Query with SQL

Docvera projects the workspace into read-only GlueSQL tables. Use SQL for reports and exploration without setting up a database server.

Find your documents

SELECT declared_id, type, path
FROM documents
ORDER BY path

The declared_id comes from the schema's identity declaration. type comes from documentType.

Explore the section tree

SELECT d.path, s.heading
FROM documents d
JOIN sections s ON d.path = s.document_id
WHERE s.level = 1
ORDER BY d.path

Section rows expose parsed headings. The full table contract, including source spans, lives in the in-memory SQL reference.

Report diagnostics

Validation results are data too:

SELECT * FROM diagnostics

Try introducing an error in the playground, then run this query in query.sql.

Understand the snapshot

Each CLI command reads the current files. The JavaScript checker maintains its own in-memory baseline; updating a document changes subsequent checks and queries.

An explicitly created SQL snapshot is independent. It does not automatically pick up later edits:

const snapshot = await checker.sqlSnapshot();
try {
  const result = await snapshot.querySql(
    'SELECT COUNT(*) AS n FROM documents'
  );
  console.log(result.rows);
} finally {
  await snapshot.close();
}

Boundaries

This is GlueSQL, not PostgreSQL or SQLite. SQL is read-only: mutation statements and multiple statements in a single query are rejected. Do not assume extensions or database-specific functions are available.

Use docvera check as your validation gate. A SQL report is not itself an enforcement policy unless your integration explicitly evaluates its result.