# Import tasks

Import tasks do more than import CSV files. The creation pages provide different flows by data shape, source, and target. Select a flow from the following table, then call its API.

| Data shape | Data source | Import target | Task creation method |
| --- | --- | --- | --- |
| Unstructured files | Files or directories in an object-storage or distributed-file-system connector | Data volume | Use [Create a connector-file import task](create-import-task.md). |
| Unstructured files | Local files or directories | Data volume | Use [Upload locally and create an import task](create-local-import-task.md). |
| Structured files | CSV, XLS, or XLSX files available on the current creation page | An existing data table, or a new data table in the specified database | Use [Create a connector-file import task](create-import-task.md) and provide `table_config`. |
| Structured files | Local CSV, XLS, or XLSX files available on the current creation page | An existing data table, or a new data table in the specified database | First [upload files](../connector-files/upload-file.md), then use [Upload locally and create an import task](create-local-import-task.md) to submit the table configuration. |
| Structured database data | Tables, collections, or other supported objects in a database connector | An existing data table, or a new data table in the specified database | Use [Create a database structured-load task](create-structured-load-task.md). |

## Confirm where data will be imported

- **Unstructured imports**: The selected data volume is the final write location. Connector-based imports specify it with `volume_id`; local imports specify it with the multipart `VolumeID` field. The current request has no target-directory field: files are written to the selected data volume.
- **Structured-file imports**: Data is written to the target table in `table_config`. For an existing table, provide `table_id`; to create a table, provide `database_id`, `new_table: true`, and `create_table`. A data volume is not the target location for data rows in this task type.
- **Structured database loads**: Data is written to the target table in `structured_load_config.target`. The request cannot include `volume_id` and has no target-directory field.

A connector file or directory source path is specified in `source_config.common_file_task_config.uris`. In local uploads, the `path` in `meta` describes and filters the uploaded source path; it is not a target directory in the data volume. The current creation page does not offer **Keep directory structure**, so compressed packages are handled in a flattened form.

## Check connector file types

Before creating a task, you can call [List files](../connector-files/list-files.md). `data.files[].type` in the response is the file type recognized by the connector, and `data.files[].uri` can be used as the source URI for connector-file imports.

- The current **structured** file picker allows only `CSV` (`7`), `XLS` (`24`), and `XLSX` (`25`).
- The current **unstructured** file filter offers document, audio, video, image, web page, email, CAD, and other categories. Use a `type` value returned by the file list in `file_filter_config.file_types` or the local-upload `file_types` field.

File types indicate only that the type can be recognized from a file name. Whether a task completes still depends on the source file, connector, and import configuration. After submitting a task, get task details, task files, or run records to confirm the actual result.

## Local-upload count and size limits

The following are limits of the current creation page's local file picker. They are not a general promise about server-side processing for direct API calls.

| Import type | Maximum selection per attempt | Single-file size | Notes |
| --- | --- | --- | --- |
| Unstructured files | Up to 20 files | Up to 200 MiB | You can select local files or a folder. Each file in a selected folder counts toward the 20-file limit. |
| Structured files | 1 file | Up to 200 MiB | The current file picker allows only CSV, XLS, or XLSX. |

## Configurable options for unstructured imports

After selecting a data volume, unstructured imports can use the following options as needed.

| Page feature | Request field | Effect |
| --- | --- | --- |
| Connector source / local upload | `/task` JSON request or `/connectors/upload` multipart request | Determines whether files are read from a connector or uploaded from the local computer. |
| One-time / scheduled import | `load_mode_config` | A connector source can run once or scan a directory every minute, hour, or day. Scheduled tasks can select only directories. |
| File-type filter | `file_filter_config.file_types` or `file_types` | Select document, audio, video, image, web page, email, CAD, and other types. When **Other** is selected, files are not filtered by type. |
| Path regular expression | `file_filter_config.path_regex` or `path_regex` | Processes only files whose source paths match the regular expression. |
| Decompression strategy | `unzip_keep_structure` | The current page handles compressed packages in a flattened form. |
| Deduplication strategy | `dedup.by`, `dedup.strategy` | Detects duplicate files by file name and/or MD5, then skips or overwrites them. |

```{toctree}
:hidden:
:maxdepth: 1

List tasks <list-import-tasks>
Get task details <get-import-task>
Import task object fields <import-task-object-fields>
Create a task <create-import-task>
Upload locally and create an import task <create-local-import-task>
Create a database structured-load task <create-structured-load-task>
Resolve a structured-load backfill start <resolve-structured-backfill-start>
Update a task <update-import-task>
Delete a task <delete-import-task>
Pause a task <pause-import-task>
Resume a task <resume-import-task>
List task files <list-import-task-files>
List run records <list-import-task-runs>
Retry failed files <retry-failed-import-files>
```