CredCore - Tusk Liquid

Resolve datapoint titles to `column_slug` plus a plain-language `definition` describing what the datapoint means in the source credit agreement and how the stored cell is shaped (units, structure, conventions). Use before designing a `query_sql` extraction so the regex / parser is built against documented semantics, not assumptions.

lookup_columns

Execute a single read-only SELECT / WITH … SELECT over the deals DB. Queries a read-only PostgreSQL database of credit-agreement datapoints aggregated and curated by CredCore analysts from publicly filed credit agreements. Not a live upstream API; data is a curated snapshot. Schema lives in the server INSTRUCTIONS. Tables: deals, datapoints, facts. ``facts.search_vector`` provides indexed full-text search. PostgreSQL regex is POSIX-based, and the server is PostgreSQL 14 — ``regexp_substr``, ``regexp_like``, and ``regexp_count`` do not exist. Use ``substring(text from 'pattern')`` for extraction (a pattern with a group returns the first group, so write ``(?:...)`` unless you want that), ``text ~ pattern`` / ``text ~* pattern`` for matching, and ``regexp_replace(text, pattern, replacement, 'g')`` for replacement.

query_sql

Full-text search across `facts.raw_markdown` via PostgreSQL text search. `query` is passed to ``plainto_tsquery('english', ...)``, so callers do not need to know PostgreSQL tsquery syntax. Plain tokens use AND semantics and English stemming. Pass empty `query` plus a filter (section_name / datapoint_name / deal_pattern) to enumerate a filtered subset without any text constraint.

search_values