Architecture & Anti-Drift Safeguards

This document explains the architectural separation between the OWNd OpenWebNet protocol library and the MyHOME Home Assistant custom integration, along with the multi-tiered Anti-Drift Sentinel System designed to ensure both codebases never diverge.

---

1. Architectural Overview: The Split Model

Starting with MyHOME v2.0 and OWNd 2.0, the OpenWebNet protocol stack and the Home Assistant integration are cleanly decoupled into two dedicated repositories:

graph TD
    subgraph "Core Protocol Layer (OWNd)"
        A[OWNd Python Library] --> A1[OpenWebNet Frame Encoders / Decoders]
        A --> A2[Socket & Transport Handlers]
        A --> A3[Authentication Nonce / HMAC / SHA-256]
        A --> A4[Hardware Gateway Profiles]
        A --> A5[Event / Command Session Schedulers]
    end

    subgraph "Home Automation Layer (MyHOME)"
        B[MyHOME Integration] --> B1[Config Flow & Options UI]
        B --> B2[Platform Entities Light, Climate, Cover, Sensor, etc.]
        B --> B3[HA Device & Entity Registries]
        B --> B4[Diagnostics & In-Band Bus Monitor WebSocket]
    end

    A -->|Published via PyPI: OWNd==2.0.0b6| B

Why Decouple?

1. Single Source of Truth: Protocol decoding, dimension parsing, and frame syntax rules exist in one authoritative library rather than duplicated or vendored across multiple projects.
2. Reusability: Other automation frameworks, standalone CLI tools, diagnostic bridges, and testing scripts can leverage OWNd without pulling in Home Assistant dependencies.
3. Independent Release Cadence: Protocol fixes and newly decoded WHO dimensions can be tested and released on PyPI independently.

---

2. The Drift Problem

When a core protocol library and a downstream consumer live in separate repositories, three critical divergence risks arise:

1. Breaking Contract Changes: A parameter change, field renaming, or return type modification in OWNd passes all OWNd unit tests but breaks MyHOME entities or listeners.
2. Parser Regressions: A change to a regular expression or frame parsing logic in OWNd causes downstream entity state updates or device triggers to silently fail.
3. Dependency Desynchronization: MyHOME pins a specific release in manifest.json, but development branches assume newer unreleased features (or vice versa).

To permanently prevent these issues, the project implements a 3-Pillar Anti-Drift Architecture.

---

3. The 3-Pillar Anti-Drift Architecture

graph TD
    subgraph "Pillar 1: Shift-Left Downstream Canary (OWNd)"
        O1[OWNd PR or Commit] --> O2[Build Candidate OWNd Wheel]
        O2 --> O3[Checkout MyHOME integration]
        O3 --> O4[Run full MyHOME 1,600+ test suite]
        O4 -->|Any failure| O5[Block OWNd PR from Merging]
        O4 -->|All green| O6[Allow OWNd Merge]
    end

    subgraph "Pillar 2: Upstream Canary CI (MyHOME)"
        M1[Nightly Cron 04:00 UTC] --> M2[Install git+master of OWNd]
        M2 --> M3[Run MyHOME Test Suite & Enforcers]
        M3 -->|Alert on failure| M4[Proactive Warning Before PyPI Release]
    end

    subgraph "Pillar 3: Automated Release Bump"
        R1[OWNd PyPI Release] --> R2[repository_dispatch Webhook]
        R2 --> R3[Auto-Bump manifest.json & PR in MyHOME]
    end

Pillar 1: Downstream Integration Canary in OWNd (Shift-Left Sentinel)

The most effective safeguard is Shift-Left Testing: stopping breaking changes before they are ever merged into OWNd.

In OpenWebNet-HA/OWNd/.github/workflows/ci.yml, every PR and push to master triggers a downstream verification job:

• The runner builds and installs the candidate OWNd wheel.
• It clones the active development branch of OpenWebNet-HA/MyHOME (v2-phase1-architecture or master).
• It runs the complete automated unit test suite of MyHOME with strict 100.0% line coverage enforcement.

Important

A pull request to OWNd cannot merge if it breaks any behavior, parser, or assumption in MyHOME.

---

Pillar 2: Upstream Canary CI in MyHOME (Nightly Sentinel)

To detect upstream changes before they are tagged and released to PyPI, MyHOME runs a nightly scheduled workflow (.github/workflows/ownd-smoke.yml):

• Runs daily at 04:00 UTC and on manual workflow_dispatch.
• Installs the cutting-edge development head of OWNd:

  pip install git+https://github.com/OpenWebNet-HA/OWNd.git@master
  

• Runs the complete test suite. If an unreleased commit in OWNd triggers a deprecation warning, subtle behavioral divergence, or test failure, the team is alerted immediately.

---

Pillar 3: Release Auto-Bump & Pin Enforcer

To keep production and development dependencies in lock-step:

1. Strict Version Pinning: custom_components/myhome/manifest.json pins exact releases:

   {
     "requirements": [
       "OWNd==2.0.0b6"
     ]
   }
   

2. PyPI Release Webhook: When OWNd tags and publishes a new release to PyPI (e.g. 2.0.0b7), a GitHub Actions repository_dispatch event notifies MyHOME. A dedicated workflow automatically updates manifest.json, regenerates the lockfile/specs, verifies 100% coverage, and opens a pre-validated PR.

---

4. Test Coverage & Quality Enforcers

Both repositories enforce automated zero-tolerance quality gates:

Quality Gate	Standard	Enforced By
Statement Coverage	Strict 100.0% (0 missing lines across all integration modules)	pytest --cov --cov-report=term-missing
Linting & Formatting	0 Ruff Violations	ruff check .
Home Assistant Standards	Gold / Platinum Scale	scripts/verify_ha_standards.py
Upstream Compatibility	dev, beta, stable	.github/workflows/ha-upstream-compat.yml

By combining Shift-Left testing in OWNd with nightly canary builds and automated release bumping in MyHOME, protocol drift is structurally impossible.