v5.20.0

Features

• Fund data objects improvements: Overhauled fund entity data objects for better performance, cohesion, and memory safety. Funds are now more efficient to work with and less prone to memory issues in large datasets.
• XBRL fact_id in DataFrame output: The fact_id field is now exposed in the XBRL facts DataFrame, making it easier to uniquely identify and cross-reference individual XBRL facts.

Fixes

• Improved SGML parser error diagnostics: The SGML parser now includes diagnostic details in "Unknown format" errors, making it much easier to identify and debug malformed or unexpected SGML input.
• BDC test suite reliability: Switched Business Development Company (BDC) tests from ARCC to Blue Owl Capital after ARCC started returning empty responses from the SEC, restoring reliable test coverage.
• ETF integration test correctness: Corrected ETF integration test assertions to accurately reflect the actual library implementation, eliminating false failures.
• AI skill verification test: Added the funds domain to the AI skill verification test suite to ensure fund-related skills are properly covered.
• Test collection stability: Eliminated test collection timeouts and removed references to defunct fund tickers that were causing slow or hanging test runs.

Documentation & CI

• Fund entity guide: Added a comprehensive guide for working with Fund entities, updated the data-objects index, and created a dedicated fund AI skill definition to improve discoverability.
• README accuracy: Adjusted the comparison table in the README to more accurately reflect library capabilities.
• Smoke test workflow: Disabled automatic smoke test workflow execution; the workflow is now kept as a manual-only trigger to reduce unnecessary CI runs.