Dockerfile

# Base image
FROM python:3.12-slim-bookworm AS base

# Set shared environment variables
ENV POETRY_VERSION=1.8.4 \
    POETRY_NO_INTERACTION=1 \
    POETRY_VIRTUALENVS_CREATE=true \
    POETRY_VIRTUALENVS_IN_PROJECT=true \
    POETRY_CACHE_DIR=/tmp/poetry_cache \
    PYTHONDONTWRITEBYTECODE=1 \
    LANG=en_US.UTF-8 \
    LANGUAGE=en_US:en \
    LC_ALL=en_US.UTF-8 \
    PORT=7860 \
    NODE_ENV=production

# Create user first (HF Spaces requirement)
RUN useradd -m -u 1000 user

# Install system dependencies and set up locales
RUN apt-get update && apt-get install -y \
    curl \
    git \
    gcc \
    python3-dev \
    libgmp-dev \
    libmpfr-dev \
    libmpc-dev \
    nodejs \
    npm \
    postgresql \
    postgresql-contrib \
    locales \
    nginx \
    netcat-openbsd \
    net-tools \
    procps \
    psmisc \
    && rm -rf /var/lib/apt/lists/* \
    && pip install --no-cache-dir "poetry==${POETRY_VERSION}" \
    && sed -i '/en_US.UTF-8/s/^# //g' /etc/locale.gen \
    && locale-gen

# Configure nginx
RUN rm /etc/nginx/sites-enabled/default || true
COPY <<EOF /etc/nginx/nginx.conf
user user;
pid /run/nginx.pid;
worker_processes auto;

events {
    worker_connections 1024;
}

http {
    include /etc/nginx/mime.types;
    default_type application/octet-stream;
    sendfile on;
    keepalive_timeout 120;
    client_body_temp_path /var/lib/nginx/body;
    proxy_temp_path /var/lib/nginx/proxy;
    fastcgi_temp_path /var/lib/nginx/fastcgi;
    uwsgi_temp_path /var/lib/nginx/uwsgi;
    scgi_temp_path /var/lib/nginx/scgi;
    
    # Added buffer size configurations
    proxy_buffer_size 128k;
    proxy_buffers 4 256k;
    proxy_busy_buffers_size 256k;
    
    # Added timeout configurations
    proxy_connect_timeout 120s;
    proxy_read_timeout 120s;
    proxy_send_timeout 120s;

    upstream frontend {
        server 127.0.0.1:3000;
    }

    upstream backend {
        server 127.0.0.1:5001;
    }

    server {
        listen 7860;
        server_name _;
        client_max_body_size 100M;
        
        # Increased buffer sizes
        proxy_buffer_size 128k;
        proxy_buffers 4 256k;
        proxy_busy_buffers_size 256k;
        
        location / {
            proxy_pass http://frontend;
            proxy_http_version 1.1;
            proxy_set_header Upgrade "$http_upgrade";
            proxy_set_header Connection "upgrade";
            proxy_set_header Host "$host";
            proxy_cache_bypass "$http_upgrade";
            proxy_set_header X-Real-IP "$remote_addr";
            proxy_set_header X-Forwarded-For "$proxy_add_x_forwarded_for";
            proxy_set_header X-Forwarded-Proto "$scheme";
            proxy_read_timeout 300s;
            proxy_connect_timeout 120s;
            proxy_send_timeout 120s;
        }

        location /api {
            proxy_pass http://backend;
            proxy_http_version 1.1;
            proxy_set_header Upgrade "$http_upgrade";
            proxy_set_header Connection "upgrade";
            proxy_set_header Host "$host";
            proxy_cache_bypass "$http_upgrade";
            proxy_set_header X-Real-IP "$remote_addr";
            proxy_set_header X-Forwarded-For "$proxy_add_x_forwarded_for";
            proxy_set_header X-Forwarded-Proto "$scheme";
            proxy_read_timeout 300s;
            proxy_connect_timeout 120s;
            proxy_send_timeout 120s;
        }

        access_log /var/log/nginx/access.log;
        error_log /var/log/nginx/error.log debug;
    }
}
EOF

# Set up directories and permissions
RUN mkdir -p /var/log/nginx /var/run/postgresql /var/lib/postgresql/data /var/log/postgresql /app/api /app/web /data/storage && \
    mkdir -p /var/lib/nginx/body /var/lib/nginx/proxy /var/lib/nginx/fastcgi /var/lib/nginx/uwsgi /var/lib/nginx/scgi && \
    chown -R root:root /etc/nginx && \
    chmod -R 755 /etc/nginx && \
    touch /run/nginx.pid && \
    chown -R root:root /run/nginx.pid && \
    chown -R user:user /var/lib/nginx && \
    chmod -R 755 /var/lib/nginx && \
    chmod -R 777 /run && \
    chmod 700 /var/lib/postgresql/data && \
    chmod 777 /data /app && \
    chown -R user:user /var/run/postgresql /var/lib/postgresql/data /var/log/postgresql /app /data /var/log/nginx

# Create and configure entrypoint script
COPY <<-'EOT' /app/entrypoint.sh
#!/bin/bash
set -e
echo "===== Application Startup at $(date "+%Y-%m-%d %H:%M:%S") ====="

# Function to check if a service is ready
check_service() {
    local service=$1
    local url=$2
    local max_attempts=$3
    local wait_time=$4
    
    echo "Checking $service..."
    for i in $(seq 1 $max_attempts); do
        if curl -s "$url" >/dev/null 2>&1; then
            echo "$service is ready"
            return 0
        fi
        echo "Waiting for $service (attempt $i/$max_attempts)..."
        sleep $wait_time
    done
    echo "$service failed to start"
    return 1
}

# Function to check if a port is in use
wait_for_port() {
    local port=$1
    local service=$2
    local max_attempts=$3
    local wait_time=$4
    local host=${5:-localhost}

    echo "Waiting for $service on $host:$port..."
    for i in $(seq 1 $max_attempts); do
        if nc -z $host $port; then
            echo "$service is listening on $host:$port"
            return 0
        fi
        echo "Waiting for $service (attempt $i/$max_attempts)..."
        sleep $wait_time
    done
    
    # If we get here, the service failed to start
    echo "=== $service Startup Failure ==="
    echo "Service failed to bind to $host:$port after $max_attempts attempts"
    echo "Checking process status:"
    ps aux | grep -i "$service" || true
    echo "Checking port status:"
    netstat -tulpn | grep ":$port" || true
    echo "Checking logs:"
    if [ "$service" = "Frontend" ]; then
        tail -n 50 /app/web/output.log || true
    fi
    return 1
}

# Function to start the frontend server
start_frontend() {
    cd /app/web
    echo "Cleaning up any existing Node.js processes..."
    # Kill any existing node processes more thoroughly
    pkill -9 -f "node server.js" || true
    # Wait a moment to ensure ports are released
    sleep 2
    # Check if port is still in use
    if netstat -tulpn | grep ":3000" > /dev/null; then
        echo "Port 3000 is still in use. Attempting to force cleanup..."
        # Find and kill any process using port 3000
        fuser -k 3000/tcp || true
        sleep 2
    fi
    
    # Ensure the output log directory exists
    touch output.log
    echo "Starting Next.js server at $(date)" > output.log
    
    # Configure Next.js to listen on all interfaces
    export HOSTNAME="0.0.0.0"
    export NODE_ENV=production
    export PORT=3000
    
    # Start the server with proper error handling
    node server.js >> output.log 2>&1 &
    local pid=$!
    echo "Frontend server started with PID: $pid"
    
    # Wait a bit to ensure process is stable and check its status
    sleep 5
    if ! kill -0 $pid 2>/dev/null; then
        echo "Frontend server failed to start. Last 50 lines of logs:"
        tail -n 50 output.log
        return 1
    fi
    
    # Verify the port is actually being used by our process
    if ! netstat -tulpn | grep ":3000.*$pid" > /dev/null; then
        echo "Frontend server is running but not bound to port 3000. Last 50 lines of logs:"
        tail -n 50 output.log
        return 1
    fi
    
    return 0
}

# Function to monitor and restart services if needed
monitor_services() {
    local restart_count=0
    local max_restarts=3
    
    while true; do
        if ! pgrep -f "gunicorn" > /dev/null; then
            echo "API server died"
            tail -n 100 /app/api/gunicorn.error.log || true
            exit 1
        fi
        
        if ! pgrep -f "node server.js" > /dev/null; then
            echo "Frontend server died. Attempting restart..."
            if [ $restart_count -lt $max_restarts ]; then
                restart_count=$((restart_count + 1))
                echo "Restart attempt $restart_count of $max_restarts"
                # Force cleanup before restart
                pkill -9 -f "node server.js" || true
                fuser -k 3000/tcp || true
                sleep 2
                if start_frontend; then
                    echo "Frontend server restarted successfully"
                    # Reset counter on successful restart
                    restart_count=0
                else
                    echo "Failed to restart frontend server"
                    tail -n 100 /app/web/output.log || true
                    exit 1
                fi
            else
                echo "Maximum frontend restart attempts reached"
                tail -n 100 /app/web/output.log || true
                exit 1
            fi
        fi
        
        if ! pgrep -f "nginx" > /dev/null; then
            echo "Nginx died"
            tail -n 100 /var/log/nginx/error.log || true
            exit 1
        fi
        sleep 5
    done
}

# Initialize PostgreSQL database if not already initialized
if [ ! -f "$PGDATA/PG_VERSION" ]; then
    echo "Initializing PostgreSQL database..."
    initdb --username=user --pwfile=<(echo "$DB_PASSWORD") --auth=md5 --encoding=UTF8
    
    # Configure PostgreSQL
    echo "local all all trust" > "$PGDATA/pg_hba.conf"
    echo "host all all 127.0.0.1/32 md5" >> "$PGDATA/pg_hba.conf"
    echo "host all all ::1/128 md5" >> "$PGDATA/pg_hba.conf"
    echo "host all all 0.0.0.0/0 md5" >> "$PGDATA/pg_hba.conf"
    
    echo "listen_addresses = '*'" >> "$PGDATA/postgresql.conf"
    echo "max_connections = 100" >> "$PGDATA/postgresql.conf"
    echo "shared_buffers = 128MB" >> "$PGDATA/postgresql.conf"
    echo "work_mem = 16MB" >> "$PGDATA/postgresql.conf"
    echo "maintenance_work_mem = 128MB" >> "$PGDATA/postgresql.conf"
    echo "effective_cache_size = 512MB" >> "$PGDATA/postgresql.conf"
fi

# Start PostgreSQL with detailed logging
echo "Starting PostgreSQL server..."
pg_ctl start -D "$PGDATA" -l /var/log/postgresql/postgresql.log -o "-c logging_collector=on -c log_directory='/var/log/postgresql' -c log_filename='postgresql-%Y-%m-%d_%H%M%S.log' -c log_statement='all'" -w

# Wait for PostgreSQL to start and show logs if there are issues
max_tries=30
count=0
echo "Checking database connection..."
until pg_isready -h localhost -p 5432; do
    if [ $count -eq 0 ]; then
        echo "PostgreSQL logs:"
        tail -n 50 /var/log/postgresql/postgresql.log
    fi
    echo "Waiting for database connection... (${count}/${max_tries})"
    sleep 2
    count=$((count+1))
    if [ $count -gt $max_tries ]; then
        echo "Failed to connect to database after ${max_tries} attempts"
        echo "Last 100 lines of PostgreSQL logs:"
        tail -n 100 /var/log/postgresql/postgresql.log
        exit 1
    fi
done

# Create database if it doesn't exist
if ! psql -lqt | cut -d \| -f 1 | grep -qw dify; then
    echo "Creating database dify..."
    createdb -U user dify
fi

echo "Database connection successful"

# Start application services
cd /app/api && poetry run python -m flask db upgrade

# Start API server
echo "Starting API server..."
cd /app/api && poetry run python -m gunicorn app:app \
    --bind ${DIFY_BIND_ADDRESS:-127.0.0.1}:${DIFY_PORT:-5001} \
    --worker-class gevent \
    --workers 1 \
    --timeout 300 \
    --preload \
    --access-logfile - \
    --error-logfile - &

# Wait for API to be ready
echo "Waiting for API server to be ready..."
wait_for_port 5001 "API" 30 2 "127.0.0.1"

# Start frontend server
echo "Starting frontend server..."
start_frontend

# Get container IP
CONTAINER_IP=$(hostname -i || ip route get 1 | awk '{print $7}' || echo "127.0.0.1")
echo "Container IP: $CONTAINER_IP"

# Wait for frontend to be ready
echo "Waiting for frontend server to be ready..."
if ! wait_for_port 3000 "Frontend" 60 5 "0.0.0.0"; then
    echo "Frontend server failed to start. Last 50 lines of frontend logs:"
    tail -n 50 /app/web/output.log
    exit 1
fi

# Start nginx with debug logging
echo "Starting nginx..."
nginx -g "daemon off; error_log /var/log/nginx/error.log debug;" &

# Wait for nginx to be ready
wait_for_port 7860 "Nginx" 30 2 "0.0.0.0"

echo "All services are running. Starting monitoring..."

# Start monitoring services
monitor_services
EOT

# Set permissions for entrypoint script
RUN chmod +x /app/entrypoint.sh && \
    chown user:user /app/entrypoint.sh

# Switch to user for remaining operations
USER user

# Set environment for user
ENV HOME=/home/user \
    PATH=/usr/lib/postgresql/15/bin:/home/user/.local/bin:$PATH \
    PGDATA=/var/lib/postgresql/data

# Pull official images
FROM langgenius/dify-web:latest AS web
FROM langgenius/dify-api:latest AS api

# Final stage
FROM base

# Set up directory structure
WORKDIR /app
RUN mkdir -p api web /data/storage

# Copy from official images
COPY --from=web --chown=user:user /app/web /app/web/
COPY --from=api --chown=user:user /app/api /app/api/

# Install API dependencies using Poetry
WORKDIR /app/api
COPY --from=api --chown=user /app/api/pyproject.toml /app/api/poetry.lock /app/api/poetry.toml ./
RUN poetry install --no-root --no-dev

# Create symlink for persistent storage
RUN ln -s /data/storage /app/api/storage

# Set environment variables
ENV FLASK_APP=app.py \
    EDITION=SELF_HOSTED \
    DEPLOY_ENV=PRODUCTION \
    MODE=api \
    LOG_LEVEL=INFO \
    DEBUG=false \
    FLASK_DEBUG=false \
    SECRET_KEY=sk-9f73s3ljTXVcMT3Blb3ljTqtsKiGHXVcMT3BlbkFJLK7U \
    CONSOLE_API_URL=https://${SPACE_ID}.hf.space \
    CONSOLE_WEB_URL=https://${SPACE_ID}.hf.space \
    SERVICE_API_URL=https://${SPACE_ID}.hf.space \
    APP_WEB_URL=https://${SPACE_ID}.hf.space \
    DIFY_PORT=5001 \
    DIFY_BIND_ADDRESS=127.0.0.1 \
    DB_USERNAME=user \
    DB_PASSWORD=difyai123456 \
    DB_HOST=localhost \
    DB_PORT=5432 \
    DB_DATABASE=dify \
    PYTHONPATH=/app/api \
    STORAGE_PATH=/data/storage

EXPOSE 7860

WORKDIR /app

CMD ["./entrypoint.sh"]