`Refactor Dockerfile to support multi-arch/mode builds and Spaces compatibility`

Dockerfile

@@ -1,3 +1,4 @@
+
 # FROM python:3.10-slim
 
 # # Install system dependencies
@@ -127,53 +128,96 @@
 
 # CMD ["gunicorn", "--bind", "0.0.0.0:7860", "--workers", "1", "--threads", "8", "--timeout", "0", "ai_med_extract.app:app"]
 
-# Ultra-minimal Spaces-optimized Dockerfile
-FROM python:3.11-slim
-
-# Minimize layers and maximize cache hits
-RUN apt-get update \
-    && apt-get install -y --no-install-recommends \
-        tzdata \
-        tesseract-ocr \
-        poppler-utils \
-        ffmpeg \
-    && ln -fs /usr/share/zoneinfo/Etc/UTC /etc/localtime \
+
+# Multi-arch/mode Dockerfile that supports both GPU and CPU builds via build-args.
+# Defaults to CPU runtime suitable for Hugging Face Spaces Basic.
+
+ARG BASE_IMAGE=python:3.10-slim
+
+# Stage 1: builder (installs Python deps into a venv with build tooling)
+FROM ${BASE_IMAGE} AS builder
+
+ARG DEBIAN_FRONTEND=noninteractive
+ENV TZ=Etc/UTC
+
+# Install build tools only in builder
+RUN apt-get update && apt-get install -y --no-install-recommends \
+    tzdata \
+    build-essential \
+    python3 \
+    python3-pip \
+    python3-venv \
+    python3-dev \
+    tesseract-ocr \
+    poppler-utils \
+    ffmpeg \
+    && ln -fs /usr/share/zoneinfo/$TZ /etc/localtime \
     && dpkg-reconfigure -f noninteractive tzdata \
-    && rm -rf /var/lib/apt/lists/* \
-    && apt-get clean \
-    && rm -rf /tmp/*
+    && rm -rf /var/lib/apt/lists/*
+
+# Create virtual environment
+ENV VIRTUAL_ENV=/opt/venv
+RUN python3 -m venv "$VIRTUAL_ENV"
+ENV PATH="$VIRTUAL_ENV/bin:$PATH"
 
 WORKDIR /app
+COPY requirements.txt ./
 
-# Copy and install requirements in one layer for better caching
-COPY requirements.txt .
-RUN pip install --no-cache-dir --upgrade pip \
-    && pip install --no-cache-dir -r requirements.txt gunicorn
-
-# Copy only essential application files
-COPY ai_med_extract.py .
-COPY services/ai-service/src/ai_med_extract ./ai_med_extract/
-COPY services/ai-service/src/app.py .
-COPY services/ai-service/src/config_settings.py ./ai_med_extract/
-COPY services/ai-service/src/gradio_app.py ./ai_med_extract/
-COPY services/ai-service/src/wsgi.py .
-
-# Essential environment variables for Spaces compatibility
-ENV PYTHONUNBUFFERED=1 \
-    PYTHONDONTWRITEBYTECODE=1 \
-    PYTHONPATH=/app \
-    HF_HOME=/tmp \
-    XDG_CACHE_HOME=/tmp \
-    TORCH_HOME=/tmp \
-    WHISPER_CACHE=/tmp \
-    FAST_MODE=true \
-    PRELOAD_SMALL_MODELS=false
+# Upgrade pip and install dependencies
+RUN pip install --upgrade pip && \
+    pip install --prefer-binary -r requirements.txt
 
-# Create necessary directories with correct permissions
-RUN mkdir -p /tmp/uploads /tmp/huggingface /tmp/torch /tmp/whisper \
-    && chmod -R 777 /tmp
+# Stage 2: runtime (minimal runtime deps + venv from builder)
+FROM ${BASE_IMAGE} AS runtime
+
+ARG DEBIAN_FRONTEND=noninteractive
+ENV TZ=Etc/UTC
+
+# Install only runtime system packages; keep minimal
+RUN apt-get update && apt-get install -y --no-install-recommends \
+    tzdata \
+    tesseract-ocr \
+    poppler-utils \
+    ffmpeg \
+    && ln -fs /usr/share/zoneinfo/$TZ /etc/localtime \
+    && dpkg-reconfigure -f noninteractive tzdata \
+    && rm -rf /var/lib/apt/lists/*
+
+# Copy Python environment from builder
+COPY --from=builder /opt/venv /opt/venv
+ENV PATH="/opt/venv/bin:$PATH"
+
+# App
+WORKDIR /app
+COPY . .
+
+# Runtime cache dirs (kept in /tmp, auto-cleared on restart)
+ENV HF_HOME=/tmp/huggingface \
+    XDG_CACHE_HOME=/tmp \
+    TORCH_HOME=/tmp/torch \
+    WHISPER_CACHE=/tmp/whisper \
+    PYTHONUNBUFFERED=1 \
+    PYTHONPATH=/app \
+    GGUF_N_THREADS=2 \
+    GGUF_N_BATCH=64 \
+    OMP_NUM_THREADS=2 \
+    MKL_NUM_THREADS=2 \
+    NUMEXPR_NUM_THREADS=2
+
+# Ensure writable directories exist
+RUN mkdir -p /tmp/uploads /tmp/huggingface /tmp/torch /tmp/whisper && \
+    chmod -R 777 /tmp
+
+# Add entrypoint script that clears cache/models before app starts
+RUN echo '#!/bin/bash\n\
+echo "[ENTRYPOINT] Clearing Hugging Face / Torch / tmp cache..."\n\
+rm -rf /tmp/* ~/.cache/huggingface ~/.cache/torch || true\n\
+mkdir -p /tmp/uploads /tmp/huggingface /tmp/torch /tmp/whisper\n\
+chmod -R 777 /tmp\n\
+exec "$@"' > /entrypoint.sh && chmod +x /entrypoint.sh
+
+ENTRYPOINT ["/entrypoint.sh"]
 
 EXPOSE 7860
 
-# Use Gunicorn with WSGI app (Spaces standard) - single worker for stability
-CMD ["gunicorn", "--bind", "0.0.0.0:7860", "--workers", "1", "--threads", "2", "--timeout", "300", "--max-requests", "100", "--max-requests-jitter", "10", "wsgi:app"]
+CMD ["gunicorn", "--bind", "0.0.0.0:7860", "--workers", "1", "--threads", "2", "--timeout", "0", "ai_med_extract.app:app"]