Skip to content

Atomic ADR template

Use this template for atomic architecture decisions derived from Knowledge Graph nodes and operational receipts. Markdown is a projection — emit structured graph/ADR records first when using research agents.

Template

# ADR-XXXX: <short title>

Question: <single atomic question — one ownership or placement dimension per ADR>

Answer: UNKNOWN

Evidence: [graph node IDs / UP-* / RX-*]

Related ADRs: ADR-YYYY, ADR-ZZZZ

## Status

Candidate — conclusion **UNKNOWN** until evidence links close.

## Context

<RETRIEVED|OBSERVED|INFERRED tagged claims only>

## Decision

**UNKNOWN**

## Consequences

<Fill when Answer is no longer UNKNOWN>

Placement decomposition rule

Complex placement topics MUST decompose into atomic ADRs. One ADR per question dimension.

Parent topic Atomic ADR questions (examples)
Town HQ Who owns HQ (semantic)? Who writes HQ (operational)? Who consumes HQ? Is HQ replicated?
Beads / Dolt Who owns the Dolt database? Who is the multi-writer? Where is persistence authoritative?
Upstream authorities Where do RETRIEVED stubs live? Who maintains the URL index?
Mac / NAS / Oracle What remains on Mac after demotion? What is NAS substitute for each Mac SoR candidate?

Evidence linking

  • Graph node IDs: KG-* from knowledge-record.schema.yaml (NOT_FOUND — schema was never committed)
  • Upstream facts: Evidence/Capability/**/UP-*.md
  • Operational receipts: Evidence/receipts/RX-*.md
  • Inventory rows: Platform-Inventory/objects/*.yaml

Numbering

Sequential ADR numbers are assigned in ADR-INDEX.md. Do not reuse numbers assigned to accepted ADRs.