dockerfile

# Dockerfile for Advanced Iranian Legal Archive System
# Optimized for Hugging Face Spaces deployment

# Stage 1: Builder
FROM python:3.10-slim AS builder

# Install build dependencies and system packages
RUN apt-get update && apt-get install -y \
    build-essential \
    gcc \
    g++ \
    libffi-dev \
    libssl-dev \
    wget \
    curl \
    && rm -rf /var/lib/apt/lists/* \
    && apt-get clean

# Upgrade pip and install build tools
RUN pip install --no-cache-dir --upgrade pip setuptools wheel

# Create virtual environment
RUN python -m venv /opt/venv
ENV PATH="/opt/venv/bin:$PATH"

# Copy requirements and install Python dependencies
WORKDIR /build
COPY requirements.txt .
RUN pip install --no-cache-dir -r requirements.txt

# Create cache directory for models
RUN mkdir -p /app/cache/transformers /app/cache/sentence-transformers

# Pre-download Persian BERT models (Primary classification model)
RUN python -c "from transformers import AutoModel, AutoTokenizer; \
               print('Downloading ParsBERT...'); \
               AutoModel.from_pretrained('HooshvareLab/bert-base-parsbert-uncased', cache_dir='/app/cache/transformers'); \
               AutoTokenizer.from_pretrained('HooshvareLab/bert-base-parsbert-uncased', cache_dir='/app/cache/transformers')" || true

# Pre-download NER model for entity recognition
RUN python -c "from transformers import AutoModel, AutoTokenizer; \
               print('Downloading Persian NER model...'); \
               AutoModel.from_pretrained('HooshvareLab/bert-fa-base-uncased-ner', cache_dir='/app/cache/transformers'); \
               AutoTokenizer.from_pretrained('HooshvareLab/bert-fa-base-uncased-ner', cache_dir='/app/cache/transformers')" || true

# Pre-download Persian embedding model for semantic search
RUN python -c "from sentence_transformers import SentenceTransformer; \
               print('Downloading Persian embedding model...'); \
               model = SentenceTransformer('xmanii/maux-gte-persian'); \
               model.save('/app/cache/sentence-transformers/maux-gte-persian')" || true

# Pre-download multilingual sentence transformer as fallback
RUN python -c "from sentence_transformers import SentenceTransformer; \
               print('Downloading multilingual model...'); \
               model = SentenceTransformer('sentence-transformers/paraphrase-multilingual-MiniLM-L12-v2'); \
               model.save('/app/cache/sentence-transformers/paraphrase-multilingual')" || true

# Try to download FaBERT (latest SOTA Persian model)
RUN python -c "from transformers import AutoModel, AutoTokenizer; \
               print('Downloading FaBERT...'); \
               AutoModel.from_pretrained('sbunlp/fabert', cache_dir='/app/cache/transformers'); \
               AutoTokenizer.from_pretrained('sbunlp/fabert', cache_dir='/app/cache/transformers')" || echo "FaBERT download failed, will fallback to ParsBERT"

# Stage 2: Production
FROM python:3.10-slim

# Install runtime dependencies
RUN apt-get update && apt-get install -y \
    sqlite3 \
    libsqlite3-dev \
    curl \
    && rm -rf /var/lib/apt/lists/* \
    && apt-get clean

# Create non-root user for security
RUN groupadd -g 1000 appuser && useradd -r -u 1000 -g appuser appuser

# Copy virtual environment from builder stage
COPY --from=builder /opt/venv /opt/venv
ENV PATH="/opt/venv/bin:$PATH"

# Copy pre-downloaded models
COPY --from=builder /app/cache /app/cache

# Create application directories
RUN mkdir -p /app/data /app/logs /app/uploads /app/tmp && \
    chown -R appuser:appuser /app

# Set working directory
WORKDIR /app

# Copy application files
COPY --chown=appuser:appuser . .

# Environment variables for Iranian Legal Archive System
ENV PYTHONPATH=/app
ENV PYTHONUNBUFFERED=1
ENV HF_HOME=/app/cache
ENV TRANSFORMERS_CACHE=/app/cache/transformers
ENV TORCH_HOME=/app/cache/torch
ENV TOKENIZERS_PARALLELISM=false
ENV LOG_LEVEL=INFO
ENV ENVIRONMENT=production

# Gradio specific settings
ENV GRADIO_SERVER_NAME=0.0.0.0
ENV GRADIO_SERVER_PORT=7860
ENV GRADIO_SHARE=false

# Memory optimization for HF Spaces
ENV PYTORCH_CUDA_ALLOC_CONF=max_split_size_mb:128
ENV OMP_NUM_THREADS=2
ENV MKL_NUM_THREADS=2

# Switch to non-root user
USER appuser

# Create data directories with proper permissions
RUN mkdir -p data/cache_system.sqlite data/iranian_legal_archive_advanced.sqlite data/embeddings_cache.pkl data/faiss_index.bin

# Health check
HEALTHCHECK --interval=30s --timeout=30s --start-period=60s --retries=3 \
    CMD curl -f http://localhost:7860 || exit 1

# Expose Gradio port
EXPOSE 7860

# Command to run the application (matching README.md app_file)
CMD ["python", "persian_legal_scraper.py"]