Profiles#

Profiles in dlt define environment-specific configurations and secrets.
They allow you to manage separate settings for development, testing, and production using the same codebase.

Each profile provides isolated configuration, credentials, and working directories for dlt pipelines, datasets, transformations, and notebooks. You don't need to write any additional code to benefit from profiles.

Profiles are defined and managed through TOML files located in the .dlt directory.
They are compatible with the secrets.toml and config.toml files you may already know from OSS dlt.

The dltHub platform automatically uses certain profiles to deploy and run pipelines and notebooks.

Enable the workspace and profiles#

Before you start, make sure you have followed the installation instructions and enabled additional Workspace features (which also include Profiles).

dltHub Workspace is a unified environment for developing, running, and maintaining data pipelines—from local development to production.

More about dlt Workspace →

Scaffold a workspace with uvx dlthub-start@latest, then add a pipeline to it from inside the workspace:

dlthub pipeline init pokemon_api duckdb

Once initialized, the workspace exposes the extended CLI surface, including profile-aware commands:

dlthub profile
dlthub local

Define profiles#

Once your workspace is scaffolded, you'll have two familiar toml files in .dlt: secrets.toml and config.toml. They work exactly the same way as in OSS dlt. You can run your OSS dlt code without modifications.

Anything you place in those files is visible to all profiles. For example, if you place
log_level="INFO" in config.toml, it applies to all profiles. Only when you want certain settings to vary across profiles (for example, INFO level for development, WARNING for production) do you need to create profile-specific toml files.

dltHub Workspace predefines several profiles, and together with the dltHub platform, assigns them specific functions:

Profile	Description
`dev`	Default profile for local development.
`prod`	Production profile, used by the dltHub platform to run pipelines.
`tests`	Profile for automated test runs and CI/CD.
`access`	Read-only production profile for interactive notebooks on the dltHub platform.

View available profiles:

dlthub profile list

Switching profiles#

To change environments locally, pin the desired profile.
This makes it the default for subsequent dlthub local … commands:

dlthub local profile use prod

You can verify your current profile:

dlthub profile info

To unpin:

rm .dlt/profile-name

Tip

You can pin a profile with any name, not just those from the predefined list. This allows you to create as many profiles as you need.
You can also pin a profile that doesn't yet have profile-specific TOML files and add those files later.

dlthub -v local info

This command lists all expected file locations from which dlt reads profile settings.

Once pinned, you can run your pipeline as usual through the local runner:

dlthub local pipeline run pokemon_api_pipeline

The workspace automatically uses the active profile's configuration, secrets, and data locations to run the pipeline.

Switching profiles in code#

You can interact with the workspace run context, switch profiles, and inspect workspace configuration using code:

import dlt

workspace = dlt.current.workspace()

workspace.switch_profile("test")

Example: Switch destinations using profiles#

Let's walk through a setup that switches between local DuckDB (dev) and MotherDuck (prod).

Step 1. Configure the development profile#

In .dlt/dev.secrets.toml (to fully separate profiles), define your local destination:

[destination.warehouse]
destination_type = "duckdb"

Then, in your pipeline script, use destination="warehouse":

import dlt

pipeline = dlt.pipeline(
    pipeline_name='pokemon_api_pipeline',
    destination='warehouse',
    dataset_name='pokemon_api_data',
)

Run it locally:

dlthub local pipeline run pokemon_api_pipeline

Data will be stored in .dlt/data/dev/warehouse.duckdb.
Pipeline state will be stored in .dlt/state/dev/.

Step 2. Configure the production profile#

Create .dlt/prod.secrets.toml:

[destination.warehouse]
destination_type = "motherduck"
credentials = "md:///dlt_data?motherduck_token=...."

Pin and activate the profile:

dlthub local profile use prod

Test the connection (optional)#

Before running your pipeline in production, you can verify that the credentials and dataset configuration work correctly:

dlthub --debug local pipeline sync pokemon_api_pipeline --destination warehouse --dataset-name pokemon_api_data

This command connects to your destination, validates credentials, and bootstraps a local copy of pipeline state from the destination. If your credentials are invalid or the configuration is wrong, dlt will raise a detailed exception with a full stack trace—helping you debug before deployment.

If the connection succeeds but the dataset doesn't yet exist in MotherDuck, you'll see a message like:

ERROR: Pipeline pokemon_api_pipeline was not found in dataset pokemon_api_data in warehouse

This simply means the target dataset hasn't been created yet—no action is required.
Now run your pipeline script to load data into MotherDuck:

Run the pipeline with the `prod` profile#

dlthub local pipeline run pokemon_api_pipeline

Data will be stored in MotherDuck.
Pipeline state will be stored in .dlt/state/prod/.

Once the pipeline completes, open the Workspace Dashboard with:

dlthub local show

You'll see your pipeline connected to the remote MotherDuck dataset and ready for further exploration.

Schedule the pipeline to run on the dltHub platform#

Now you're ready to deploy your Workspace to the dltHub platform and schedule your pipeline to run.
Note that the dltHub platform will automatically use the prod profile you just created.

Inspecting and managing profiles#

List profiles
```
dlthub profile list
```
Show the current profile
```
dlthub profile info
```
Clean the workspace (useful in dev)
```
dlthub local clean
```

Best practices#

Use dev for local testing and experimentation.
Use prod for production jobs and runtime environments.
Keep secrets in separate <profile>.secrets.toml files—never in code.
Use named destinations (like warehouse) to simplify switching.
Commit config.toml, but exclude all .secrets.toml files.