Security Audit Dashboard

Real-time LLM input & output guard monitoring

Total Requests
-
Scanned inputs & responses
Pass Rate
-%
Queries marked safe
Blocked Requests
-
Threats blocked at boundary
Avg Risk Score
-
System-wide mean risk index

Threat Category Distribution

Recent Risk Index Trend

Recent Audit Transactions

Timestamp Prompt Response Safety status Risk Score Threat Category
Loading transactions...

Rules & Settings Editor

Customize guardrail parameters dynamically

Ollama Service Completion Config

Admin Portal Authentication Credentials

🔔 Alert Notifications Configuration

Configure real-time security alerts to Telegram and Slack when threats are blocked.

🧠 Semantic Vector Blocklist (Sentence Similarity)

Blocks user prompts that have a high semantic similarity to a database of known jailbreaks, even if they bypass regex patterns.

0.82
Blocklisted Phrase Action

🔍 Semantic Similarity Tester

Enter a prompt to calculate its similarity against the active blocklist. Does not trigger real block events.

Jailbreak Regex Rules

Rule Name Regex Pattern Description Action

Output Toxicity Keywords

Enter keywords or phrases (comma-separated) to monitor and block in model responses.