curl", "-f", "http://localhost:11434/"]
      interval: 15s
      timeout: 5s
      retries: 20
      start_period: 60s

  # ── MCP Bridge ────────────────────────────────────────────────────────────
  mcp-bridge:
    <<: *app-base
    container_name: agent-q3-mcp-bridge
    ports:
      - "8004:8004"
    command: ["python", "-m", "uvicorn", "orchestrator.services.mcp_bridge:app",
              "--host", "0.0.0.0", "--port", "8004", "--log-level", "info"]
    healthcheck:
      test: ["CMD", "curl", "-f", "http://localhost:8004/health"]
      interval: 30s
      timeout: 10s
      retries: 5

  # ── Multimodal — Kimi-Linear + Harmonic-Hermes + Ultra-Coder + Qwopus ────
  multimodal:
    <<: *app-base
    container_name: agent-q3-multimodal
    ports:
      - "8000:8000"
    command: ["python", "-m", "uvicorn", "orchestrator.main:app",
              "--host", "0.0.0.0", "--port", "8000", "--log-level", "info"]
    healthcheck:
      test: ["CMD", "curl", "-f", "http://localhost:8000/health"]
      interval: 30s
      timeout: 10s
      retries: 5

  # ── Dedicated Coder — Qwen3-Coder-53B TOTAL-RECALL ───────────────────────
  coder:
    <<: *app-base
    container_name: agent-q3-coder
    ports:
      - "8001:8001"
    command: ["python", "-m", "uvicorn", "orchestrator.services.coder:app",
              "--host", "0.0.0.0", "--port", "8001", "--log-level", "info"]
    healthcheck:
      test: ["CMD", "curl", "-f", "http://localhost:8001/health"]
      interval: 30s
      timeout: 10s
      retries: 5

  # ── Deep Research ─────────────────────────────────────────────────────────
  research:
    <<: *app-base
    container_name: agent-q3-research
    ports:
      - "8002:8002"
    command: ["python", "-m", "uvicorn", "orchestrator.services.research:app",
              "--host", "0.0.0.0", "--port", "8002", "--log-level", "info"]
    healthcheck:
      test: ["CMD", "curl", "-f", "http://localhost:8002/health"]
      interval: 30s
      timeout: 10s
      retries: 5

  # ── Monitor — Harmonic-Hermes (analysis) + Ultra-Coder (remediation) ─────
  monitor:
    <<: *app-base
    container_name: agent-q3-monitor
    ports:
      - "8003:8003"
    environment:
      OLLAMA_BASE_URL:        "http://ollama:11434"
      REDIS_URL:              "redis://redis:6379/0"
      MONITOR_MODEL:          "nicholasjmcleod/harmonic-hermes:q6_k"
      MONITOR_CODER_MODEL:    "nicholasjmcleod/ultracoder:q4_k_m"
      MONITOR_POLL_INTERVAL_SECS: "30"
      MONITOR_TARGETS:        "http://multimodal:8000,http://coder:8001,http://research:8002,http://mcp-bridge:8004,http://ollama:11434"
    command: ["python", "-m", "uvicorn", "orchestrator.services.monitor:app",
              "--host", "0.0.0.0", "--port", "8003", "--log-level", "info"]
    healthcheck:
      test: ["CMD", "curl", "-f", "http://localhost:8003/health"]
      interval: 30s
      timeout: 10s
      retries: 5

  # ── Open WebUI ────────────────────────────────────────────────────────────
  open-webui:
    image: ghcr.io/open-webui/open-webui:main
    container_name: agent-q3-webui
    ports:
      - "3000:8080"
    volumes:
      - open_webui_data:/app/backend/data
    environment:
      OLLAMA_BASE_URL:  "http://ollama:11434"
      WEBUI_SECRET_KEY: ${WEBUI_SECRET_KEY:-change-me-in-env}
      ENABLE_SIGNUP:    "true"
      DEFAULT_MODELS:   "nicholasjmcleod/kimi-linear:q6_k"
    depends_on:
      ollama:
        condition: service_healthy
    restart: unless-stopped

volumes:
  redis_data:
    driver: local
  open_webui_data:
    driver: local