Add Dockerfiles and Docker examples

.devops/cpu.Dockerfile

@@ -0,0 +1,96 @@
+# audio.cpp — CPU Dockerfile
+#
+# Usage:
+#   docker build -f .devops/cpu.Dockerfile -t local/audiocpp:full-cpu .
+
+# ============================================================
+# [BUILD] Compile all release binaries
+# ============================================================
+ARG UBUNTU_VERSION=24.04
+ARG BUILD_DATE=N/A
+ARG APP_VERSION=N/A
+ARG APP_REVISION=N/A
+ARG GCC_VERSION=14
+
+FROM docker.io/ubuntu:$UBUNTU_VERSION AS build
+
+ARG GCC_VERSION=14
+
+# Install build toolchain
+RUN apt-get update && \
+    apt-get install -y --no-install-recommends \
+        gcc-${GCC_VERSION} g++-${GCC_VERSION} make cmake libgomp1 && \
+    apt-get clean && \
+    rm -rf /var/lib/apt/lists/*
+
+ENV CC=gcc-${GCC_VERSION} CXX=g++-${GCC_VERSION}
+
+WORKDIR /app
+COPY . .
+
+# Configure
+RUN cmake -S . -B build \
+        -DCMAKE_BUILD_TYPE=Release \
+        -DENGINE_ENABLE_CUDA=OFF \
+        -DENGINE_ENABLE_VULKAN=OFF \
+        -DENGINE_ENABLE_OPENMP=ON \
+        -DENGINE_BUILD_EXAMPLES=OFF \
+        -DENGINE_BUILD_TESTS=OFF \
+        -DENGINE_BUILD_WARMBENCH=OFF && \
+    cmake --build build --parallel $(nproc) \
+        --target audiocpp_cli \
+        --target audiocpp_server \
+        --target model_perf \
+        --target miocodec_wavlm_parity
+
+# Collect all binaries + multiplexer into /app/full
+RUN mkdir -p /app/full && \
+    cp build/bin/audiocpp_cli build/bin/audiocpp_server \
+       build/bin/model_perf build/bin/miocodec_wavlm_parity /app/full/ && \
+    cp .devops/entrypoint.sh /app/full/entrypoint.sh && \
+    chmod +x /app/full/entrypoint.sh
+
+# ============================================================
+# [BASE] Shared runtime (OS + common libs)
+# ============================================================
+FROM docker.io/ubuntu:$UBUNTU_VERSION AS base
+
+ARG BUILD_DATE=N/A
+ARG APP_VERSION=N/A
+ARG APP_REVISION=N/A
+ARG IMAGE_URL=https://github.com/0xShug0/audio.cpp
+ARG IMAGE_SOURCE=https://github.com/0xShug0/audio.cpp
+
+LABEL org.opencontainers.image.created=$BUILD_DATE \
+      org.opencontainers.image.version=$APP_VERSION \
+      org.opencontainers.image.revision=$APP_REVISION \
+      org.opencontainers.image.title="audio.cpp" \
+      org.opencontainers.image.description="An all-in-one, pure C++ inference engine for audio models, powered by ggml" \
+      org.opencontainers.image.url=$IMAGE_URL \
+      org.opencontainers.image.source=$IMAGE_SOURCE
+
+# Runtime deps: OpenMP threading, curl (healthcheck), ffmpeg (audio I/O)
+RUN apt-get update && \
+    apt-get install -y --no-install-recommends \
+        libgomp1 curl ffmpeg && \
+    apt-get autoremove -y && \
+    apt-get clean -y && \
+    rm -rf /tmp/* /var/tmp/* && \
+    find /var/cache/apt/archives /var/lib/apt/lists -not -name lock -type f -delete && \
+    find /var/cache -type f -delete
+
+WORKDIR /app
+
+# ============================================================
+# [FULL] All binaries + entrypoint.sh multiplexer
+# ============================================================
+FROM base AS full
+
+COPY --from=build /app/full /app
+
+USER ubuntu
+
+HEALTHCHECK --interval=30s --timeout=10s --retries=3 --start-period=15s \
+  CMD curl -f http://localhost:8080/health || exit 1
+
+ENTRYPOINT ["/app/entrypoint.sh"]

.devops/cuda.Dockerfile

@@ -0,0 +1,102 @@
+# audio.cpp — CUDA Dockerfile
+#
+# Usage:
+#   docker build -f .devops/cuda.Dockerfile -t local/audiocpp:full-cuda .
+
+# ============================================================
+# [BUILD] Compile all release binaries with CUDA
+# ============================================================
+ARG UBUNTU_VERSION=24.04
+ARG CUDA_VERSION=12.9.0
+ARG BUILD_DATE=N/A
+ARG APP_VERSION=N/A
+ARG APP_REVISION=N/A
+ARG GCC_VERSION=14
+
+ARG BASE_CUDA_DEV_CONTAINER=docker.io/nvidia/cuda:${CUDA_VERSION}-devel-ubuntu${UBUNTU_VERSION}
+ARG BASE_CUDA_RUN_CONTAINER=docker.io/nvidia/cuda:${CUDA_VERSION}-runtime-ubuntu${UBUNTU_VERSION}
+
+FROM ${BASE_CUDA_DEV_CONTAINER} AS build
+
+ARG GCC_VERSION=14
+
+# Install build toolchain
+RUN apt-get update && \
+    apt-get install -y --no-install-recommends \
+        gcc-${GCC_VERSION} g++-${GCC_VERSION} cmake && \
+    apt-get clean && \
+    rm -rf /var/lib/apt/lists/*
+
+ENV CC=gcc-${GCC_VERSION} CXX=g++-${GCC_VERSION} CUDAHOSTCXX=g++-${GCC_VERSION}
+
+WORKDIR /app
+COPY . .
+
+# Configure
+RUN cmake -S . -B build \
+        -DCMAKE_BUILD_TYPE=Release \
+        -DENGINE_ENABLE_CUDA=ON \
+        -DENGINE_ENABLE_CUDA_GRAPHS=ON \
+        -DENGINE_ENABLE_VULKAN=OFF \
+        -DENGINE_ENABLE_OPENMP=ON \
+        -DENGINE_BUILD_EXAMPLES=OFF \
+        -DENGINE_BUILD_TESTS=OFF \
+        -DENGINE_BUILD_WARMBENCH=OFF && \
+    cmake --build build --parallel $(nproc) \
+        --target audiocpp_cli \
+        --target audiocpp_server \
+        --target model_perf \
+        --target miocodec_wavlm_parity
+
+# Collect all binaries + multiplexer into /app/full
+RUN mkdir -p /app/full && \
+    cp build/bin/audiocpp_cli build/bin/audiocpp_server \
+       build/bin/model_perf build/bin/miocodec_wavlm_parity /app/full/ && \
+    cp .devops/entrypoint.sh /app/full/entrypoint.sh && \
+    chmod +x /app/full/entrypoint.sh
+
+# ============================================================
+# [BASE] Shared runtime (NVIDIA CUDA + common libs)
+# ============================================================
+FROM ${BASE_CUDA_RUN_CONTAINER} AS base
+
+ARG BUILD_DATE=N/A
+ARG APP_VERSION=N/A
+ARG APP_REVISION=N/A
+ARG IMAGE_URL=https://github.com/0xShug0/audio.cpp
+ARG IMAGE_SOURCE=https://github.com/0xShug0/audio.cpp
+
+LABEL org.opencontainers.image.created=$BUILD_DATE \
+      org.opencontainers.image.version=$APP_VERSION \
+      org.opencontainers.image.revision=$APP_REVISION \
+      org.opencontainers.image.title="audio.cpp" \
+      org.opencontainers.image.description="An all-in-one, pure C++ inference engine for audio models, powered by ggml" \
+      org.opencontainers.image.url=$IMAGE_URL \
+      org.opencontainers.image.source=$IMAGE_SOURCE
+
+# Runtime deps: OpenMP threading, curl (healthcheck), ffmpeg (audio I/O)
+RUN apt-get update && \
+    apt-get install -y --no-install-recommends \
+        libgomp1 curl ffmpeg && \
+    apt-get autoremove -y && \
+    apt-get clean -y && \
+    rm -rf /tmp/* /var/tmp/* && \
+    find /var/cache/apt/archives /var/lib/apt/lists -not -name lock -type f -delete && \
+    find /var/cache -type f -delete
+
+WORKDIR /app
+
+# ============================================================
+# [FULL] All binaries + entrypoint.sh multiplexer
+# ============================================================
+FROM base AS full
+
+COPY --from=build /app/full /app
+
+USER ubuntu
+
+HEALTHCHECK --interval=30s --timeout=10s --retries=3 --start-period=15s \
+  CMD curl -f http://localhost:8080/health || exit 1
+
+ENTRYPOINT ["/app/entrypoint.sh"]
+

.devops/entrypoint.sh

@@ -0,0 +1,31 @@
+#!/usr/bin/env bash
+# ============================================================
+# audio.cpp Docker multiplexer
+#
+# Dispatches to the correct binary based on the first
+# argument. All remaining arguments are forwarded verbatim.
+# ============================================================
+set -e
+
+arg1="$1"
+shift || true
+
+if [[ "$arg1" == "cli" ]]; then
+    exec ./audiocpp_cli "$@"
+elif [[ "$arg1" == "server" ]]; then
+    exec ./audiocpp_server "$@"
+elif [[ "$arg1" == "perf" ]]; then
+    exec ./model_perf "$@"
+elif [[ "$arg1" == "parity" ]]; then
+    exec ./miocodec_wavlm_parity "$@"
+else
+    echo "Unknown command: $arg1"
+    echo ""
+    echo "Available commands:"
+    echo "  cli     Run audio tasks (TTS, ASR, VAD, VC, diar, etc.)"
+    echo "  server  Run the HTTP server"
+    echo "  perf    Run model performance benchmarks"
+    echo "  parity  Run Miocodec WavLM parity tests"
+    exit 1
+fi
+

.dockerignore

@@ -0,0 +1,9 @@
+.git/
+.gitignore
+.github/
+examples/
+scripts/
+docs/
+tools/
+build/
+*.md

examples/docker/cli/.gitignore

@@ -0,0 +1 @@
+output/

examples/docker/cli/README.md

@@ -0,0 +1,48 @@
+# Docker CLI
+
+Run the audio.cpp CLI tool inside a Docker container for text-to-speech.
+
+## Setup
+
+**1. Download the PocketTTS model**
+
+Get the English model from [kyutai/pocket-tts](https://huggingface.co/kyutai/pocket-tts/) on Hugging Face.
+Place the files from `languages/english/` into `../models/pocket-tts/languages/english/`:
+
+The directory should look like:
+```
+../models/pocket-tts/languages/english/
+├── model.safetensors
+├── tokenizer.model
+└── embeddings/
+    ├── alba.safetensors
+    └── ...
+```
+
+**2. Build the image**
+
+Run from repository root:
+
+CPU:
+
+```bash
+docker build -f .devops/cpu.Dockerfile -t local/audiocpp:full-cpu .
+```
+
+GPU (CUDA):
+
+```bash
+docker build -f .devops/cuda.Dockerfile -t local/audiocpp:full-cuda .
+```
+
+## Usage
+
+Run one of:
+
+```bash
+./cpu-tts.sh
+./cuda-tts.sh
+```
+
+Speech is saved to `output/speech.wav`.
+

examples/docker/cli/cpu-tts.sh

@@ -0,0 +1,22 @@
+#!/usr/bin/env bash
+set -e
+
+OUTPUT_DIR="$(cd "$(dirname "$0")" && pwd)/output"
+mkdir -p "$OUTPUT_DIR"
+MODEL_DIR="$(cd "$(dirname "$0")/../models" && pwd)"
+
+docker run --rm \
+  --user "$(id -u):$(id -g)" \
+  -v "$MODEL_DIR:/models:ro" \
+  -v "$OUTPUT_DIR:/output" \
+  local/audiocpp:full-cpu \
+  cli \
+  --task tts \
+  --family pocket_tts \
+  --model /models/pocket-tts \
+  --backend cpu \
+  --text "You are successfully running a text-to-speech model using audio.cpp, a pure C++ inference engine for audio models." \
+  --voice-id alba \
+  --out /output/speech.wav
+
+echo "Saved to: $OUTPUT_DIR/speech.wav"

examples/docker/cli/cuda-tts.sh

@@ -0,0 +1,24 @@
+#!/usr/bin/env bash
+set -e
+
+OUTPUT_DIR="$(cd "$(dirname "$0")" && pwd)/output"
+mkdir -p "$OUTPUT_DIR"
+MODEL_DIR="$(cd "$(dirname "$0")/../models" && pwd)"
+
+docker run --rm \
+  --gpus all \
+  --user "$(id -u):$(id -g)" \
+  -v "$MODEL_DIR:/models:ro" \
+  -v "$OUTPUT_DIR:/output" \
+  local/audiocpp:full-cuda \
+  cli \
+  --task tts \
+  --family pocket_tts \
+  --model /models/pocket-tts \
+  --backend cuda \
+  --device 0 \
+  --text "You are successfully running a text-to-speech model using audio.cpp, a pure C++ inference engine for audio models." \
+  --voice-id alba \
+  --out /output/speech.wav
+
+echo "Saved to: $OUTPUT_DIR/speech.wav"

examples/docker/models/.gitignore

@@ -0,0 +1,2 @@
+*
+!put_models_here

examples/docker/models/put_models_here

@@ -0,0 +1,4 @@
+# Download PocketTTS model files here:
+#   models/pocket-tts/languages/english/model.safetensors
+#   models/pocket-tts/languages/english/tokenizer.model
+#   models/pocket-tts/languages/english/embeddings/*.safetensors

examples/docker/server/.gitignore

@@ -0,0 +1 @@
+output/

examples/docker/server/README.md

@@ -0,0 +1,43 @@
+# Docker Compose Server
+
+Run the audio.cpp TTS server (as configured in server.json) in Docker.
+
+## Setup
+
+**1. Download the PocketTTS model**
+
+Get the English model from [kyutai/pocket-tts](https://huggingface.co/kyutai/pocket-tts/) on Hugging Face.
+Place the files from `languages/english/` into `../models/pocket-tts/languages/english/`:
+
+The directory should look like:
+```
+../models/pocket-tts/languages/english/
+├── model.safetensors
+├── tokenizer.model
+└── embeddings/
+    ├── alba.safetensors
+    └── ...
+```
+
+**2. Start the server**
+
+CPU:
+
+```bash
+docker compose -f cpu-server.yml up
+```
+
+GPU (CUDA):
+
+```bash
+docker compose -f cuda-server.yml up
+```
+
+## Generate speech
+
+```bash
+./tts.sh
+```
+
+This sends a request to `http://localhost:8080/v1/audio/speech` and saves the result to `output/speech.wav`.
+