Instance Settings

Locale settings control how the UI displays dates, times, and language content. Each setting supports automatic detection from the user’s browser or a fixed instance-wide default.

Language

Controls the UI language for all users of this instance.

Value	Behaviour
`Automatic`	Detected from the user’s browser preferences. Currently supports English and German. Any unrecognised browser language defaults to English.
`English`	Forces English regardless of browser preference.
`German`	Forces German regardless of browser preference.

A user can override the instance language for their session without affecting other users. The UI header includes a language switcher with the same three options. The session override takes precedence over the instance default and persists for one year.

Resolution order:

User override → Instance default → Browser preference → English

Timezone

Controls how dates and times are displayed. The instance default can be a specific timezone or Automatic for browser-based detection.

Value	Behaviour
`Automatic`	Detected from the user’s browser. Falls back to UTC.
IANA timezone	Fixed timezone applied to all date/time displays.

Common timezones available in the selector:

UTC, Europe/London, Europe/Vienna, Europe/Berlin, Europe/Kyiv, America/New_York, America/Chicago, America/Denver, America/Los_Angeles, Asia/Dubai, Asia/Singapore, Asia/Tokyo, Australia/Sydney

Each timezone is shown with its current UTC offset (e.g. Europe/Berlin (UTC+02:00)).

Time format

Controls whether times are displayed in 12-hour or 24-hour format.

Value	Behaviour	Example
`Automatic`	Detected from the user’s browser preferences. Typically 12-hour for US browsers, 24-hour for most others.	depends on browser
`12-hour`	Fixed 12-hour display.	`08:45 PM`
`24-hour`	Fixed 24-hour display.	`20:45`

Hugging Face token

A Hugging Face token authenticates requests to the Hugging Face Hub (huggingface.co). It is used internally by the platform when downloading models — specifically for audio transcription and Hugging Face custom detectors.

Without a token, model downloads are unauthenticated, which results in lower rate limits and slower download speeds. With a token, downloads are authenticated, giving you higher rate limits and faster speeds — especially important when processing audio files or running many custom detectors.

Where tokens are used

Feature	Why it needs the token
Audio transcription	Downloads Whisper or other speech-to-text models from Hugging Face Hub to transcribe audio and video assets.
Custom detectors	Downloads the model specified in the detector configuration (text classification, image classification, feature extraction, object detection).

Where to get a token

Sign up or log in at huggingface.co
Go to Settings → Access Tokens
Click Create new token
Choose Read role (writing is not needed)
Copy the token — it looks like hf_xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx

How the token is configured

There are two ways to provide a Hugging Face token:

Instance-level (Helm chart)

When deploying with Kubernetes, the operator sets a token via the Helm chart:

api:
  cliJobs:
    huggingFace:
      existingSecret: hf-token
      existingKey: token

The token value lives in a Kubernetes Secret and is injected into every CLI job pod. This is the recommended approach for production deployments.

User-configured (Settings UI)

If no instance-level token is configured in the Helm chart, any user can set their own Hugging Face token on the General tab of the Settings page. The token is encrypted at rest and cannot be viewed after saving.

Important: When an instance-level token is configured in the Helm chart, the UI field is disabled and user-configured tokens are ignored. The instance-level token always takes priority.

How it works

The token travels from where it is configured to the CLI job that downloads the model:

Instance-level (K8s Secret) ──► CLI job environment (HF_TOKEN)
                                        │
User-configured (Settings UI) ──► API decrypts ──► CLI job environment
                                        │
                              huggingface_hub library reads HF_TOKEN
                                        │
                              Authenticated model download

In both cases the token reaches the huggingface_hub Python library inside the CLI job as the HF_TOKEN environment variable. The library automatically picks it up — no code changes are needed when adding or removing a token.