Skip to main content
DuckDB is available as an Available Now source connector for organization-managed .duckdb artifacts. MantrixFlow opens source artifacts read-only, discovers their schemas, tables, views, and columns, and uses the standard SQL source pipeline for Full Table and Incremental extraction.
DuckDB currently requires a pre-provisioned managed file reference. The connection form does not upload a local file or accept a browser/server file path. If your workspace has not been given a file_... reference, contact your MantrixFlow administrator before creating the connection.

​
Connection fields

Do not enter /tmp/..., ../..., file://..., a SQLAlchemy URL, or a local filesystem path. Hosted MantrixFlow accepts only an authorized managed file reference.

​
Create the source connection

  1. Open Connections and click + New Connection.
  2. Set the role to Source.
  3. Choose DuckDB.
  4. Enter a connection name and the managed file_... reference.
  5. Keep Default Schema as main unless the artifact uses another schema.
  6. Click Test Connection.
  7. Save only after the test succeeds.
The connection test fails safely if the reference is invalid, unavailable, or not resolvable by the runtime. Internal materialized paths are never returned to the browser.

​
Read-only source behavior

MantrixFlow opens DuckDB sources with read-only driver configuration for connection tests, discovery, preview, and extraction. A source run must not create tables, write rows, install extensions, or attach another database to the customer artifact. DuckDB source and destination connections in the same pipeline must reference different artifacts.

​
Discover and preview

After creating a pipeline:
  1. Choose the DuckDB connection in the pipeline Source tab.
  2. Discover the catalog.
  3. Select a schema.table stream.
  4. Preview representative rows and confirm the inferred column types.
  5. Choose Full Table or Incremental when a suitable cursor is available.
  6. Save the source configuration.
The connector filters DuckDB system schemas and returns tables and views from the current database. The default user schema is main.

​
Full Table and Incremental

Use Full Table for snapshots, reference data, or tables without a stable cursor. Every selected source row visible in the artifact is extracted. Use Incremental for growing tables with a stable cursor such as updated_at, created_at, or a monotonic integer. A primary key is strongly recommended for deterministic downstream delivery.

​
Type compatibility

The current DuckDB source path maps common values including:
  • signed, unsigned, and huge integers;
  • floating-point and high-precision decimal values;
  • strings, Unicode, UUID, JSON, arrays, and binary values;
  • dates, times, timestamps, timezone-aware timestamps, and intervals; and
  • nullable columns.
Preview representative nested values before publishing a pipeline. Deep STRUCT, MAP, UNION, and custom enum round trips remain dependent on the selected destination’s type compatibility.

​
Troubleshooting

See DuckDB destination for destination setup and the current managed-artifact limitations.