docs/DEVELOPMENT.md

Development Guide

Project Structure

/
├── scripts/
│   ├── install/              # Installation scripts (*-install.sh)
│   │   ├── core-service-install.sh   # Your service dependencies
│   │   ├── filebrowser-install.sh    # File browser
│   │   └── persistence-install.sh    # Data persistence
│   └── start/                # Service startup scripts (*-start.sh)
│       ├── core-service-start.sh     # Your main service (IMPLEMENT THIS)
│       ├── filebrowser-start.sh      # File browser startup
│       └── persistence-start.sh      # Persistence service
├── configs/                  # Configuration template files (copied to /home/user/config/ at runtime)
├── Dockerfile               # Container build configuration
└── docker-entrypoint.sh    # Main container entry point

Core Service Implementation

The core service is implemented in scripts/start/core-service-start.sh and contains your main application logic.

Basic Implementation

#!/bin/bash
set -e

# Log function
log() {
    echo "[$(date '+%Y-%m-%d %H:%M:%S')] [CORE-SERVICE] $*"
}

# Activate Python virtual environment
source /opt/venv/bin/activate

# Configuration from environment variables
SERVICE_PORT="${SERVICE_PORT:-7860}"
MODEL_NAME="${MODEL_NAME:-your-default-model}"

# Set up directories and cache
mkdir -p /home/user/models /home/user/cache
export HF_HOME=/home/user/cache
export TRANSFORMERS_CACHE=/home/user/cache

# Start your service
log "Starting service on port $SERVICE_PORT..."
exec python /home/user/your_app.py

Common Frameworks

Gradio Interface:

# In core-service-install.sh
pip install gradio

# In core-service-start.sh
exec python -c "
import gradio as gr
def predict(text): return f'Echo: {text}'
gr.Interface(predict, 'text', 'text').launch(server_name='0.0.0.0', server_port=7860)
"

FastAPI Service:

# In core-service-install.sh
pip install fastapi uvicorn

# In core-service-start.sh
exec uvicorn your_app:app --host 0.0.0.0 --port 7860

Streamlit Dashboard:

# In core-service-install.sh
pip install streamlit

# In core-service-start.sh
exec streamlit run your_app.py --server.port 7860 --server.address 0.0.0.0

Dependencies Installation

Add your specific dependencies to scripts/install/core-service-install.sh:

# =============================================================================
# CUSTOM DEPENDENCIES SECTION
# =============================================================================

log "Installing service dependencies..."
source /opt/venv/bin/activate

# Example: AI/ML service
pip install --no-cache-dir \
    torch \
    transformers \
    gradio \
    numpy

# Example: System libraries
apt-get install -y --no-install-recommends \
    libssl-dev \
    ffmpeg

Common dependency patterns:

• AI/ML: torch transformers huggingface-hub
• Web APIs: fastapi uvicorn
• Dashboards: streamlit plotly
• Computer vision: opencv-python pillow

Environment Variables

# Core configuration
SERVICE_PORT="${SERVICE_PORT:-7860}"
MODEL_NAME="${MODEL_NAME:-microsoft/DialoGPT-medium}"
HF_TOKEN="${HF_TOKEN:-}"
DEBUG_MODE="${DEBUG_MODE:-false}"

# Custom variables
YOUR_API_KEY="${YOUR_API_KEY:-}"
YOUR_CONFIG_OPTION="${YOUR_CONFIG_OPTION:-default}"

Optional Services

Enable optional services via environment variables:

Service	Enable	Purpose
File Browser	FILEBROWSER_ENABLED=true	Web-based file management
Persistence	PERSISTENCE_ENABLED=true	Automatic backup/restore with HF Datasets
Cloudflare Tunnel	CLOUDFLARED_ENABLED=true	Secure external access
FRP Client	FRPC_ENABLED=true	Reverse proxy for complex networking

Required variables:

• Persistence: HF_TOKEN, DATASET_ID
• Cloudflare: CLOUDFLARED_TUNNEL_TOKEN
• FRP: FRPC_SERVER_ADDR, FRPC_AUTH_TOKEN

Development Workflow

Local Development

# Build and test
docker build -t my-hf-service .
docker run -p 7860:7860 \
  -e SERVICE_PORT=7860 \
  -e MODEL_NAME=your-model \
  -e DEBUG_MODE=true \
  my-hf-service

Testing

# Test core service
docker run -p 7860:7860 my-hf-service

# Test with optional services
docker run -p 7860:7860 \
  -e FILEBROWSER_ENABLED=true \
  -e PERSISTENCE_ENABLED=true \
  -e HF_TOKEN=your_token \
  -e DATASET_ID=username/dataset \
  my-hf-service

Deployment to HF Spaces

1. Push your changes to repository
2. Create a new Docker Space on Hugging Face
3. Set environment variables in Space settings
4. HF Spaces automatically builds and deploys

Adding Custom Services

1. Create Installation Script

Create scripts/install/your-service-install.sh:

#!/bin/bash
set -e

log() {
    echo "[$(date '+%Y-%m-%d %H:%M:%S')] [YOUR-SERVICE] $*"
}

log "Installing your service..."
# Installation steps here
log "Installation completed"

2. Create Startup Script

Create scripts/start/your-service-start.sh:

#!/bin/bash
set -e

log() {
    echo "[$(date '+%Y-%m-%d %H:%M:%S')] [YOUR-SERVICE] $*"
}

if [[ "${YOUR_SERVICE_ENABLED:-false}" == "true" ]]; then
    log "Starting your service..."
    exec your-service-binary
else
    log "Service disabled"
    exit 0
fi

The system automatically discovers and starts services based on the *_ENABLED environment variable pattern.

Troubleshooting

Common Issues

Problem	Solution
Service not starting	Check logs and verify environment variables are set
Port conflicts	Ensure SERVICE_PORT matches exposed port (7860)
Model loading failures	Verify model name and HF_TOKEN permissions
Permission denied	Check file ownership: chown user:user /home/user/script/start/*

Debugging Commands

# Check running processes
ps aux | grep -E "(python|your-service)"

# Check environment variables
env | grep -E "(SERVICE_|MODEL_|HF_)"

# Check logs
tail -f /tmp/*.log

# Test connectivity
curl -I http://localhost:7860

Best Practices

Code Organization

• Keep application code in /home/user/
• Use /home/user/models/ for model files
• Use /home/user/cache/ for temporary files
• Use /home/user/data/ for persistent data

Development

• Use environment variables for configuration
• Provide sensible defaults
• Implement proper error handling
• Use HF Spaces secrets for sensitive data
• Test locally before deploying
• Document custom environment variables