Projects
Drawing Intelligence
Alloovium automatically extracts structured metadata from construction drawings on upload — no configuration required. Sheet numbers, revisions, disciplines, and title block data are indexed for search and AI queries.
Overview
When a drawing PDF is uploaded to a project, Alloovium reads the title block and extracts structured metadata — discipline, sheet number, revision, title, date, and scale. This happens automatically as part of the standard processing pipeline.
Once indexed, drawings are queryable by any of these fields in both the project search and the AI assistant. The assistant can answer questions like "what is the latest revision of A-101?" or "list all structural drawings issued after June" without any manual tagging.
No routine required
Drawing intelligence is not a routine step. It runs automatically on every drawing uploaded to a project — there is nothing to configure.A drawing open in the preview with its extracted title-block metadata in the details panel: sheet number A-101, revision C, discipline Architectural, and date shown as structured fields beside the title block they were read from.
Video · ~2 min
Drawings that know what they are
Upload a small drawing set and watch title-block extraction fill in sheet numbers, revisions, and disciplines automatically — then ask the assistant for the latest revision of a sheet and jump straight to it.
What Gets Extracted
Alloovium reads the title block from each drawing page and extracts the following fields where present:
| Field | Example | Notes |
|---|---|---|
| Sheet number | A-101, S-03, M-204 | Indexed for search; used to group revisions of the same sheet |
| Discipline | Architectural, Structural, Mechanical, Civil | Indexed for search; inferred from the sheet number prefix if not explicit |
| Drawing title | Ground Floor Plan, Column Schedule | Indexed for search; from the title block title field |
| Revision | Rev C, Rev 3, P4 | Read from the title block; used to link successive revisions of a sheet |
| Date, scale, and other fields | 14 Apr 2025, 1:100 | Read where present, but not currently exposed as separate search filters |
Scanned drawings
Scanned drawing PDFs are processed using GPU-accelerated OCR before title block extraction. Accuracy is high for standard title blocks, but may be lower for handwritten annotations or non-standard layouts.Searching Drawings
Drawing metadata is searchable directly in the project and through the AI assistant. Both support natural language queries using any extracted field.
Project search
Type a sheet number (e.g. "A-101"), discipline ("structural"), or keyword from a drawing title. Results are filtered to matching drawings with revision context shown.
AI assistant
Ask in plain language — "What does drawing S-04 show?", "Are there any civil drawings for the eastern retaining wall?", or "Show me the latest revision of the floor plans."
Filter by discipline
In the project drawing view, use the discipline filter to narrow results to Architectural, Structural, Mechanical, Electrical, or Civil drawings.
Best results
For best AI results, upload full drawing sets rather than individual sheets. The assistant uses cross-sheet context — for example, it can correlate a structural column grid from one sheet with a detail on another.Limitations
| Limitation | Detail |
|---|---|
| Non-standard title blocks | Extraction accuracy drops on highly customised or handwritten title blocks |
| Raster-only PDFs | Very low-resolution scans (below ~150 DPI) may produce lower-accuracy OCR results |
| Multi-page sheets | Each page is processed independently; multi-page assemblies are treated as separate sheets |