Skip to content
banshee

Schema-aware features

With [database] configured and a db-feature build, banshee introspects information_schema / pg_catalog for tables, columns, primary keys, functions, and samples JSONB columns to infer their shape. The result is cached at .banshee/schema-cache.json (keyed by connection + schema, with a TTL), so only the first run touches the database.

The cached schema drives:

  • column / table completion,
  • JSONB key completion (resource->'…'),
  • RF01 / RF02 reference checks and “did you mean?” suggestions,
  • AM04’s SELECT * expansion.

Reference checks (RF*) only run when a schema is available, so they never produce false positives offline. The .banshee/ cache is plain JSON and can be read without the db feature, so embedders stay async-free.