> For clean Markdown of any page, append .md to the page URL.
> For a complete documentation index, see https://docs.umnai.com/llms.txt.
> For AI client integration (Claude Code, Cursor, etc.), connect to the MCP server at https://docs.umnai.com/_mcp/server.

# Ingestion

Ingestion reads source files, applies a data schema and creates a datasource for onboarding. It connects the data in your storage bucket with the field definitions you reviewed in the schema.

The configuration, job and datasource have separate roles. Saving a configuration records the settings. Running a job applies them to source files. The datasource is the resulting data resource.

## How ingestion works

#### Storage and schema

A storage connector provides access to the files. The data schema defines how each column should be interpreted.

#### Configuration

The ingestion configuration combines the schema with settings for interpreting observation metadata, such as observation identifiers and train or test labels.

#### Job

The ingestion job selects the source path and runs the configuration using processing compute. Processing continues after the job has been submitted.

#### Datasource

A successful job creates or updates the datasource. Check that the output is **AVAILABLE** before using it for onboarding.

![Ingestion resources and their relationships](/_fern-files/_img/927a0cc8012b4e59aa22eb50d50c554f9ca919ed8fe98e1fc5790e62bdfa4b87.webp)

## Initial and incremental ingestion

An **Initial** job creates a new datasource. An **Increment** job adds data to an existing datasource using its schema. An optional **Increment label** helps identify the update.

An ingestion increment does not automatically update a dataset or retrain a model. Onboarding and training remain separate stages, with their own jobs and outputs.

## Find ingestion resources in the app

Under **Data Lab > Datasources**, the configuration, job and datasource tabs let you move from saved settings to execution history and the resulting data. Use the resource references and lineage to confirm which files and settings produced the datasource.

## Next steps

Create an [ingestion configuration](/ui-guides/data/ingestion-configuration), then run an [ingestion job](/ui-guides/data/ingestion-jobs).