feat(hermes-workspace): add combined entrypoint script with gateway health check and graceful shutdown

ai/compose.yml

@@ -112,7 +112,22 @@ services:
       - /mnt/HoardingCow_docker_data/Ollama/ollama:/root/.ollama
     environment:
       - OLLAMA_VULKAN=0
+      - HSA_OVERRIDE_GFX_VERSION=9.0.6
+      - HCC_AMDGPU_TARGET=gfx906
+      - HIP_VISIBLE_DEVICES=0,1
+      - ROCR_VISIBLE_DEVICES=0,1
+      - HSA_ENABLE_SDMA=0 
       - OLLAMA_HOST=0.0.0.0
+      - OLLAMA_DEBUG=1
+      - OLLAMA_FLASH_ATTENTION=1
+      - OLLAMA_NUM_PARALLEL=2
+    devices:
+      # Map the render nodes and KFD for ROCm to work inside the container
+      - /dev/kfd:/dev/kfd
+      - /dev/dri:/dev/dri
+    group_add:
+      - "303"
+      - "26"
 
 networks:
   ai_net:
@@ -122,40 +137,47 @@ networks:
     driver: bridge
     name: ai_backend
     
-  llama-cpp-hermes:
+  # llama_cpp_devstral:
-    image: llama-cpp:rocm-gfx906
+  #   image: ghcr.io/ggml-org/llama.cpp:server-rocm
-    container_name: llama-cpp-hermes
+  #   container_name: llama_cpp_devstral
-    restart: unless-stopped
+  #   restart: unless-stopped
-    networks:
+  #   networks:
-      - ai_backend
+  #     - ai_backend
-    ports:
+  #   ports:
-      - "127.0.0.1:8300:8080"
+  #     - "8300:8080"
-    ipc: host
+  #   ipc: host
-    devices:
+  #   devices:
-      - /dev/kfd:/dev/kfd
+  #     - "/dev/kfd:/dev/kfd"
-      - /dev/dri:/dev/dri
+  #     - "/dev/dri:/dev/dri"
-    group_add:
+  #   group_add:
-      - "303"
+  #     - "303" # video
-      - "26"
+  #     - "26"  # render
-    environment:
+  #   environment:
-      - HSA_OVERRIDE_GFX_VERSION=9.0.6
+  #     HSA_OVERRIDE_GFX_VERSION: 9.0.6
-      - HSA_ENABLE_SDMA=0
+  #     HIP_VISIBLE_DEVICES: 0,1
-      - HIP_VISIBLE_DEVICES=0,1
+  #     LLAMA_CACHE: /models
-      - LLAMA_CACHE=/models
+  #   volumes:
-    volumes:
+  #     - /mnt/HoardingCow_docker_data/Llama_cpp/models:/models
-      - /mnt/HoardingCow_docker_data/Llama_cpp/models:/models
+  #     - /mnt/HoardingCow_docker_data/Llama_cpp/devstral-agent.jinja:/template.jinja
-      - /mnt/HoardingCow_docker_data/Ollama/ollama/models/blobs/sha256-17823599694fa3503ef54bf748d5078c6ce881f4d01616cafa255dc05d215a08:/model.gguf:ro
+  #   command: >
-    command: >
+  #     -hf unsloth/Devstral-Small-2-24B-Instruct-2512-GGUF:Devstral-Small-2-24B-Instruct-2512-Q8_0.gguf
-      -m /model.gguf
+  #     -a devstral-2-small-llama_cpp
-      --host 0.0.0.0
+  #     --chat-template-file /template.jinja
-      --port 8080
+  #     --host 0.0.0.0
-      --gpu-layers 99
+  #     --port 8080
-      --ctx-size 163840
+  #     --n-gpu-layers 99
-      -ctk f16 -ctv f16
+  #     --ctx-size 163840
-      --flash-attn on
+  #     --batch-size 4096
-      --split-mode layer
+  #     --ubatch-size 4096
-      --no-mmap
+  #     --cache-type-k f16
-      --n-predict -1
+  #     --cache-type-v f16
+  #     --cache-reuse 256
+  #     --flash-attn on
+  #     --context-shift
+  #     --split-mode layer
+  #     --no-mmap
+  #     --n-predict -1
+  #     --parallel 2
 
   # vllm:
   #   image: nalanzeyu/vllm-gfx906:v0.9.0-rocm6.3

ai/hermes-workspace/entrypoint-combined.sh

@@ -0,0 +1,154 @@
+#!/bin/bash
+# ── Hermes Workspace Combined Entrypoint ──
+# Waits for the Hermes gateway container (hermes:8642) to become healthy,
+# then starts the Hermes Workspace web UI in the foreground.
+# Supports graceful shutdown via SIGTERM/SIGINT.
+# ──────────────────────────────────────────
+
+set -euo pipefail
+
+# ── Configuration ──────────────────────────────────────────────
+GATEWAY_HOST="${GATEWAY_HOST:-hermes}"
+GATEWAY_PORT="${GATEWAY_PORT:-8642}"
+GATEWAY_URL="http://${GATEWAY_HOST}:${GATEWAY_PORT}"
+
+HEALTH_ENDPOINT="${HEALTH_ENDPOINT:-/health}"
+MAX_RETRIES="${HEALTH_MAX_RETRIES:-60}"
+RETRY_INTERVAL="${HEALTH_RETRY_INTERVAL:-2}"
+
+WORKSPACE_DIR="${WORKSPACE_DIR:-/workspace}"
+WORKSPACE_ENTRY="${WORKSPACE_ENTRY:-server-entry.js}"
+
+PID_FILE="${PID_FILE:-/tmp/workspace.pid}"
+
+# ── Logging ────────────────────────────────────────────────────
+log_info()  { echo "[$(date '+%Y-%m-%d %H:%M:%S')] [INFO]  $*"; }
+log_warn()  { echo "[$(date '+%Y-%m-%d %H:%M:%S')] [WARN]  $*"; }
+log_error() { echo "[$(date '+%Y-%m-%d %H:%M:%S')] [ERROR] $*"; }
+
+# ── Graceful Shutdown ──────────────────────────────────────────
+_workspace_pid=""
+_shutting_down=false
+
+cleanup() {
+    if [ "$_shutting_down" = true ]; then
+        return
+    fi
+    _shutting_down=true
+
+    log_info "Shutdown signal received, cleaning up..."
+
+    # Stop workspace process if running
+    if [ -n "$_workspace_pid" ] && kill -0 "$_workspace_pid" 2>/dev/null; then
+        log_info "Stopping workspace (PID: $_workspace_pid)..."
+        kill -TERM "$_workspace_pid" 2>/dev/null || true
+
+        # Give it time to shut down gracefully
+        local wait_sec=10
+        while kill -0 "$_workspace_pid" 2>/dev/null && [ "$wait_sec" -gt 0 ]; do
+            sleep 1
+            wait_sec=$((wait_sec - 1))
+        done
+
+        # Force kill if still running
+        if kill -0 "$_workspace_pid" 2>/dev/null; then
+            log_warn "Workspace did not shut down gracefully, force killing..."
+            kill -KILL "$_workspace_pid" 2>/dev/null || true
+        fi
+    fi
+
+    # Clean up PID file
+    [ -f "$PID_FILE" ] && rm -f "$PID_FILE"
+
+    log_info "Shutdown complete."
+    exit 0
+}
+
+# Trap termination signals for graceful shutdown
+trap cleanup SIGTERM SIGINT
+
+# ── Gateway Health Check ───────────────────────────────────────
+wait_for_gateway() {
+    local url="${GATEWAY_URL}${HEALTH_ENDPOINT}"
+    local retries="$MAX_RETRIES"
+    local interval="$RETRY_INTERVAL"
+    local attempt=0
+
+    log_info "Waiting for Hermes gateway at ${GATEWAY_URL}..."
+    log_info "Max retries: ${retries}, interval: ${interval}s"
+
+    while [ "$attempt" -lt "$retries" ]; do
+        attempt=$((attempt + 1))
+
+        if curl -fsS "${url}" >/dev/null 2>&1; then
+            log_info "Gateway is healthy after ${attempt} attempt(s) (${GATEWAY_URL})"
+            return 0
+        fi
+
+        if [ "$attempt" -lt "$retries" ]; then
+            log_info "Gateway not ready yet (attempt ${attempt}/${retries}), retrying in ${interval}s..."
+            sleep "$interval"
+        fi
+    done
+
+    log_error "Gateway did not become healthy after ${retries} attempts ($((retries * interval))s)"
+    return 1
+}
+
+# ── Workspace Startup ──────────────────────────────────────────
+start_workspace() {
+    local entry="${WORKSPACE_DIR}/${WORKSPACE_ENTRY}"
+
+    if [ ! -d "$WORKSPACE_DIR" ]; then
+        log_error "Workspace directory not found: ${WORKSPACE_DIR}"
+        return 1
+    fi
+
+    if [ ! -f "$entry" ]; then
+        log_error "Workspace entry point not found: ${entry}"
+        return 1
+    fi
+
+    log_info "Starting Hermes Workspace web UI..."
+    log_info "  Directory: ${WORKSPACE_DIR}"
+    log_info "  Entry:     ${entry}"
+
+    cd "$WORKSPACE_DIR"
+
+    # Start workspace in background so we can trap signals
+    node --max-old-space-size=2048 "${entry}" &
+    _workspace_pid=$!
+    echo "$_workspace_pid" > "$PID_FILE"
+
+    log_info "Workspace started (PID: ${_workspace_pid})"
+
+    # Wait for workspace process
+    wait "$_workspace_pid"
+    local exit_code=$?
+
+    log_info "Workspace exited with code ${exit_code}"
+    return "$exit_code"
+}
+
+# ── Main ───────────────────────────────────────────────────────
+main() {
+    log_info "=== Hermes Workspace Combined Entrypoint ==="
+    log_info "Gateway:   ${GATEWAY_URL}"
+    log_info "Workspace: ${WORKSPACE_DIR}/${WORKSPACE_ENTRY}"
+    log_info "PID file:  ${PID_FILE}"
+
+    # Wait for gateway to be healthy
+    if ! wait_for_gateway; then
+        log_warn "Proceeding without confirmed gateway health..."
+    fi
+
+    # Start the workspace
+    start_workspace
+    local exit_code=$?
+
+    log_info "Entrypoint exiting with code ${exit_code}"
+    return "$exit_code"
+}
+
+# Run main; exit with its return code
+main "$@"

ai/llama-cpp/Dockerfile

@@ -1,30 +0,0 @@
-# llama-cpp-rocm6/Dockerfile
-# Custom llama.cpp server with ROCm 6.1 + gfx906 (MI50) support.
-# Build: docker build -t llama-cpp:rocm-gfx906 .
-
-FROM rocm/dev-ubuntu-22.04:6.1.2-complete AS builder
-RUN apt-get update && DEBIAN_FRONTEND=noninteractive apt-get install -y curl git build-essential pkg-config cmake make && rm -rf /var/lib/apt/lists/*
-ARG LLAMACPP_VERSION=b9596
-RUN git clone --depth 1 --branch ${LLAMACPP_VERSION} https://github.com/ggml-org/llama.cpp.git /build
-WORKDIR /build
-ENV HIP_PATH=/opt/rocm ROCM_PATH=/opt/rocm PATH=/opt/rocm/bin:/opt/rocm/llvm/bin:${PATH} CMAKE_PREFIX_PATH=/opt/rocm
-RUN mkdir build && cd build && \
-    cmake .. -DGGML_HIP=ON -DCMAKE_BUILD_TYPE=Release \
-      -DAMDGPU_TARGETS="gfx906:xnack-" \
-      -DCMAKE_POSITION_INDEPENDENT_CODE=ON \
-      -DGGML_CUDA=OFF -DGGML_VULKAN=OFF -DGGML_METAL=OFF \
-      -DBUILD_SHARED_LIBS=OFF && \
-    cmake --build . --target llama-server -- -j $(nproc)
-
-FROM ubuntu:24.04
-RUN apt-get update && DEBIAN_FRONTEND=noninteractive apt-get install -y \
-    ca-certificates curl libstdc++6 libgomp1 libopenblas0 \
-    libnuma1 libelf1 libdrm2 libdrm-amdgpu1 \
-    && rm -rf /var/lib/apt/lists/*
-COPY --from=builder /opt/rocm/lib/ /opt/rocm/lib/
-COPY --from=builder /opt/rocm/share/ /opt/rocm/share/
-COPY --from=builder /build/build/bin/llama-server /usr/local/bin/llama-server
-RUN echo /opt/rocm/lib > /etc/ld.so.conf.d/rocm.conf && ldconfig
-ENV HSA_OVERRIDE_GFX_VERSION=9.0.6 HCC_AMDGPU_TARGET=gfx906 HSA_ENABLE_SDMA=0
-EXPOSE 8080
-ENTRYPOINT ["/usr/local/bin/llama-server"]