M.I.M.I.R - Multi-agent Intelligent Memory & Insight Repository

# NornicDB AMD64 CUDA # Uses pre-built llama-cuda-libs for fast builds (~2 min vs ~15 min) # # Prereq (one-time): docker build -f docker/Dockerfile.llama-cuda -t timothyswt/llama-cuda-libs:7285 . # # Build: docker build -f docker/Dockerfile.amd64-cuda -t nornicdb-amd64-cuda . # With BGE model embedded: # docker build -f docker/Dockerfile.amd64-cuda --build-arg EMBED_MODEL=true -t nornicdb-amd64-cuda-bge . # Headless (no UI): # docker build -f docker/Dockerfile.amd64-cuda --build-arg HEADLESS=true -t nornicdb-amd64-cuda-headless . ARG LLAMA_CUDA_IMAGE=timothyswt/llama-cuda-libs:7285 # ============================================================================= # Stage 1: UI (skipped in headless mode) # ============================================================================= FROM node:20-alpine AS ui ARG HEADLESS=false WORKDIR /ui COPY ui/package*.json ./ RUN if [ "$HEADLESS" != "true" ]; then \ npm ci 2>/dev/null || npm install --legacy-peer-deps; \ fi COPY ui/ . RUN if [ "$HEADLESS" != "true" ]; then \ npm run build; \ else \ mkdir -p dist && echo "Headless mode - UI skipped" > dist/HEADLESS; \ fi # ============================================================================= # Stage 2: Pre-built CUDA libs # ============================================================================= FROM ${LLAMA_CUDA_IMAGE} AS llama # ============================================================================= # Stage 3: Go build # ============================================================================= FROM nvidia/cuda:12.6.3-devel-ubuntu22.04 AS builder ARG HEADLESS=false ENV GO_VERSION=1.25.5 RUN apt-get update && apt-get install -y wget git gcc g++ libgomp1 && \ wget -q https://go.dev/dl/go${GO_VERSION}.linux-amd64.tar.gz && \ tar -C /usr/local -xzf go${GO_VERSION}.linux-amd64.tar.gz && rm go*.tar.gz ENV PATH="/usr/local/go/bin:${PATH}" CUDA_HOME=/usr/local/cuda WORKDIR /build # Copy llama artifacts COPY --from=llama /output/lib/*.a /build/lib/llama/ COPY --from=llama /output/include/*.h /build/lib/llama/ # Go dependencies COPY go.mod go.sum ./ RUN go mod download # Source + UI (only copy UI dist if not headless) COPY . . COPY --from=ui /ui/dist ./ui/dist # Build with CUDA + localllm (with or without UI based on HEADLESS arg) RUN COMMIT_HASH=$(git rev-parse --short HEAD 2>/dev/null || echo "unknown") && \ BUILD_TIME=$(date -u +%Y%m%d-%H%M%S) && \ if [ "$HEADLESS" = "true" ]; then \ echo "Building headless (no UI)..." && \ CGO_ENABLED=1 go build -tags "cuda localllm noui" \ -ldflags="-s -w -X main.buildTime=${BUILD_TIME} -X main.commit=${COMMIT_HASH}" \ -o nornicdb ./cmd/nornicdb; \ else \ echo "Building with UI..." && \ CGO_ENABLED=1 go build -tags "cuda localllm" \ -ldflags="-s -w -X main.buildTime=${BUILD_TIME} -X main.commit=${COMMIT_HASH}" \ -o nornicdb ./cmd/nornicdb; \ fi # Build APOC plugin RUN echo "Building APOC plugin..." && \ mkdir -p apoc/built-plugins && \ cd apoc/plugin-src/apoc && go build -buildmode=plugin -o ../../../apoc/built-plugins/apoc.so apoc_plugin.go && \ echo "✓ Built plugin:" && ls -lh /build/apoc/built-plugins/*.so # ============================================================================= # Stage 4: Runtime # ============================================================================= FROM nvidia/cuda:12.6.3-runtime-ubuntu22.04 WORKDIR /app RUN apt-get update && apt-get install -y --no-install-recommends \ ca-certificates tzdata wget libgomp1 && rm -rf /var/lib/apt/lists/* && \ mkdir -p /data /app/models COPY --from=builder /build/nornicdb /app/ COPY --from=builder /build/apoc/built-plugins /app/plugins/ COPY docker/entrypoint.sh /app/entrypoint-real.sh COPY docker/cuda-fallback-wrapper.sh /app/entrypoint.sh RUN chmod +x /app/entrypoint.sh /app/entrypoint-real.sh # Model embedding: only copy when EMBED_MODEL=true (uses BuildKit mount to avoid layer bloat) ARG EMBED_MODEL=false ARG HEADLESS=false RUN --mount=type=bind,source=models,target=/models,ro \ if [ "$EMBED_MODEL" = "true" ]; then \ if [ -f /models/bge-m3.gguf ]; then \ cp /models/bge-m3.gguf /app/models/ && \ echo "✓ Embedded bge-m3.gguf model ($(du -h /app/models/bge-m3.gguf | cut -f1))"; \ else \ echo "ERROR: EMBED_MODEL=true but models/bge-m3.gguf not found" && exit 1; \ fi; \ else \ echo "→ BYOM mode (no embedded model)"; \ fi && \ if [ "$HEADLESS" = "true" ]; then \ echo "✓ Headless mode enabled (no UI)"; \ fi EXPOSE 7474 7687 HEALTHCHECK --interval=30s --timeout=10s --start-period=10s --retries=3 \ CMD wget --spider -q http://localhost:7474/health || exit 1 # Set NORNICDB_HEADLESS based on build arg ENV NORNICDB_DATA_DIR=/data \ NORNICDB_HTTP_PORT=7474 \ NORNICDB_BOLT_PORT=7687 \ NORNICDB_EMBEDDING_PROVIDER=local \ NORNICDB_EMBEDDING_MODEL=bge-m3 \ NORNICDB_EMBEDDING_DIMENSIONS=1024 \ NORNICDB_MODELS_DIR=/app/models \ NORNICDB_EMBEDDING_GPU_LAYERS=-1 \ NORNICDB_NO_AUTH=true \ NORNICDB_GPU_ENABLED=true \ NORNICDB_HEADLESS=${HEADLESS} \ NORNICDB_PLUGINS_DIR=/app/plugins \ NVIDIA_VISIBLE_DEVICES=all \ NVIDIA_DRIVER_CAPABILITIES=compute,utility ENTRYPOINT ["/app/entrypoint.sh"]