Skip to main content

​
What is the Drawing Index?

The Drawing Index is a structured knowledge graph of your construction drawing set. When you upload a PDF, Bedrock segments every page into sheets and blocks, classifies content, and extracts text for search. Use the Vision API to detect symbols, generate revision overlays, and analyze changes. All results are stored as queryable entities with typed relationships.

​
The search engine analogy

Enterprise search tools like Elasticsearch take unstructured documents, parse them, and build an index so you can query instantly. Bedrock does the same thing for construction drawings, the densest unstructured documents in the built world, making them incredibly difficult for LLM and AI agents to reason through.
Search engineBedrock
Ingest documentsUpload drawing PDFs
Tokenize and parseSegment sheets into blocks, detect features
Build inverted indexBuild knowledge graph with relations and grids
Query with filtersQuery any entity by type, label, grid location
The result: instead of flipping through hundreds of PDF pages, you query the Drawing Index and get structured results that can be fed into your automation workflows or agentic systems.

​
How it works

1

Upload

Submit construction drawing PDFs via the CMS API. Any size, any discipline.
2

Parse

Bedrock’s vision pipeline segments each page into sheets and blocks, classifies block types (plan view, schedule, legend, title block), and extracts text for indexing. Use the Vision API to run symbol detection, overlay generation, and change analysis on demand.
3

Query

Access the structured Drawing Index through the REST API for your application or the MCP Server for AI agents (Claude, ChatGPT, Microsoft Copilot).

​
Entity hierarchy

The Drawing Index organizes construction documents into a hierarchy: Organization → Project → File → Drawing → Sheet → Block → Feature. Every query traverses this tree. Drawings contain Sheets (pages). Each Sheet contains Blocks (plan views, schedules, legends). Each Block contains Features (rooms, doors, symbols). Relations connect entities across the graph — a door to its schedule row, a symbol to its legend definition. A shared Grid coordinate system aligns everything across disciplines.

​
A simple query

To see the power of the Drawing Index, consider this question: “How many duplex receptacles are on sheet E-201?” Without the Drawing Index, someone has to open the PDF, navigate to E-201, and count symbols by hand. With the Drawing Index: 
// query_feature({ sheet_number: "E-201", type: "duplex_receptacle" })
{
  "count": 17,
  "items": [
    { "type": "duplex_receptacle", "label": null, "parent": "Room 201", "grid": "A-2" },
    { "type": "duplex_receptacle", "label": null, "parent": "Room 201", "grid": "A-3" },
    { "type": "duplex_receptacle", "label": null, "parent": "Room 202", "grid": "B-2" }
  ]
}
Every feature carries its parent context (which room it is in) and grid location (where it sits on the building coordinate system).

​
Deep dives

Relations

Cross-references and connections: the edges of the knowledge graph.

Grid System

The universal coordinate system for cross-discipline coordination.