> For clean Markdown of any page, append .md to the page URL. > For a complete documentation index, see https://docs.nvidia.com/sdgm/llms.txt. > For AI client integration (Claude Code, Cursor, etc.), connect to the MCP server at https://docs.nvidia.com/sdgm/_mcp/server. # kumoai.connector > Connector and SourceTable interfaces for accessing raw data The `kumoai.connector` module provides interfaces for connecting Kumo to backing data sources and inspecting raw table data. `SourceTable` objects returned by connectors are used to create [`Table`](/sdk/kumoai-graph#table) and [`Graph`](/sdk/kumoai-graph#graph) objects for downstream ML. *** ## Uploading Your Own Data Kumo supports uploading local tables directly. Files larger than 1 GB are supported via automatic partitioning. Tables must be a single Parquet or CSV file on your local machine. Use `FileUploadConnector.upload()` to upload and `FileUploadConnector.delete()` to remove a table. ### `FileUploadConnector` A connector for uploading local files to the Kumo data plane. #### `upload()` Uploads a Parquet or CSV file to Kumo. The name to assign to the uploaded table. Local path to the Parquet or CSV file. Whether to automatically partition files larger than `partition_size_mb`. Target partition size in megabytes when `auto_partition` is enabled. **Returns** `None` (after uploading, access the table via `connector.table(name)`) #### `delete()` Deletes a previously uploaded table. The name of the uploaded table to delete. #### `source_type` `property` **Returns** `DataSourceType` — The source type for this connector. #### `name` `property` **Returns** `str` — The connector name. *** ## Connector Connectors link Kumo to data stored in an external data warehouse or object store. The SDK supports Amazon S3, Google Cloud Storage (GCS), Azure Data Lake Storage (ADLS), AWS Glue, Snowflake, BigQuery, and Databricks. All connectors inherit the following methods from the base `Connector` class: #### `table_names()` **Returns** `List[str]` — Names of all tables accessible through this connector. #### `has_table()` The table name to check. **Returns** `bool` — `True` if the table exists in the connector. #### `table()` Returns a `SourceTable` for the named table. Use the returned object to inspect raw data or construct a [`Table`](/sdk/kumoai-graph#table). The table name. **Returns** `SourceTable` **Raises** `ValueError` if `name` does not exist in the connector. *** ### `SnowflakeConnector` Connects Kumo to a Snowflake data warehouse. #### `get_by_name()` `classmethod` Retrieves a `SnowflakeConnector` by its registered name. The registered connector name. **Returns** `SnowflakeConnector` #### `source_type` `property` **Returns** `DataSourceType` #### `name` `property` **Returns** `str` *** ### `BigQueryConnector` Connects Kumo to a Google BigQuery data warehouse. #### `get_by_name()` `classmethod` The registered connector name. **Returns** `BigQueryConnector` #### `source_type` `property` **Returns** `DataSourceType` #### `name` `property` **Returns** `str` *** ### `DatabricksConnector` Connects Kumo to a Databricks data lakehouse. #### `get_by_name()` `classmethod` The registered connector name. **Returns** `DatabricksConnector` #### `source_type` `property` **Returns** `DataSourceType` #### `name` `property` **Returns** `str` *** ### `S3Connector` Connects Kumo to an Amazon S3 object store. #### `source_type` `property` **Returns** `DataSourceType` *** ### `GCSConnector` Connects Kumo to a Google Cloud Storage (GCS) bucket. #### `get_by_name()` `classmethod` Retrieves a `GCSConnector` by its registered name. The registered connector name. **Returns** `GCSConnector` #### `source_type` `property` **Returns** `DataSourceType` #### `name` `property` **Returns** `str` *** ### `ADLSConnector` Connects Kumo to an Azure Data Lake Storage (ADLS) container. #### `get_by_name()` `classmethod` Retrieves an `ADLSConnector` by its registered name. The registered connector name. **Returns** `ADLSConnector` #### `source_type` `property` **Returns** `DataSourceType` #### `name` `property` **Returns** `str` *** ### `GlueConnector` Connects Kumo to an AWS Glue catalog. Supports tables in partitioned Parquet format, authenticated via IAM permissions on the Glue catalog and its underlying S3 data location. #### `get_by_name()` `classmethod` Retrieves a `GlueConnector` by its registered name. The registered connector name. **Returns** `GlueConnector` #### `source_type` `property` **Returns** `DataSourceType` #### `name` `property` **Returns** `str` *** ## Source Data Tables accessed from connectors are represented as `SourceTable` objects, with column metadata represented as `SourceColumn` objects. ### `SourceTable` A reference to a table stored behind a backing `Connector`. Use it to inspect raw data before constructing a [`Table`](/sdk/kumoai-graph#table) for ML. ```python connector = kumoai.SnowflakeConnector.get_by_name("my-connector") src = connector.table("orders") src.head(10) ``` The name of the table in the backing connector. The connector containing this table. #### `columns` `property` **Returns** `List[SourceColumn]` — Column metadata for all columns in this table. #### `column_dict` `property` **Returns** `Dict[str, SourceColumn]` — Column metadata keyed by column name. #### `head()` Returns the first `num_rows` rows by reading from the backing connector. Number of rows to return. Returns all rows if fewer are available. **Returns** `pd.DataFrame` #### `add_llm()` Enriches the table with LLM-generated embeddings, adding a new embedding column. The embedding model name (e.g. `"text-embedding-3-small"`). API key for the embedding model provider. A template string defining which columns to embed (e.g. `"{title} {description}"`). Output directory path (S3 or local) where the enriched table is written. Name of the new embedding column in the output table. Name of the output table. Embedding dimensionality. Uses the model's default if not specified. If `True`, returns an `LLMSourceTableFuture` immediately rather than blocking. **Returns** `Union[SourceTable, LLMSourceTableFuture]` *** ### `SourceTableFuture` Represents an ongoing asynchronous `SourceTable` generation process. #### `result()` Blocks until complete and returns the `SourceTable`. **Returns** `SourceTable` #### `status()` **Returns** `JobStatus` — Current status of the job. #### `future()` **Returns** `concurrent.futures.Future[SourceTable]` — The underlying future object. *** ### `LLMSourceTableFuture` Extends `SourceTableFuture` with cancellation support for LLM enrichment jobs. #### `cancel()` Cancels the running LLM enrichment job. **Returns** `JobStatus` *** ### `SourceColumn` Metadata for a single column in a `SourceTable`, including its name and inferred data type.