> ## Documentation Index
> Fetch the complete documentation index at: https://docs.horizonmanaged.com/llms.txt
> Use this file to discover all available pages before exploring further.

# Drawings

> Versioned site drawings with AI floor-plan analysis, isometric 3D views, and low-voltage takeoff

The Drawings module is the system of record for a client site's technical
drawings — floor plans, network and riser diagrams, cable paths, rack
elevations, conduit layouts, and as-builts. Each drawing is scoped to a company
(and optionally a survey, bid, or project), holds the uploaded file with full
revision history, and can be read by Claude Vision into structured room data — to
render a 3D isometric view and to count low-voltage devices for bid takeoff.

Open it at `/drawings`.

## When to use it

* Store and version-track client site drawings.
* Associate a drawing with a company, survey, bid, or project for traceability.
* Turn a floor-plan image into a 3D isometric view of rooms, doors, and features.
* Auto-count low-voltage devices and cable runs for bid takeoff.

## Upload and versioning

Upload files up to 64 MB across 11 drawing types (floor plan, network diagram,
riser, cable path, as-built, site plan, rack elevation, conduit layout, and
more), with metadata like building, floor, area, scale, sheet number, and
revision. Each upload becomes a numbered version — the latest shows on the
detail page, and you can remove individual versions.

## AI floor-plan analysis and 3D

Run **Analyze floor plan** on an image drawing and Claude Vision extracts rooms,
doors, corridors, and safety features. From that, NEXOS CORE renders a
server-side **isometric 3D view** (`/drawings/{id}/3d`) — no plugins, just SVG.

## Quantity takeoff

For drawings attached to a bid, an AI takeoff counts low-voltage devices and
cable runs and feeds the [bid engine](/bidengine)'s reconciliation. When the
drawing is scoped to a bid, the takeoff is gated to only the trades that bid
actually covers.

## Good to know

* **AI features are image-only** — PNG/JPG/WebP work; PDF, DWG, and SVG files are
  stored but can't be analyzed.
* AI output is the model's best estimate — treat extracted dimensions, positions,
  and quantities as a starting point to verify, not gospel.
* The 3D view is empty until you run Analyze.
* **Drawing annotations (pins, notes, measurements) aren't available yet.**
