Production Deployment

Docker is the recommended way to run BuEM in production. The repository ships a Dockerfile and docker-compose.yml that handle the conda environment, CBC solver binary, and Gunicorn entry-point.

Build and Run

# Build the image
docker build -t buem:latest .

# Start via Compose (recommended)
docker compose up -d

# Verify
curl http://localhost:5000/api/health

Environment Variables

Set these in docker-compose.yml or pass with -e:

Variable	Description
`BUEM_WEATHER_DIR`	Path to weather CSV directory inside the container
`BUEM_CBC_EXE`	Path to CBC binary (MILP solver)
`BUEM_RESULTS_DIR`	Directory where result JSON files are stored
`BUEM_LOG_FILE`	Log file path (default `logs/buem_api.log`)

Gunicorn Tuning

The entry-point script (entrypoint.sh) launches Gunicorn. Adjust workers and timeout as needed:

gunicorn -w 4 -b 0.0.0.0:5000 --timeout 120 buem.apis.api_server:app

A rule of thumb is 2 * CPU_CORES + 1 workers. Each worker can process one building request at a time.

Reverse Proxy

Place Nginx (or another proxy) in front of Gunicorn for TLS termination and static-file serving:

upstream buem {
    server 127.0.0.1:5000;
}

server {
    listen 80;
    location / {
        proxy_pass http://buem;
        proxy_set_header Host $host;
        proxy_set_header X-Real-IP $remote_addr;
        proxy_read_timeout 120s;
    }
    location /api/health {
        access_log off;
        proxy_pass http://buem;
    }
}

Health Checks

Use the /api/health endpoint for load-balancer or orchestrator probes:

# docker-compose snippet
healthcheck:
  test: ["CMD", "curl", "-f", "http://localhost:5000/api/health"]
  interval: 30s
  timeout: 10s
  retries: 3