# methodology/ml-overlay.yml
name: ml
description: >
  Machine learning overlay — injects ML domain knowledge into existing
  pipeline steps for model training, serving, experiment tracking, evaluation,
  observability, and MLOps patterns.
project-type: ml

# ---------------------------------------------------------------------------
# step-overrides
# ---------------------------------------------------------------------------
# ML training/serving routinely uses Docker for reproducibility (pinned
# CUDA/Python) and model-serving containers — keep staging-environments
# available even under the mvp preset. It self-skips at runtime when
# tech-stack.md declares no containerized services.
step-overrides:
  staging-environments: { enabled: true, conditional: "if-needed" }

# ---------------------------------------------------------------------------
# knowledge-overrides
# ---------------------------------------------------------------------------
# Map ML knowledge entries into existing pipeline steps so that machine
# learning domain expertise is injected during prompt assembly.
# Skips UX/design steps and database steps.
knowledge-overrides:
  # Foundational
  create-prd:
    append: [ml-requirements]
  user-stories:
    append: [ml-requirements]
  coding-standards:
    append: [ml-conventions]
  project-structure:
    append: [ml-project-structure]
  dev-env-setup:
    append: [ml-dev-environment]
  git-workflow:
    append: [ml-conventions]

  # Architecture & Design
  # training-patterns + serving-patterns → system-architecture
  system-architecture:
    append: [ml-architecture, ml-training-patterns, ml-serving-patterns]
  tech-stack:
    append: [ml-architecture]
  adrs:
    append: [ml-architecture]
  domain-modeling:
    append: [ml-architecture]
  security:
    append: [ml-security]
  # experiment-tracking → operations; observability → operations
  operations:
    append: [ml-experiment-tracking, ml-observability]

  # Testing
  # model-evaluation → create-evals
  tdd:
    append: [ml-testing]
  add-e2e-testing:
    append: [ml-testing]
  create-evals:
    append: [ml-testing, ml-model-evaluation]
  story-tests:
    append: [ml-testing]

  # Reviews (mirror authoring steps)
  review-architecture:
    append: [ml-architecture, ml-training-patterns, ml-serving-patterns]
  review-security:
    append: [ml-security]
  # observability → review-operations
  review-operations:
    append: [ml-experiment-tracking, ml-observability]
  review-testing:
    append: [ml-testing, ml-model-evaluation]

  # Planning
  implementation-plan:
    append: [ml-architecture]