unified open-webui and tools setup script

setup.sh

#!/bin/bash

# Exit immediately if a command exits with a non-zero status.
set -e

# --- WSL Docker Credential Helper Fix ---
# In WSL, Docker can sometimes fail trying to find the Windows credential helper.
# This creates a config file to disable it, fixing the "docker-credential-desktop.exe not found" error.
echo "ℹ️ Applying WSL Docker credential helper fix..."
mkdir -p ~/.docker
echo '{"credsStore": ""}' > ~/.docker/config.json
echo "✅ Docker credential helper fix applied."


# --- Docker Compose Command Detection ---
if docker compose version &> /dev/null; then
    DOCKER_COMPOSE_CMD="docker compose"
    echo "✅ Found 'docker compose' (V2). Using it."
elif docker-compose version &> /dev/null; then
    DOCKER_COMPOSE_CMD="docker-compose"
    echo "✅ Found 'docker-compose' (V1). Using it."
else
    echo "❌ Neither 'docker compose' nor 'docker-compose' found. Please install Docker Compose."
    exit 1
fi

# --- Argument Parsing for Update Flag ---
UPDATE_FLAG=false
if [[ "$1" == "--update" || "$1" == "-u" ]]; then
  UPDATE_FLAG=true
  echo "✅ Update flag detected. The script will check for and pull the latest container images."
fi

# --- Generate Secure Random Credentials ---
echo "⚙️ Generating secure, random credentials..."
LITELLM_MASTER_KEY=$(openssl rand -hex 32)
DB_USER="user_$(openssl rand -hex 4)"
DB_PASS=$(openssl rand -hex 16)


# --- GPU Prerequisite Check (Informational) ---
if ! command -v nvidia-smi &> /dev/null
then
    echo "--------------------------------- WARNING ----------------------------------"
    echo "NVIDIA drivers ('nvidia-smi') not found. The GPU containers (LocalAI, vLLM)"
    echo "will fail to start without proper NVIDIA drivers and the NVIDIA Container Toolkit."
    echo "Press Ctrl+C to abort or Enter to continue at your own risk."
    echo "--------------------------------------------------------------------------"
    read -p ""
fi


# --- Interactive Port Configuration ---
echo "-----------------------------------------------------"
echo "Please specify the HOST ports for the services."
echo "Using a more robust method to set default values."
echo "-----------------------------------------------------"

read -e -p "Enter port for Open WebUI [8088]: " -i "8088" WEBUI_PORT
echo "-> Open WebUI Port set to: ${WEBUI_PORT}"

read -e -p "Enter port for LiteLLM Proxy [4000]: " -i "4000" LITELLM_PORT
echo "-> LiteLLM Port set to: ${LITELLM_PORT}"

read -e -p "Enter port for vLLM API [8000]: " -i "8000" VLLM_PORT
echo "-> vLLM Port set to: ${VLLM_PORT}"

read -e -p "Enter port for LocalAI API [8082]: " -i "8082" LOCALAI_PORT
echo "-> LocalAI Port set to: ${LOCALAI_PORT}"

read -e -p "Enter port for SearXNG [8081]: " -i "8081" SEARXNG_PORT
echo "-> SearXNG Port set to: ${SEARXNG_PORT}"

read -e -p "Enter port for Ollama API [11434]: " -i "11434" OLLAMA_PORT
echo "-> Ollama Port set to: ${OLLAMA_PORT}"

read -e -p "Enter port for Qdrant REST API [6333]: " -i "6333" QDRANT_REST_PORT
echo "-> Qdrant REST Port set to: ${QDRANT_REST_PORT}"

read -e -p "Enter port for Qdrant gRPC API [6334]: " -i "6334" QDRANT_GRPC_PORT
echo "-> Qdrant gRPC Port set to: ${QDRANT_GRPC_PORT}"

read -e -p "Enter port for Tika API (on host) [9998]: " -i "9998" TIKA_PORT
echo "-> Tika Port set to: ${TIKA_PORT}"


# --- Create .env File for Docker Compose ---
echo ""
echo "Creating .env file for service configuration..."
cat <<EOF > .env
# This file is auto-generated by the setup script.

# Service Ports
WEBUI_PORT=${WEBUI_PORT}
LITELLM_PORT=${LITELLM_PORT}
VLLM_PORT=${VLLM_PORT}
LOCALAI_PORT=${LOCALAI_PORT}
SEARXNG_PORT=${SEARXNG_PORT}
OLLAMA_PORT=${OLLAMA_PORT}
QDRANT_REST_PORT=${QDRANT_REST_PORT}
QDRANT_GRPC_PORT=${QDRANT_GRPC_PORT}
TIKA_PORT=${TIKA_PORT}

# Database Credentials
DB_USER=${DB_USER}
DB_PASS=${DB_PASS}
EOF


# --- Pre-configuration for SearXNG ---
echo "Creating SearXNG configuration directory..."
mkdir -p searxng
SECRET_KEY=$(openssl rand -hex 32)
echo "Generating SearXNG settings.yml with a new secret key..."
# Note: The base_url here is for externally facing links generated by SearXNG.
cat <<EOF > searxng/settings.yml
use_default_settings: true
server:
  secret_key: "${SECRET_KEY}"
  base_url: "http://localhost:${SEARXNG_PORT}"
redis:
  host: 'redis'
  port: 6379
  db: 0
EOF


# --- Pre-configuration for LiteLLM ---
echo "Generating LiteLLM configuration..."
# Note: api_base URLs use the Docker service name and INTERNAL port for communication.
cat <<EOF > litellm_config.yaml
# LiteLLM Configuration - Auto-generated by setup script
model_list:
  - model_name: llama3-from-ollama
    litellm_params:
      model: ollama/llama3
      api_base: http://ollama:11434
  - model_name: opt-125m-from-vllm
    litellm_params:
      model: openai/facebook/opt-125m
      api_base: http://vllm:8000
  - model_name: localai-embedding-model
    litellm_params:
      model: text-embedding-ada-002
      api_base: http://localai:8080
      api_key: "not-needed"

router_settings:
  allow_fallbacks: true

litellm_settings:
  set_verbose: false
  cache: true
  cache_params:
    type: redis
    host: "redis"
    port: 6379

general_settings:
  master_key: "${LITELLM_MASTER_KEY}"
EOF

# --- Install Custom Tika on Host ---
echo ""
echo "--- Installing Custom Apache Tika on Host ---"
echo "Downloading Tika installation script from Gist..."
curl -o tika-install.sh https://gist.githubusercontent.com/sycomix/51a1ca264abe77673636b7201401dbd6/raw/376ae996987c5301877907d65245298d56f815f8/tika-install.sh
chmod +x tika-install.sh
echo "WARNING: The Tika installer script is about to run."
echo "It may require sudo privileges to install dependencies and will compile from source."
read -p "Press Enter to continue..."
./tika-install.sh
echo "✅ Custom Tika installation process complete."


echo "Configuration complete. Creating docker-compose.yml..."

# --- Create the docker-compose.yml file ---
# Note: The top-level 'version' tag has been removed as it is obsolete in modern Docker Compose.
cat <<EOF > docker-compose.yml
services:
  # --- High-Performance Inference Engines ---
  vllm:
    image: vllm/vllm-openai:latest
    container_name: vllm
    ports:
      - "\${VLLM_PORT}:8000"
    volumes:
      - huggingface_cache:/root/.cache/huggingface
    command: ["--host", "0.0.0.0", "--model", "facebook/opt-125m"]
    deploy: &gpu_deploy
      resources:
        reservations:
          devices:
            - driver: nvidia
              count: all
              capabilities: [gpu]
    networks:
      - webui-network
    restart: unless-stopped

  # --- General Purpose / Multi-Model Engines ---
  ollama:
    image: ollama/ollama
    container_name: ollama
    ports:
      - "\${OLLAMA_PORT}:11434"
    volumes:
      - ollama_data:/root/.ollama
    deploy: *gpu_deploy
    networks:
      - webui-network
    restart: unless-stopped

  localai:
    image: localai/localai:latest-aio-gpu-nvidia-cuda-12
    container_name: localai
    ports:
      - "\${LOCALAI_PORT}:8080"
    volumes:
      - localai_data:/build/models
    deploy: *gpu_deploy
    networks:
      - webui-network
    restart: unless-stopped

  # --- Central Management and UI ---
  litellm:
    image: ghcr.io/berriai/litellm:main-stable
    container_name: litellm
    ports:
      - "\${LITELLM_PORT}:4000"
    environment:
      # Connect to postgres using the service name 'postgres' and internal port 5432
      - DATABASE_URL=postgresql://\${DB_USER}:\${DB_PASS}@postgres:5432/litellm_db
      # Force prisma to use the correct binary engine to avoid "wolfi" warnings
      - PRISMA_CLI_BINARY_TARGETS=debian-openssl-3.0.x
    volumes:
      - ./litellm_config.yaml:/app/config.yaml:ro
    command: ["--config", "/app/config.yaml", "--host", "0.0.0.0"]
    depends_on:
      postgres:
        condition: service_healthy
    networks:
      - webui-network
    restart: unless-stopped

  open-webui:
    image: ghcr.io/open-webui/open-webui:dev
    container_name: open-webui
    ports:
      - "\${WEBUI_PORT}:8080"
    environment:
      - VECTOR_DB=qdrant
      # Connect to qdrant using the service name 'qdrant' and internal port 6333
      - QDRANT_URI=http://qdrant:6333
      - ENABLE_WEB_SEARCH=true
      - WEB_SEARCH_ENGINE=searxng
      # Connect to searxng using the service name 'searxng' and internal port 8080
      - SEARXNG_URL=http://searxng:8080
      # Enable and configure Tika for document parsing on the HOST
      - TIKA_ENABLED=true
      - TIKA_URL=http://host.docker.internal:\${TIKA_PORT}
    volumes:
      - open_webui_data:/app/backend/data
    depends_on:
      - litellm
    networks:
      - webui-network
    restart: unless-stopped

  # --- Supporting Services ---
  postgres:
    image: postgres:16
    container_name: postgres
    environment:
      POSTGRES_USER: \${DB_USER}
      POSTGRES_PASSWORD: \${DB_PASS}
      POSTGRES_DB: litellm_db
    volumes:
      - postgres_data:/var/lib/postgresql/data
    networks:
      - webui-network
    healthcheck:
      test: ["CMD-SHELL", "pg_isready -U \${DB_USER} -d litellm_db"]
      interval: 10s
      timeout: 5s
      retries: 5
    restart: unless-stopped

  qdrant:
    image: qdrant/qdrant
    container_name: qdrant
    ports:
      - "\${QDRANT_REST_PORT}:6333"
      - "\${QDRANT_GRPC_PORT}:6334"
    volumes:
      - qdrant_data:/qdrant/storage
    networks:
      - webui-network
    restart: unless-stopped

  redis:
    image: redis:7
    container_name: redis
    volumes:
      - redis_data:/data
    networks:
      - webui-network
    restart: unless-stopped

  searxng:
    image: docker.io/searxng/searxng:latest
    container_name: searxng
    ports:
      # The host port is configurable, but the internal container port is 8080.
      - "\${SEARXNG_PORT}:8080"
    volumes:
      - ./searxng:/etc/searxng:rw
    depends_on:
      - redis
    networks:
      - webui-network
    restart: unless-stopped

volumes:
  huggingface_cache:
  ollama_data:
  localai_data:
  open_webui_data:
  qdrant_data:
  redis_data:
  postgres_data:

networks:
  webui-network:
    driver: bridge
EOF

# --- Verify Settings and Start Services ---
echo ""
echo "-----------------------------------------------------"
echo "✅ The following settings have been written to the .env file:"
echo "-----------------------------------------------------"
cat .env
echo "-----------------------------------------------------"
read -p "Press Enter to continue and start the containers..."

# --- Check for updates if flag is set ---
if [ "$UPDATE_FLAG" = true ]; then
  echo ""
  echo "🚀 Pulling the latest versions of all container images. This may take a moment..."
  ${DOCKER_COMPOSE_CMD} pull
  echo "✅ Image update check complete."
fi

echo ""
# Conditionally remove volumes based on the update flag
if [ "$UPDATE_FLAG" = true ]; then
  echo "Stopping any old containers before updating (preserving data volumes)..."
  ${DOCKER_COMPOSE_CMD} down
else
  echo "Stopping any old containers and removing data volumes for a clean start..."
  ${DOCKER_COMPOSE_CMD} down --volumes
fi

echo "Starting all services in detached mode..."
${DOCKER_COMPOSE_CMD} up -d

echo ""
echo "--------------------------------------------------------------------------"
echo "✅ Deployment Complete!"
echo ""
echo "Access Services at: (These URLs use the host ports you configured)"
echo "- Open WebUI:        http://localhost:${WEBUI_PORT}"
echo "- LiteLLM UI/API:    http://localhost:${LITELLM_PORT}"
echo "- vLLM API:          http://localhost:${VLLM_PORT}"
echo "- LocalAI API:       http://localhost:${LOCALAI_PORT}"
echo "- Ollama API:        http://localhost:${OLLAMA_PORT}"
echo "- Qdrant REST API:   http://localhost:${QDRANT_REST_PORT}"
echo "- Tika API (on host): http://localhost:${TIKA_PORT}"
echo "- SearXNG:           http://localhost:${SEARXNG_PORT}"
echo ""
echo "------------------------- IMPORTANT INFORMATION --------------------------"
echo "🔐 SAVE THESE RANDOMLY GENERATED SECRETS! 🔐"
echo "--------------------------------------------------------------------------"
echo "🔑 LiteLLM Master Key: ${LITELLM_MASTER_KEY}"
echo "👤 Database User:      ${DB_USER}"
echo "🔑 Database Password:  ${DB_PASS}"
echo "--------------------------------------------------------------------------"
echo ""
echo "--------------------------- IMPORTANT NEXT STEPS -------------------------"
echo "1. CHOOSE A MODEL FOR vLLM"
echo "   - Open the 'docker-compose.yml' file."
echo "   - In the 'vllm' service, change 'facebook/opt-125m' to your desired model."
echo "   - Apply changes with: ${DOCKER_COMPOSE_CMD} up -d --force-recreate"
echo ""
echo "2. CONFIGURE MODELS IN LITELLM"
echo "   - Open the 'litellm_config.yaml' file to add/edit models."
echo "   - Restart LiteLLM to apply changes: ${DOCKER_COMPOSE_CMD} restart litellm"