Parse floor plans and elevations into structured data for takeoff and automation.

Floor plans, elevations, and site plans arrive as DWG or DXF files produced by architects and engineers in AutoCAD, MicroStation, or similar tools. The data needed for takeoff, area calculations, door/window counts, and document automation is all there — but it is encoded in a binary or text format that requires a CAD library to read reliably across format versions.

Building and maintaining that library in-house is a significant ongoing engineering cost. Open-source options handle common cases but break on real-world drawings with non-standard layer conventions, external references, or older DWG versions. CADLens provides a hosted parsing service with a stable, versioned JSON output schema.

Submit a drawing to POST /v1/parse. Once the job is COMPLETED, the result contains three data objects directly useful for construction-tech workflows:

• vectorJson — every entity in the drawing: polylines (walls, room boundaries, site perimeters), block inserts (doors, windows, fixtures), text (room labels, dimensions, annotations), arcs, hatches (filled regions like concrete or insulation), and circles.
• layersJson — layer names, colors, and per-layer entity counts. Use this to identify which layers carry wall geometry, which carry dimensions, and which carry mechanical/electrical content you may want to exclude from a floor-plan takeoff.
• metadata — drawing units and bounding box (width, height) of the drawing extents. Essential for converting pixel coordinates to real-world measurements.

A rendered PNG preview is also available, useful for displaying the drawing to the end user before triggering a takeoff or automation workflow.

A typical integration in a construction-tech platform:

1. A user or subcontractor uploads a DWG or DXF. Your backend posts it to POST /v1/parse with Authorization: Bearer <api_key>. For drawings under ~2 MB add wait=true. For larger drawing sets, supply a webhook_url so CADLens calls your endpoint on completion.
2. Poll GET /v1/jobs/{job_id} until status is COMPLETED (or handle the webhook callback). Status values: PENDING → PROCESSING → COMPLETED or FAILED.
3. Fetch GET /v1/jobs/{job_id}/result. Filter vectorJson by layer name to isolate wall geometry, room boundaries, or fixture inserts.
4. Run area calculations, door/window counts, or text extraction on the filtered entities to populate your takeoff or document automation output.

Construction-tech applications typically use the following fields from the CADLens result:

• entities[].type === "polyline" — wall segments and room boundary lines. Closed polylines represent complete room or space boundaries usable for area calculation.
• entities[].vertices — { x, y } point array. Use for perimeter length (sum of segment distances) and polygon area (shoelace formula for closed polylines).
• entities[].type === "insert" — block inserts for doors, windows, columns, and fixture symbols. Each insert carries blockName, insertionPoint { x, y }, scale, and rotation. Count and locate openings by block name.
• entities[].type === "text" — room labels, area tags, elevation marks, and general annotations. Each text entity carries value (the string content), insertionPoint, height, and rotation.
• entities[].layer — cross-reference with layersJson to filter by discipline (architectural, structural, mechanical) using layer-naming conventions.
• metadata.units and metadata.boundingBox — convert drawing coordinates to real-world measurements and validate overall drawing extents against known site or floor dimensions.

A simplified extract from /v1/jobs/{job_id}/result for a single-storey floor plan DWG (schema version 2024-01):

{
  "schemaVersion": "2024-01",
  "metadata": {
    "units": "mm",
    "boundingBox": { "width": 18000.0, "height": 12000.0 }
  },
  "layersJson": [
    { "name": "A-WALL",  "colorHex": "#000000", "entityCount": 48 },
    { "name": "A-DOOR",  "colorHex": "#FF0000", "entityCount": 12 },
    { "name": "A-TEXT",  "colorHex": "#0000FF", "entityCount": 22 },
    { "name": "A-DIMS",  "colorHex": "#888888", "entityCount": 64 }
  ],
  "vectorJson": [
    {
      "type": "polyline",
      "layer": "A-WALL",
      "closed": true,
      "vertices": [
        { "x": 0.0,     "y": 0.0 },
        { "x": 5400.0,  "y": 0.0 },
        { "x": 5400.0,  "y": 4200.0 },
        { "x": 0.0,     "y": 4200.0 }
      ]
    },
    {
      "type": "insert",
      "layer": "A-DOOR",
      "blockName": "DOOR-900",
      "insertionPoint": { "x": 2100.0, "y": 0.0 },
      "scale": { "x": 1.0, "y": 1.0 },
      "rotation": 90
    },
    {
      "type": "text",
      "layer": "A-TEXT",
      "value": "LIVING ROOM",
      "insertionPoint": { "x": 2700.0, "y": 2100.0 },
      "height": 200,
      "rotation": 0
    }
  ]
}

From this: one room boundary with perimeter 2 × (5400 + 4200) = 19,200 mm and area 5400 × 4200 = 22.68 m²; one 900 mm door insert at the south wall; the room labelled "LIVING ROOM". Scale the pattern across all closed polylines on A-WALL to build a full room schedule.

CADLens extracts 2-D drawing entities. Construction drawings bring specific challenges to be aware of:

• External references (XREFs) — DWG files frequently reference external drawing files for base plans, structural grids, or site plans. CADLens parses the model space of the submitted file; it does not resolve or fetch XREF attachments. For full geometry you will need to submit a DWG with XREFs bound in, or submit each file separately.
• Block inserts vs. exploded geometry — doors and windows stored as block inserts are returned as insert entities with insertion point and block name, not as the underlying line geometry. If your workflow needs the actual linework inside a door symbol, the file must have the blocks exploded in the CAD tool before submission.
• Layer naming conventions — CADLens returns layer names exactly as stored in the file. Mapping layer names to disciplines (A- for architectural, S- for structural, M- for mechanical) is your application's responsibility. There is no built-in semantic layer classifier.
• 3D geometry — elevation and section drawings with 3D geometry or solids are parsed for their 2-D entity representation. ACIS/SAT solid bodies are not yet extracted. Complex 3D models are better served by a BIM API.