Adr 030 dashboard architecture

Decision

Implement the Governance Health and Intelligence Dashboard using a layered architecture and a deterministic heuristic health model.

Context

• Source: specs/030-governance-health-intelligence-dashboard/spec.md
• Decision classification: system

Problem

The CAS portal lacks a central landing experience that synthesizes governance health. Users must read individual ADRs, Architecture Work items, and Alignment Reports to understand the system state, which is inefficient for high-level monitoring and risk identification.

Rationale

1. Layered Decoupling: Separating data access (data_access.py), health logic (health_model.py), and the web interface ensures the health engine can be tested independently of the UI.
2. Heuristic Model (Phase 1): Using simple, transparent heuristics (e.g., Red status if high-severity findings exist) ensures trust and allows for incremental refinement.
3. Read-Only Intelligence: The dashboard serves as a consumption layer. It derives insights but never becomes a new source of truth, preserving the ADR as the authoritative record.
4. Information Scent: Providing summaries with direct links to artifacts (Traceability) ensures users can verify insights without friction.

Next Step

• Implement the cas_runner/dashboard/ module structure.
• Define specific heuristic triggers for Green/Amber/Red states in the health model.
• Replace the portal homepage with the synthesized dashboard view.

Traceability

Artefact	ID
Specification	030-governance-health-intelligence-dashboard
Pattern	heuristic-driven-arbitration
Pattern	contract-first-architecture
Pattern	structured-output-via-schema