Extracting data from inspection reports
A regulator inspects a permitted site and writes up what it found, on a form that differs by agency. The permit holder has to answer each finding by its deadline, and a lender reads the same report as diligence. Environmental teams track thousands a year, arriving as downloads from public databases.
A compliance manager needs the permit number, the inspection date, each violation with the condition it cites and the date the fix is due. A compliance system opens an action per finding and escalates the ones that run past their deadline.
What makes inspection reports hard
Findings cite their conditions
A violation names the permit condition it breaches, what the inspector saw, the action required and the date it falls due.
Free text inside a ruled form
A general data grid gives way to narrative background and long cells that mix bold, colored and plain text.
Photographs carry the evidence
Landscape pages of site photographs follow the findings, each captioned and stamped with a date, time and position.
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 compliance inspection report, end to end
About this document: A Washington State Department of Ecology inspection under a construction stormwater permit. Seven pages of mixed orientation: three portrait pages of general data, background and the enforcement notice, then four landscape pages of site photographs with GPS-stamped captions. Page 1 is featured. The only person named is the inspector, acting officially.
Processing objective: The objective is to read the record of the inspection itself: which permit, when it was inspected, whether the visit was announced and what the site discharges to. The inspection date is returned in ISO form, read out of a line that also carries the entry and exit times.
Demonstration fields: Permit number, Inspection date, Inspection type, Receiving waters, Weather at inspection
The document
4 of 7 pages
Page 1shown below
Page 2
Page 4
Page 5
Where each value came from
permit_number
{
"value": "WAR313805",
"page": 1,
"range": {
"end": 691,
"start": 663
},
"box": {
"xmin": 0.72339,
"ymin": 0.22447,
"xmax": 0.94294,
"ymax": 0.27038
}
}inspection_date
Printed on the page as Inspection Date and Entry/Exit Time: December 4, 2024, 11:28/11:47, normalized on extraction.
{
"value": "2024-12-04",
"page": 1,
"range": {
"end": 563,
"start": 488
},
"box": {
"xmin": 0.49539,
"ymin": 0.16858,
"xmax": 0.72339,
"ymax": 0.22449
}
}inspection_type
{
"value": "Unannounced",
"page": 1,
"range": {
"end": 603,
"start": 572
},
"box": {
"xmin": 0.72324,
"ymin": 0.1686,
"xmax": 0.94294,
"ymax": 0.22448
}
}receiving_waters
{
"value": "Groundwater",
"page": 1,
"range": {
"end": 654,
"start": 622
},
"box": {
"xmin": 0.49552,
"ymin": 0.22448,
"xmax": 0.72351,
"ymax": 0.27039
}
}weather
{
"value": "39 degrees Fahrenheit, mostly cloudy with a WSW wind at 1 mph.",
"page": 1,
"range": {
"end": 1600,
"start": 1507
},
"box": {
"xmin": 0.05376,
"ymin": 0.86593,
"xmax": 0.83255,
"ymax": 0.88326
}
}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.
