feat: simplified multi-platform Docker builds (#12)

* fix: simplify Docker builds for testing

- Disable minimal and whisper builds
- Remove TARGET platform variables
- Build only for linux/amd64 initially
- Use simplest build configuration first

* fix: remove suffix from Docker tags causing invalid format

- Remove matrix.suffix from all tag types
- Fixes 'invalid reference format' error in PR builds

* fix: remove problematic SHA tag format

- SHA tag with prefix was causing invalid tag format
- Keep simple tags only for now

* feat: add arm64 platform support back

- Re-enable arm64 in both docker-build.yml and release.yml workflows
- Add platform build arguments to Dockerfile (TARGETPLATFORM, TARGETOS, TARGETARCH)
- Use dynamic linking for better multi-platform compatibility
- Include opus runtime library in final image for arm64 support

* docs: clarify Docker buildx platform argument usage

- Update comments to clarify that TARGETPLATFORM, TARGETOS, and TARGETARCH
  are automatically set by Docker buildx based on the --platform flag
- Apply consistent comment style across all Dockerfiles
- No functional changes, just documentation improvements

* refactor: remove unnecessary TARGETPLATFORM/OS/ARCH args

- Docker buildx automatically handles cross-compilation via --platform flag
- Removed explicit ARG declarations as they're not needed
- Simplified Dockerfiles across the board
- Build still works correctly for both linux/amd64 and linux/arm64

* feat: enable minimal Docker build for amd64

- Switch from main Dockerfile to minimal build in workflows
- Minimal build creates ~12MB image using scratch base
- Static linking only works for amd64 (opus library limitation on arm64)
- Add -minimal suffix to Docker tags
- Update labels to indicate minimal build without ffmpeg

* feat: enable whisper Docker build for amd64

- Switch from minimal to whisper build in docker-build workflow
- Whisper build includes whisper.cpp binaries for local transcription
- Successfully builds for amd64 platform
- Includes ffmpeg and whisper model support

* feat: add arm64 support to whisper Docker build

- Switch from static to dynamic linking for multi-platform compatibility
- Include opus runtime library in final image
- Enables whisper build for both linux/amd64 and linux/arm64 platforms

* feat: enable all Docker builds with optimized caching

- Enable all three Dockerfiles in CI matrix
- Add main and whisper builds to release workflow
- Optimize whisper build caching:
  - Move ARG before dependencies for better layer caching
  - Split clone and build into separate layers
  - Add build cache mount for whisper compilation
  - Add opus runtime library for dynamic linking
- Main build: linux/amd64,linux/arm64 with ffmpeg
- Minimal build: linux/amd64 only (static linking)
- Whisper build: linux/amd64,linux/arm64 with whisper.cpp

* fix: remove cache mount from whisper build

The cache mount was causing issues with finding the built binaries.
Keep the layer separation for caching but remove the problematic mount.

* feat: set TRANSCRIBER_TYPE=whisper as default in whisper Docker image

The whisper image includes whisper.cpp binaries, so it makes sense to
default to using the whisper transcriber instead of the mock one.

* fix: remove audio configuration env vars from Dockerfiles

Audio processing defaults should be defined in code, not in Dockerfiles.
Keep only TRANSCRIBER_TYPE=whisper in whisper image as it's image-specific.

* refactor: simplify Docker strategy to two images

- Remove Dockerfile.minimal (complexity not worth 38MB savings)
- Remove ffmpeg from main image (not used in codebase)
- Simplify to two images: standard and whisper
- Both images now support linux/amd64 and linux/arm64
- Standard image ~15-20MB (was ~50MB with ffmpeg)
- Better user experience with simpler choice

* chore: remove debug ls command from Dockerfile.whisper

Remove the 'ls -la build/bin/' debugging command as suggested by
code review. The command was only used for debugging during development.

* fix: remove Dockerfile.minimal from docker-lint workflow

The minimal Dockerfile was removed but the CI workflow still tried
to lint it, causing the docker-lint job to fail.

* fix: resolve docker lint warnings in Dockerfile.whisper

- Quote nproc to prevent word splitting (SC2046)
- Consolidate multiple RUN commands to reduce layers (DL3059)
- DL3018 already properly ignored with hadolint directive

* fix: remove trailing spaces from hadolint ignore directive

The hadolint ignore comment had trailing spaces which prevented
it from being recognized properly.

Author: fank

.github/workflows/ci.yml

@@ -119,8 +119,9 @@ jobs:
           dockerfile: Dockerfile
           failure-threshold: warning
 
-      - name: Run Hadolint on Dockerfile.minimal
+      - name: Run Hadolint on Dockerfile.whisper
         uses: hadolint/[email protected]
         with:
-          dockerfile: Dockerfile.minimal
-          failure-threshold: warning
+          dockerfile: Dockerfile.whisper
+          failure-threshold: warning
+      

.github/workflows/docker-build.yml

@@ -25,16 +25,14 @@ jobs:
     strategy:
       matrix:
         include:
+          # Standard Dockerfile
           - dockerfile: Dockerfile
             suffix: ""
             platforms: linux/amd64,linux/arm64
-          - dockerfile: Dockerfile.minimal
-            suffix: "-minimal"
+          # Whisper Dockerfile with whisper.cpp for local transcription
+          - dockerfile: Dockerfile.whisper
+            suffix: "-whisper"
             platforms: linux/amd64,linux/arm64
-          # Whisper Dockerfile disabled - experimental and requires fixes
-          # - dockerfile: Dockerfile.whisper
-          #   suffix: "-whisper"
-          #   platforms: linux/amd64,linux/arm64
 
     steps:
       - name: Checkout repository
@@ -66,19 +64,13 @@ jobs:
             ${{ env.REGISTRY }}/${{ env.IMAGE_NAME }}
           tags: |
             # For branches
-            type=ref,event=branch,suffix=${{ matrix.suffix }}
-            # For PRs
-            type=ref,event=pr,suffix=${{ matrix.suffix }}
+            type=ref,event=branch
+            # For PRs  
+            type=ref,event=pr
             # For tags
-            type=ref,event=tag,suffix=${{ matrix.suffix }}
+            type=ref,event=tag
             # Latest tag for main branch
-            type=raw,value=latest${{ matrix.suffix }},enable={{is_default_branch}}
-            # Version tags
-            type=semver,pattern={{version}},suffix=${{ matrix.suffix }}
-            type=semver,pattern={{major}}.{{minor}},suffix=${{ matrix.suffix }}
-            type=semver,pattern={{major}},suffix=${{ matrix.suffix }}
-            # SHA
-            type=sha,prefix={{branch}}-,suffix=${{ matrix.suffix }}
+            type=raw,value=latest,enable={{is_default_branch}}
           flavor: |
             latest=false
       
@@ -134,13 +126,6 @@ jobs:
           
           docker manifest push ${{ env.REGISTRY }}/${{ env.IMAGE_NAME }}:${TAG}-multiarch
           
-          # Create manifest for minimal image
-          docker manifest create ${{ env.REGISTRY }}/${{ env.IMAGE_NAME }}:${TAG}-minimal-multiarch \
-            --amend ${{ env.REGISTRY }}/${{ env.IMAGE_NAME }}:${TAG}-minimal-linux-amd64 \
-            --amend ${{ env.REGISTRY }}/${{ env.IMAGE_NAME }}:${TAG}-minimal-linux-arm64
-          
-          docker manifest push ${{ env.REGISTRY }}/${{ env.IMAGE_NAME }}:${TAG}-minimal-multiarch
-          
           # Create manifest for whisper image
           docker manifest create ${{ env.REGISTRY }}/${{ env.IMAGE_NAME }}:${TAG}-whisper-multiarch \
             --amend ${{ env.REGISTRY }}/${{ env.IMAGE_NAME }}:${TAG}-whisper-linux-amd64 \

.github/workflows/release.yml

@@ -40,8 +40,8 @@ jobs:
           username: ${{ github.actor }}
           password: ${{ secrets.GITHUB_TOKEN }}
 
-      # Build and push normal image
-      - name: Build and push normal Docker image
+      # Build and push standard Docker image
+      - name: Build and push standard Docker image
         uses: docker/build-push-action@v6
         with:
           context: .
@@ -53,27 +53,27 @@ jobs:
             ${{ env.REGISTRY }}/${{ env.IMAGE_NAME }}:latest
           labels: |
             org.opencontainers.image.title=Discord Voice MCP
-            org.opencontainers.image.description=Discord voice channel transcription with MCP integration
+            org.opencontainers.image.description=Discord voice transcription MCP server
             org.opencontainers.image.version=${{ github.event.release.tag_name || github.event.inputs.tag }}
             org.opencontainers.image.source=${{ github.server_url }}/${{ github.repository }}
             org.opencontainers.image.revision=${{ github.sha }}
           cache-from: type=gha
           cache-to: type=gha,mode=max
 
-      # Build and push minimal image
-      - name: Build and push minimal Docker image
+      # Build and push whisper image with local transcription
+      - name: Build and push whisper Docker image
         uses: docker/build-push-action@v6
         with:
           context: .
-          file: Dockerfile.minimal
+          file: Dockerfile.whisper
           platforms: linux/amd64,linux/arm64
           push: true
           tags: |
-            ${{ env.REGISTRY }}/${{ env.IMAGE_NAME }}:${{ github.event.release.tag_name || github.event.inputs.tag }}-minimal
-            ${{ env.REGISTRY }}/${{ env.IMAGE_NAME }}:minimal
+            ${{ env.REGISTRY }}/${{ env.IMAGE_NAME }}:${{ github.event.release.tag_name || github.event.inputs.tag }}-whisper
+            ${{ env.REGISTRY }}/${{ env.IMAGE_NAME }}:whisper
           labels: |
-            org.opencontainers.image.title=Discord Voice MCP (Minimal)
-            org.opencontainers.image.description=Minimal Discord voice MCP without ffmpeg
+            org.opencontainers.image.title=Discord Voice MCP (Whisper)
+            org.opencontainers.image.description=Discord voice MCP with whisper.cpp for local transcription
             org.opencontainers.image.version=${{ github.event.release.tag_name || github.event.inputs.tag }}
             org.opencontainers.image.source=${{ github.server_url }}/${{ github.repository }}
             org.opencontainers.image.revision=${{ github.sha }}

Dockerfile

@@ -1,11 +1,6 @@
 # Build stage
 FROM golang:1.24-alpine3.21 AS builder
 
-# Build arguments for target platform
-ARG TARGETPLATFORM
-ARG TARGETOS
-ARG TARGETARCH
-
 # Install build dependencies
 # hadolint ignore=DL3018
 RUN apk add --no-cache git gcc musl-dev pkgconfig opus-dev
@@ -19,17 +14,18 @@ RUN go mod download
 # Copy source code
 COPY . .
 
-# Build static binary with CGO
-RUN CGO_ENABLED=1 GOOS=${TARGETOS} GOARCH=${TARGETARCH} \
-    go build -a -tags netgo -ldflags '-w -s -extldflags "-static"' \
+# Build binary with CGO
+# Docker buildx automatically handles cross-compilation via --platform flag
+# Using dynamic linking as static opus lib not available for all architectures
+RUN CGO_ENABLED=1 go build -ldflags '-w -s' \
     -o discord-voice-mcp ./cmd/discord-voice-mcp
 
-# Final stage - using alpine for ffmpeg support
+# Final stage
 FROM alpine:3.20
 
-# Install only ffmpeg (needed for audio processing)
+# Install opus runtime library (required for dynamic linking)
 # hadolint ignore=DL3018
-RUN apk add --no-cache ffmpeg
+RUN apk add --no-cache opus
 
 WORKDIR /app
 
@@ -42,14 +38,9 @@ USER mcp
 
 # Note: No ports exposed as this uses stdin/stdout for MCP protocol
 
-# Audio processing configuration (defaults)
-ENV AUDIO_BUFFER_DURATION_SEC=2 \
-    AUDIO_SILENCE_TIMEOUT_MS=1500 \
-    AUDIO_MIN_BUFFER_MS=100
-
 # Run the binary
 CMD ["./discord-voice-mcp"]
 
-# Expected image size: ~50MB (vs 2.35GB for Node.js version!)
+# Expected image size: ~15-20MB
 # Binary size: ~15MB
-# Alpine + ffmpeg: ~35MB
+# Alpine + opus: ~5MB

Dockerfile.minimal

@@ -1,11 +1,6 @@
 # Build stage for Go binary
 FROM golang:1.24-alpine3.21 AS go-builder
 
-# Build arguments for target platform
-ARG TARGETPLATFORM
-ARG TARGETOS
-ARG TARGETARCH
-
 # Install build dependencies
 # hadolint ignore=DL3018
 RUN apk add --no-cache git gcc musl-dev pkgconfig opus-dev
@@ -19,35 +14,37 @@ RUN go mod download
 # Copy source code
 COPY . .
 
-# Build static binary with CGO
-RUN CGO_ENABLED=1 GOOS=${TARGETOS} GOARCH=${TARGETARCH} \
-    go build -a -tags netgo -ldflags '-w -s -extldflags "-static"' \
+# Build binary with CGO
+# Using dynamic linking as static opus lib not available for all architectures
+RUN CGO_ENABLED=1 go build -ldflags '-w -s' \
     -o discord-voice-mcp ./cmd/discord-voice-mcp
 
 # Build stage for whisper.cpp  
 FROM alpine:3.20 AS whisper-builder
 
+# Pin to a specific version for reproducible builds
+ARG WHISPER_CPP_VERSION=v1.7.3
+
 # Install build dependencies for whisper.cpp (including cmake which is now required)
 # hadolint ignore=DL3018
 RUN apk add --no-cache git make g++ bash cmake
 
-# Clone and build whisper.cpp
+# Clone whisper.cpp (separate layer for better caching)
 WORKDIR /build
-# Pin to a specific version for reproducible builds
-ARG WHISPER_CPP_VERSION=v1.7.3
-RUN git clone --depth 1 --branch ${WHISPER_CPP_VERSION} https://github.com/ggerganov/whisper.cpp.git && \
-    cd whisper.cpp && \
-    cmake -B build -DGGML_CCACHE=OFF . && \
-    cmake --build build --config Release -- -j$(nproc) && \
-    ls -la build/bin/ && \
+RUN git clone --depth 1 --branch ${WHISPER_CPP_VERSION} https://github.com/ggerganov/whisper.cpp.git
+
+# Build whisper.cpp (this layer only rebuilds if source changes)
+WORKDIR /build/whisper.cpp
+RUN cmake -B build -DGGML_CCACHE=OFF . && \
+    cmake --build build --config Release -- -j"$(nproc)" && \
     chmod +x build/bin/*
 
 # Final stage
 FROM alpine:3.20
 
-# Install runtime dependencies
-# hadolint ignore=DL3018  
-RUN apk add --no-cache ffmpeg libstdc++ libgomp
+# Install runtime dependencies (including opus for dynamic linking)
+# hadolint ignore=DL3018
+RUN apk add --no-cache ffmpeg libstdc++ libgomp opus
 
 WORKDIR /app
 
@@ -60,20 +57,14 @@ COPY --from=whisper-builder /build/whisper.cpp/build/bin/server /usr/local/bin/w
 # Copy whisper shared libraries from correct directories
 COPY --from=whisper-builder /build/whisper.cpp/build/src/libwhisper.so* /usr/local/lib/
 COPY --from=whisper-builder /build/whisper.cpp/build/ggml/src/libggml*.so* /usr/local/lib/
-# Update library cache
-RUN ldconfig /usr/local/lib
-
-# Create directories for models
-RUN mkdir -p /models
-
-# Create non-root user
-RUN adduser -D -u 1000 mcp
+# Update library cache, create directories, and add user
+RUN ldconfig /usr/local/lib && \
+    mkdir -p /models && \
+    adduser -D -u 1000 mcp
 USER mcp
 
-# Audio processing configuration (defaults)
-ENV AUDIO_BUFFER_DURATION_SEC=2 \
-    AUDIO_SILENCE_TIMEOUT_MS=1500 \
-    AUDIO_MIN_BUFFER_MS=100
+# Set default transcriber type for whisper image
+ENV TRANSCRIBER_TYPE=whisper
 
 # Run the binary
 CMD ["./discord-voice-mcp"]