68dbc014da
19 corrections from handoffs/otopcua-corrections-2026-04-17.md: Inaccuracies fixed: - A1: OPC UA-native equipment requires OpcUaClient gateway driver (~hours config), not "no driver build" - A2: "single endpoint" is per-node (non-transparent redundancy), not per-cluster; no VIP planned Missing constraints added: - B1: ACL surface (EquipmentAcl table, Admin UI, NodeManager enforcement) as Year 1 deliverable before Tier 1 cutover - B2: schemas-repo creation on OtOpcUa critical path with FANUC CNC pilot - B3: Certificate-distribution as pre-cutover step (per-node ApplicationUri trust-pinning) Architectural decisions incorporated: - C1: 8 committed core drivers (added TwinCAT/Beckhoff, split AB Legacy) - C2: Three-tier driver stability model (A/B/C with out-of-process for Galaxy and FOCAS) - C3: Polly v8+ resilience with default-no-retry on writes - C4: Multi-identifier equipment model (5 IDs: UUID, EquipmentId, MachineCode, ZTag, SAPID) - C5: Consumer cutover plan needs an owner (flagged) - C6: Per-building cluster implications at Warsaw clarified TBDs resolved: - D1: Pilot equipment class = FANUC CNC - D2: Schemas repo format = JSON Schema (.json), Protobuf derived - D3: ACL definitions in central config DB alongside driver/topology - D4: Enterprise shortname still unresolved (flagged as pre-cutover blocker) New TBDs added: - E1: UUID generation authority (OtOpcUa vs external system) - E2: Aveva System Platform IO pattern validation (Year 1/2 research) - E3: Site-wide vs per-cluster consumer addressing at Warsaw - E4: Cluster endpoint wording (resolved via A2)
3-Year Plan: Shopfloor IT/OT Transformation
A 3-year plan for transforming and enhancing shopfloor IT/OT interfaces and data collection — bridging IT and OT layers, improving operator interfaces, and upgrading data collection pipelines.
Vision
A stable, single point of integration between shopfloor OT and enterprise IT.
Three Pillars (binary at end of plan)
- Unification — 100% of sites on the standardized stack (OtOpcUa + ScadaBridge + Redpanda + SnowBridge + Snowflake/dbt).
- Analytics / AI Enablement — machine data in Snowflake with a ≤15-minute analytics SLO; at least one "not possible before" use case in production.
- Legacy Retirement — zero remaining bespoke IT/OT integration paths outside ScadaBridge.
Key Architecture
Layer 1 Equipment (PLCs, controllers, instruments)
Layer 2 OtOpcUa (unified site-level OPC UA — single session per equipment, two namespaces)
Layer 3 SCADA (Aveva System Platform + Ignition)
Layer 4 ScadaBridge (sole IT/OT crossing point)
─── IT/OT Boundary ───
Enterprise IT (Camstar, Delmia, Snowflake, Power BI, SnowBridge)
The plan also declares a Unified Namespace (UNS) composed of OtOpcUa + Redpanda + canonical model in schemas repo + dbt curated layer, with a 5-level naming hierarchy standard (Enterprise → Site → Area → Line → Equipment).
Plan Documents
| File | Purpose |
|---|---|
current-state.md |
Snapshot of today's systems, integrations, and pain points |
goal-state.md |
Target end-state: architecture, components, success criteria, UNS, canonical model |
roadmap.md |
7 workstreams x 3 years migration grid |
STATUS.md |
Working-session bookmark — where we left off, pending items |
Component Detail Files
| File | Purpose |
|---|---|
current-state/legacy-integrations.md |
Pillar 3 denominator: 3 legacy IT/OT integrations to retire |
current-state/equipment-protocol-survey.md |
Year 1 protocol survey template (also produces UNS hierarchy snapshot) |
goal-state/digital-twin-management-brief.md |
Digital twin management conversation brief (completed) |
Output Generation
| File | Purpose |
|---|---|
outputs/README.md |
How to regenerate PPTX + PDF from plan source |
outputs/presentation-spec.md |
18-slide mixed-stakeholder deck structure |
outputs/longform-spec.md |
Faithful-typeset PDF structure |
Trigger: regenerate outputs, regenerate presentation, or regenerate longform in a Claude Code session.
Conventions
- Everything in markdown — no proprietary formats.
_TBD_marks gaps that still need filling.- Decisions are captured inline with rationale; don't re-open without new information.
Working with Claude Code
This repo includes a CLAUDE.md file that orients Claude Code sessions to the plan's structure and conventions. Resume work with:
claude --resume <session-id>
or start a new session in this directory — CLAUDE.md and STATUS.md provide enough context to pick up where the last session left off.