VerisyntaxEvidence-oriented language
In developmentVerisyntax is a local prototype. Features, syntax, and documentation may change.

Verisyntax — User Manual (prototype 0.1)

Supported import files

The standalone IDE accepts:

  • CSV with a header row;
  • a top-level JSON array of flat objects;
  • plain UTF-8 text, with one physical line per row;
  • XLSX with an explicitly selected sheet.

CSV/JSON/XLSX fields must currently match the source's logical column name or declared physical mapping. A physical mapping looks like this:

column id: integer from "Customer No"

Available scalar types are boolean, integer, real, text, category, date, and datetime. Numeric columns may add a quoted unit, and any column may declare period grain as day, week, month, quarter, or year. Add optional only when a missing value may become SQL NULL:

column country: text optional
column market: category
column revenue: real unit "USD" period month

For plain text, the source must contain exactly one column:

source article_lines:
  database local
  table "article_lines"
  column line: text
  provenance row

Every import is previewed and bound to the saved program hash, retained file hash, mapping hash, and—when merging—the exact merge-plan hash. Typed destination rows and provenance records commit in one transaction. A failure rolls back both.

Bounded local PDF and DOCX text extraction is available in the C++ core and CLI. It records source/text hashes and warns when a PDF has no text layer. It is not yet connected to the import and claim-review UI, and it does not treat extracted text as approved evidence automatically.