Dockerfile.rocm

# Dockerfile.rocm
# ProofBridge Liner — AMD MI300X / ROCm 7 variant
#
# Target: DigitalOcean FX-AM7-gpu-mi300x1-192gb (Ubuntu 24.04, ROCm 7 pre-installed)
#
# Strategy: lightweight Node.js 20 image — ROCm libraries are mounted
# from the HOST at /opt/rocm (already installed by the DO marketplace image).
# This avoids pulling the 20+ GB rocm/dev base image and gives access to
# the exact ROCm 7 version the host drivers expect.
#
# Build:
#   docker build -f Dockerfile.rocm -t proofbridge-liner:rocm .
#
# Run (with MI300X):
#   docker run --rm \
#     --device /dev/kfd \
#     --device /dev/dri \
#     --group-add video \
#     --group-add render \
#     -v /opt/rocm:/opt/rocm:ro \
#     -e ROCM_PATH=/opt/rocm \
#     -e DASHBOARD_PORT=7860 \
#     -p 7860:7860 \
#     proofbridge-liner:rocm

FROM node:20-bookworm-slim

LABEL org.opencontainers.image.title="ProofBridge Liner (MI300X)"
LABEL org.opencontainers.image.description="Ghost-Risk Safety Kernel — AMD Instinct MI300X, 192 GB HBM3"
LABEL rocm.version="7"
LABEL amd.gpu="MI300X"

# ── System deps needed to call into mounted ROCm libs ─────────────
RUN apt-get update && apt-get install -y --no-install-recommends \
      libstdc++6 \
      libnuma1 \
      libelf1 \
      curl \
      ca-certificates \
    && apt-get clean && rm -rf /var/lib/apt/lists/*

WORKDIR /app

# ── Node deps (production only) ───────────────────────────────────
COPY package.json ./
RUN npm install --omit=dev

# ── App source ────────────────────────────────────────────────────
COPY . .

# ── Entrypoint: probe GPU, then launch dashboard ──────────────────
COPY <<'EOF' /app/entrypoint.sh
#!/bin/sh
set -e

echo "╔═══════════════════════════════════════╗"
echo "║  ProofBridge Liner — MI300X Kernel    ║"
echo "╚═══════════════════════════════════════╝"

# Report GPU if ROCm is mounted
if [ -x "${ROCM_PATH:-/opt/rocm}/bin/rocm-smi" ]; then
  "${ROCM_PATH}/bin/rocm-smi" --showproductname --showmeminfo vram 2>/dev/null \
    && echo "[rocm] AMD MI300X detected — 192 GB HBM3 available" \
    || echo "[rocm] rocm-smi ran but no GPU reported"
else
  echo "[rocm] /opt/rocm not mounted — running CPU-only mode"
fi

exec node dashboard/server.js
EOF
RUN chmod +x /app/entrypoint.sh

# ── Environment ───────────────────────────────────────────────────
ENV NODE_ENV=production
ENV DASHBOARD_PORT=7860
ENV DASHBOARD_HOST=0.0.0.0

# ROCm path (populated at runtime by -v /opt/rocm:/opt/rocm:ro)
ENV ROCM_PATH=/opt/rocm
ENV PATH="${ROCM_PATH}/bin:${PATH}"
ENV LD_LIBRARY_PATH="${ROCM_PATH}/lib:${ROCM_PATH}/lib64:${LD_LIBRARY_PATH:-}"

# MI300X-specific HIP tuning env vars
ENV HIP_VISIBLE_DEVICES=0
ENV GPU_MAX_HW_QUEUES=8
ENV HSA_ENABLE_SDMA=0

EXPOSE 7860

ENTRYPOINT ["/app/entrypoint.sh"]