Redact Usage
Configure usage data redaction in your self-hosted Deepgram deployment to enhance privacy and comply with security requirements.
Configure usage data redaction in your self-hosted Deepgram deployment to enhance privacy and comply with security requirements.
redact_usage boolean
Deepgram’s Redact Usage feature allows you to automatically redact sensitive customer data from usage records and logs in your self-hosted deployment. When enabled, arbitrary string values in redaction-eligible fields are replaced with [REDACTED] to enhance privacy and ensure compliance with security requirements.
To enable Redact Usage, add the redact_usage configuration to your self-hosted API container’s api.toml file:
Default behavior: If redact_usage is not present, it is treated as true (redaction ON).
Only set redact_usage=false in non-production environments and for the minimum time necessary for debugging.
When redact_usage=true, your self-hosted API container automatically redacts sensitive values before they’re included in usage records and logs:
keyterm or mappings in replace will appear as [REDACTED][REDACTED]Deepgram also provides a fallback redaction step at the server-side usage reporting endpoint to ensure any redaction-eligible values are redacted before ingestion into usage/billing systems, providing defense-in-depth protection.
This feature affects metadata/parameters recorded for observability and usage—not model behavior or audio content processing.
When redact_usage=true, certain parameters will show as [REDACTED], which can limit Deepgram’s ability to diagnose issues solely from your usage records.
If you need support:
redact_usage=false to capture a reproducible example, then revert to true after debuggingredact_usage=true in production for enhanced privacy and complianceredact_usage=false only as long as necessary, ideally in a staging environmentRedact Usage provides:
If you can’t upgrade immediately, server-side fallback redaction will still protect usage/billing ingestion.