Extracting data from special inspection reports
Special inspection agencies report on the structural work a building code singles out, such as soil compaction or welding, each agency on its own forms. The engineer of record and the building official receive a report after every inspection, dozens to hundreds per project, as emailed PDFs and permit portal uploads.
Code compliance needs each test's location, measured result, specified minimum and outcome. The engineer of record compares each result with the specification. The building department's permit system holds the final report, which the official needs before issuing a certificate of occupancy.
What makes special inspection reports hard
Wide tables on scanned pages
A field density report packs many narrow columns under stacked multi-line headers onto a landscape page, often printed and then scanned.
Results cite other pages
Each test row cites a Proctor lab number, and the maximum density it is compared against sits on that lab's summary sheet.
Results as tables and curves
Laboratory sheets set compaction results in small boxed tables beside a moisture-density curve plotted from the same test.
Parse, extract, verify
Parse
Parse turns the document into reading-order Markdown and a hierarchy of blocks (text, tables, figures), and grounds each block to its exact location on the page.
Extract
You define the fields in a JSON schema, and Extract returns their values in a consistent, structured format, each linked back to where it appears in the Markdown.
Verify
Each value links back to where it appears, down to the character range in the Markdown, and the page and bounding box in the source document, so every result can be verified.
One special inspection report, end to end
About this document: A building pad certification package from ProTeX, a special inspection firm, for the mass grading of a subdivision in Oro Valley, Arizona. The town posts it in its public planning records. Its 6 pages hold the engineer's certification letter, a landscape field density table and the scanned Soils Summary lab sheets the table cites. Page 2 is featured. Individuals' names and signatures were replaced with stand-ins.
Processing objective: The objective is to read one compaction test as a complete calculation: where it was taken, the field density measured, the laboratory maximum it is compared against, and the percent compaction and result that follow. Each value grounds to its own cell of the table.
Demonstration fields: Lot tested, Field dry density (lb/ft3), Max dry density (lb/ft3), Percent compaction, Test result
The document
3 of 6 pages
Page 1
Page 2shown below
Page 3
Where each value came from
field_density_report.tests[0].lot
{
"value": "Lot 39",
"page": 2,
"range": {
"end": 3248,
"start": 3242
},
"box": {
"xmin": 0.25365,
"ymin": 0.31814,
"xmax": 0.40521,
"ymax": 0.34755
}
}field_density_report.tests[0].field_dry_density_pcf
{
"value": 101.9,
"page": 2,
"range": {
"end": 3289,
"start": 3284
},
"box": {
"xmin": 0.52614,
"ymin": 0.31816,
"xmax": 0.58669,
"ymax": 0.34755
}
}field_density_report.tests[0].max_dry_density_pcf
{
"value": 104.1,
"page": 2,
"range": {
"end": 3315,
"start": 3310
},
"box": {
"xmin": 0.64665,
"ymin": 0.31817,
"xmax": 0.70032,
"ymax": 0.34755
}
}field_density_report.tests[0].percent_compaction
{
"value": 98,
"page": 2,
"range": {
"end": 3338,
"start": 3336
},
"box": {
"xmin": 0.75273,
"ymin": 0.31818,
"xmax": 0.7909,
"ymax": 0.34755
}
}field_density_report.tests[0].result
{
"value": "Pass",
"page": 2,
"range": {
"end": 3399,
"start": 3395
},
"box": {
"xmin": 0.93441,
"ymin": 0.3182,
"xmax": 0.97271,
"ymax": 0.34755
}
}How to try it yourself
Everything on this page is reproducible. The sample, its schema, and the Parse and Extract API outputs are published. You can repeat the same two calls from whichever surface you already build on.
- Claude Skilldescribe the job; the agent writes it
- CLIno code at all
- MCPexpose ADE docs as an agent tool
- REST APIany language
- Pythonlandingai-ade
- TypeScriptlandingai-ade
Documents into data
Accurate
The best accuracy on the difficult documents other tools get wrong, such as messy scans, complicated tables, checkboxes and handwriting.
Auditable
Every answer cites its sources. Values come back with an audit trail (page, bounding box, character range) so a human can review them.
Agentic
Applies multi-step reasoning, and plans before it acts. Handles documents it has never seen before and adapts to variations.
API-first
Composable APIs that drop into any tech stack and can be called from anywhere. Easy to build on and embed in your pipelines and products.
Affordable
Cost-competitive rates with complete price transparency. Spend controls before you run, and every job reports its exact cost.
Enterprise-grade
Built for production and scale from day one. Fully observable and monitored, running securely in our cloud or yours.
Built for high-volume, high-throughput production
- Zero data retention
- Priority rate limits
- Enhanced support
- VPC and on-prem deployment
Available on Team and Enterprise plans. VPC and on-prem deployment require an Enterprise plan.
