Cinestar — Install

macOS

Download the latest Cinestar.app from the SourceForge Files section.
Move Cinestar.app to /Applications.
If macOS blocks the app (Gatekeeper), clear the quarantine attributes:
```
sudo xattr -c -r /Applications/Cinestar.app
```
If the app is still in Downloads:
```
sudo xattr -c -r ~/Downloads/Cinestar.app
```
Right-click the app and choose Open the first time to allow it through Gatekeeper.

Notes: The first run may take longer while models initialize. All processing runs locally.

Build from source (optional)

Install Node.js 18+ and npm.
Clone the source repository.

Inside the project root run:

npm install
npm run dev    # start Electron in dev mode
# or
npm run build  # produce a packaged app

AI Services Setup

Recommended Setup

Ollama local install is required for AI services. All processing runs locally on your machine.

Cinestar uses local AI services: Ollama (for BGE embeddings, Moondream vision, and Llama 3.2) for all AI processing.

Ollama Setup (Required)

Install Ollama:

📥 Download Ollama installer or use command line:

# macOS/Linux - Quick install
curl -fsSL https://ollama.com/install.sh | sh

# Windows - Download .exe from ollama.com/download

Pull the required AI models:

These models will be downloaded automatically on first pull (~2-4GB total)

# Embedding model for semantic search
ollama pull bge-large

# Vision model for image/video captioning  
ollama pull moondream:v2

# Language model for general tasks
ollama pull llama3.2:3b

Verify models are ready:

# List installed models
ollama list

# Test a model (optional)
ollama run llama3.2:3b "Hello, are you working?"

Launch Cinestar app - Ollama will automatically start as a system service when needed.

💡

Quick Links:
• BGE-Large model page
• Moondream model page
• Llama 3.2 model page

✅ Benefits: Better performance, easier model management, direct access to Ollama CLI, faster startup times, automatic model updates.

Health checks

# Check Ollama
ollama list

# Test Ollama API
curl http://localhost:11434/api/tags

Manage Ollama

# Ollama runs as a system service and typically doesn't need manual stopping
# To restart Ollama (if needed):
# macOS: Quit Ollama from menu bar
# Linux: systemctl restart ollama

Docker Compose Configuration

Complete docker-compose.yml file for running AI services:

# Drillbit AI Services
# Docker Compose setup for the AI services required by Drillbit desktop app
# The Drillbit app itself runs as a native desktop application, not in Docker

version: '3.8'

x-ollama: &ollama
  image: ollama/ollama:${OLLAMA_DOCKER_TAG:-latest}
  volumes:
    - ${OLLAMA_MODELS:-~/.ollama}:/root/.ollama
  environment: &env
    OLLAMA_KEEP_ALIVE: ${OLLAMA_KEEP_ALIVE:--1}
    OLLAMA_LOAD_TIMEOUT: ${OLLAMA_LOAD_TIMEOUT:-5m}
    OLLAMA_MAX_LOADED_MODELS: ${OLLAMA_MAX_LOADED_MODELS:-3}
    OLLAMA_NUM_PARALLEL: ${OLLAMA_NUM_PARALLEL:-5}
    OLLAMA_FLASH_ATTENTION: ${OLLAMA_FLASH_ATTENTION:-0}
    OLLAMA_DEBUG: ${OLLAMA_DEBUG:-0}
    OLLAMA_HOST: 0.0.0.0
  restart: unless-stopped
  deploy:
    resources:
      limits:
        memory: 6G
      reservations:
        memory: 4G

services:
  # Primary Ollama instance
  ollama-1:
    <<: *ollama
    container_name: drillbit-ollama-1
    environment:
      <<: *env
      CUDA_VISIBLE_DEVICES: ${CUDA_DEVICE_1:-0}
    healthcheck:
      test: ["CMD", "ollama", "list"]
      interval: 15s
      timeout: 10s
      retries: 5
      start_period: 30s

  # Secondary Ollama instance (optional, for load balancing)
  ollama-2:
    <<: *ollama
    container_name: drillbit-ollama-2
    environment:
      <<: *env
      CUDA_VISIBLE_DEVICES: ${CUDA_DEVICE_2:-1}
    healthcheck:
      test: ["CMD", "ollama", "list"]
      interval: 15s
      timeout: 10s
      retries: 5
      start_period: 30s
    profiles: ["multi-gpu"]  # Only start with --profile multi-gpu

  # Load balancer for multiple Ollama instances
  ollama-lb:
    container_name: drillbit-ollama-lb
    image: nginx:alpine
    volumes:
      - ./nginx.conf:/etc/nginx/nginx.conf:ro
    ports:
      - "${OLLAMA_LB_PORT:-9001}:11434"
    depends_on:
      ollama-1:
        condition: service_healthy
      ollama-2:
        condition: service_healthy
    restart: unless-stopped
    profiles: ["multi-gpu"]  # Only start with --profile multi-gpu

  # Main Ollama service (single instance)
  ollama:
    <<: *ollama
    container_name: drillbit-ollama
    ports:
      - "${OLLAMA_PORT:-11434}:11434"
    healthcheck:
      test: ["CMD", "ollama", "list"]
      interval: 15s
      timeout: 10s
      retries: 5
      start_period: 30s

  # Whisper transcription service
  whisper:
    image: onerahmet/openai-whisper-asr-webservice:latest
    container_name: drillbit-whisper
    ports:
      - "${WHISPER_PORT:-9000}:9000"
    environment:
      - ASR_MODEL=${WHISPER_MODEL:-base.en}
      - ASR_ENGINE=${WHISPER_ENGINE:-faster_whisper}
    restart: unless-stopped
    healthcheck:
      test: ["CMD", "curl", "-f", "http://localhost:9000/"]
      interval: 30s
      timeout: 10s
      retries: 3
    deploy:
      resources:
        limits:
          memory: 4G
        reservations:
          memory: 2G

volumes:
  ollama_models:
    driver: local

Nginx Load Balancer Configuration

Complete nginx.conf file for load balancing AI services:

events {
    worker_connections 2048;
}

http {
    # Specialized Ollama upstreams for different tasks
    
    # Embedding service (search, BGE-large, llama3.2:3b for text)
    upstream embed_backend {
        server ollama-embed:11434 max_fails=3 fail_timeout=30s;
    }
    
    # Vision/Captioning service (moondream:v2)
    upstream caption_backend {
        server ollama-caption:11434 max_fails=3 fail_timeout=30s;
    }

    # Whisper transcription service
    upstream whisper_backend {
        server whisper:9000 max_fails=3 fail_timeout=30s;
    }

    # Unified load balancer on port 9001
    server {
        listen 9001;
        listen [::]:9001;

        # Embedding service routes (search, BGE-large, tinyllama)
        location /embed/ {
            proxy_pass http://embed_backend/;
            proxy_set_header Host $host;
            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;

            # Optimized timeouts for embedding generation
            proxy_connect_timeout 30s;
            proxy_send_timeout 60s;
            proxy_read_timeout 60s;

            # Handle streaming responses
            proxy_buffering off;
            proxy_cache off;
        }
        
        # Vision/Captioning service routes (moondream:v2)
        location /caption/ {
            proxy_pass http://caption_backend/;
            proxy_set_header Host $host;
            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;

            # Extended timeouts for vision model processing
            proxy_connect_timeout 60s;
            proxy_send_timeout 300s;
            proxy_read_timeout 300s;

            # Handle streaming responses
            proxy_buffering off;
            proxy_cache off;
        }

        # Whisper ASR routes
        location /asr {
            proxy_pass http://whisper_backend;
            proxy_set_header Host $host;
            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;
            
            # Increase timeouts for transcription requests
            proxy_connect_timeout 300s;
            proxy_send_timeout 300s;
            proxy_read_timeout 300s;
            
            # Increase buffer sizes for large audio files
            client_max_body_size 100M;
            
            # Handle large file uploads
            proxy_buffering off;
            proxy_request_buffering off;
            
            # Keep connections alive
            proxy_http_version 1.1;
            proxy_set_header Connection "";
        }

        # Health check endpoint
        location /health {
            access_log off;
            return 200 "nginx load balancer healthy\n";
            add_header Content-Type text/plain;
        }
    }
}

Install Cinestar

macOS

Build from source (optional)

AI Services Setup

Docker Compose Configuration

Nginx Load Balancer Configuration