Overview
The Input Sanitizer API returns a redacted result, safe finding metadata, a health score, and an allow, review, or block verdict. It is designed to stop secrets, unsafe files, PII, and high-confidence prompt injection before model access.
What it does
- Redacts sensitive information such as tokens, credentials, and payment cards
- Detects and removes high-confidence prompt injection
- Sanitizes text, JSON, and supported PDF content
- Returns health scores and allow, review, or block verdicts