Overview
Semantic Router provides content safety features to protect against malicious inputs, sensitive data exposure, and adversarial attacks at the routing layer.
Core Concepts​
PII Detection​
Automatically detects and protects personally identifiable information in user queries.
Jailbreak Protection​
Detects and blocks adversarial prompts and prompt injection attempts.
Key Features​
- Real-time Protection: Analyzes requests before they reach LLM endpoints
- Model-specific Policies: Configure different PII policies for different models
- Automatic Filtering: Models that don't meet security requirements are filtered out
- Comprehensive Logging: Complete audit trail of all security decisions