TextIn xParse Document Parse

Problem

Document parsing often gets stuck on fragmented formats, page-level retrieval, repeated multi-file submissions, and unclear free/paid quota boundaries. Feeding files directly into generic scripts or models usually lacks stable doc_id, navigation cache, task state, and quota feedback, making results hard to reproduce.

How It Works

TextIn xParse Document Parse delegates the execution core to the installed xparse-cli; the skill only orchestrates calls and enforces constraints. Use parse for a single document or URL. Use the durable Task Runtime for multiple local files, persistent task_id, status checks, exports, or recovery. The default --api auto checks quota and prefers free allowance; use --api paid only after explicit user approval. For local documents, a complete parse writes the navigation cache, then use get_doc_info, get_outline, search_text, read_pages, and read_content for targeted extraction. On failure, inspect xparse_error.v1 and follow retryable plus next_action, allowing at most one retry for the same logical action.

Boundaries

This skill focuses on parsing, navigation, and batch task control, not downstream business processing. The free endpoint usually covers PDFs and images, while Office, HTML, OFD, and similar formats may require paid routing. Multi-document tasks should not be split into parallel parse calls to bypass quota. States such as waiting_paid_authorization or waiting_funds require external confirmation before task resume.