Authentication and licensing

Most real deployments use licensed models exported from the DENKweit Vision AI Hub. To run them, you need a personal access token and a suitable license source.

Step 1: Create a personal access token

1. Open the Integration Page.
2. Create a service user or select an existing one.
3. Generate a PAT for that service user.
4. Store the PAT securely. It will only be shown once.

Treat the PAT like a password.

Step 2: Choose a license mode

HubLicenseSource

Use this when the target system can stay online.

Python

from denkflow import HubLicenseSource

license_source = HubLicenseSource.from_pat(
    "YOUR-PAT",
    license_id="YOUR-LICENSE-ID",
)

You can also let Pipeline.from_denkflow(...) build this implicitly by passing pat=.

from denkflow import Pipeline

pipeline = Pipeline.from_denkflow("model.denkflow", pat="YOUR-PAT")

C / C++

DenkflowHubLicenseSource* license_source = NULL;
handle_error(
    denkflow_hub_license_source_from_pat(&license_source, "YOUR-PAT", "YOUR-LICENSE-ID", NULL),
    "denkflow_hub_license_source_from_pat"
);

DenkflowPipeline* pipeline = NULL;
handle_error(
    denkflow_pipeline_from_denkflow(&pipeline, "model.denkflow", (void*)license_source),
    "denkflow_pipeline_from_denkflow"
);

OneTimeLicenseSource

Use this when the deployment must continue working offline after initial activation.

Python

from denkflow import HubLicenseSource

hub_source = HubLicenseSource.from_pat(
    "YOUR-PAT",
    license_id="YOUR-LICENSE-ID",
)
license_source = hub_source.to_one_time_license_source()

C / C++

DenkflowHubLicenseSource* hub_source = NULL;
handle_error(
    denkflow_hub_license_source_from_pat(&hub_source, "YOUR-PAT", "YOUR-LICENSE-ID", NULL),
    "denkflow_hub_license_source_from_pat"
);

DenkflowOneTimeLicenseSource* license_source = NULL;
handle_error(
    denkflow_hub_license_source_to_one_time_license_source(&license_source, &hub_source),
    "denkflow_hub_license_source_to_one_time_license_source"
);

Notes:

• Offline activation must be enabled for your license.
• The first activation needs online connectivity.
• The persisted offline state is stored in DENKFLOW_DATA_DIRECTORY.
• If you add new licensed models later, refresh the license while online:

Python

license_source.refresh()

C / C++

handle_error(
    one_time_license_source_refresh(license_source),
    "one_time_license_source_refresh"
);

Step 3: Keep licensing stable in production

For reliable deployments:

1. Keep SDK persistent data on a volume that survives restarts. In Docker, the usual approach is to mount a host directory onto the default data path (for Linux as root, /root/.config/denkflow — the same as $HOME/.config/denkflow). You only need to set DENKFLOW_DATA_DIRECTORY if you choose a non-default path inside the container.
2. Keep /etc/machine-id stable in Docker deployments.
3. Reuse the same persistent volume across restarts.

Which option should you use?

Use HubLicenseSource when:

• the system stays online
• you want the simplest rollout

Use OneTimeLicenseSource when:

• the system must run offline after first activation
• you can guarantee stable persistent storage

Pitfalls

• Using a PAT without access to the required license.
• Persistent state is not actually persistent: the SDK data directory (default or whatever DENKFLOW_DATA_DIRECTORY points to) must be writable and survive restarts — in Docker, typically by mounting a host volume at /root/.config/denkflow (or at your custom path if you set DENKFLOW_DATA_DIRECTORY).
• The machine identity changes between restarts: keep /etc/machine-id stable, especially in Docker deployments.
• Adding new licensed models while offline without refreshing the offline license state first.