Platform Activity

0
Projects Protected
0
Total Queries Validated

Query Analytics

?
Safety Controls
These tables show all guardrails that validate user inputs and AI outputs.
  • Guardrails - The safety check being applied
  • Policy Violation Rate - How often this guardrail blocks/filters
  • Primary Safety Triggers - LLM-generated pattern analysis of violations
  • Examples - Hover to see top 5 blocked queries from logs
Note: Clicking "Summarize Patterns" will make 1 LLM API call to analyze failure patterns.

Safety Controls

Input Safety Controls

Guardrails Policy Violation Rate Primary Safety Triggers Examples

Output Safety Controls

Guardrails Policy Violation Rate Primary Safety Triggers Examples
🛡️ How RAI Works
1
User Input Received
The chatbot receives a user query that needs to be validated before processing.
2
Input Safety Check
RAI validates the input against 6 guardrails: PII, Toxicity, Jailbreak, Code Injection, Illegal Activity, and Fairness.
3
LLM Processing
If input passes all checks, it's sent to the LLM for generating a response.
4
Output Safety Check
RAI validates the LLM response against 4 guardrails: Toxicity, Fairness, Groundedness, and Relevance.
5
Safe Response Delivered
Only validated, safe responses are delivered to the user. Violations are blocked with a safe message.
🚀 How to Onboard Application
1
Create New Application
Click "New Application" and provide a unique name for your chatbot or AI application.
2
Configure Guardrails
Enable/disable specific safety checks and choose between LLM or Library-based validation methods.
3
Integrate RAI API
Use the /check-input and /check-output endpoints in your application with your username.
4
Monitor & Analyze
View logs, track policy violations, and analyze failure patterns from the dashboard.
5
Optimize Configuration
Refine your guardrail settings based on insights to balance safety and user experience.