# Engineering Context for Shipbuilding & Offshore

In shipbuilding and offshore, very large 3D models, system hierarchy, spatial density, and repeated review grow at the same time.

> Distributed design review only works when HQ, yard, and partners share the same current model and the same object identity.

## Very large models and dense spaces are the default.

Smart3D-scale models, discipline and system hierarchy, dense overlaps of structure, piping, and supports, local/On-prem environments, and distributed design review are required together. Full republish alone cannot keep daily current-state sharing on pace.

## Where shipbuilding and offshore review stalls today

- Full extract and distribution of very large models lag, so HQ and yard review different revisions.
- Inside system and assembly hierarchy, object identity and surrounding structure, piping, and supports are hard to read together.
- When 1D lists, 2D drawings, and 3D models are not synchronized, multidisciplinary review breaks.
- Cloud-first AI does not fit environments with local and On-prem constraints.

## Why spatial context and distributed review matter

3D Tiles and Spatial Graph from S3DX, shared-identity review in L3D Vue, and PIDEX where process systems or P&IDs apply must work together. A local AI Agent should run on that context — not invent a separate reality.

## From large model to distributed review

Incremental spatial publishing and a shared-identity workspace keep yard and HQ on the same current context.

1. **Smart3D / large model** — Source engineering model
2. **S3DX incremental publish** — 3D Tiles + Spatial Graph
3. **Object identity · system hierarchy** — Assembly and discipline context
4. **Spatial Resolver (where applicable)** — Surrounding structure, piping, supports
5. **L3D Vue distributed review** — Synchronized 1D · 2D · 3D review
6. **Evidence Snapshot** — Preserve review evidence
7. **Local AI Agent (future)** — Engineering Context → GraphRAG → local SLM

## Four recurring shipbuilding & offshore workflows

### A. Large-model spatial review

Smart3D → S3DX → 3D Tiles + Spatial Graph → L3D Vue for current large-model spatial review.

- Smart3D
- S3DX
- 3D Tiles + Spatial Graph
- L3D Vue

### B. System / Assembly Context Review

Read system hierarchy, object identity, and surrounding structure, piping, and supports to set review scope.

- System hierarchy
- Object identity
- Surrounding structure / piping / support
- Review

### C. 1D / 2D / 3D synchronized review

Engineering Graph, P&ID or drawing, and 3D model stay synchronized in L3D Vue under shared identity.

- Engineering Graph
- P&ID / drawing
- 3D model
- L3D Vue

### D. Local AI Agent (future)

nanochat as a future Alpha targets Engineering Context, GraphRAG, a local SLM, OID-backed responses, and viewer actions.

- Engineering Context
- GraphRAG
- Local SLM
- OID-backed response → viewer action

## What this workflow actually uses

Primary focus is spatial publishing and the review workspace. Process, local agent, and deliverables join only where they apply.

### Primary

- **S3DX**: Keeps 3D Tiles and Spatial Graph current through incremental publish of very large models.
- **L3D Vue**: Engineering Context Workspace for distributed 1D · 2D · 3D design review.

### Secondary

- **PIDEX**: Connects process context where process systems or P&IDs apply.
- **nanochat** (future): Future local Engineering Agent layer (design stage / planned Alpha).
- **L3D Emit**: Turns model views into downstream engineering deliverables where relevant.

## What AI does — and what engineers close

AI understands, analyzes, identifies candidates, and supports review. Final judgment and approval stay with authorized engineers.

We do not claim automated approval, yard disposition, or 100% savings or mapping. Focus stays on current-model review, spatial context, and evidence. The nanochat use case is future tense.