# docker/docker-compose.prod.yml - Production deployment
# Used by: Production deployment with `docker compose -f docker/docker-compose.prod.yml up`
#
# This file is the SINGLE SOURCE OF TRUTH for production container configuration.
# All values use environment variable substitution from .env or .env.production files.
# Named volumes persist data between container restarts.
#
# Requirements: 0.4, 11.1, 11.2, 11.3, 15.3, 16.3
#
# Prerequisites:
# 1. Copy .env.production.example to .env.production and configure
# 2. (Optional) Build Docker image: docker build -f packages/server/Dockerfile -t thought-server .
#
# ============================================================================
# PROFILES REFERENCE
# ============================================================================
#
# QUICK START PROFILES (recommended):
# - all : Everything with local Ollama (postgres + redis + ollama + server + ui)
# - all-remote : Everything with remote Ollama (postgres + redis + server + ui)
#
# INFRASTRUCTURE PROFILES:
# - infra : PostgreSQL + Redis only (for external server connections)
# - infra-ollama : PostgreSQL + Redis + local Ollama (for local dev server)
#
# SERVER PROFILES (choose ONE):
# - mcp-local : MCP server only with local Ollama
# - mcp-remote : MCP server only with remote Ollama
# - rest-local : REST API server only with local Ollama
# - rest-remote : REST API server only with remote Ollama
# - server-local : Full server (MCP + REST) with local Ollama
# - server-remote : Full server (MCP + REST) with remote Ollama
#
# UI PROFILE:
# - ui : Mind Palace UI (can combine with any server profile)
#
# ============================================================================
# USAGE EXAMPLES
# ============================================================================
#
# Start all services (local Ollama):
# docker compose -f docker/docker-compose.prod.yml --profile all up -d
#
# Start all services (remote Ollama):
# docker compose -f docker/docker-compose.prod.yml --profile all-remote up -d
#
# Start MCP server only (local Ollama):
# docker compose -f docker/docker-compose.prod.yml --profile mcp-local up -d
#
# Start MCP server only (remote Ollama):
# docker compose -f docker/docker-compose.prod.yml --profile mcp-remote up -d
#
# Start REST API server only (local Ollama):
# docker compose -f docker/docker-compose.prod.yml --profile rest-local up -d
#
# Start REST API server only (remote Ollama):
# docker compose -f docker/docker-compose.prod.yml --profile rest-remote up -d
#
# Start full server (local Ollama):
# docker compose -f docker/docker-compose.prod.yml --profile server-local up -d
#
# Start full server (remote Ollama):
# docker compose -f docker/docker-compose.prod.yml --profile server-remote up -d
#
# Start infrastructure only (for external server):
# docker compose -f docker/docker-compose.prod.yml --profile infra up -d
#
# Start infrastructure + local Ollama:
# docker compose -f docker/docker-compose.prod.yml --profile infra-ollama up -d
#
# Start UI only (assumes server is running elsewhere):
# docker compose -f docker/docker-compose.prod.yml --profile ui up -d
#
# Start full server + UI (local Ollama):
# docker compose -f docker/docker-compose.prod.yml --profile server-local --profile ui up -d
#
# ============================================================================
# MCP CLIENT CONNECTION
# ============================================================================
#
# By default, the thought container runs in MCP_STANDBY_MODE. This means:
# - The container stays alive without starting the MCP server process
# - MCP clients connect via: docker exec -i thought-server node dist/index.js
# - Each client connection spawns a dedicated MCP server process
# - All environment variables are pre-configured inside the container
#
# Configure your MCP client (e.g., .kiro/settings/mcp.json):
# {
# "mcpServers": {
# "thought": {
# "command": "docker",
# "args": ["exec", "-i", "thought-server", "node", "dist/index.js"],
# "env": {},
# "disabled": false
# }
# }
# }
#
# To disable standby mode and run the MCP server directly (legacy behavior):
# Set MCP_STANDBY_MODE=false in .env.production
#
# ============================================================================
# REST API ACCESS
# ============================================================================
#
# By default, the REST API is enabled and accessible at:
# http://localhost:${REST_API_PORT:-3000}/api/v1/
#
# Configure CORS_ORIGINS in .env.production to allow your Mind Palace UI origin.
# To disable REST API: Set REST_API_ENABLED=false in .env.production
#
# ============================================================================
# MODEL AUTO-PULL
# ============================================================================
#
# The Ollama container automatically pulls required models on startup:
# - EMBEDDING_MODEL (default: nomic-embed-text) - Required for memory embeddings
# - LLM_MODEL (default: llama3.2:1b) - Required for AI-augmented reasoning
# Models are pulled before the container is marked healthy, ensuring dependent
# services wait until models are ready. Set SKIP_MODEL_PULL=true to disable.
services:
# ============================================================================
# INFRASTRUCTURE SERVICES
# ============================================================================
# PostgreSQL with pgvector extension for production
postgres:
image: pgvector/pgvector:pg16
container_name: thought-postgres
restart: unless-stopped
environment:
POSTGRES_USER: ${POSTGRES_USER:-thought}
POSTGRES_PASSWORD: ${POSTGRES_PASSWORD:-thought_secret}
POSTGRES_DB: ${POSTGRES_DB:-thought}
POSTGRES_INITDB_ARGS: "-E UTF8"
ports:
- "${POSTGRES_PORT:-5432}:5432"
volumes:
- postgres_data:/var/lib/postgresql/data
- ../packages/server/scripts/db/init.sql:/docker-entrypoint-initdb.d/01-init.sql:ro
- ../packages/server/scripts/db/enable-pgvector.sql:/docker-entrypoint-initdb.d/02-enable-pgvector.sql:ro
healthcheck:
test: ["CMD-SHELL", "pg_isready -U ${POSTGRES_USER:-thought} -d ${POSTGRES_DB:-thought}"]
interval: 10s
timeout: 5s
retries: 5
start_period: 10s
networks:
- thought-network
profiles:
- infra
- infra-ollama
- mcp-local
- mcp-remote
- rest-local
- rest-remote
- server-local
- server-remote
- all
- all-remote
# Redis for distributed caching and pub/sub
redis:
image: redis:alpine
container_name: thought-redis
restart: unless-stopped
command: redis-server --appendonly yes ${REDIS_PASSWORD:+--requirepass ${REDIS_PASSWORD}}
ports:
- "${REDIS_PORT:-6379}:6379"
volumes:
- redis_data:/data
healthcheck:
test: ["CMD", "redis-cli", "ping"]
interval: 10s
timeout: 5s
retries: 5
networks:
- thought-network
profiles:
- infra
- infra-ollama
- mcp-local
- mcp-remote
- rest-local
- rest-remote
- server-local
- server-remote
- all
- all-remote
# ============================================================================
# OLLAMA SERVICE (LOCAL)
# ============================================================================
# Ollama for embedding generation (local instance)
ollama:
image: ollama/ollama:latest
container_name: thought-ollama
restart: unless-stopped
entrypoint: ["/bin/sh", "/ollama-entrypoint.sh"]
environment:
EMBEDDING_MODEL: ${EMBEDDING_MODEL:-nomic-embed-text}
LLM_MODEL: ${LLM_MODEL:-llama3.2:1b}
OLLAMA_WAIT_TIMEOUT: ${OLLAMA_WAIT_TIMEOUT:-60}
SKIP_MODEL_PULL: ${SKIP_MODEL_PULL:-false}
ports:
- "${OLLAMA_PORT:-11434}:11434"
volumes:
- ollama_models:/root/.ollama
- ./ollama-entrypoint.sh:/ollama-entrypoint.sh:ro
healthcheck:
test: ["CMD-SHELL", "ollama list | grep -q '${EMBEDDING_MODEL:-nomic-embed-text}'"]
interval: 10s
timeout: 10s
retries: 30
start_period: 120s
networks:
- thought-network
profiles:
- infra-ollama
- mcp-local
- rest-local
- server-local
- all
# ============================================================================
# THOUGHT SERVER SERVICES
# ============================================================================
# MCP Server only (local Ollama) - stdio mode, no REST API
thought-mcp-local:
build:
context: ..
dockerfile: packages/server/Dockerfile
container_name: thought-server
restart: unless-stopped
depends_on:
postgres:
condition: service_healthy
redis:
condition: service_healthy
ollama:
condition: service_healthy
environment:
DATABASE_URL: postgresql://${POSTGRES_USER:-thought}:${POSTGRES_PASSWORD:-thought_secret}@postgres:5432/${POSTGRES_DB:-thought}
DB_HOST: postgres
DB_PORT: 5432
DB_NAME: ${POSTGRES_DB:-thought}
DB_USER: ${POSTGRES_USER:-thought}
DB_PASSWORD: ${POSTGRES_PASSWORD:-thought_secret}
DB_POOL_SIZE: ${DB_POOL_SIZE:-20}
OLLAMA_HOST: http://ollama:11434
EMBEDDING_MODEL: ${EMBEDDING_MODEL:-nomic-embed-text}
EMBEDDING_DIMENSION: ${EMBEDDING_DIMENSION:-768}
EMBEDDING_TIMEOUT: ${EMBEDDING_TIMEOUT:-60000}
LLM_MODEL: ${LLM_MODEL:-llama3.2:1b}
LLM_TIMEOUT: ${LLM_TIMEOUT:-60000}
LLM_TEMPERATURE: ${LLM_TEMPERATURE:-0.7}
REDIS_HOST: redis
REDIS_PORT: 6379
REDIS_PASSWORD: ${REDIS_PASSWORD:-}
REDIS_DB: ${REDIS_DB:-0}
REDIS_ENABLED: ${REDIS_ENABLED:-true}
NODE_ENV: production
LOG_LEVEL: ${LOG_LEVEL:-WARN}
LOG_FORMAT: ${LOG_FORMAT:-json}
CACHE_TTL: ${CACHE_TTL:-300}
MAX_PROCESSING_TIME: ${MAX_PROCESSING_TIME:-30000}
ENABLE_CACHE: ${ENABLE_CACHE:-true}
ENABLE_MONITORING: ${ENABLE_MONITORING:-true}
ENABLE_BIAS_DETECTION: ${ENABLE_BIAS_DETECTION:-true}
ENABLE_EMOTION_DETECTION: ${ENABLE_EMOTION_DETECTION:-true}
ENABLE_METACOGNITION: ${ENABLE_METACOGNITION:-true}
MCP_STANDBY_MODE: ${MCP_STANDBY_MODE:-true}
REST_API_ENABLED: "false"
stdin_open: true
tty: true
networks:
- thought-network
profiles:
- mcp-local
# MCP Server only (remote Ollama) - stdio mode, no REST API
thought-mcp-remote:
build:
context: ..
dockerfile: packages/server/Dockerfile
container_name: thought-server
restart: unless-stopped
depends_on:
postgres:
condition: service_healthy
redis:
condition: service_healthy
environment:
DATABASE_URL: postgresql://${POSTGRES_USER:-thought}:${POSTGRES_PASSWORD:-thought_secret}@postgres:5432/${POSTGRES_DB:-thought}
DB_HOST: postgres
DB_PORT: 5432
DB_NAME: ${POSTGRES_DB:-thought}
DB_USER: ${POSTGRES_USER:-thought}
DB_PASSWORD: ${POSTGRES_PASSWORD:-thought_secret}
DB_POOL_SIZE: ${DB_POOL_SIZE:-20}
OLLAMA_HOST: ${OLLAMA_HOST:-http://localhost:11434}
EMBEDDING_MODEL: ${EMBEDDING_MODEL:-nomic-embed-text}
EMBEDDING_DIMENSION: ${EMBEDDING_DIMENSION:-768}
EMBEDDING_TIMEOUT: ${EMBEDDING_TIMEOUT:-60000}
LLM_MODEL: ${LLM_MODEL:-llama3.2:1b}
LLM_TIMEOUT: ${LLM_TIMEOUT:-60000}
LLM_TEMPERATURE: ${LLM_TEMPERATURE:-0.7}
REDIS_HOST: redis
REDIS_PORT: 6379
REDIS_PASSWORD: ${REDIS_PASSWORD:-}
REDIS_DB: ${REDIS_DB:-0}
REDIS_ENABLED: ${REDIS_ENABLED:-true}
NODE_ENV: production
LOG_LEVEL: ${LOG_LEVEL:-WARN}
LOG_FORMAT: ${LOG_FORMAT:-json}
CACHE_TTL: ${CACHE_TTL:-300}
MAX_PROCESSING_TIME: ${MAX_PROCESSING_TIME:-30000}
ENABLE_CACHE: ${ENABLE_CACHE:-true}
ENABLE_MONITORING: ${ENABLE_MONITORING:-true}
ENABLE_BIAS_DETECTION: ${ENABLE_BIAS_DETECTION:-true}
ENABLE_EMOTION_DETECTION: ${ENABLE_EMOTION_DETECTION:-true}
ENABLE_METACOGNITION: ${ENABLE_METACOGNITION:-true}
MCP_STANDBY_MODE: ${MCP_STANDBY_MODE:-true}
REST_API_ENABLED: "false"
stdin_open: true
tty: true
networks:
- thought-network
profiles:
- mcp-remote
# REST API Server only (local Ollama) - HTTP API, no MCP
thought-rest-local:
build:
context: ..
dockerfile: packages/server/Dockerfile
container_name: thought-rest-server
restart: unless-stopped
depends_on:
postgres:
condition: service_healthy
redis:
condition: service_healthy
ollama:
condition: service_healthy
environment:
DATABASE_URL: postgresql://${POSTGRES_USER:-thought}:${POSTGRES_PASSWORD:-thought_secret}@postgres:5432/${POSTGRES_DB:-thought}
DB_HOST: postgres
DB_PORT: 5432
DB_NAME: ${POSTGRES_DB:-thought}
DB_USER: ${POSTGRES_USER:-thought}
DB_PASSWORD: ${POSTGRES_PASSWORD:-thought_secret}
DB_POOL_SIZE: ${DB_POOL_SIZE:-20}
OLLAMA_HOST: http://ollama:11434
EMBEDDING_MODEL: ${EMBEDDING_MODEL:-nomic-embed-text}
EMBEDDING_DIMENSION: ${EMBEDDING_DIMENSION:-768}
EMBEDDING_TIMEOUT: ${EMBEDDING_TIMEOUT:-60000}
LLM_MODEL: ${LLM_MODEL:-llama3.2:1b}
LLM_TIMEOUT: ${LLM_TIMEOUT:-60000}
LLM_TEMPERATURE: ${LLM_TEMPERATURE:-0.7}
REDIS_HOST: redis
REDIS_PORT: 6379
REDIS_PASSWORD: ${REDIS_PASSWORD:-}
REDIS_DB: ${REDIS_DB:-0}
REDIS_ENABLED: ${REDIS_ENABLED:-true}
NODE_ENV: production
LOG_LEVEL: ${LOG_LEVEL:-WARN}
LOG_FORMAT: ${LOG_FORMAT:-json}
CACHE_TTL: ${CACHE_TTL:-300}
MAX_PROCESSING_TIME: ${MAX_PROCESSING_TIME:-30000}
ENABLE_CACHE: ${ENABLE_CACHE:-true}
ENABLE_MONITORING: ${ENABLE_MONITORING:-true}
ENABLE_BIAS_DETECTION: ${ENABLE_BIAS_DETECTION:-true}
ENABLE_EMOTION_DETECTION: ${ENABLE_EMOTION_DETECTION:-true}
ENABLE_METACOGNITION: ${ENABLE_METACOGNITION:-true}
MCP_STANDBY_MODE: "false"
REST_API_ENABLED: "true"
REST_API_PORT: ${REST_API_PORT:-3000}
CORS_ORIGINS: ${CORS_ORIGINS:-http://localhost:5173}
ports:
- "${REST_API_PORT:-3000}:3000"
networks:
- thought-network
profiles:
- rest-local
# REST API Server only (remote Ollama) - HTTP API, no MCP
thought-rest-remote:
build:
context: ..
dockerfile: packages/server/Dockerfile
container_name: thought-rest-server
restart: unless-stopped
depends_on:
postgres:
condition: service_healthy
redis:
condition: service_healthy
environment:
DATABASE_URL: postgresql://${POSTGRES_USER:-thought}:${POSTGRES_PASSWORD:-thought_secret}@postgres:5432/${POSTGRES_DB:-thought}
DB_HOST: postgres
DB_PORT: 5432
DB_NAME: ${POSTGRES_DB:-thought}
DB_USER: ${POSTGRES_USER:-thought}
DB_PASSWORD: ${POSTGRES_PASSWORD:-thought_secret}
DB_POOL_SIZE: ${DB_POOL_SIZE:-20}
OLLAMA_HOST: ${OLLAMA_HOST:-http://localhost:11434}
EMBEDDING_MODEL: ${EMBEDDING_MODEL:-nomic-embed-text}
EMBEDDING_DIMENSION: ${EMBEDDING_DIMENSION:-768}
EMBEDDING_TIMEOUT: ${EMBEDDING_TIMEOUT:-60000}
LLM_MODEL: ${LLM_MODEL:-llama3.2:1b}
LLM_TIMEOUT: ${LLM_TIMEOUT:-60000}
LLM_TEMPERATURE: ${LLM_TEMPERATURE:-0.7}
REDIS_HOST: redis
REDIS_PORT: 6379
REDIS_PASSWORD: ${REDIS_PASSWORD:-}
REDIS_DB: ${REDIS_DB:-0}
REDIS_ENABLED: ${REDIS_ENABLED:-true}
NODE_ENV: production
LOG_LEVEL: ${LOG_LEVEL:-WARN}
LOG_FORMAT: ${LOG_FORMAT:-json}
CACHE_TTL: ${CACHE_TTL:-300}
MAX_PROCESSING_TIME: ${MAX_PROCESSING_TIME:-30000}
ENABLE_CACHE: ${ENABLE_CACHE:-true}
ENABLE_MONITORING: ${ENABLE_MONITORING:-true}
ENABLE_BIAS_DETECTION: ${ENABLE_BIAS_DETECTION:-true}
ENABLE_EMOTION_DETECTION: ${ENABLE_EMOTION_DETECTION:-true}
ENABLE_METACOGNITION: ${ENABLE_METACOGNITION:-true}
MCP_STANDBY_MODE: "false"
REST_API_ENABLED: "true"
REST_API_PORT: ${REST_API_PORT:-3000}
CORS_ORIGINS: ${CORS_ORIGINS:-http://localhost:5173}
ports:
- "${REST_API_PORT:-3000}:3000"
networks:
- thought-network
profiles:
- rest-remote
# Full Server (MCP + REST API) with local Ollama
thought-server-local:
build:
context: ..
dockerfile: packages/server/Dockerfile
container_name: thought-server
restart: unless-stopped
depends_on:
postgres:
condition: service_healthy
redis:
condition: service_healthy
ollama:
condition: service_healthy
environment:
DATABASE_URL: postgresql://${POSTGRES_USER:-thought}:${POSTGRES_PASSWORD:-thought_secret}@postgres:5432/${POSTGRES_DB:-thought}
DB_HOST: postgres
DB_PORT: 5432
DB_NAME: ${POSTGRES_DB:-thought}
DB_USER: ${POSTGRES_USER:-thought}
DB_PASSWORD: ${POSTGRES_PASSWORD:-thought_secret}
DB_POOL_SIZE: ${DB_POOL_SIZE:-20}
OLLAMA_HOST: http://ollama:11434
EMBEDDING_MODEL: ${EMBEDDING_MODEL:-nomic-embed-text}
EMBEDDING_DIMENSION: ${EMBEDDING_DIMENSION:-768}
EMBEDDING_TIMEOUT: ${EMBEDDING_TIMEOUT:-60000}
LLM_MODEL: ${LLM_MODEL:-llama3.2:1b}
LLM_TIMEOUT: ${LLM_TIMEOUT:-60000}
LLM_TEMPERATURE: ${LLM_TEMPERATURE:-0.7}
REDIS_HOST: redis
REDIS_PORT: 6379
REDIS_PASSWORD: ${REDIS_PASSWORD:-}
REDIS_DB: ${REDIS_DB:-0}
REDIS_ENABLED: ${REDIS_ENABLED:-true}
NODE_ENV: production
LOG_LEVEL: ${LOG_LEVEL:-WARN}
LOG_FORMAT: ${LOG_FORMAT:-json}
CACHE_TTL: ${CACHE_TTL:-300}
MAX_PROCESSING_TIME: ${MAX_PROCESSING_TIME:-30000}
ENABLE_CACHE: ${ENABLE_CACHE:-true}
ENABLE_MONITORING: ${ENABLE_MONITORING:-true}
ENABLE_BIAS_DETECTION: ${ENABLE_BIAS_DETECTION:-true}
ENABLE_EMOTION_DETECTION: ${ENABLE_EMOTION_DETECTION:-true}
ENABLE_METACOGNITION: ${ENABLE_METACOGNITION:-true}
MCP_STANDBY_MODE: ${MCP_STANDBY_MODE:-true}
REST_API_ENABLED: ${REST_API_ENABLED:-true}
REST_API_PORT: ${REST_API_PORT:-3000}
CORS_ORIGINS: ${CORS_ORIGINS:-http://localhost:5173}
stdin_open: true
tty: true
ports:
- "${REST_API_PORT:-3000}:3000"
networks:
- thought-network
profiles:
- server-local
- all
# Full Server (MCP + REST API) with remote Ollama
thought-server-remote:
build:
context: ..
dockerfile: packages/server/Dockerfile
container_name: thought-server
restart: unless-stopped
depends_on:
postgres:
condition: service_healthy
redis:
condition: service_healthy
environment:
DATABASE_URL: postgresql://${POSTGRES_USER:-thought}:${POSTGRES_PASSWORD:-thought_secret}@postgres:5432/${POSTGRES_DB:-thought}
DB_HOST: postgres
DB_PORT: 5432
DB_NAME: ${POSTGRES_DB:-thought}
DB_USER: ${POSTGRES_USER:-thought}
DB_PASSWORD: ${POSTGRES_PASSWORD:-thought_secret}
DB_POOL_SIZE: ${DB_POOL_SIZE:-20}
OLLAMA_HOST: ${OLLAMA_HOST:-http://localhost:11434}
EMBEDDING_MODEL: ${EMBEDDING_MODEL:-nomic-embed-text}
EMBEDDING_DIMENSION: ${EMBEDDING_DIMENSION:-768}
EMBEDDING_TIMEOUT: ${EMBEDDING_TIMEOUT:-60000}
LLM_MODEL: ${LLM_MODEL:-llama3.2:1b}
LLM_TIMEOUT: ${LLM_TIMEOUT:-60000}
LLM_TEMPERATURE: ${LLM_TEMPERATURE:-0.7}
REDIS_HOST: redis
REDIS_PORT: 6379
REDIS_PASSWORD: ${REDIS_PASSWORD:-}
REDIS_DB: ${REDIS_DB:-0}
REDIS_ENABLED: ${REDIS_ENABLED:-true}
NODE_ENV: production
LOG_LEVEL: ${LOG_LEVEL:-WARN}
LOG_FORMAT: ${LOG_FORMAT:-json}
CACHE_TTL: ${CACHE_TTL:-300}
MAX_PROCESSING_TIME: ${MAX_PROCESSING_TIME:-30000}
ENABLE_CACHE: ${ENABLE_CACHE:-true}
ENABLE_MONITORING: ${ENABLE_MONITORING:-true}
ENABLE_BIAS_DETECTION: ${ENABLE_BIAS_DETECTION:-true}
ENABLE_EMOTION_DETECTION: ${ENABLE_EMOTION_DETECTION:-true}
ENABLE_METACOGNITION: ${ENABLE_METACOGNITION:-true}
MCP_STANDBY_MODE: ${MCP_STANDBY_MODE:-true}
REST_API_ENABLED: ${REST_API_ENABLED:-true}
REST_API_PORT: ${REST_API_PORT:-3000}
CORS_ORIGINS: ${CORS_ORIGINS:-http://localhost:5173}
stdin_open: true
tty: true
ports:
- "${REST_API_PORT:-3000}:3000"
networks:
- thought-network
profiles:
- server-remote
- all-remote
# ============================================================================
# UI SERVICE
# ============================================================================
# Mind Palace UI
ui:
build:
context: ..
dockerfile: packages/ui/Dockerfile
args:
VITE_API_URL: ${VITE_API_URL:-http://localhost:3000}
container_name: thought-ui
restart: unless-stopped
environment:
NODE_ENV: production
ports:
- "${UI_PORT:-5173}:80"
networks:
- thought-network
profiles:
- ui
- all
- all-remote
volumes:
postgres_data:
driver: local
redis_data:
driver: local
ollama_models:
driver: local
networks:
thought-network:
driver: bridge
