DocIntel — Document Intelligence

Route: /docintel
Purpose: Import BIM files and project documents into the Purple8 Graph automatically

9.1 Supported Formats

Format	Content Extracted
IFC	Full geometry, elements, properties, relationships, spaces
PDF	Text, tables (via OCR), drawing annotations
DWG/DXF	2D geometry, layers, blocks (converted to elements)
DOCX	Specifications, reports — extracted as document nodes
XLSX	Room schedules, cost plans, door/window schedules
S3	Any of the above, fetched from AWS S3 bucket
SharePoint	Documents from SharePoint libraries (Office 365)
Confluence	Wiki pages and attachments from Confluence

9.2 The Pipeline

Upload / Connect
     ↓
QUEUED — job is in the queue
     ↓
PARSING — file read, format detected, structure extracted
     ↓
EXTRACTING — elements classified, properties normalised, relationships mapped
     ↓
EMITTING — graph nodes created, edges wired, LOD assigned
     ↓
DONE — available in BIM Viewport

9.3 Uploading a File

1. Click Browse or drag a file into the upload zone
2. Select the file type if not auto-detected
3. Optionally set a Project ID to scope the elements
4. Click Upload
5. The document appears in the pipeline list with status QUEUED
6. Refresh after a few seconds — status updates automatically

9.4 Connecting Cloud Sources

For SharePoint, S3, or Confluence:

1. Click the Connectors tab in the upload panel
2. Select connector type
3. Enter the required credentials (URL, API key, etc.)
4. Browse and select the document or folder
5. Click Connect — the same pipeline runs asynchronously

9.5 After Import

Once DONE:

• All elements appear in BIM Viewport under your project
• You can run Algorithms on the imported data
• The AI Chat can query the imported elements
• Compliance checks run on the imported floor plans

9.6 IFC Import Tips

• Use IFC2x3 or IFC4 format — IFC4X3 is supported but experimental
• Ensure IfcProject and IfcBuilding are defined at the top level
• IfcSpace elements map to space nodes; IfcWall to wall nodes, etc.
• Properties in property sets (Pset_*) are extracted as node properties
• Large IFC files (>50MB) may take 2-5 minutes to fully process

---