feat: fix Dockerfile for hanzo-engine, add compose.yml and GHCR CI

- Dockerfile now builds hanzo-engine binary (not just mistralrs-server)
- Build context includes both engine/ and ml/ for workspace deps
- CPU-only build (--no-default-features) for linux/amd64 cloud
- compose.yml for local dev with HF cache mount
- build-image.yml workflow on self-hosted hanzo-deploy-linux-amd64
- Image: ghcr.io/hanzoai/engine

Author: hanzo-dev

.github/workflows/build-image.yml

@@ -0,0 +1,64 @@
+name: Build Docker Image
+
+on:
+  push:
+    branches: [main]
+    tags: ["v*"]
+  workflow_dispatch:
+
+concurrency:
+  group: docker-${{ github.ref }}
+  cancel-in-progress: false
+
+permissions:
+  contents: read
+  packages: write
+
+jobs:
+  build:
+    name: Build and Push
+    runs-on: hanzo-deploy-linux-amd64
+
+    steps:
+      - uses: actions/checkout@v4
+        with:
+          path: engine
+
+      - uses: actions/checkout@v4
+        with:
+          repository: hanzoai/ml
+          path: ml
+
+      - name: Set up Docker Buildx
+        uses: docker/setup-buildx-action@v3
+
+      - name: Log in to GHCR
+        uses: docker/login-action@v3
+        with:
+          registry: ghcr.io
+          username: ${{ github.actor }}
+          password: ${{ secrets.GITHUB_TOKEN }}
+
+      - name: Extract metadata
+        id: meta
+        uses: docker/metadata-action@v5
+        with:
+          images: ghcr.io/hanzoai/engine
+          tags: |
+            type=ref,event=branch
+            type=semver,pattern={{version}}
+            type=semver,pattern={{major}}.{{minor}}
+            type=sha,prefix=
+            type=raw,value=latest,enable={{is_default_branch}}
+
+      - name: Build and push
+        uses: docker/build-push-action@v6
+        with:
+          context: .
+          file: engine/Dockerfile
+          push: true
+          platforms: linux/amd64
+          tags: ${{ steps.meta.outputs.tags }}
+          labels: ${{ steps.meta.outputs.labels }}
+          cache-from: type=gha
+          cache-to: type=gha,mode=max

Dockerfile

@@ -3,12 +3,18 @@
 # Stage 1: Build environment
 FROM rust:latest AS builder
 
-# Set working directory and copy files
-WORKDIR /engine
-COPY . .
+# Set working directory
+WORKDIR /build
+
+# Copy ml dependency (must be present at ../ml relative to engine)
+COPY ml/ /build/ml/
+
+# Copy engine source
+COPY engine/ /build/engine/
 
 # Build the project in release mode, excluding the specified workspace
-RUN cargo build --release --workspace --exclude mistralrs-pyo3
+WORKDIR /build/engine
+RUN cargo build --release --workspace --exclude mistralrs-pyo3 --no-default-features
 
 
 # Stage 2: Minimal runtime environment
@@ -29,11 +35,19 @@ RUN <<HEREDOC
 HEREDOC
 
 # Copy the built binaries from the builder stage
-COPY --chmod=755 --from=builder /engine/target/release/mistralrs-bench /usr/local/bin/
-COPY --chmod=755 --from=builder /engine/target/release/mistralrs-server /usr/local/bin/
-COPY --chmod=755 --from=builder /engine/target/release/mistralrs-web-chat /usr/local/bin/
+COPY --chmod=755 --from=builder /build/engine/target/release/hanzo-engine /usr/local/bin/
+COPY --chmod=755 --from=builder /build/engine/target/release/mistralrs-server /usr/local/bin/
+COPY --chmod=755 --from=builder /build/engine/target/release/mistralrs-bench /usr/local/bin/
 # Copy chat templates for users running models which may not include them
-COPY --from=builder /engine/chat_templates /chat_templates
+COPY --from=builder /build/engine/chat_templates /chat_templates
 
 ENV HUGGINGFACE_HUB_CACHE=/data \
-    PORT=80
+    PORT=36900 \
+    RUST_LOG=info
+
+EXPOSE 36900
+
+HEALTHCHECK --interval=30s --timeout=10s --start-period=60s --retries=3 \
+    CMD curl -f http://localhost:36900/health || exit 1
+
+ENTRYPOINT ["hanzo-engine"]

compose.yml

@@ -0,0 +1,16 @@
+services:
+  engine:
+    image: ghcr.io/hanzoai/engine:latest
+    ports:
+      - "36900:36900"
+    volumes:
+      - ~/.cache/huggingface:/data
+    environment:
+      - RUST_LOG=info
+      - PORT=36900
+    healthcheck:
+      test: ["CMD", "curl", "-f", "http://localhost:36900/health"]
+      interval: 30s
+      timeout: 10s
+      start_period: 60s
+      retries: 3