Supported types

SQL Server        SQLite
-----------        ------
int, bigint   -->  INTEGER
nvarchar      -->  TEXT

The package stores values using SQLite affinities chosen for reliable transport, not a one-to-one type mapping. The original SQL Server type name is kept alongside every column, so import can convert each value back into what SqlBulkCopy expects for its real type.

The matrix

Category	SQL Server types	SQLite storage
Integer and boolean	bigint, int, smallint, tinyint, bit	INTEGER
Floating point	float, real	REAL
Text	char, varchar, text, nchar, nvarchar, ntext	TEXT
Date/time	date, datetime, datetime2, datetimeoffset, smalldatetime, time	TEXT
Numeric, text-preserved	decimal, numeric, money, smallmoney	TEXT
XML, text-preserved	xml	TEXT
JSON, text-preserved	native json (SQL Server 2025 and Azure SQL)	TEXT
Vector embeddings	native vector (float32 base type, GA on Azure SQL Database, Azure SQL Managed Instance, SQL Server 2025, and Fabric SQL DB; preview float16 base type also supported)	TEXT (JSON array)
Binary	binary, varbinary, image	BLOB
Identifiers	uniqueidentifier	TEXT
Server-generated	timestamp, rowversion	BLOB (captured for inspection, skipped on import; SQL Server generates a fresh value on the target)

XML columns round-trip through SQLite TEXT. If a package is edited and an XML value is no longer valid XML, import fails with SQL Server's own XML conversion error.

Vector columns

vector columns are stored as SQLite TEXT JSON arrays, with the base type and dimension count recorded alongside the column so import can reconstruct the native value.

• float32, the GA base type, round-trips bit-for-bit through the native binary SqlVector<float> transport.
• The preview float16 base type round-trips through its JSON representation instead. Import requires the target database to have PREVIEW_FEATURES enabled, and export emits a warning whenever a float16 column is present.

Either way, import requires the matching vector(N[, float16]) column to already exist on the target; SqlDataPack does not create it for you.

JSON columns

Native SQL Server json columns are stored as SQLite TEXT and imported back into SQL Server json columns.

JSON that already lives in nvarchar, varchar, or another text column is not treated specially: it round-trips as ordinary text, the same as any other string value.

If a package is edited and a native JSON value is no longer valid JSON, import fails with SQL Server's own JSON validation error.

Unsupported types

sql_variant, geography, geometry, and hierarchyid are not supported. A table carrying one of these fails export preflight rather than exporting a column SqlDataPack cannot faithfully convert.

Exclude the column with ExcludeColumns (schema.table.column) to export the rest of the table:

options.ExcludeColumns = ["dbo.Stores.Location"]; // a geography column

See Options for ExcludeColumns, and Troubleshooting for what the preflight failure looks like and how to work through it.