v5.14.0

Performance

• SGML Parser 10x Faster — Rewrote SGML parser from line-by-line to offset scanning with lazy content references. Parse times drop from 52ms to 5.5ms for large filings (Apple 10-K, 9.3MB). Peak memory reduced 275x (23.4MB to 0.1MB). Form 4 parsing is 71x faster (96ms to 1.3ms) by removing a network call from header parsing.

• SGML Max File Size — Raised max content size from 200MB to 500MB to handle large filings with embedded images.

Improved

• Entity Classification — `is_individual` / `is_company` now uses a 9-signal priority chain for more accurate classification of SEC entities. Catches holding companies, old/inactive companies, and institutional investors that were previously misclassified as individuals. (#624)
  • Insider transaction flags (`insiderTransactionForIssuerExists`) used as the strongest company signal
  • Name-based heuristics detect company keywords (INC, CORP, LLC, FUND, etc.) with word-boundary matching to avoid false positives on personal names
  • Ampersand (`&`) in entity names detected as a company/partnership signal
  • Expanded recognized company forms from ~50 to 94 (adds small business, amendments, investment company, and proxy forms)

Fixed

• XBRL API Docs — Corrected financials API examples in StatementType quick reference and quickstart docs. (#580, #619)

• Facts Query — Added `is_dimensioned` column to facts query DataFrame output. (#612)