Turn TAB and commissioning reports into rows of design versus measured values per terminal, so out-of-tolerance readings are easy to filter.
Every item comes back with the page and quote it was taken from, checked against the document, so you can verify any value in seconds. Scanned PDFs work too: the Terminal reads scanned pages with AI vision, and the API can too.
extractor_id.curl -X POST https://api.jsonit.io/v1/extract-file \
-H "Authorization: Bearer sk_live_your_key" \
-F "file=@tab-report-level-3.pdf" \
-F "accuracy=high" \
-F "evidence=true" \
-F 'schema={"type":"object","properties":{"readings":{"type":"array","items":{"type":"object","properties":{"tag":{"type":"string"},"design":{"type":["number","null"]},"measured":{"type":["number","null"]},"unit":{"type":["string","null"]},"percent_of_design":{"type":["number","null"]},"remarks":{"type":["string","null"]}}}}}}'