Extracting data from product data sheets
Manufacturers publish a data sheet for every part they sell, each with its own table layout and parameter names. Component engineers and distributors take them in by the thousand, as PDF downloads from manufacturer sites and attachments to supplier emails.
A component library needs the part number, the maximum ratings, and each characteristic's minimum, typical and maximum values with unit and test conditions. A PLM system or parametric search index stores them, and engineers filter candidate parts against a design's limits.
What makes product data sheets hard
Units sit in merged cells
A characteristics table prints each unit once, in a cell merged across the rows of related parameters that share it.
Values hold under stated conditions
A minimum, typical or maximum value holds only under the test conditions printed beside it, such as junction temperature and gate voltage.
Performance lives in curves
Typical behavior is plotted as curves, such as on-state resistance against junction temperature, set in a grid across several pages.
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 product data sheet, end to end
About this document: Wolfspeed's preliminary data sheet for a silicon carbide half-bridge power module rated at 3300 V, as published on the manufacturer's site. Its 9 portrait pages run from characteristic tables through performance curves to a package drawing. Page 1 is featured. It names the part, prints its headline ratings and opens the maximum ratings table.
Processing objective: The objective is to identify the part and read the limits a designer screens on first. The maximum ratings come from a table whose unit cells span several rows. Each value grounds to its own cell or line on the page.
Demonstration fields: Part number, Rated drain-source voltage, Rated on-state resistance, Continuous drain current at 25 °C, Power dissipation
The document
4 of 9 pages
Page 1shown below
Page 3
Page 4
Page 6
Where each value came from
product.part_number
{
"value": "HAB900C33LM4",
"page": 1,
"range": {
"end": 82,
"start": 70
},
"box": {
"xmin": 0.08822,
"ymin": 0.14717,
"xmax": 0.36018,
"ymax": 0.17563
}
}headline_ratings.rated_drain_source_voltage
{
"value": "3300 V",
"page": 1,
"range": {
"end": 159,
"start": 136
},
"box": {
"xmin": 0.71878,
"ymin": 0.15913,
"xmax": 0.89714,
"ymax": 0.17778
}
}headline_ratings.rated_on_resistance
{
"value": "2 mΩ",
"page": 1,
"range": {
"end": 185,
"start": 160
},
"box": {
"xmin": 0.72091,
"ymin": 0.18102,
"xmax": 0.88444,
"ymax": 0.20055
}
}maximum_ratings.continuous_drain_current_25c_a
{
"value": 1200,
"page": 1,
"range": {
"end": 1739,
"start": 1735
},
"box": {
"xmin": 0.54085,
"ymin": 0.68816,
"xmax": 0.59384,
"ymax": 0.71096
}
}maximum_ratings.power_dissipation_w
{
"value": 7282,
"page": 1,
"range": {
"end": 2477,
"start": 2473
},
"box": {
"xmin": 0.48792,
"ymin": 0.79048,
"xmax": 0.54083,
"ymax": 0.82458
}
}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.
