Known Limitations
- Auto-discovery extract (no schema) is not reliable on production. Always pass a
schemaorschema_idtotalonic_extract. - Schema definition: prefer full JSON Schema for now. The flat key-type map is documented as supported but the server-side normaliser does not translate it yet.
- `is_not_empty` filter currently underreports. Use specific operators (
eq,gt,contains, etc.) against known values instead. - `schema_id` on `talonic_extract` requires the UUID form. Other endpoints accept either UUID or
SCH-XXXXXXXX, but/v1/extractcurrently only accepts UUIDs. Pass the UUID fromtalonic_list_schemas.