noteflow/docker-bake.hcl

# docker-bake.hcl
# Docker Buildx Bake configuration for NoteFlow
#
# Usage:
#   docker buildx bake                    # Build default targets (CPU server)
#   docker buildx bake prod               # Build all production images
#   docker buildx bake dev                # Build all dev images
#   docker buildx bake server             # Build CPU server only
#   docker buildx bake server-gpu         # Build CUDA GPU server only
#   docker buildx bake server-rocm        # Build ROCm GPU server only
#   docker buildx bake --print            # Show build plan without building
#   docker buildx bake --push prod        # Build and push production images
#
# Registry configuration:
#   REGISTRY=git.baked.rocks/vasceannie docker buildx bake --push prod

# =============================================================================
# Variables
# =============================================================================

variable "REGISTRY" {
  default = "git.baked.rocks/vasceannie"
}

variable "IMAGE_NAME" {
  default = "noteflow"
}

variable "TAG" {
  default = "latest"
}

variable "PYTHON_VERSION" {
  default = "3.12"
}

variable "CUDA_VERSION" {
  default = "12.4.1"
}

variable "ROCM_VERSION" {
  default = "6.4.1"
}

variable "ROCM_PYTORCH_RELEASE" {
  default = "2.6.0"
}

variable "SPACY_MODEL_URL" {
  default = "https://github.com/explosion/spacy-models/releases/download/en_core_web_sm-3.8.0/en_core_web_sm-3.8.0-py3-none-any.whl"
}

# =============================================================================
# Functions - Single repo tagging scheme
# =============================================================================

# Base image path: REGISTRY/IMAGE_NAME
function "image_base" {
  params = []
  result = REGISTRY != "" ? "${REGISTRY}/${IMAGE_NAME}" : IMAGE_NAME
}

# CPU server tags: :TAG and :latest
function "tags_cpu" {
  params = []
  result = TAG == "latest" ? [
    "${image_base()}:latest"
  ] : [
    "${image_base()}:${TAG}",
    "${image_base()}:latest"
  ]
}

# GPU (CUDA) server tags: :TAG-gpu and :latest-gpu
function "tags_gpu" {
  params = []
  result = TAG == "latest" ? [
    "${image_base()}:latest-gpu"
  ] : [
    "${image_base()}:${TAG}-gpu",
    "${image_base()}:latest-gpu"
  ]
}

# ROCm server tags: :TAG-rocm and :latest-rocm
function "tags_rocm" {
  params = []
  result = TAG == "latest" ? [
    "${image_base()}:latest-rocm"
  ] : [
    "${image_base()}:${TAG}-rocm",
    "${image_base()}:latest-rocm"
  ]
}

# Dev tags (not published, local only)
function "tags_dev" {
  params = [suffix]
  result = ["${image_base()}:dev${suffix}"]
}

# =============================================================================
# Groups - Organized by use case
# =============================================================================

group "default" {
  targets = ["server"]
}

# Production images (publishable)
group "prod" {
  targets = ["server", "server-gpu", "server-rocm"]
}

# Development images (local only)
group "dev" {
  targets = ["server-dev", "server-gpu-dev", "server-rocm-dev"]
}

# All server variants
group "servers" {
  targets = ["server", "server-gpu", "server-rocm"]
}

# GPU variants only
group "servers-gpu" {
  targets = ["server-gpu", "server-rocm"]
}

# CI targets with GHA caching
group "ci" {
  targets = ["server-ci", "server-gpu-ci", "server-rocm-ci"]
}

# =============================================================================
# Base Targets (inherited)
# =============================================================================

target "_common" {
  context = "."
  labels = {
    "org.opencontainers.image.source" = "https://github.com/noteflow/noteflow"
    "org.opencontainers.image.vendor" = "NoteFlow"
  }
}

target "_server-common" {
  inherits = ["_common"]
  dockerfile = "docker/server.Dockerfile"
  args = {
    PYTHON_VERSION = PYTHON_VERSION
    SPACY_MODEL_URL = SPACY_MODEL_URL
  }
  cache-from = [
    "type=registry,ref=${image_base()}:cache"
  ]
  cache-to = [
    "type=inline"
  ]
}

target "_server-gpu-common" {
  inherits = ["_common"]
  dockerfile = "docker/server-gpu.Dockerfile"
  args = {
    PYTHON_VERSION = PYTHON_VERSION
    CUDA_VERSION = CUDA_VERSION
    SPACY_MODEL_URL = SPACY_MODEL_URL
  }
  cache-from = [
    "type=registry,ref=${image_base()}:cache-gpu"
  ]
  cache-to = [
    "type=inline"
  ]
}

target "_server-rocm-common" {
  inherits = ["_common"]
  dockerfile = "docker/Dockerfile.rocm"
  args = {
    ROCM_VERSION = ROCM_VERSION
    ROCM_PYTORCH_RELEASE = ROCM_PYTORCH_RELEASE
    SPACY_MODEL_URL = SPACY_MODEL_URL
  }
  cache-from = [
    "type=registry,ref=${image_base()}:cache-rocm"
  ]
  cache-to = [
    "type=inline"
  ]
}

# =============================================================================
# Production Server Targets
# =============================================================================

# CPU server (multi-arch: amd64 + arm64)
target "server" {
  inherits = ["_server-common"]
  target = "server"
  tags = tags_cpu()
  platforms = ["linux/amd64", "linux/arm64"]
  labels = {
    "org.opencontainers.image.title" = "NoteFlow Server"
    "org.opencontainers.image.description" = "NoteFlow gRPC server - CPU multi-arch build"
  }
}

# CUDA GPU server (amd64 only)
target "server-gpu" {
  inherits = ["_server-gpu-common"]
  target = "server"
  tags = tags_gpu()
  platforms = ["linux/amd64"]
  labels = {
    "org.opencontainers.image.title" = "NoteFlow Server (GPU)"
    "org.opencontainers.image.description" = "NoteFlow gRPC server - NVIDIA CUDA GPU build"
    "ai.noteflow.cuda.version" = CUDA_VERSION
  }
}

# ROCm GPU server (amd64 only)
target "server-rocm" {
  inherits = ["_server-rocm-common"]
  target = "server"
  tags = tags_rocm()
  platforms = ["linux/amd64"]
  labels = {
    "org.opencontainers.image.title" = "NoteFlow Server (ROCm)"
    "org.opencontainers.image.description" = "NoteFlow gRPC server - AMD ROCm GPU build"
    "ai.noteflow.rocm.version" = ROCM_VERSION
  }
}

# =============================================================================
# Development Server Targets (local only, not published)
# =============================================================================

# CPU dev server
target "server-dev" {
  inherits = ["_server-common"]
  target = "dev"
  tags = tags_dev("")
  labels = {
    "org.opencontainers.image.title" = "NoteFlow Server Dev"
    "org.opencontainers.image.description" = "NoteFlow development server with all extras"
  }
}

# CUDA GPU dev server
target "server-gpu-dev" {
  inherits = ["_server-gpu-common"]
  target = "dev"
  tags = tags_dev("-gpu")
  platforms = ["linux/amd64"]
  labels = {
    "org.opencontainers.image.title" = "NoteFlow Server Dev (GPU)"
    "org.opencontainers.image.description" = "NoteFlow development server - NVIDIA CUDA GPU build"
    "ai.noteflow.cuda.version" = CUDA_VERSION
  }
}

# ROCm GPU dev server
target "server-rocm-dev" {
  inherits = ["_server-rocm-common"]
  target = "dev"
  tags = tags_dev("-rocm")
  platforms = ["linux/amd64"]
  labels = {
    "org.opencontainers.image.title" = "NoteFlow Server Dev (ROCm)"
    "org.opencontainers.image.description" = "NoteFlow development server - AMD ROCm GPU build"
    "ai.noteflow.rocm.version" = ROCM_VERSION
  }
}

# =============================================================================
# CI/CD Targets (GHA caching)
# =============================================================================

target "server-ci" {
  inherits = ["server"]
  cache-from = ["type=gha"]
  cache-to = ["type=gha,mode=max"]
}

target "server-gpu-ci" {
  inherits = ["server-gpu"]
  cache-from = ["type=gha"]
  cache-to = ["type=gha,mode=max"]
}

target "server-rocm-ci" {
  inherits = ["server-rocm"]
  cache-from = ["type=gha"]
  cache-to = ["type=gha,mode=max"]
}