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

# Redact Usage

`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.

## Enable Feature

To enable Redact Usage, add the `redact_usage` configuration to your self-hosted API container's `api.toml` file:

```toml
# Enable usage redaction (default: true)
redact_usage = true

# Disable usage redaction (for debugging only)
# redact_usage = false
```

**Default behavior**: If `redact_usage` is not present, it is treated as `true` (redaction ON).

> **Warning**
>
> Only set `redact_usage=false` in non-production environments and for the minimum time necessary for debugging.

## How It Works

### Client-Side Redaction

When `redact_usage=true`, your self-hosted API container automatically redacts sensitive values before they're included in usage records and logs:

* Query parameters such as `keyterm` or `mappings` in `replace` will appear as `[REDACTED]`
* App/agent settings that include customer-provided lists (e.g., keyterms) will appear as `[REDACTED]`

### Server-Side Fallback

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.

> **Warning**
>
> This feature affects metadata/parameters recorded for observability and usage—not model behavior or audio content processing.

## Troubleshooting

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:

* Provide the request ID from the API response along with sanitized sample requests; or
* Temporarily set `redact_usage=false` to capture a reproducible example, then revert to `true` after debugging

## Best Practices

* **Keep `redact_usage=true` in production** for enhanced privacy and compliance
* **For debugging**, use `redact_usage=false` only as long as necessary, ideally in a staging environment
* **Upgrade to the latest self-hosted release** (250731 or later) to get client-side redaction by default

## Special Considerations

### Privacy and Compliance

Redact Usage provides:

* **Stronger privacy by default**: Reduces the chance that customer-supplied terms are stored in readable form
* **Defense-in-depth**: Redaction at the source plus server-side fallback ensures consistency
* **Compliance readiness**: Aligns with least-data-retained practices across environments

If you can't upgrade immediately, server-side fallback redaction will still protect usage/billing ingestion.

> Configure usage data redaction in your self-hosted Deepgram deployment to enhance privacy and comply with security requirements.