Dockerfile.hhtld

# Dockerfile.hhtld — HHTL-D TTS inference container
#
# Downloads pre-baked codebooks from GitHub release, runs inference.
# Multi-arch: builds on amd64 (AVX2/AVX-512) and arm64 (NEON/Pi 4).
#
# Build:
#   docker build -f Dockerfile.hhtld \
#     --build-arg RELEASE_TAG=v0.1.0 \
#     -t ghcr.io/adaworldapi/lance-graph-tts:v0.1.0 .
#
# Run:
#   docker run --rm ghcr.io/adaworldapi/lance-graph-tts:v0.1.0 \
#     --text "Hello, this is Ada speaking."
#
# Pi 4:
#   docker run --rm --platform linux/arm64 \
#     ghcr.io/adaworldapi/lance-graph-tts:v0.1.0 \
#     --text "Guten Morgen, Jan."
#
# RAM: ~75 MB (11 MB model + 64 MB KV cache at 512 tokens)

# ═══════════════════════════════════════════════════════════════════
# Stage 1: Build
# ═══════════════════════════════════════════════════════════════════
FROM debian:bookworm-slim AS builder

RUN apt-get update && apt-get install -y --no-install-recommends \
    curl ca-certificates gcc libc6-dev pkg-config libssl-dev \
    protobuf-compiler cmake \
    && rm -rf /var/lib/apt/lists/*

ENV RUSTUP_HOME=/usr/local/rustup \
    CARGO_HOME=/usr/local/cargo \
    PATH=/usr/local/cargo/bin:$PATH
RUN curl --proto '=https' --tlsv1.2 -sSf https://sh.rustup.rs | \
    sh -s -- -y --default-toolchain 1.94.0 --profile minimal

WORKDIR /build

# Copy workspace (lance-graph + ndarray must both be present)
COPY . /build/lance-graph
# ndarray expected at /build/ndarray (sibling to lance-graph)
# If building in CI, clone it here:
RUN if [ ! -d /build/ndarray ]; then \
      echo "WARNING: ndarray not found, attempting clone..." && \
      curl -sL "https://github.com/AdaWorldAPI/ndarray/archive/refs/heads/main.tar.gz" | \
      tar xz && mv ndarray-* /build/ndarray || true; \
    fi

WORKDIR /build/lance-graph

# Build the decode example (the inference binary)
RUN cargo build --release --example tts_17b_hhtld_decode \
    --manifest-path crates/thinking-engine/Cargo.toml 2>&1 || \
    echo "Build failed — ndarray dependency may be missing"

# Also build the cascade runner if available
RUN cargo build --release --example tts_cascade_runner \
    --manifest-path crates/thinking-engine/Cargo.toml 2>&1 || true

# ═══════════════════════════════════════════════════════════════════
# Stage 2: Runtime
# ═══════════════════════════════════════════════════════════════════
FROM debian:bookworm-slim AS runtime

RUN apt-get update && apt-get install -y --no-install-recommends \
    ca-certificates curl \
    && rm -rf /var/lib/apt/lists/*

ARG RELEASE_TAG=latest
ARG REPO=AdaWorldAPI/lance-graph
ENV RELEASE_TAG=${RELEASE_TAG}
ENV CODEBOOK_DIR=/opt/codebooks
ENV MODEL_NAME=qwen3-tts-1.7b

WORKDIR /opt

# Download codebooks from GitHub release
RUN mkdir -p ${CODEBOOK_DIR} && \
    echo "Downloading HHTL-D codebooks (${RELEASE_TAG})..." && \
    if [ "${RELEASE_TAG}" = "latest" ]; then \
      DOWNLOAD_URL="https://github.com/${REPO}/releases/latest/download/qwen3-tts-1.7b-hhtld.tar.gz"; \
    else \
      DOWNLOAD_URL="https://github.com/${REPO}/releases/download/${RELEASE_TAG}/qwen3-tts-1.7b-hhtld.tar.gz"; \
    fi && \
    curl -L --progress-bar "${DOWNLOAD_URL}" | tar xz -C ${CODEBOOK_DIR} && \
    echo "Codebooks downloaded:" && \
    du -sh ${CODEBOOK_DIR}/* 2>/dev/null || echo "Download may have failed — release not yet published"

# Copy built binaries from builder
COPY --from=builder /build/lance-graph/target/release/examples/tts_17b_hhtld_decode /usr/local/bin/hhtld-decode 2>/dev/null || true
COPY --from=builder /build/lance-graph/target/release/examples/tts_cascade_runner /usr/local/bin/hhtld-cascade 2>/dev/null || true

# Entrypoint script
COPY --chmod=755 <<'ENTRY' /usr/local/bin/entrypoint.sh
#!/bin/bash
set -euo pipefail

CODEBOOK_DIR="${CODEBOOK_DIR:-/opt/codebooks}"

echo "═══ HHTL-D TTS (${MODEL_NAME:-qwen3-tts-1.7b}) ═══"
echo ""

# Check codebooks
if [ -f "${CODEBOOK_DIR}/manifest.json" ]; then
    RATIO=$(python3 -c "import json; m=json.load(open('${CODEBOOK_DIR}/manifest.json')); print(f'{m[\"compression_ratio\"]:.0f}')" 2>/dev/null || echo "?")
    GROUPS=$(python3 -c "import json; m=json.load(open('${CODEBOOK_DIR}/manifest.json')); print(len(m['groups']))" 2>/dev/null || echo "?")
    echo "  Codebooks: ${GROUPS} palette groups, ${RATIO}:1 compression"
    echo "  Size:      $(du -sh ${CODEBOOK_DIR} | cut -f1)"
elif [ -f "${CODEBOOK_DIR}/model_hhtld.safetensors" ]; then
    echo "  Single-file mode: model_hhtld.safetensors"
    echo "  Size: $(du -h ${CODEBOOK_DIR}/model_hhtld.safetensors | cut -f1)"
else
    echo "  ERROR: No codebooks found at ${CODEBOOK_DIR}"
    echo "  Expected: manifest.json or model_hhtld.safetensors"
    echo ""
    echo "  To download manually:"
    echo "    curl -L https://github.com/AdaWorldAPI/lance-graph/releases/latest/download/qwen3-tts-1.7b-hhtld.tar.gz | tar xz -C ${CODEBOOK_DIR}"
    exit 1
fi

echo ""

# Route to appropriate binary
if [ "${1:-}" = "--decode" ] || [ "${1:-}" = "--validate" ]; then
    shift
    exec hhtld-decode "${CODEBOOK_DIR}/model_hhtld.safetensors" "$@"
elif [ "${1:-}" = "--cascade" ]; then
    shift
    exec hhtld-cascade "$@"
elif [ "${1:-}" = "--info" ]; then
    echo "Manifest:"
    cat "${CODEBOOK_DIR}/manifest.json" | python3 -m json.tool 2>/dev/null || cat "${CODEBOOK_DIR}/manifest.json"
elif [ "${1:-}" = "--shell" ]; then
    exec /bin/bash
else
    # Default: show info and run decode validation
    echo "Usage:"
    echo "  --decode     Run HHTL-D decoder/validator"
    echo "  --cascade    Run HHTL cascade inference"
    echo "  --info       Show manifest"
    echo "  --shell      Interactive shell"
    echo ""
    if command -v hhtld-decode &>/dev/null; then
        echo "Running decoder validation..."
        exec hhtld-decode "${CODEBOOK_DIR}/model_hhtld.safetensors"
    else
        echo "No binary found — build may have failed (ndarray dependency)."
        echo "Use --shell to inspect manually."
    fi
fi
ENTRY

ENTRYPOINT ["entrypoint.sh"]