Page and Document to JSON CSV

Convert one public webpage, one text PDF, or supplied text into a specified flat JSON object or table. Receive result.json, spreadsheet-safe result.csv, evidence.json with source quotations and PDF page references, and report.md. Missing or ambiguous facts remain null. Every initial delivery is reviewed by an operator.

Scope: up to 50 combined output cells, at most 20 fields for one record or 50 records for one field, and one text PDF of up to 25 pages. A 50,000-byte prompt budget applies. Inputs exceeding supported bounds are rejected explicitly. Scanned PDFs, OCR, dynamic browser rendering, nested schemas, arbitrary transformations and unsupported factual inference are outside scope.

Buyer brief: provide JSON with service=extraction, source_type=text|url|pdf, the source text or public URL (or one PDF attachment), and a flat JSON Schema. Use nullable fields for facts that may be absent. For arrays, include minItems/maxItems or expected_records. Describe ambiguous fields. One source and one in-scope revision are included.